[Skip to Content]
[Skip to Content Landing]
Figure 1.
A schematic representation of the stepwise progress of a nevus to a melanoma. Five distinct evolutionary stages, including benign nevus, dysplastic nevus, radial growth-phase (RGP) melanoma, vertical growth-phase (VGP) melanoma, and metastatic melanoma, are represented. Estimated frequencies are depicted as percentages; unknown ones, as question marks.

A schematic representation of the stepwise progress of a nevus to a melanoma. Five distinct evolutionary stages, including benign nevus, dysplastic nevus, radial growth-phase (RGP) melanoma, vertical growth-phase (VGP) melanoma, and metastatic melanoma, are represented. Estimated frequencies are depicted as percentages; unknown ones, as question marks.

Figure 2.
Multiple signal requirements in metastasis. Subsequent to neoplastic transformation (eg, through activation of oncogenes, inactivation of tumor suppressors, viral transformation), the neoplastic cell needs to acquire optimal and directional cell migration to achieve metastasis. Multiple signaling and regulatory pathways that spatially and temporally integrate these processes characterize the metastatic phenotype. They can be broadly classified into the phases of initiation or activation of movement and regulation of speed and directionality. The events in these 2 phases are driven in response to a variety of factors, including gradients of chemokines, growth factors, or extracellular matrix molecules. These factors engage cell surface receptors, initiating a cascade of events inside and outside the neoplastic cell.

Multiple signal requirements in metastasis. Subsequent to neoplastic transformation (eg, through activation of oncogenes, inactivation of tumor suppressors, viral transformation), the neoplastic cell needs to acquire optimal and directional cell migration to achieve metastasis. Multiple signaling and regulatory pathways that spatially and temporally integrate these processes characterize the metastatic phenotype. They can be broadly classified into the phases of initiation or activation of movement and regulation of speed and directionality. The events in these 2 phases are driven in response to a variety of factors, including gradients of chemokines, growth factors, or extracellular matrix molecules. These factors engage cell surface receptors, initiating a cascade of events inside and outside the neoplastic cell.

Figure 3.
Differential gene profiling of clinical melanoma metastases. A, Significance analysis of microarrays (SAM) plot of human cancer complementary DNA (cDNA) array data. In this data set we chose the limit as 0.675, producing 174 significant genes (spots indicate significant genes among the up-regulated [red] and down-regulated [green] genes). B, Unsupervised hierarchical clustering and supervised functional categorization of significant genes from human cancer cDNA array data. Up-regulated and down-regulated clusters are shown as red and green dendrograms, respectively. The tables give gene names, gene identification number, fold change, and the biological function of the cluster components. ATP indicates adenosine triphosphate. C, SAM plot of human apoptosis cDNA array data. In this data set we chose the limit as 0.467, producing 102 significant genes (spots indicate significant genes among the up-regulated [red] and down-regulated [green] genes). D, Unsupervised hierarchical clustering and supervised functional categorization of gene expression of significant genes from human apoptosis cDNA array data. Up-regulated and down-regulated clusters are shown as red and green dendrograms, respectively. Grb indicates growth factor receptor–bound protein; G1, G1 phase.

Differential gene profiling of clinical melanoma metastases. A, Significance analysis of microarrays (SAM) plot of human cancer complementary DNA (cDNA) array data. In this data set we chose the limit as 0.675, producing 174 significant genes (spots indicate significant genes among the up-regulated [red] and down-regulated [green] genes). B, Unsupervised hierarchical clustering and supervised functional categorization of significant genes from human cancer cDNA array data. Up-regulated and down-regulated clusters are shown as red and green dendrograms, respectively. The tables give gene names, gene identification number, fold change, and the biological function of the cluster components. ATP indicates adenosine triphosphate. C, SAM plot of human apoptosis cDNA array data. In this data set we chose the limit as 0.467, producing 102 significant genes (spots indicate significant genes among the up-regulated [red] and down-regulated [green] genes). D, Unsupervised hierarchical clustering and supervised functional categorization of gene expression of significant genes from human apoptosis cDNA array data. Up-regulated and down-regulated clusters are shown as red and green dendrograms, respectively. Grb indicates growth factor receptor–bound protein; G1, G1 phase.

Figure 4.
Real-time reverse transcriptase polymerase chain reaction (PCR) detection of c-met using TaqMan gene expression assay (Applied Biosystems Inc, Foster City, Calif). To confirm complementary DNA arrays, ABI prism 7300 (Applied Biosystems Inc) was performed. The bar graph shows expression of normalized c-met compared with normal human melanocytes (NHM) in 3 different melanoma cell lines (BLM, MV3, and M13) and 4 melanoma metastases. c-met was up-regulated in most investigated samples, with NHM used as the calibrator.

Real-time reverse transcriptase polymerase chain reaction (PCR) detection of c-met using TaqMan gene expression assay (Applied Biosystems Inc, Foster City, Calif). To confirm complementary DNA arrays, ABI prism 7300 (Applied Biosystems Inc) was performed. The bar graph shows expression of normalized c-met compared with normal human melanocytes (NHM) in 3 different melanoma cell lines (BLM, MV3, and M13) and 4 melanoma metastases. c-met was up-regulated in most investigated samples, with NHM used as the calibrator.

Figure 5.
A schematic integration of the complementary DNA (cDNA) microarray data to construct pathways. Our hypothesis was that growth factor receptor–bound protein (Grb10) may have a critical regulatory function in melanoma metastasis by promoting insulin-like growth factor I receptor (IGF-IR)–independent Raf/mitogen-activated protein kinase (MAPK) signaling to elicit cell cycle progression, proliferation, and metastasis through the presence of c-met/Grb10–mediated activation of the Raf/(MEK)/extracellular-regulated kinase (ERK) pathway in certain subsets of cutaneous melanoma metastases. The genes marked by an asterisk have been identified using cDNA arrays (depicted in Figure 3B). HGFR indicates hepatocyte growth factor receptor; HGF/SF, hepatocyte growth factor/scatter factor.

A schematic integration of the complementary DNA (cDNA) microarray data to construct pathways. Our hypothesis was that growth factor receptor–bound protein (Grb10) may have a critical regulatory function in melanoma metastasis by promoting insulin-like growth factor I receptor (IGF-IR)–independent Raf/mitogen-activated protein kinase (MAPK) signaling to elicit cell cycle progression, proliferation, and metastasis through the presence of c-met/Grb10–mediated activation of the Raf/(MEK)/extracellular-regulated kinase (ERK) pathway in certain subsets of cutaneous melanoma metastases. The genes marked by an asterisk have been identified using cDNA arrays (depicted in Figure 3B). HGFR indicates hepatocyte growth factor receptor; HGF/SF, hepatocyte growth factor/scatter factor.

Figure 6.
Western blot analysis of growth factor receptor–bound protein 10 (Grb10) and phosphorylated Grb10 (P-Grb10). To confirm protein expression, protein extracts of normal human melanocytes (NHM), different melanoma cell lines, and 4 melanoma metastases were probed with Grb10 and P-Grb10 antibodies. The Grb10 and P-Grb10 were up-regulated in most samples compared with NHM. Reprinted with permission from Mirmohammadsadegh et al. Copyright 2004, Karger Publishers, Basel, Switzerland.

Western blot analysis of growth factor receptor–bound protein 10 (Grb10) and phosphorylated Grb10 (P-Grb10). To confirm protein expression, protein extracts of normal human melanocytes (NHM), different melanoma cell lines, and 4 melanoma metastases were probed with Grb10 and P-Grb10 antibodies. The Grb10 and P-Grb10 were up-regulated in most samples compared with NHM. Reprinted with permission from Mirmohammadsadegh et al.45 Copyright 2004, Karger Publishers, Basel, Switzerland.

Figure 7.
Western blot analysis of extracellular-regulated kinase 1/2 (ERK1/2) and phosphorylated ERK1/2 (P-ERK1/2). The ERK1/2 and P-ERK1/2 (double bands corresponding to p42 and p44) were up-regulated in all samples compared with normal human melanocytes (NHM).

