A gene expression signature associated with survival in metastatic melanoma
© Mandruzzato et al; licensee BioMed Central Ltd. 2006
Received: 06 October 2006
Accepted: 27 November 2006
Published: 27 November 2006
Current clinical and histopathological criteria used to define the prognosis of melanoma patients are inadequate for accurate prediction of clinical outcome. We investigated whether genome screening by means of high-throughput gene microarray might provide clinically useful information on patient survival.
Forty-three tumor tissues from 38 patients with stage III and stage IV melanoma were profiled with a 17,500 element cDNA microarray. Expression data were analyzed using significance analysis of microarrays (SAM) to identify genes associated with patient survival, and supervised principal components (SPC) to determine survival prediction.
SAM analysis revealed a set of 80 probes, corresponding to 70 genes, associated with survival, i.e. 45 probes characterizing longer and 35 shorter survival times, respectively. These transcripts were included in a survival prediction model designed using SPC and cross-validation which allowed identifying 30 predicting probes out of the 80 associated with survival.
The longer-survival group of genes included those expressed in immune cells, both innate and acquired, confirming the interplay between immunological mechanisms and the natural history of melanoma. Genes linked to immune cells were totally lacking in the poor-survival group, which was instead associated with a number of genes related to highly proliferative and invasive tumor cells.
Although cutaneous melanoma is a relatively rare tumor, its incidence is rising sharply, with minimal progress made in its treatment . When lymph node or distant metastases are present, clinical outcome is poor but highly variable. In fact, patients in the same TNM stage can have very different clinical outcomes, which is particularly true for TNM stage III. Currently, the most useful prognostic factors in metastatic disease are the metastatic site (e.g. subcutaneous vs visceral localization) and lactate dehyrogenase (LDH) plasma levels [2–4].
Despite significant efforts to identify independent predictors of melanoma outcome, no generally accepted histopathological or molecular marker defines disease subsets with clinically different outcomes [5–7]. Understanding differences in clinical behavior is important not only in the design and analysis of clinical trials, but also in planning different therapeutic strategies, such as adjuvant treatment.
In recent years, microarray technology has been extensively used in cancer research to obtain gene expression profiles aimed at identifying tumor classes, disease-related genes and new markers for predicting clinical outcome. Several studies also demonstrated that gene expression signatures can be used as a tool to predict survival of cancer patients [8–11]. As regards melanoma, analyses of gene expression have shed new light on the progression from local to metastatic disease as well as on melanoma immune responsiveness [12–14].
In this study, we correlated the gene-expression profile of tumors with overall survival in a cohort of patients with stage III and IV melanoma to determine whether survival among patients is reflected by specific sets of expressed genes. By using the significance analysis of microarrays (SAM), we identified 80 differentially expressed probes. We also generated a survival prediction model that resulted in 30 survival-related probes, all included in the set identified by SAM.
Patients and tissue collection
From 1997 to 2000, we collected 43 fresh metastatic melanoma biopsies from 38 patients with stage III and IV melanoma who underwent surgery as a part of the diagnostic work-up or therapeutic strategy. For three of these patients, we collected two biopsies, and in one case we obtained three biopsies. Immediately after surgery, half specimen was fixed in formalin and processed for routine histology, and the other half was snap-frozen in liquid nitrogen and stored at -80°C until use for RNA extraction. This study was examined and approved by the Ethics Committee of the local Health and Social Services (Azienda Ospedaliera, Padova) in accordance with the ethical standards laid down in the 1964 Declaration of Helsinki.
RNA extraction, amplification and labeling
Total RNA was extracted from frozen material by homogenizing the sample in TRIZOL reagent (Invitrogen, CA), according to the manufacturer's instructions, amplified into anti-sense RNA (aRNA) and reverse transcribed into fluorescent-labeled cDNA for hybridization to a custom-made 17,500-gene cDNA-based array as previously described [15, 16]. After amplification, the quality of aRNA was tested with the Agilent Bioanalyzer 2000 (Agilent Technologies, Palo Alto, CA). Total RNA from peripheral blood mononuclear cells pooled from 6 normal donors was extracted and amplified to prepare reference aRNA to be cohybridized in all experiments with test aRNA. cDNA targets were labeled with Cy3 for reference material and Cy5 for test material. The 32 × 24 × 23 (17,500 spots) human cDNA microarray was prepared in the Immunogenetics Section of Transfusion Medicine, Clinical Center, National Institute of Health, Bethesda. Clones for printing of 17 k cDNA array included a combination from a RG_HsKG_031901 7 k clone set and 10,000 clones from the RG_Hs_seq_ver_070700 40 k clone set (Research Genetics, Huntsville, AL). The cDNA clones include 12,072 uniquely named genes, 875 duplicates of named genes; the remainder consisted of expression sequence tags (complete gene list and printing layout are available . The hybridization was carried out at 65°C for 16–18 hours, and the slides were then washed and scanned on a Gene Pix 4000 scanner at variable PMT to obtain optimized signal intensities with minimum (< 1% spots) intensity saturation.