Western blot analysis of extracellular-regulated kinase 1/2 (ERK1/2) and phosphorylated ERK1/2 (P-ERK1/2). The ERK1/2 and P-ERK1/2 (double bands corresponding to p42 and p44) were up-regulated in all samples compared with normal human melanocytes (NHM).

Table 1. 
Different Microarrays
Different Microarrays
Table 2. 
A Summary of Various Steps Involved in Microarray Analysis
A Summary of Various Steps Involved in Microarray Analysis
1.
Fodor  SPRead  JLPirrung  MCStryer  LLu  ATSolas  D Light-directed spatially addressable chemical synthesis Science 1991;251767- 773
PubMedArticle
2.
Fodor  SPRava  RPHuang  XCPease  ACHolmes  CPAdams  CL Multiplexed biochemical assays with biological chips Nature 1993;364555- 556
PubMedArticle
3.
Schena  MShalon  DDavis  RWBrown  PO Quantitative monitoring of gene-expression patterns with a complementary-DNA microarray Science 1995;270467- 470
PubMedArticle
4.
Lockhart  DJDong  HByrne  MC  et al.  Expression monitoring by hybridization to high-density oligonucleotide arrays Nat Biotechnol 1996;141675- 1680
PubMedArticle
5.
Eickhoff  HKonthur  ZLueking  A  et al.  Protein array technology: the tool to bridge genomics and proteomics Adv Biochem Eng Biotechnol 2002;77103- 112
PubMed
6.
Schraml  PKononen  JBubendorf  L  et al.  Tissue microarrays for gene amplification surveys in many different tumor types Clin Cancer Res 1999;51966- 1975
PubMed
7.
Rao  JSeligson  DHemstreet  GP Protein expression analysis using quantitative fluorescence image analysis on tissue microarray slides Biotechniques 2002;32924- 932
PubMed
8.
Pinkel  DSegraves  RSudar  D  et al.  High resolution analysis of DNA copy number variation sing comparative genomic hybridization to microarrays Nat Genet 1998;20207- 211
PubMedArticle
9.
Snijders  AMNowak  NSegraves  R  et al.  Assembly of microarrays for genome-wide measurement of DNA copy number Nat Genet 2001;29263- 264
PubMedArticle
10.
Marks  RWhiteman  D Sunburn and melanoma: how strong is the evidence? BMJ 1994;30875- 76
PubMedArticle
11.
Whiteman  DCWhiteman  CAGreen  AC Childhood sun exposure as a risk factor for melanoma: a systematic review of epidemiologic studies Cancer Causes Control 2001;1269- 82
PubMedArticle
12.
Clark  WH  JrFrom  LBernardino  EAMihm  MC The histogenesis and biologic behavior of primary human malignant melanomas of the skin Cancer Res 1969;29705- 727
PubMed
13.
Breslow  A Thickness, cross-sectional areas and depth of invasion in the prognosis of cutaneous melanoma Ann Surg 1970;172902- 908
PubMedArticle
14.
Breslow  A Tumor thickness, level of invasion and node dissection in stage I cutaneous melanoma Ann Surg 1975;182572- 575
PubMedArticle
15.
Balch  CMMurad  TMSoong  SJIngalls  ALHalpern  NBMaddox  WA A multifactorial analysis of melanoma: prognostic histopathological features comparing Clark’s and Breslow’s staging methods Ann Surg 1978;188732- 742
PubMedArticle
16.
Armstrong  SAStaunton  JESilverman  LB  et al.  MLL translocations specify a distinct gene expression profile that distinguishes a unique leukemia Nat Genet 2002;3041- 47
PubMedArticle
17.
Golub  TRSlonim  DKTamayo  P  et al.  Molecular classification of cancer: class discovery and class prediction by gene expression monitoring Science 1999;286531- 537
PubMedArticle
18.
Bittner  MMeltzer  PChen  Y  et al.  Molecular classification of cutaneous malignant melanoma by gene expression profiling Nature 2000;406536- 540
PubMedArticle
19.
Bhattacharjee  ARichards  WGStaunton  J  et al.  Classification of human lung carcinomas by mRNA expression profiling reveals distinct adenocarcinoma subclasses Proc Natl Acad Sci U S A 2001;9813790- 13795
PubMedArticle
20.
Pomeroy  SLTamayo  PGaasenbeek  M  et al.  Prediction of central nervous system embryonal tumour outcome based on gene expression Nature 2002;415436- 442
PubMedArticle
21.
Kamb  AGruis  NWeaver-Feldhaus  J  et al.  A cell cycle regulator potentially involved in genesis of many tumor types Science 1994;264436- 444
PubMedArticle
22.
Cairns  PPolascik  TEby  Y  et al.  Frequency of homozygous deletion at p16/CDKN2 in primary human tumours Nat Genet 1995;11210- 212
PubMedArticle
23.
Sherr  CJ Tumor surveillance via the ARF-p53 pathway Genes Dev 1998;122984- 2991
PubMedArticle
24.
Rizos  HDarmanian  APHolland  EAMann  GJKefford  RF Mutations in the INK4a/ARF melanoma susceptibility locus functionally impair p14ARF J Biol Chem 2001;27641424- 41434
PubMedArticle
25.
Walker  GJHayward  NK Pathways to melanoma development: lessons from the mouse J Invest Dermatol 2002;119783- 792
PubMedArticle
26.
Bardeesy  NBastian  BCHezel  APinkel  DDePinho  RAChin  L Dual inactivation of RB and p53 pathways in RAS-induced melanomas Mol Cell Biol 2001;212144- 2153
PubMedArticle
27.
Gorden  AOsman  IGai  W  et al.  Analysis of BRAF and N-RAS mutations in metastatic melanoma tissues Cancer Res 2003;633955- 3957
PubMed
28.
Cruz  FRubin  BPWilson  D  et al.  Absence of BRAF and NRAS mutations in uveal melanoma Cancer Res 2003;635761- 5766
PubMed
29.
Weber  AHengge  URUrbanik  D  et al.  Absence of mutations of the BRAF gene and constitutive activation of extracellular-regulated kinase in malignant melanomas of the uvea Lab Invest 2003;831771- 1776
PubMedArticle
30.
Clark  EAGolub  TRLander  ESHynes  RO Genomic analysis of metastasis reveals an essential role for RhoC Nature 2000;406532- 535
PubMedArticle
31.
Weeraratna  ATJiang  YHostetter  G  et al.  Wnt5a signaling directly affects cell motility and invasion of metastatic melanoma Cancer Cell 2002;1279- 288
PubMedArticle
32.
Tusher  VGTibshirani  RChu  G Significance analysis of microarrays applied to the ionizing radiation response Proc Natl Acad Sci U S A 2001;985116- 5121[published correction appears inProc Natl Acad Sci U S A 2001;9810515
PubMedArticle
33.
Dysvik  BJonassen  I J-Express: exploring gene expression data using Java Bioinformatics 2001;17369- 370
PubMedArticle
34.
SOURCE, Stanford University Web site Available athttp://source.stanford.edu/cgi-bin/source/sourceSearchAccessed April 24, 2004
35.
KEGG, Kyoto Encyclopedia of Genes and Genomes Available athttp://www.genome.jp/kegg/kegg.htmlAccessed April 24, 2004
36.
 Signaling pathway database Available athttp://www.grt.kyushu-u.ac.jp/spad/index.htmlAccessed April 24, 2004
37.