Before statistical analysis, data were loaded and filtered using the software package BRB Array Tools . Specifically, a statistical significance criterion based on the variance was used to filter log expression variations and remove those genes whose expression log ratios did not differ significantly from the median value (p-value ≤ 0.01). Genes were also excluded if the percentage of missing expression values was greater than 50%, resulting in a total of 4,845 probes used for further analysis.
Significance analysis of microarrays (SAM), [19, 20] was applied to identify genes correlated with patient overall survival. Briefly, SAM computes a score for each gene that measures the strength of transcript correlation with survival. This score is the maximum-likelihood score statistic from Cox's proportional hazards model (Cox score). A threshold value was chosen to give a reasonably low false positive rate, as estimated by repeatedly permuting the survival times and counting the number of genes that were significant at each threshold. Missing data were handled using the K-nearest neighbors imputer (k = 10) of the SAM imputation engine.
Survival prediction was calculated using supervised principal components (SPC). This analysis is similar to conventional principal component analysis although SPC uses a subset of predictors that are selected based on their association with outcome (see additional file 1 for a detailed description).
SPC were applied using the R superpc package .
The following antibodies were used: HLA Class II (DR) (Clone LN-3, Novocastra, Newcastle upon Tyne, UK, dilution 1:50) and IL-4R (clone 25463, R&D Systems, Abingdon, UK).
Tissue sections of 4-μm were cut from formalin fixed, paraffin-embedded tissue blocks and rehydrated. Endogenous peroxidase activity was blocked with incubation of the slides in 3% H202 in methanol and then, primary antibodies were applied at room temperature for 45 minutes. Binding sites of the primary antibodies were visualized using the Envision Plus detection System (Dako, Glostrup, Denmark) for HLA-DR and the Vectstain ABC kit (Vector laboratories, Burlingame, USA) for IL-4R.
Sections were developed using 3-amino-9-ethylcarbazole (Biogenex, San Ramon, USA), the slides were counterstained with hematoxylin and mounted in aqueous medium.
Immunostaining for IL4R and HLA-DR was separately evaluated in tumor cells and in intratumoral and peritumoral inflammatory cells and scored semi-quantitatively as 0 (no staining), 1+ (0–25%), 2+ (26–50%), 3+ (> 50%).
Clinical characteristics of the 38 patients with melanoma considered in the present study.
Patients (N = 38)
Identification of a gene set correlated with survival
Top 45probes whose increase in expression is significantly associated with longer survival.
Decay accelerating factor for complement
immunoglobulin lambda locus
Tumor necrosis factor, alpha-induced protein 3
Major histocompatibility complex, class II, DR alpha
Haematopoietic cell-specific Lyn substrate 1
Immunoglobulin lambda-like polypeptide 1
major histocompatibility complex, class II, DR beta 4
major histocompatibility complex, class II, DR beta 4
Heterogeneous nuclear ribonucleoprotein U
Nuclear factor of kappa light polypeptide gene enhancer in B-cells inhibitor, zeta
Tumor necrosis factor, alpha-induced protein 3
Neutrophil cytosolic factor 4, 40 kDa
Tumor necrosis factor, alpha-induced protein 3
lymphotoxin beta (TNF superfamily, member 3)
IL-4 receptor alpha chain
Axin 2 (conductin, axil)
Similar to serine/threonine kinase 4
Ras-related C3 otulinum toxin substrate 2
Jun B proto-oncogene
CD1D antigen, d polypeptide
Bruton agammaglobulinemia tyrosine kinase
Lysosomal associated multispanning membrane protein 5
chromosome 14 open reading frame 117
similar to bK246H3.1 (immunoglobulin lambda-like polypeptide 1, pre-B-cell specific)
Superoxide dismutase 2, mitochondrial
Major histocompatibility complex, class II, DQ alpha 1
NACHT, leucine rich repeat and PYD (pyrin domain) containing 1
Thymosin, beta 4, X-linked
CD2 antigen (p50)
major histocompatibility complex, class II, DR beta 1
Dual specificity phosphatase 5
Mitogen-activated protein kinase 13
E74-like factor 4
Rho GTPase activating protein 15
T cell receptor alpha chain
IL2-inducible T-cell kinase
Top 35 probes whose increase in expression is significantly associated with shorter survival.