 Cell signaling networks database Available athttp://geo.nihs.go.jp/csndb.htmlAccessed April 24, 2004
38.
 Human protein reference database Available athttp://www.hprd.orgAccessed April 24, 2004
39.
 Biocarta Available athttp://www.biocarta.com/genes/allpathways.aspAccessed April 24, 2004
40.
Furge  KAZhang  YWVande Woude  GF Met receptor tyrosine kinase: enhanced signaling through adapter proteins Oncogene 2000;195582- 5589
PubMedArticle
41.
Jeffers  MRong  SWoude  GF Hepatocyte growth factor/scatter factor-Met signaling in tumorigenicity and invasion/metastasis J Mol Med 1996;74505- 513
PubMedArticle
42.
Halaban  RRubin  JSFunasaka  Y  et al.  Met and hepatocyte growth factor/scatter factor signal transduction in normal melanocytes and melanoma cells Oncogene 1992;72195- 2206
PubMed
43.
Otsuka  TTakayama  HSharp  R  et al.  c-Met autocrine activation induces development of malignant melanoma and acquisition of the metastatic phenotype Cancer Res 1998;585157- 5167
PubMed
44.
Natali  PGNicotra  MRDi Renzo  MF  et al.  Expression of the c-Met/HGF receptor in human melanocytic neoplasms: demonstration of the relationship to malignant melanoma tumour progression Br J Cancer 1993;68746- 750
PubMedArticle
45.
Mirmohammadsadegh  ABaer  ANambiar  SBardenheuer  WHengge  UR Rapid identification of dysregulated genes in cutaneous malignant melanoma metastases using cDNA technology Cells Tissues Organs 2004;177119- 123
PubMedArticle
46.
Ooi  JYajnik  VImmanuel  D  et al.  The cloning of Grb10 reveals a new family of SH2 domain proteins Oncogene 1995;101621- 1630
PubMed
47.
Nantel  AMohammad-Ali  KSherk  JPosner  BIThomas  DY Interaction of the Grb10 adapter protein with the Raf1 and MEK1 kinases J Biol Chem 1998;27310475- 10484
PubMedArticle
48.
Jahn  TSeipel  PUrschel  SPeschel  CDuyster  J Role for the adaptor protein Grb10 in the activation of Akt Mol Cell Biol 2002;22979- 991
PubMedArticle
49.
Morrione  APlant  PValentinis  B  et al.  mGrb10 interacts with Nedd4 J Biol Chem 1999;27424094- 24099
PubMedArticle
50.
O’Neill  TJRose  DWPillay  TSHotta  KOlefsky  JMGustafson  TA Interaction of a GRBIR splice variant (a human GRB10 homolog) with the insulin and insulin-like growth factor I receptors: evidence for a role in mitogenic signaling J Biol Chem 1996;27122506- 22513
PubMedArticle
51.
Wang  JDai  HYousaf  N  et al.  Grb10, a positive, stimulatory signaling adapter in platelet-derived growth factor BB–, insulin-like growth factor I–, and insulin-mediated mitogenesis Mol Cell Biol 1999;196217- 6228
PubMed
52.
Liu  FRoth  RA GrbIR: a SH2-domain–containing protein that binds to the insulin receptor and inhibits its function Proc Natl Acad Sci U S A 1995;9210287- 10291
PubMedArticle
53.
Liu  FRoth  RA Binding of SH2 containing proteins to the insulin receptor: a new way for modulating insulin signalling Mol Cell Biochem 1998;18273- 78
PubMedArticle
54.
Stein  EGGustafson  TAHubbard  SR The BPS domain of Grb10 inhibits the catalytic activity of the insulin and IGF1 receptors FEBS Lett 2001;493106- 111
PubMedArticle
55.
Morrione  AValentinis  BResnicoff  MXu  SBaserga  R The role of mGrb10alpha in insulin-like growth factor I–mediated growth J Biol Chem 1997;27226382- 26387
PubMedArticle
56.
Langlais  PDong  LQRamos  FJ  et al.  Negative regulation of insulin-stimulated mitogen-activated protein kinase signaling by Grb10 Mol Endocrinol 2003;18350- 358
PubMedArticle
57.
Tschentscher  FHusing  JHolter  T  et al.  Tumor classification based on gene expression profiling shows that uveal melanomas with and without monosomy 3 represent two distinct entities Cancer Res 2003;632578- 2584
PubMed
58.
Alonso  SROrtiz  PPollan  M  et al.  Progression in cutaneous malignant melanoma is associated with distinct expression profiles: a tissue microarray–based study Am J Pathol 2004;164193- 203
PubMedArticle
59.
St Croix  BRago  CVelculescu  V  et al.  Genes expressed in human tumor endothelium Science 2000;2891197- 1202
PubMedArticle
60.
Macaulay  VMSalisbury  AJBohula  EAPlayford  MPSmorodinsky  NIShiloh  Y Downregulation of the type 1 insulin-like growth factor receptor in mouse melanoma cells is associated with enhanced radiosensitivity and impaired activation of Atm kinase Oncogene 2001;204029- 4040
PubMedArticle
61.
Lee  MLKuo  FCWhitmore  GASklar  J Importance of replication in microarray gene expression studies: statistical methods and evidence from repetitive cDNA hybridizations Proc Natl Acad Sci U S A 2000;979834- 9839
PubMedArticle
62.
Yang  YHBuckley  MJSpeed  TP Analysis of cDNA microarray images Brief Bioinform 2001;2341- 349
PubMedArticle
63.
DeRisi  JPenland  LBrown  PO  et al.  Use of a cDNA microarray to analyse gene expression patterns in human cancer Nat Genet 1996;14457- 460
PubMedArticle
64.
Eisen  MBSpellman  PTBrown  POBotstein  D Cluster analysis and display of genome-wide expression patterns Proc Natl Acad Sci U S A 1998;9514863- 14868
PubMedArticle
65.
Alon  UBarkai  NNotterman  DA  et al.  Patterns of gene expression revealed by clustering analysis of tumor and normal colon tissues probed by oligonucleotide arrays Proc Natl Acad Sci U S A 1999;966745- 6750
PubMedArticle
66.
Chen  YDougherty  ERBittner  ML Ratio-based decisions and the quantitative analysis of cDNA microarray images J Biomed Opt 1997;2367- 374Article
67.
Newton  MAKenziorski  CMRichmond  CSBlattner  FRTsui  KW On differential variability of expression ratios: improving statistical inference about gene expression changes from microarray data J Comput Biol 2001;837- 52
PubMedArticle
68.
Antoniadis  ALambert-Lacroix  SLeblanc  F Effective dimension reduction methods for tumor classification using gene expression data Bioinformatics 2003;19563- 570
PubMedArticle
69.
Kerr  MKChurchill  GA Bootstrapping cluster analysis: assessing the reliability of conclusions from microarray experiments Proc Natl Acad Sci U S A 2001;988961- 8965
PubMedArticle
70.
Chuaqui  RFBonner  RFBest  CJ  et al.  Post-analysis follow-up and validation of microarray experiments Nat Genet 2002;32509- 514
PubMedArticle
71.
Rajeevan  MSRanamukhaarachchi  DGVernon  SDUnger  ER Use of real-time quantitative PCR to validate the results of cDNA array and differential display PCR technologies Methods 2001;25443- 451
PubMedArticle
Study
February 2005

Signaling Networks in Cutaneous Melanoma Metastasis Identified by Complementary DNA Microarrays

Author Affiliations

Author Affiliations: Department of Dermatology (Drs Nambiar, Mirmohammadsadegh, Doroudi, Marini, Ruzicka, and Hengge and Ms Gustrau) and Institute of Clinical Oncology (Dr Roeder), Heinrich-Heine-University, Düsseldorf, Germany.

Arch Dermatol. 2005;141(2):165-173. doi:10.1001/archderm.141.2.165
Abstract

Background  Melanoma is a complex multigenic disease, susceptibility to which is determined by several parallel and stepwise progressive pathways affecting growth control, differentiation, cell adhesion, and survival. Melanoma and human cancers in general undergo a continuous development from benign to malignant states, as most thoroughly documented in the multistep mole-to-melanoma transition.

Objective  To examine how high-throughput microarrays are being used in expression profiling to identify regulated genes, patterns, and pathways that may lead to functional characterization and tumor subclassification.

Design  Ten melanoma metastases were analyzed by DNA array technology for important regulated candidate genes, with subsequent confirmation by real-time reverse transcription polymerase chain reaction.

Results  Hepatocyte growth factor receptor c-met, growth factor receptor–bound protein 10, B-raf proto-oncogene, and several mitogen-activated protein kinase kinase genes were significantly up-regulated in melanoma metastases and several melanoma cell lines relative to normal human melanocytes (P = .03). Among the up-regulated genes, phosphorylated growth factor receptor–bound protein 10 is known to serve a molecular switch turning on the mitogen-activated protein kinase pathway in response to hepatocyte growth factor receptor binding.

Conclusions  As suggested by the DNA arrays, we found the mitogen-activated protein kinase kinase/extracellular-regulated kinase pathway to be activated in most of the cutaneous melanoma metastasis specimens. These findings are in the context of the current microarray technology in melanoma research. Additional steps are needed to gain insights into the pluralistic signaling milieu of this malignancy as we enter the postgenomic era.

Microarray platforms share in essence the same principle—a miniaturized slide that carries multiple probes. They differ, however, in the nature of the arrayed probe and the way the probes are created (Table 1). There are several variants of the DNA microarray technology for analysis of transcription profiles, including complementary DNA (cDNA) (>250 base pairs long), which are immobilized to a solid surface such as glass or nylon membrane using robotic spotting, and oligonucleotides (20-80 mer oligodeoxynucleotides), which are synthesized in situ (on chip) or by conventional methods followed by on-chip immobilization.14 Protein/peptide arrays are solid-phase ligand-binding assays that use immobilized proteins/peptides on surfaces, including glass, membranes, microtiter wells, mass spectrometer plates, and beads or other particles (Table 1). Protein arrays enable the parallel screening of thousands of protein interactions that encompass protein-antibody, protein-protein, protein-ligand, and protein-drug interactions; enzyme-substrate screening; and multianalytic diagnostic assays.5 Tissue microarrays contain tissues arrayed on modified glass slides. This microarray technology allows the rapid visualization of molecular targets in hundreds of tissue specimens at the same time, at the DNA, the RNA, or the protein level (Table 1). In particular, the signal is visualized on a per-cell basis (in situ).6,7 By revealing the cellular localization, prevalence, and clinical significance of candidate genes, tissue microarrays are ideally suited for genomic-based diagnostic and drug-target discovery. In recent years, comparative genomic hybridization has also emerged as a novel array-based platform for genome-wide comprehensive analysis of chromosomal imbalances.8,9 Thus, array technologies represent a spectrum of high-throughput means to identify molecular targets associated with biological and clinical phenotypes by comparing samples representative of distinct pathophysiological states. However, irrespective of the type of the array platform, an appropriate experimental design, accurately annotated clinical specimens, and proper controls are key to the successful use.

STEPWISE TUMORIGENESIS MODEL OF MELANOMA DEVELOPMENT AND PROGRESSION

Melanoma is an ideal model to study the molecular mechanisms of tumor progression, as it develops in about half of the cases through a series of architectural and phenotypically distinct stages and becomes progressively aggressive, culminating in metastasis (Figure 1).

Five distinct stages have been proposed in the evolution of melanoma on the basis of histological criteria: (1) common acquired and congenital nevi without dysplastic changes; (2) dysplastic nevi with structural and architectural atypia; (3) radial growth-phase melanoma; (4) vertical growth-phase melanoma; and (5) metastatic melanoma. Considerable progress has been made in understanding the biological, pathological, and immunological aspects of human melanoma progression. Epidemiological and experimental studies have suggested that intense exposures to UV radiation during early childhood may lead to melanoma in adults,10,11 but molecular and genetic studies have shown few autosomal abnormalities, infrequent mutational spectra, and very few epistatic and epigenetic mechanisms. At the cellular level, it has become clear that dysregulated homeostatic control of the cutaneous microenvironment occurs through alterations in the expression of specific proteins. These include growth factors and their receptors, adhesion molecules and their ligands, proteases and their substrates, and transcription factors and their target genes. As in most other human tumors, there are significant alterations in the regulatory network involved in signal transduction in human melanoma.

The prognosis of melanoma is based on histological criteria such as tumor thickness (measured by the Breslow index), level of invasion (Clark level), presence of ulceration, and number of mitoses per square millimeter.1215 However, these variables do not provide a precise prognosis in all cases; thin melanomas may develop metastases, and thick melanomas may remain localized for many years. For these reasons, the search for other prognostic factors is ongoing. Traditional techniques like Northern blot analysis, differential display, and serial analysis of gene expression have proved inadequate to provide a comprehensive image of these pathophysiological state transitions. A global view of the malignant genome, transcriptome, and proteome will be valuable, not only for understanding this neoplastic transformation but also for having profound effects in improving diagnosis, predicting clinical outcome, and responding to therapeutic intervention of a given candidate. Gene expression profiling could define, at the molecular level, the clinical and histopathological phenotypes of given tumors. Moreover, the tumor subclasses are likely to reflect the basic differences in cell biology of tumor cells. Expression profiling and pattern identification using various available microarray platforms are becoming novel tools for identifying new cancer classes (class discovery; eg, the discovery of a novel class of lymphoblastic leukemia16) or for assigning tumors to known classes (class prediction; eg, human acute leukemia,17 cutaneous malignant melanoma,18 lung carcinomas,19 and embryonal tumors of the central nervous system20).

APPLYING MICROARRAYS FOR FUNCTIONAL CLASSIFICATION OF GENES AND PATHWAY DISCOVERY

Biological signaling pathways interact with each other to form complex networks. If we want to understand cellular behavior and its responses to external signals or to influence cellular function in a predictable manner, we have to understand the pathways through which these signals are mediated into and within the cell. In most cases, changes in cellular behavior involve the abrogation or the block of transcriptional events, which are specific for each signal in its cellular context. Hence, expression profiling can be used for the functional classification of dysregulated genes. This is often referred to as “guilt by association.” This method is based on the observation that genes with related expression patterns—genes that presumably are coregulated—are likely to be functionally related and involved in the same biological processes or physiological pathways. Disruption of the INK4a/ARF tumor suppressor locus is a common genetic alteration in many different types of human cancers, including malignant melanoma.2124 Activation of receptor tyrosine kinase pathways as well as Ras and phosphatidylinositol 3-kinase has been described in subsets of human melanomas.25 Using a mouse model for cutaneous melanoma, Bardeesy et al26 investigated whether activated Ras can cooperate with p53 loss to cause melanoma, whether such melanomas are biologically comparable to those arising in INK4a (delta2/3−/−) mice, and whether tumor-associated mutations emerge in the p16(INK4a)-Rb pathway in those melanomas. These studies validated the role of p53 inactivation in melanoma development and suggested that the Rb and p53 pathways function to suppress melanocyte transformation in vivo, which had previously been disputed. Similarly, the BRAF (B-raf proto-oncogene) mutation is common in cutaneous melanoma metastases. These mutations include exons 11 and 15 and suggest that anti-Ras/Raf strategies may be effective for treating metastatic melanoma27 ; however, these mutations were absent in uveal melanoma.28

Earlier genomic and proteomic investigations by our group showed that mitogen-activated protein kinase kinase (MEK)–1, MEK2, and MEK3, as well as extracellular-regulated kinase (ERK) 1/2 (p44/42) were highly up-regulated in subsets of uveal melanoma metastases without the prerequirement of activating BRAF mutations.29 Results from our recent studies suggest that subgroups of cutaneous malignant melanoma metastases have alternate pathways of regulating extracellular matrix and thereby show variation in their metastatic potential (S.N., A.M., A.G., U.R.H., unpublished data, December 2002). Earlier expression profiling studies have shown a putative role for RhoC30 and Wnt5a31 in cell motility and invasiveness of melanoma. However, further investigation is required to unravel when and how the metastatic cell makes such decisions (Figure 2) and what its impact would be on diagnosis and therapy.

For elucidating signal transduction pathways in melanoma, we analyzed differential gene expression between normal human melanocytes (NHMs) and 10 cutaneous melanoma metastases of different patients with the use of cDNA array technology focusing on oncogenesis and apoptosis.