MCM3 minichromosome maintenance deficient 3
Gap junction protein, beta 2, 26 kDa (connexin 26)
Phosphorylase, glycogen; liver
Tripartite motif-containing 32
Chromosome 20 open reading frame 129
Dopachrome tautomerase (tyrosine-related protein 2)
EF-hand domain family, member D1
Chondroitin sulfate proteoglycan 4 (melanoma-associated)
Myosin phosphatase-Rho interacting protein
Kynurenine 3-monooxygenase & Opsin 3*
SRY (sex determining region Y)-box 8
Growth arrest-specific 1
Solute carrier family 16, member 1
Myosin, heavy polypeptide 10, non-muscle
Farnesyl-diphosphate farnesyltransferase 1
Regulator of G-protein signalling 12
Protein kinase C, epsilon
A disintegrin-like and metalloprotease with thrombospondin type 1 motif, 5
Neogenin homolog 1 (chicken)
Phosphorylase, glycogen; liver
CHK1 checkpoint homolog
Neogenin homolog 1 (chicken)
Hypothetical protein FLJ10656
Kynurenine 3-monooxygenase & Opsin 3*
cDNA FLJ31683 fis
Chromosome 14 open reading frame 109
EST, Weakly similar to S10889 proline-rich protein
Serine threonine kinase 39
DNA microarray technology allows for the development of predictive models using gene expression profiles to study the relationship between prognosis and molecular features of the tumor. New statistical methods have been developed to address high dimensionality and low sample size issues characterizing microarray data [21, 23, 24]. To predict survival in this group of melanoma patients, we applied SPC, a survival prediction model that was designed and validated using a leave-one-out cross-validation procedure . Specifically, the model has been constructed applying the singular value decomposition procedure to the matrix X' consisting of only those genes whose Cox scores are greater than some threshold θ whose optimal value has been determined through a leave-one-out cross-validation procedure. In details, starting from a set G of possible values of θ, for each θ in G, one sample is left out, and those genes with absolute Cox scores greater than θ in the other n-1 samples are used to calculate for the left out sample. A Cox proportional hazards model is then fit to (i.e., to the first supervised principal component of ) and the chi-square statistic for the log-rank test associated with this model calculated. The procedure is repeated for each value of θ and any sample in the dataset. The overall likelihood ratio test indicated that the association between left-out samples and the supervised predictor constructed on is optimized at a value of θ = 1.7 at a p-value = 0.0002 (Supplementary Fig. S1). Thus, the survival prediction model has been constructed using the first principal component of those genes having a Cox score greater than 1.7 and resulted in 30 survival-related probes (Fig. 1, in black the transcripts used to build the SPC survival predictor).
Fig. 2, panel A illustrates the survival probability of the 43 samples, grouped according to the median value of = (θ = 1.7), i.e., the predicted values obtained selecting the first 30 probes corresponding to the threshold θ = 1.7, with short and long survivors having a value of higher and lower than the median, respectively.
Although these 30 predicting probes are all among the 80 transcript identified by SAM as associated to survival (Fig. 1), it has to be noted that class predictors constructed using a small number of variables may heavily depend on any one variable and produce spuriously high prediction strengths. However, the cross-validated predictor can also be constructed using a large number of transcripts without loosing statistical significance. Indeed, the overall likelihood ratio test indicates that the association between left-out samples and supervised predictor constructed on is still significant (p-value < 5%) for lower values of the threshold θ . As an examples, selecting a value of θ = 0.6 resulted in 1500 predicting probes which partitioned the samples according to the predicted survival curves of Fig. 2, panel B ( = (θ = 0.6)).
To adjust for prognostic factors, we also applied a multidimensional Cox model where the predicted value of the first principal component obtained by SPC using the first 30 and 1,500 probes (e.g., and ) has been analyzed together with other prognostic factors, such as LDH plasma level and the number of metastases. Results from this multidimensional analysis indicate that the transcriptional patterns are significantly associated with survival time (p-value = 0.02 for and p-value = 0.06 for ) also in the presence of other prognostic factors. On the contrary, LDH and the number of metastasis did not result statistically correlated with survival (p-value = 0.16 and p-value = 0.21 for LDH and p-value = 0.87 and p-value = 0.91 for number of metastasis considering and , respectively).