METHODS
CELL CULTURE

Normal human melanocytes were obtained from PromoCell (Heidelberg, Germany) and cultured in phorbol myristate acetate–free medium (PromoCell). Cell culture was maintained in a 37°C incubator and a moist atmosphere of 5% carbon dioxide. The human melanoma cell lines BLM, M13, MV3, and SK-Mel-28 were cultured in Dulbecco Modified Eagle Medium (Gibco BRL Life Technologies, Taufkirchen, Germany) supplemented with 10% fetal calf serum, 100 U/mL of penicillin, and 100 pg/mL of streptomycin.

TUMOR SPECIMENS

The cutaneous melanoma metastases came from different body sites and had diameters greater than 5 mm. Biopsy specimens were snap frozen in liquid nitrogen and homogenized using a dismembrator (Braun AG, Melsungen, Germany). One half of the homogenate was transferred into ice-cold lysis buffer containing 25mM N-2-hydroxyethylpiperazine-N′-2-ethanesulfonic acid (pH, 7.9), 50mM sodium fluoride, 15mM Triton X-100, 5mM EDTA, 100mM sodium chloride, and 1 tablet of protease inhibitor cocktail per 10mL of buffer (Roche, Mannheim, Germany) for Western blot analysis. The other half of the pulverized specimen was immediately transferred into TriZol reagent (Invitrogen, Karlsruhe, Germany) for total RNA isolation according to the manufacturer’s instructions. Contaminating DNA was removed by treatment with DNase I (Invitrogen). The RNA quality was controlled by agarose gel electrophoresis.

MICROARRAY HYBRIDIZATION

Several membrane arrays containing apoptosis- and cancer-related genes (Atlas human apoptosis cDNA array and Atlas human cancer cDNA array 1.2K; Clontech, Heidelberg, Germany) were hybridized comparing NHMs, different melanoma cell lines, and patient-derived metastases. After DNase treatment, 1 μg of total RNA was reverse transcribed using superscript (Invitrogen) and SMART cDNA primers (Clontech). We used 100 ng of SMART cDNA for long-distance reverse transcription polymerase chain reaction (RT-PCR) with SMART primers. The PCR reaction consisted of 21 cycles and was within the linear amplification range. The reaction mix was removed from the PCR product with the use of Microcon-100 columns (Millipore, Tokyo, Japan). We used random-primed labeling for 100 ng of SMART-PCR products with phosphorus 33-α deoxycytidine triphosphate (ICN, Eschwege, Germany). The 33P-labeled probes were purified with the use of Microcon-100 columns. Prehybridization and hybridization of membrane microarrays were performed in DIG Easy Hyb solution (Roche) at 42°C.

MICROARRAY IMAGE AND DATA ANALYSIS

The radioactive intensity of each spot was detected by a phosphoimager (BAS-1500; Fuji, Raytest GmbH, Straubenhardt, Germany) and recorded with AIDA software (Raytest, Germany). To compare the results, normalization of each spot was performed. As reference spots, 9 different housekeeping genes were used. The ratio of gene expression between the experimental samples and NHM was computed. No gene regulation was present at a ratio of 1, and an up- or a down-regulation was observed when the normalized ratio was greater than or equal to or less than 2, respectively.

STATISTICAL EVALUATION

We used significance analysis of microarrays,32 which identifies genes with statistically significant changes in expression by assimilating a set of gene-specific unpaired t tests. Each gene is assigned a score on the basis of its change relative to the standard deviation of repeated measurements for that gene’s expression level. Genes with scores greater than a threshold are deemed potentially significant. The percentage of such genes identified by chance is the false discovery rate. To estimate the false discovery rate, nonsense genes are identified by analyzing permutations of the measurements. The threshold can be adjusted to identify smaller or larger sets of genes, and false discovery rates are calculated for each set. Subsequently, J-Express version 1.1,33 a Java software application, was used to perform the hierarchical agglomerative cluster analysis to group similar data (average linkage).

SUPERVISED DATA ANALYSIS

We also performed supervised data analyses, including data mining, functional categorization, and pathway construction. The statistically significant genes defined within meaningful clusters were queried for in the SOURCE databank, which contains gene reports that include aliases, chromosomal location, functional descriptions, gene ontology annotations, gene expression data, and links to external databases. Functional categorization based on gene ontology annotations and an analysis of the cause-consequence relationship were performed using available information on signaling pathways in cancer.3439 Finally, the results of expression profiling were mapped to known biological pathways to gain further functional insights.

QUANTITATIVE PCR BASED ON THE USE OF FLUOROGENIC PROBES

Briefly, in a 5′ nuclease assay, a specific oligonucleotide probe that is labeled with a reporter fluorescent dye at the 5′ end and a dark quencher at the 3′ end is annealed to the target sequence located between the 2 primer-binding sites. The 5′ exonuclease-mediated release of the reporter dye from the probe during the extension phase results in an increase in the fluorescent signal of the reporter dye. This is proportional to the amount of amplicon produced and is monitored in real time during PCR amplification using the PRISM 7300 sequence detector (Applied Biosystems Inc, Foster City, Calif). The primers and probes used for detection of c-met and 18S were ordered as TaqMan gene expression assay (Applied Biosystems Inc), and the PCR analysis was carried out in a 25-μL mixture. All samples were amplified simultaneously in triplicates. The data were analyzed by using a standard curve with 18S as the endogenous control.

WESTERN BLOT ANALYSIS

Cells were washed twice with ice-cold phosphate-buffered saline and lysed in lysis buffer. The samples were pretreated with ultrasound, and cell extracts were centrifuged at 14.000g for 20 minutes at 4°C. The protein concentration of the cellular extracts was determined using the advanced protein assay reagent (TEBU, Frankfurt, Germany). Twenty micrograms of the protein extract was electrophoresed on 4% to 12% NuPage Bis-Tris-Glycin gels (Invitrogen) for 2 hours at 120 V. Proteins were blotted onto polyvinylidene fluoride membranes (Bio-Rad, Munich, Germany) at 25 V for 90 minutes using a tank-blot system. The membranes were blocked with 5% nonfat dry milk powder in TBST buffer consisting of 10mM Tris hydrochloride (pH, 7.5), 150mM sodium chloride, and 0.05% polysorbate 20 (Tween 20) overnight at 4°C; washed 3 times with TBST buffer; and incubated with the appropriate primary anti–human rabbit polyclonal antibody (growth factor receptor–bound protein 10 [Grb10], phosphorylated Grb10, p42/44, and phosphorylated p42/44, all from Santa Cruz, Heidelberg, Germany) in TBST buffer with 1% nonfat dry milk overnight at 4°C. The membranes were washed 3 times with TBST buffer and incubated with the respective secondary antibody (antirabbit IgG, dilution 1:5000; Cell Signaling Technology, Beverly, Mass) and visualized using the chemiluminescence detection system (Pierce Biotechnology Inc, Rockford, Ill) following the supplier’s instructions.

RESULTS

For elucidating signal transduction pathways in melanoma, we analyzed differential gene expression between NHMs and cutaneous melanoma metastases using cDNA array technology focusing on oncogenesis and apoptosis. A selection of genes that demonstrate altered expression in a consistent pattern was selected after significance analysis of microarrays and unsupervised hierarchical clustering (Figure 3). In particular, c-met, GRB10, mitogen-activated protein kinase kinase 6 (MAPKK6), MAPKK7, E2F transcription factor 3, and several other promitotic genes were up-regulated in melanoma metastases (Figure 3). The c-met proto-oncogene encodes the receptor for the hepatocyte growth factor (HGF)/scatter factor, which is known to mediate mitogenic, migratory, and invasive effects in several different cell types.40 The ligand for c-met is HGF/scatter factor, and although normal HGF/scatter factor–Met signaling is required for embryonic development, abnormal c-met signaling has been strongly implicated in tumorigenesis, particularly in the development of invasive and metastatic phenotypes.41,42 After ligand binding and autophosphorylation, c-met transmits intercellular signals using a unique multisubstrate docking site contained within the C-terminal end of the receptor. This docking site mediates the binding of several adapter proteins such as Grb2, SHC, Crk/CRKL, and the large adapter protein Gab1. These adapter proteins in turn recruit several signal-transducing proteins to form an intricate signaling complex.40 The c-met gene is expressed at late stages in melanoma progression, suggesting that the presence of the Met/HGF receptor may contribute to the acquisition of an invasive phenotype.43,44

To confirm c-met expression at the transcript level, we used real-time RT-PCR in several melanoma cell lines (MV3, BLM, and M13) and in cutaneous melanoma metastases (Figure 4). In contrast, insulinlike growth factor I (IGF-1) receptor; IGF-2; IGF-binding proteins 2, 4, and 6; and other apoptotic genes were down-regulated (Figure 3).

Next, we analyzed the cause-consequence relationship of these genes within and across clusters using available published information on signaling pathways in cancer (Figure 5). Genes defined within clusters with meaningful profiles were queried for the SOURCE gene reports, which include aliases, chromosomal location, functional descriptions, gene ontology annotations, gene expression data, and several pathway-related databases.3439 This information is valuable to understand interrelations between different pathways and elute to translational research.

Given the up-regulation of GRB10 and its potential significance in melanoma pathogenesis, we confirmed the up-regulation of Grb10 in 4 analyzed melanoma cell lines (Sk-Mel 28, MV3, BLM, and M13) and 3 of 4 cutaneous melanoma metastases using Western blot analysis (Figure 6). Growth factor receptor–bound protein 10 belongs to a small family of proteins, including Grb7 and Grb14.46 These molecules are commonly referred to as adapter proteins, because they have no catalytic function but contain numerous protein-binding motifs; thus, they are predicted to mediate interactions between disparate proteins. It is thought that Grb10 brings together activated receptors with downstream signaling components of several pathways (eg, MEK-1 and -2 and Raf-147 and Akt48) and may also interact with protein ubiquitination.49 Whether Grb10 acts positively or negatively on cellular signaling has not been resolved.4951 Several studies have described the binding of Grb10 to activated tyrosine kinase receptors, such as c-met,50 the insulin receptor, the IGF-I receptor,50,52,53 and the receptors for epidermal growth factor.46 These interactions are mediated by the Src homology (SH)2 and the “between pleckstrin-homology and SH2” domains of Grb10. Our data suggest that transcriptional activation of c-met is correlated with phosphorylated Grb10 protein in the melanoma cell lines BLM, MV3, and M13 and in melanoma metastases. This may ultimately lead to activation of the Raf/MEK/ERK pathway as evidenced from elevated phosphorylated ERK1/2 protein levels (Figure 7), suggesting that HGF/c-met/Grb10/Raf/MEK/ERK may act as a pathway involved in cutaneous melanoma metastasis.43,44,47 Although phosphorylated Grb10 was associated with increased signaling through the MEK/ERK pathway, it is known to inhibit IGF receptor signaling,54,55 which is in accordance with our microarray data (Figure 3D).

In addition, phosphorylated Grb10 has also been linked to the phosphatidylinositol 3-kinase/protein kinase B/Akt signal transduction cascade,47 to the regulation of cellular responses to insulin, to IGF-I, and possibly to other hormones.49,53 However, published studies with overexpression of Grb10 in cultured cells have provided conflicting data of stimulatory vs inhibitory actions on mitogenesis and other hormone responses.5052,5456 The conflicting evidence does not appear to be attributable to different Grb10 isoforms or specific cell types and thus, at present, leaves the physiological role of Grb10 unclear.

COMMENT

Recent research on global transcript analysis has identified previously unrecognized subtypes of cutaneous melanoma and predicted phenotypic characteristics that may be of importance to disease progression.18 Bittner et al18 reported the discovery of 2 subsets of melanomas identified by mathematical analysis of gene expression. A series of uveal melanoma showed several highly repressed genes that were inversely correlated with the same genes from the major cluster of cutaneous melanoma samples, proving the different nature of these 2 entities. However, because the patients in this study were uniformly characterized by poor prognosis, future work should determine whether these new subsets discriminate patients with distinct natural histories. Recent studies further divided uveal melanoma into 2 distinct, previously unrecognized entities that were indistinguishable by clinicopathologic features.57 Similarly, patterns of gene expression that correlated with progression to metastatic disease have recently been defined.30 In particular, 2 sets of differentially expressed genes were identified. The first set is involved in extracellular matrix assembly and the second one in regulating the actin-based cytoskeleton.30

Along these lines, tissue microarray studies have identified expression profiles that distinguish specific progression stages in melanoma. One study has defined a combination of biological markers associated with shorter overall survival in patients with vertical growth-phase melanoma and provided prognosis prediction with 4 antibodies (Ki67, p16INK4a, p21CIP1, and Bcl-6) in an independent series of 72 vertical growth-phase melanoma specimens.58 Two potential problems with the application of microarray technology to primary human tumor specimens are tissue availability and cellular heterogeneity of samples. Archival formalin-fixed, paraffin-embedded tissue is a vast source of clinical material; however, standard processing methods limit the use of such tissue in high-throughput protein analysis. To overcome this limitation, more efficient methods for using formalin-fixed, paraffin-embedded tissue need to be developed. Second, the presence of variable levels of normal cells such as stroma, blood vessels, and lymphocytes could mask expression patterns attributable to cancer cells. For instance, the macular components of nevi and radial growth-phase tumors are very thin and contain limited numbers of tumor cells. Especially, the flat nature of these lesions usually predisposes them to contamination with nontumor cells. This has been overcome with the advent of laser capture microdissection. However, from an alternative perspective, the reductionist laser-capture microdissection approach of analyzing highly homogenous cell populations could miss important interactions between the carrier and stroma cells. According to this line of reasoning, an accurate analysis of cancer pathology must address the full mixture and interactions of cancer cells, surrounding stroma, infiltrating inflammatory cells, and vascular elements (ie, interstitial, intracellular, and intercellular compartments). These so-called normal cells associated with a neoplasm are clearly distinct from populations of normal cells residing in nontumor regions. Support for this contention is provided by the recent finding that tumor endothelial cells exhibit a unique signature.59 Hence, an argument for the genomic analysis of more heterogeneous cancer tissue samples is gaining momentum at present. Therefore, we have purposely used melanoma metastases that contain the melanoma cells within their natural stroma.

Given the subsets of cutaneous melanoma metastases with phosphorylated Grb10 and transcriptional up-regulation of c-met, B-RAF, Raf-1, MAPKK6, MAPKK7, and MEK2 and down-regulation of IGF-I receptor, IGF-I, and IGF-II, we suggest that Grb10 may have a critical regulatory function in melanoma metastasis in promoting IGF-I receptor–independent Raf/MAPK signaling to elicit cell cycle progression, proliferation, and metastasis.43,44,47,54,55 In addition, the presence of c-met/Grb10–mediated activation of the Raf/MEK/ERK pathway seems to override IGF-I receptor–mediated Raf/MAPK signaling in certain subsets of cutaneous melanoma metastases.43,44,47

Moreover, because IGF-I receptor down-regulation enhances the sensitivity to cytotoxic drugs and irradiation,60 this subset of cutaneous melanoma metastases could be more responsive to chemotherapy compared with melanoma metastases, where the IGF-I receptor plays a more central role. Further experiments will be required to address these issues in detail to uncover subsets of melanoma metastases with alternative pathway preferences leading to invasion and metastasis and to decipher the critical role of adaptor molecules as decision makers in the selection of those pathways. Microarray technology offers a rapid, high-throughput technology that may help to answer clinically important questions about cancer, such as which tumors will behave aggressively, which will remain indolent, and which are likely to respond to selected therapies.