Validation of microarray data
We validated at the protein level by immunohistochemistry (IHC) the results obtained from gene expression analysis. Of the first 80 probes, 5 coded for HLA class II molecules, which are constitutively expressed on antigen presenting cells and B cells, on non-immune cell types under inflammatory conditions and by melanoma cells [25, 26]. Although we selected for microarray analysis melanoma biopsies that had a high percentage of tumor cells, the contribution of genes expressed at low level by cells present in tumor infiltrate cannot be ruled out. Moreover, 30 of 43 biopsies were infiltrated metastatic lymph nodes and thus a small amount of residual lymphoid tissue, expressing HLA class II molecules, could have been present.
HLA-DR and IL4R expression determined by IHC as graded by the scoring system (1+ to 3+).
Patients with shorter survival
Patients with longer survival
# of patients
# of patients
non tumor cells*
non tumor cells*
Up-regulation of HLA class II molecules can be induced by a number of molecules, including IL-4. This study demonstrated that increased expression of the IL-4R alpha chain gene was correlated with longer survival (Table 2). To confirm this finding, we investigated the presence of this protein in all available tumor biopsies by IHC. In 11 of 27 lesions, tumor cells expressed IL-4R (Fig. 3), while in 6 lesions a heterogeneous positive staining was observed also on inflammatory cells permeating the neoplasia (Table 4). Combining the staining on tumor and infiltrating cells, a higher incidence of IL-4R positive lesions was observed in patients with longer survival, therefore confirming the findings of the transcriptional analysis (Table 4).
CD4 and CD8 staining also revealed the presence of a mild to moderate T cell infiltrate inside melanoma cell nests in the majority of cases; usually intratumoral CD4+ and CD8+ T cells were co-localized within the lesion (data not shown).
Association of the selected genes with survival
Eighty probes identified by SAM consisted of 40 genes whose increased expression was associated with longer survival, and 30 genes whose increased expression was associated with shorter survival. Genes involved in immune response and inflammation such as HLA class II, TRA@, LTB, TNFAIP3, IL-4R, IGLL1, CD1D, CD2, ITK, SOD2, DAF and GZMK, were associated selectively with longer survival, confirming once more the important interplay between the host immune system and malignant cells in melanoma.
Conversely, analysis of tumor progression-associated genes revealed an increased expression of those implicated in cell adhesion, motility and proliferation. For example, gene GJB2, encoding the gap junction protein connexin 26 (Cx26) a sub-type of gap junction proteins; gene CSPG4, coding for the melanoma-associated chondroitin sulfate proteoglycan 4; the disintegrin-like and metalloprotease genes with thrombospondin type 1 (ADAMTS5); the minichromosome maintenance 3 (MCM3) and the DCT genes, all whose expression by tumor cells is frequently associated with more invasive and metastatic behavior [27–33].
In this study, we used gene expression profiling and survival data from advanced melanoma patients and, by SAM class comparison, identified a set of 80 probes, corresponding to 70 genes, associated with survival. The survival-associated group of genes highlighted characteristics of melanoma cells, but also of non-malignant cells present in the tumor tissue. In fact, while specimens were selected on the basis of histological diagnosis, and characterized by a high proportion of neoplastic tissue, the use of whole tumor tissue enabled us to detect changes in gene expression associated with stromal and infiltrating immune/inflammatory cells.
Five of the probes selected using SAM encode HLA class II molecules. This MHC class II signature has already been associated with a favorable outcome in patients with large B-cell lymphoma , while in the liver microenvironment HLA class II overexpression is associated with a metastasis-inclined phenotype , suggesting that the prognostic genes may differ among different tumor types. In human melanoma, the association of these molecules with survival is controversial. Some authors reported that enhanced HLA class II expression was associated with an unfavorable prognosis [25, 35], while others found a positive correlation with overall survival [36, 37]. Our data are in line with the latter findings and the IHC analysis confirmed that the expression of these molecules in the melanoma cells was almost exclusively associated with longer survival (Table 4). CD4+ and CD8+ cells were found in the same lesions, and, as reported by others, a close correlation between MHC class II expression and presence of T cells was observed (data not shown). This phenomenon may be the result of an inflammatory process at the tumor site in which the cytokines released, such as IFN and TNF, induce MHC class II up-regulation. IL-4 may also be produced in an inflammatory milieu, and can up-regulate the expression of these molecules . It has already been shown that melanoma cells may express IL-4 receptors and that IL-4 has antiproliferative and/or apoptotic effects in these cells . Interestingly, the gene for the IL-4R alpha chain was included in the longer-survival set of genes, suggesting that its presence might be related to these cytotoxic pathways. As predicted by SAM, IHC revealed a positive staining for IL-4R, confirming at the protein level that enhanced expression was more evident in patients with longer survival (Table 4). IHC studies indicated also that IL-4R staining was localized in melanoma cells, as well as in interstitial inflammatory cells and in residual lymphoid tissue. Despite the intrinsic limitations of this study, these findings highlight the role of CD4+ T-lymphocytes in the anti-melanoma immune response. Interestingly, the CD4 probe was among the first 225 probes selected by SAM, while the probe for CD8 was not included in the list of the first 3,000 differentially expressed genes.
The role of the immune system in survival of metastatic melanoma patients is also evident when other genes are considered. In addition to the MHC class II genes, other genes associated with survival direct the anti-tumor activity exerted by effector cells of the immune system such as T cells (TRA@, GZMK), NKT cells (CD1d) and B cells (IGLC2), confirming the notion that the clinical course of melanoma is closely associated with innate and acquired immune responses. It is of interest that genes known to be involved in the immune response were present exclusively in the longer-survival group, thus confirming an immune-response signature associated with favorable prognosis. The finding that genes associated with immune cell activation are part of the molecular signature that herald a good prognosis in stage III and IV melanoma patients is remarkable, since it highlights, at a molecular level, the potential efficacy of the immune response also in advanced melanoma. Thus, melanoma survival is associated not only with the intrinsic biology of tumor cells, but also with an efficient immune response that encompasses different facets of the immune system.
Of the genes correlated with shorter survival, some were already known to be over-expressed in different types of cancer, such as MCM3, BCHE [32, 40–42], and some have previously been implicated in melanoma progression, like CSPG4  and Cx26  or, more generally, associated with tumor invasiveness, like ADAMTS5 .
The survival prediction model, designed using SPC, identified 30 survival-related probes, which were all among the 80 probes identified by SAM. Although we have not yet tested this predictive model on an independent series of patients, the results of the cross-validation were quite encouraging. Identifying a subset of genes that predict cancer patients' survival is an important goal of microarray research since it could ultimately lead to the development of prognostic tools that, by identifying patients with different clinical outcome, might lead to tailor therapeutic strategies on a single patient basis. This might be particularly relevant for patients with stage III melanoma, as the lack of definitively convincing results on interferon-alpha based adjuvant treatment might depend upon inadequate patient selection .
In conclusion, this study applied high-throughput gene microarrays to screen the transcriptome in the search for genes correlated with patient survival. The bioinformatics data analysis allowed us to identify a number of genes related to cell proliferation and invasiveness that enable tumor cells to progress to a more aggressive phenotype. If validated by subsequent independent studies, these genes might serve as predictors of survival, and thus help clinicians to identify patients at higher risk of disease progression. Moreover, the transcripts associated to the longer-survival group included genes expressed in immune cells, both innate and acquired, confirming the importance of the interplay between immunological surveillance and the natural history of melanoma. These results highlight how critical the relationship is between melanoma cell aggressiveness and the activity of innate and/or acquired immune cells for the ultimate tumor outcome.
We thank Dr. Alessandra Luchini for helpful discussions and suggestions, and Andrea Azzalini for assistance with graphics. This work was supported by grants from the Italian Association for Cancer Research (AIRC), the Italian Ministry of Research (Contract MIUR-FIRB RBAU01935A) and Italy-U.S. Cooperation Program for the Therapy of Cancer (Grant 530/F-A4).