Expression profiling using microarray platforms has several limitations and pitfalls that need to be considered during experimental design and analysis, to avoid false-positive and false-negative results (Table 2). Use of adequate technical and biological replicates is a prerequisite to any microarray experiment.61 Few replicates will identify a smaller percentage of differentially expressed genes. The inclusion of dye swap labeling is critical. The analysis of the microarray results represents the next challenge. First, the estimation of the differential expression of each gene from the digital image is performed.62 Exploratory graphical methods such as hierarchical clustering are already being used to assist in identifying structures within the data,6365 but further methods are needed, including seriation and multidimensional scaling. Appropriate statistical models also need to be refined.66,67 Issues of quality control arise, and the final inference needs to take into account the propagating of uncertainties that arise at each analytical step. Constructing a meaningful classifier based on microarray gene expression data has recently emerged as an important problem for cancer classification. Difficulties arise from the fact that each sample contains expression data of a vast number of genes that may interact with one another. The selection of a small number of critical genes is fundamental to correctly analyze the vast amount of data. It is essential to use a multivariate approach for data. Therefore, effective dimension reduction methods and multivariate approaches are key to tumor classification.68 Clustering has been the primary analytical tool used to define disease subtypes from microarray experiments of cancer. Assessing cluster reliability, therefore, poses another level of complication in analyzing microarray data.69 An issue of equal importance is the need for downstream validation of identified targets.70 Several techniques such as Northern blot analysis and real-time RT-PCR are being used to validate microarray experiments.71

The appropriate use of high-throughput DNA assay platforms offers an efficient means of identifying patterns of regulated genes and associated pathways, leading to functional characterization and tumor subclassification. If we transfer the power of recent biotechnology achievements to individual patients, we will not only benefit from an era of unprecedented opportunity for addressing scientific questions to uncover the molecular basis of individual cancers but also improve the treatment and most likely the outcomes of our patients with melanoma.

Back to top
Article Information

Correspondence: Sandeep Nambiar, PhD, Department of Dermatology, Heinrich-Heine-University, Moorenstrasse 5, D-40225 Düsseldorf, Germany (sandeep.nambiar@uni-duesseldorf.de).

Accepted for Publication: November 8, 2004.

Financial Disclosure: None.