- Tsao H, Atkins MB, Sober AJ: Management of cutaneous melanoma. N Engl J Med. 2004, 351: 998-1012. 10.1056/NEJMra041245.View ArticlePubMedGoogle Scholar
- Homsi J, Kashani-Sabet M, Messina JL, Daud A: Cutaneous melanoma: prognostic factors. Cancer Control. 2005, 12: 223-229.PubMedGoogle Scholar
- Balch CM, Buzaid AC, Soong SJ, Atkins MB, Cascinelli N, Coit DG, Fleming ID, Gershenwald JE, Houghton A, Kirkwood JM, McMasters KM, Mihm MF, Morton DL, Reintgen DS, Ross MI, Sober A, Thompson JA, Thompson JF: Final version of the American Joint Committee on Cancer staging system for cutaneous melanoma. J Clin Oncol. 2001, 19: 3635-3648.PubMedGoogle Scholar
- Balch CM, Soong SJ, Gershenwald JE, Thompson JF, Reintgen DS, Cascinelli N, Urist M, McMasters KM, Ross MI, Kirkwood JM, Atkins MB, Thompson JA, Coit DG, Byrd D, Desmond R, Zhang Y, Liu PY, Lyman GH, Morabito A: Prognostic factors analysis of 17,600 melanoma patients: validation of the American Joint Committee on Cancer melanoma staging system. J Clin Oncol. 2001, 19: 3622-3634.PubMedGoogle Scholar
- Weyers W, Euler M, Diaz-Cascajo C, Schill WB, Bonczkowitz M: Classification of cutaneous malignant melanoma: a reassessment of histopathologic criteria for the distinction of different types. Cancer. 1999, 86: 288-299. 10.1002/(SICI)1097-0142(19990715)86:2<288::AID-CNCR13>3.0.CO;2-S.View ArticlePubMedGoogle Scholar
- Byers HR, Bhawan J: Pathologic parameters in the diagnosis and prognosis of primary cutaneous melanoma. Hematol Oncol Clin North Am. 1998, 12: 717-735. 10.1016/S0889-8588(05)70020-4.View ArticlePubMedGoogle Scholar
- Busam KJ: The use and application of special techniques in assessing melanocytic tumours. Pathology. 2004, 36: 462-469. 10.1080/00313020412331283824.View ArticlePubMedGoogle Scholar
- van de Vijver MJ, He YD, van't Veer LJ, Dai H, Hart AA, Voskuil DW, Schreiber GJ, Peterse JL, Roberts C, Marton MJ, Parrish M, Atsma D, Witteveen A, Glas A, Delahaye L, van der Velde T, Bartelink H, Rodenhuis S, Rutgers ET, Friend SH, Bernards R: A gene-expression signature as a predictor of survival in breast cancer. N Engl J Med. 2002, 347: 1999-2009. 10.1056/NEJMoa021967.View ArticlePubMedGoogle Scholar
- Beer DG, Kardia SL, Huang CC, Giordano TJ, Levin AM, Misek DE, Lin L, Chen G, Gharib TG, Thomas DG, Lizyness ML, Kuick R, Hayasaka S, Taylor JM, Iannettoni MD, Orringer MB, Hanash S: Gene-expression profiles predict survival of patients with lung adenocarcinoma. Nat Med. 2002, 8: 816-824.PubMedGoogle Scholar
- Lee JS, Chu IS, Heo J, Calvisi DF, Sun Z, Roskams T, Durnez A, Demetris AJ, Thorgeirsson SS: Classification and prediction of survival in hepatocellular carcinoma by gene expression profiling. Hepatology. 2004, 40: 667-676. 10.1002/hep.20375.View ArticlePubMedGoogle Scholar
- Rosenwald A, Wright G, Chan WC, Connors JM, Campo E, Fisher RI, Gascoyne RD, Muller-Hermelink HK, Smeland EB, Giltnane JM, Hurt EM, Zhao H, Averett L, Yang L, Wilson WH, Jaffe ES, Simon R, Klausner RD, Powell J, Duffey PL, Longo DL, Greiner TC, Weisenburger DD, Sanger WG, Dave BJ, Lynch JC, Vose J, Armitage JO, Montserrat E, Lopez-Guillermo A, Grogan TM, Miller TP, LeBlanc M, Ott G, Kvaloy S, Delabie J, Holte H, Krajci P, Stokke T, Staudt LM: The use of molecular profiling to predict survival after chemotherapy for diffuse large-B-cell lymphoma. N Engl J Med. 2002, 346: 1937-1947. 10.1056/NEJMoa012914.View ArticlePubMedGoogle Scholar
- Carr KM, Bittner M, Trent JM: Gene-expression profiling in human cutaneous melanoma. Oncogene. 2003, 22: 3076-3080. 10.1038/sj.onc.1206448.View ArticlePubMedGoogle Scholar
- Wang E, Miller LD, Ohnmacht GA, Mocellin S, Perez-Diez A, Petersen D, Zhao Y, Simon R, Powell JI, Asaki E, Alexander HR, Duray PH, Herlyn M, Restifo NP, Liu ET, Rosenberg SA, Marincola FM: Prospective molecular profiling of melanoma metastases suggests classifiers of immune responsiveness. Cancer Res. 2002, 62: 3581-3586.PubMed CentralPubMedGoogle Scholar
- Haqq C, Nosrati M, Sudilovsky D, Crothers J, Khodabakhsh D, Pulliam BL, Federman S, Miller JR, Allen RE, Singer MI, Leong SP, Ljung BM, Sagebiel RW, Kashani-Sabet M: The gene expression signatures of melanoma progression. Proc Natl Acad Sci U S A. 2005, 102: 6092-6097. 10.1073/pnas.0501564102.PubMed CentralView ArticlePubMedGoogle Scholar
- Wang E, Panelli MC, Zavaglia K, Mandruzzato S, Hu N, Taylor PR, Seliger B, Zanovello P, Freedman RS, Marincola FM: Melanoma-restricted genes. J Transl Med. 2004, 2: 34-10.1186/1479-5876-2-34.PubMed CentralView ArticlePubMedGoogle Scholar
- Basil CF, Zhao Y, Zavaglia K, Jin P, Panelli MC, Voiculescu S, Mandruzzato S, Lee HM, Seliger B, Freedman RS, Taylor PR, Hu N, Zanovello P, Marincola FM, Wang E: Common cancer biomarkers. Cancer Res. 2006, 66: 2953-2961. 10.1158/0008-5472.CAN-05-3433.View ArticlePubMedGoogle Scholar
- Tusher VG, Tibshirani R, Chu G: Significance analysis of microarrays applied to the ionizing radiation response. Proc Natl Acad Sci U S A. 2001, 98: 5116-5121. 10.1073/pnas.091062498.PubMed CentralView ArticlePubMedGoogle Scholar
- Bair E, Tibshirani R: Semi-supervised methods to predict patient survival from gene expression data. PLoS Biol. 2004, 2: E108-10.1371/journal.pbio.0020108.PubMed CentralView ArticlePubMedGoogle Scholar
- Gui J, Li H: Penalized Cox regression analysis in the high-dimensional and low-sample size settings, with applications to microarray gene expression data. Bioinformatics. 2005, 21: 3001-3008. 10.1093/bioinformatics/bti422.View ArticlePubMedGoogle Scholar
- Nguyen DV, Rocke DM: Partial least squares proportional hazard regression for application to DNA microarray survival data. Bioinformatics. 2002, 18: 1625-1632. 10.1093/bioinformatics/18.12.1625.View ArticlePubMedGoogle Scholar
- van Duinen SG, Ruiter DJ, Broecker EB, van der Velde EA, Sorg C, Welvaart K, Ferrone S: Level of HLA antigens in locoregional metastases and clinical course of the disease in patients with melanoma. Cancer Res. 1988, 48: 1019-1025.PubMedGoogle Scholar
- Taramelli D, Fossati G, Mazzocchi A, Delia D, Ferrone S, Parmiani G: Classes I and II HLA and melanoma-associated antigen expression and modulation on melanoma cells isolated from primary and metastatic lesions. Cancer Res. 1986, 46: 433-439.PubMedGoogle Scholar
- Jamieson S, Going JJ, D'Arcy R, George WD: Expression of gap junction proteins connexin 26 and connexin 43 in normal human breast and in breast tumours. J Pathol. 1998, 184: 37-43. 10.1002/(SICI)1096-9896(199801)184:1<37::AID-PATH966>3.0.CO;2-D.View ArticlePubMedGoogle Scholar
- Ito A, Koma Y, Uchino K, Okada T, Ohbayashi C, Tsubota N, Okada M: Increased expression of connexin 26 in the invasive component of lung squamous cell carcinoma: Significant correlation with poor prognosis. Cancer Lett. 2006, 234 (2): 239-48. 10.1016/j.canlet.2005.03.049.View ArticlePubMedGoogle Scholar
- Bumol TF, Reisfeld RA: Unique glycoprotein-proteoglycan complex defined by monoclonal antibody on human melanoma cells. Proc Natl Acad Sci U S A. 1982, 79: 1245-1249. 10.1073/pnas.79.4.1245.PubMed CentralView ArticlePubMedGoogle Scholar
- Olsen EB, Trier K, Eldov K, Ammitzboll T: Glycosaminoglycans in human breast cancer. Acta Obstet Gynecol Scand. 1988, 67: 539-542.View ArticlePubMedGoogle Scholar
- Held-Feindt J, Paredes EB, Blomer U, Seidenbecher C, Stark AM, Mehdorn HM, Mentlein R: Matrix-degrading proteases ADAMTS4 and ADAMTS5 (disintegrins and metalloproteinases with thrombospondin motifs 4 and 5) are expressed in human glioblastomas. Int J Cancer. 2006, 118: 55-61. 10.1002/ijc.21258.View ArticlePubMedGoogle Scholar