References
1.
Fodor  SPRead  JLPirrung  MCStryer  LLu  ATSolas  D Light-directed spatially addressable chemical synthesis Science 1991;251767- 773
PubMedArticle
2.
Fodor  SPRava  RPHuang  XCPease  ACHolmes  CPAdams  CL Multiplexed biochemical assays with biological chips Nature 1993;364555- 556
PubMedArticle
3.
Schena  MShalon  DDavis  RWBrown  PO Quantitative monitoring of gene-expression patterns with a complementary-DNA microarray Science 1995;270467- 470
PubMedArticle
4.
Lockhart  DJDong  HByrne  MC  et al.  Expression monitoring by hybridization to high-density oligonucleotide arrays Nat Biotechnol 1996;141675- 1680
PubMedArticle
5.
Eickhoff  HKonthur  ZLueking  A  et al.  Protein array technology: the tool to bridge genomics and proteomics Adv Biochem Eng Biotechnol 2002;77103- 112
PubMed
6.
Schraml  PKononen  JBubendorf  L  et al.  Tissue microarrays for gene amplification surveys in many different tumor types Clin Cancer Res 1999;51966- 1975
PubMed
7.
Rao  JSeligson  DHemstreet  GP Protein expression analysis using quantitative fluorescence image analysis on tissue microarray slides Biotechniques 2002;32924- 932
PubMed
8.
Pinkel  DSegraves  RSudar  D  et al.  High resolution analysis of DNA copy number variation sing comparative genomic hybridization to microarrays Nat Genet 1998;20207- 211
PubMedArticle
9.
Snijders  AMNowak  NSegraves  R  et al.  Assembly of microarrays for genome-wide measurement of DNA copy number Nat Genet 2001;29263- 264
PubMedArticle
10.
Marks  RWhiteman  D Sunburn and melanoma: how strong is the evidence? BMJ 1994;30875- 76
PubMedArticle
11.
Whiteman  DCWhiteman  CAGreen  AC Childhood sun exposure as a risk factor for melanoma: a systematic review of epidemiologic studies Cancer Causes Control 2001;1269- 82
PubMedArticle
12.
Clark  WH  JrFrom  LBernardino  EAMihm  MC The histogenesis and biologic behavior of primary human malignant melanomas of the skin Cancer Res 1969;29705- 727
PubMed
13.
Breslow  A Thickness, cross-sectional areas and depth of invasion in the prognosis of cutaneous melanoma Ann Surg 1970;172902- 908
PubMedArticle
14.
Breslow  A Tumor thickness, level of invasion and node dissection in stage I cutaneous melanoma Ann Surg 1975;182572- 575
PubMedArticle
15.
Balch  CMMurad  TMSoong  SJIngalls  ALHalpern  NBMaddox  WA A multifactorial analysis of melanoma: prognostic histopathological features comparing Clark’s and Breslow’s staging methods Ann Surg 1978;188732- 742
PubMedArticle
16.
Armstrong  SAStaunton  JESilverman  LB  et al.  MLL translocations specify a distinct gene expression profile that distinguishes a unique leukemia Nat Genet 2002;3041- 47
PubMedArticle
17.
Golub  TRSlonim  DKTamayo  P  et al.  Molecular classification of cancer: class discovery and class prediction by gene expression monitoring Science 1999;286531- 537
PubMedArticle
18.
Bittner  MMeltzer  PChen  Y  et al.  Molecular classification of cutaneous malignant melanoma by gene expression profiling Nature 2000;406536- 540
PubMedArticle
19.
Bhattacharjee  ARichards  WGStaunton  J  et al.  Classification of human lung carcinomas by mRNA expression profiling reveals distinct adenocarcinoma subclasses Proc Natl Acad Sci U S A 2001;9813790- 13795
PubMedArticle
20.
Pomeroy  SLTamayo  PGaasenbeek  M  et al.  Prediction of central nervous system embryonal tumour outcome based on gene expression Nature 2002;415436- 442
PubMedArticle
21.
Kamb  AGruis  NWeaver-Feldhaus  J  et al.  A cell cycle regulator potentially involved in genesis of many tumor types Science 1994;264436- 444
PubMedArticle
22.
Cairns  PPolascik  TEby  Y  et al.  Frequency of homozygous deletion at p16/CDKN2 in primary human tumours Nat Genet 1995;11210- 212
PubMedArticle
23.
Sherr  CJ Tumor surveillance via the ARF-p53 pathway Genes Dev 1998;122984- 2991
PubMedArticle
24.
Rizos  HDarmanian  APHolland  EAMann  GJKefford  RF Mutations in the INK4a/ARF melanoma susceptibility locus functionally impair p14ARF J Biol Chem 2001;27641424- 41434
PubMedArticle
25.
Walker  GJHayward  NK Pathways to melanoma development: lessons from the mouse J Invest Dermatol 2002;119783- 792
PubMedArticle
26.
Bardeesy  NBastian  BCHezel  APinkel  DDePinho  RAChin  L Dual inactivation of RB and p53 pathways in RAS-induced melanomas Mol Cell Biol 2001;212144- 2153
PubMedArticle
27.
Gorden  AOsman  IGai  W  et al.  Analysis of BRAF and N-RAS mutations in metastatic melanoma tissues Cancer Res 2003;633955- 3957
PubMed
28.
Cruz  FRubin  BPWilson  D  et al.  Absence of BRAF and NRAS mutations in uveal melanoma Cancer Res 2003;635761- 5766
PubMed
29.
Weber  AHengge  URUrbanik  D  et al.  Absence of mutations of the BRAF gene and constitutive activation of extracellular-regulated kinase in malignant melanomas of the uvea Lab Invest 2003;831771- 1776
PubMedArticle
30.
Clark  EAGolub  TRLander  ESHynes  RO Genomic analysis of metastasis reveals an essential role for RhoC Nature 2000;406532- 535
PubMedArticle
31.
Weeraratna  ATJiang  YHostetter  G  et al.  Wnt5a signaling directly affects cell motility and invasion of metastatic melanoma Cancer Cell 2002;1279- 288
PubMedArticle
32.
Tusher  VGTibshirani  RChu  G Significance analysis of microarrays applied to the ionizing radiation response Proc Natl Acad Sci U S A 2001;985116- 5121[published correction appears inProc Natl Acad Sci U S A 2001;9810515
PubMedArticle
33.
Dysvik  BJonassen  I J-Express: exploring gene expression data using Java Bioinformatics 2001;17369- 370
PubMedArticle
34.
SOURCE, Stanford University Web site Available athttp://source.stanford.edu/cgi-bin/source/sourceSearchAccessed April 24, 2004
35.
KEGG, Kyoto Encyclopedia of Genes and Genomes Available athttp://www.genome.jp/kegg/kegg.htmlAccessed April 24, 2004
36.
 Signaling pathway database Available athttp://www.grt.kyushu-u.ac.jp/spad/index.htmlAccessed April 24, 2004
37.
 Cell signaling networks database Available athttp://geo.nihs.go.jp/csndb.htmlAccessed April 24, 2004
38.
 Human protein reference database Available athttp://www.hprd.orgAccessed April 24, 2004
39.
 Biocarta Available athttp://www.biocarta.com/genes/allpathways.aspAccessed April 24, 2004
40.
Furge  KAZhang  YWVande Woude  GF Met receptor tyrosine kinase: enhanced signaling through adapter proteins Oncogene 2000;195582- 5589
PubMedArticle
41.
Jeffers  MRong  SWoude  GF Hepatocyte growth factor/scatter factor-Met signaling in tumorigenicity and invasion/metastasis J Mol Med 1996;74505- 513
PubMedArticle
42.
Halaban  RRubin  JSFunasaka  Y  et al.  Met and hepatocyte growth factor/scatter factor signal transduction in normal melanocytes and melanoma cells Oncogene 1992;72195- 2206
PubMed
43.
Otsuka  TTakayama  HSharp  R  et al.  c-Met autocrine activation induces development of malignant melanoma and acquisition of the metastatic phenotype Cancer Res 1998;585157- 5167
PubMed
44.
Natali  PGNicotra  MRDi Renzo  MF  et al.  Expression of the c-Met/HGF receptor in human melanocytic neoplasms: demonstration of the relationship to malignant melanoma tumour progression Br J Cancer 1993;68746- 750
PubMedArticle
45.
Mirmohammadsadegh  ABaer  ANambiar  SBardenheuer  WHengge  UR Rapid identification of dysregulated genes in cutaneous malignant melanoma metastases using cDNA technology Cells Tissues Organs 2004;177119- 123
PubMedArticle
46.
Ooi  JYajnik  VImmanuel  D  et al.  The cloning of Grb10 reveals a new family of SH2 domain proteins Oncogene 1995;101621- 1630
PubMed
47.
Nantel  AMohammad-Ali  KSherk  JPosner  BIThomas  DY Interaction of the Grb10 adapter protein with the Raf1 and MEK1 kinases J Biol Chem 1998;27310475- 10484
PubMedArticle
48.
Jahn  TSeipel  PUrschel  SPeschel  CDuyster  J Role for the adaptor protein Grb10 in the activation of Akt Mol Cell Biol 2002;22979- 991
PubMedArticle
49.
Morrione  APlant  PValentinis  B  et al.  mGrb10 interacts with Nedd4 J Biol Chem 1999;27424094- 24099
PubMedArticle
50.
O’Neill  TJRose  DWPillay  TSHotta  KOlefsky  JMGustafson  TA Interaction of a GRBIR splice variant (a human GRB10 homolog) with the insulin and insulin-like growth factor I receptors: evidence for a role in mitogenic signaling J Biol Chem 1996;27122506- 22513
PubMedArticle
51.
Wang  JDai  HYousaf  N  et al.  Grb10, a positive, stimulatory signaling adapter in platelet-derived growth factor BB–, insulin-like growth factor I–, and insulin-mediated mitogenesis Mol Cell Biol 1999;196217- 6228
PubMed
52.
Liu  FRoth  RA GrbIR: a SH2-domain–containing protein that binds to the insulin receptor and inhibits its function Proc Natl Acad Sci U S A 1995;9210287- 10291
PubMedArticle
53.
Liu  FRoth  RA Binding of SH2 containing proteins to the insulin receptor: a new way for modulating insulin signalling Mol Cell Biochem 1998;18273- 78
PubMedArticle
54.
Stein  EGGustafson  TAHubbard  SR The BPS domain of Grb10 inhibits the catalytic activity of the insulin and IGF1 receptors FEBS Lett 2001;493106- 111
PubMedArticle
55.
Morrione  AValentinis  BResnicoff  MXu  SBaserga  R The role of mGrb10alpha in insulin-like growth factor I–mediated growth J Biol Chem 1997;27226382- 26387
PubMedArticle
56.
Langlais  PDong  LQRamos  FJ  et al.  Negative regulation of insulin-stimulated mitogen-activated protein kinase signaling by Grb10 Mol Endocrinol 2003;18350- 358
PubMedArticle
57.
Tschentscher  FHusing  JHolter  T  et al.  Tumor classification based on gene expression profiling shows that uveal melanomas with and without monosomy 3 represent two distinct entities Cancer Res 2003;632578- 2584
PubMed
58.
Alonso  SROrtiz  PPollan  M  et al.  Progression in cutaneous malignant melanoma is associated with distinct expression profiles: a tissue microarray–based study Am J Pathol 2004;164193- 203
PubMedArticle
59.
St Croix  BRago  CVelculescu  V  et al.  Genes expressed in human tumor endothelium Science 2000;2891197- 1202
PubMedArticle
60.
Macaulay  VMSalisbury  AJBohula  EAPlayford  MPSmorodinsky  NIShiloh  Y Downregulation of the type 1 insulin-like growth factor receptor in mouse melanoma cells is associated with enhanced radiosensitivity and impaired activation of Atm kinase Oncogene 2001;204029- 4040
PubMedArticle
61.
Lee  MLKuo  FCWhitmore  GASklar  J Importance of replication in microarray gene expression studies: statistical methods and evidence from repetitive cDNA hybridizations Proc Natl Acad Sci U S A 2000;979834- 9839
PubMedArticle
62.
Yang  YHBuckley  MJSpeed  TP Analysis of cDNA microarray images Brief Bioinform 2001;2341- 349
PubMedArticle
63.
DeRisi  JPenland  LBrown  PO  et al.  Use of a cDNA microarray to analyse gene expression patterns in human cancer Nat Genet 1996;14457- 460
PubMedArticle
64.
Eisen  MBSpellman  PTBrown  POBotstein  D Cluster analysis and display of genome-wide expression patterns Proc Natl Acad Sci U S A 1998;9514863- 14868
PubMedArticle
65.
Alon  UBarkai  NNotterman  DA  et al.  Patterns of gene expression revealed by clustering analysis of tumor and normal colon tissues probed by oligonucleotide arrays Proc Natl Acad Sci U S A 1999;966745- 6750
PubMedArticle
66.
Chen  YDougherty  ERBittner  ML Ratio-based decisions and the quantitative analysis of cDNA microarray images J Biomed Opt 1997;2367- 374Article
67.
Newton  MAKenziorski  CMRichmond  CSBlattner  FRTsui  KW On differential variability of expression ratios: improving statistical inference about gene expression changes from microarray data J Comput Biol 2001;837- 52
PubMedArticle
68.
Antoniadis  ALambert-Lacroix  SLeblanc  F Effective dimension reduction methods for tumor classification using gene expression data Bioinformatics 2003;19563- 570
PubMedArticle
69.
Kerr  MKChurchill  GA Bootstrapping cluster analysis: assessing the reliability of conclusions from microarray experiments Proc Natl Acad Sci U S A 2001;988961- 8965
PubMedArticle
70.
Chuaqui  RFBonner  RFBest  CJ  et al.  Post-analysis follow-up and validation of microarray experiments Nat Genet 2002;32509- 514
PubMedArticle
71.
Rajeevan  MSRanamukhaarachchi  DGVernon  SDUnger  ER Use of real-time quantitative PCR to validate the results of cDNA array and differential display PCR technologies Methods 2001;25443- 451
PubMedArticle
×