- Ha SA, Shin SM, Namkoong H, Lee H, Cho GW, Hur SY, Kim TE, Kim JW: Cancer-associated expression of minichromosome maintenance 3 gene in several human cancers and its involvement in tumorigenesis. Clin Cancer Res. 2004, 10: 8386-8395. 10.1158/1078-0432.CCR-04-1029.View ArticlePubMedGoogle Scholar
- Nishioka E, Funasaka Y, Kondoh H, Chakraborty AK, Mishima Y, Ichihashi M: Expression of tyrosinase, TRP-1 and TRP-2 in ultraviolet-irradiated human melanomas and melanocytes: TRP-2 protects melanoma cells from ultraviolet B induced apoptosis. Melanoma Res. 1999, 9: 433-443. 10.1097/00008390-199910000-00002.View ArticlePubMedGoogle Scholar
- Budhu A, Forgues M, Ye QH, Jia HL, He P, Zanetti KA, Kammula US, Chen Y, Qin LX, Tang ZY, Wang XW: Prediction of venous metastases, recurrence, and prognosis in hepatocellular carcinoma based on a unique immune response signature of the liver microenvironment. Cancer Cell. 2006, 10: 99-111. 10.1016/j.ccr.2006.06.016.View ArticlePubMedGoogle Scholar
- Alonso SR, Ortiz P, Pollan M, Perez-Gomez B, Sanchez L, Acuna MJ, Pajares R, Martinez-Tello FJ, Hortelano CM, Piris MA, Rodriguez-Peralto JL: Progression in cutaneous malignant melanoma is associated with distinct expression profiles: a tissue microarray-based study. Am J Pathol. 2004, 164: 193-203.PubMed CentralView ArticlePubMedGoogle Scholar
- Bernsen MR, Hakansson L, Gustafsson B, Krysander L, Rettrup B, Ruiter D, Hakansson A: On the biological relevance of MHC class II and B7 expression by tumour cells in melanoma metastases. Br J Cancer. 2003, 88: 424-431. 10.1038/sj.bjc.6600703.PubMed CentralView ArticlePubMedGoogle Scholar
- Piras F, Colombari R, Minerba L, Murtas D, Floris C, Maxia C, Corbu A, Perra MT, Sirigu P: The predictive value of CD8, CD4, CD68, and human leukocyte antigen-D-related cells in the prognosis of cutaneous malignant melanoma with vertical growth phase. Cancer. 2005, 104: 1246-1254. 10.1002/cncr.21283.View ArticlePubMedGoogle Scholar
- Kelly-Welch AE, Hanson EM, Boothby MR, Keegan AD: Interleukin-4 and interleukin-13 signaling connections maps. Science. 2003, 300: 1527-1528. 10.1126/science.1085458.View ArticlePubMedGoogle Scholar
- Obiri NI, Siegel JP, Varricchio F, Puri RK: Expression of high-affinity IL-4 receptors on human melanoma, ovarian and breast carcinoma cells. Clin Exp Immunol. 1994, 95: 148-155.PubMed CentralView ArticlePubMedGoogle Scholar
- Brass N, Racz A, Heckel D, Remberger K, Sybrecht GW, Meese EU: Amplification of the genes BCHE and SLC2A2 in 40% of squamous cell carcinoma of the lung. Cancer Res. 1997, 57: 2290-2294.PubMedGoogle Scholar
- Soreq H, Lapidot-Lifson Y, Zakut H: A role for cholinesterases in tumorigenesis?. Cancer Cells. 1991, 3: 511-516.PubMedGoogle Scholar
- Ishimi Y, Okayasu I, Kato C, Kwon HJ, Kimura H, Yamada K, Song SY: Enhanced expression of Mcm proteins in cancer cells derived from uterine cervix. Eur J Biochem. 2003, 270: 1089-1101. 10.1046/j.1432-1033.2003.03440.x.View ArticlePubMedGoogle Scholar
- Yang J, Price MA, Neudauer CL, Wilson C, Ferrone S, Xia H, Iida J, Simpson MA, McCarthy JB: Melanoma chondroitin sulfate proteoglycan enhances FAK and ERK activation by distinct mechanisms. J Cell Biol. 2004, 165: 881-891. 10.1083/jcb.200403174.PubMed CentralView ArticlePubMedGoogle Scholar
- Kirkwood JM, Manola J, Ibrahim J, Sondak V, Ernstoff MS, Rao U: A pooled analysis of eastern cooperative oncology group and intergroup trials of adjuvant high-dose interferon for melanoma. Clin Cancer Res. 2004, 10: 1670-1677. 10.1158/1078-0432.CCR-1103-3.View ArticlePubMedGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.