Genomic profiling of malignant melanoma using tiling-resolution arrayCGH

Abstract

Malignant melanoma is an aggressive, heterogeneous disease where new biomarkers for diagnosis and clinical outcome are needed. We searched for chromosomal aberrations that characterize its pathogenesis using 47 different melanoma cell lines and tiling-resolution bacterial artificial chromosome-arrays for comparative genomic hybridization. Major melanoma genes, including BRAF, NRAS, CDKN2A, TP53, CTNNB1, CDK4 and PTEN, were examined for mutations. Distinct copy number alterations were detected, including loss or gain of whole chromosomes but also minute amplifications and homozygous deletions. Most common overlapping regions with losses were mapped to 9p24.3–q13, 10 and 11q14.1-qter, whereas copy number gains were most frequent on chromosomes 1q, 7, 17q and 20q. Amplifications were delineated to oncogenes such as MITF (3p14), CCND1 (11q13), MDM2 (12q15), CCNE1 (19q12) and NOTCH2 (1p12). Frequent findings of homozygous deletions on 9p21 and 10q23 confirmed the importance of CDKN2A and PTEN. Pair-wise comparisons revealed distinct sets of alterations, for example, mutually exclusive mutations in BRAF and NRAS, mutual mutations in BRAF and PTEN, concomitant chromosome 7 gain and 10 loss and concomitant chromosome 15q22.2–q26.3 gain and 20 gain. Moreover, alterations of the various melanoma genes were associated with distinct chromosomal imbalances suggestive of specific genomic programs in melanoma development.

Introduction

Cutaneous malignant melanoma (CMM) is an aggressive form of skin cancer with increasing incidence in the Western world (Tucker and Goldstein, 2003). CMM has a heterogeneous and unpredictable clinical course with a potential for aggressive growth and refractoriness to available chemotherapy. The risk of CMM is influenced by both genetic and environmental factors, and the incidence varies between populations depending on pigmentation, skin type and sun exposure. An established risk factor for CMM is a family history of the disease. Two high-penetrance susceptibility genes, CDKN2A (primarily the p16INK4A sequence) and CDK4, have been identified, but germ-line mutations in these genes account for merely one-third of high-risk families, implicating the existence of additional melanoma genes (Hayward, 2003).

Several studies have addressed the genetic events in sporadic CMM development. Somatic inactivation of CDKN2A (p16INK4A and p14ARF) is frequently detected in melanoma cell lines but less commonly in primary tumors (Cachia et al., 2000). More recent reports show biallelic CDKN2A deletions in 45% of CMM metastases emphasizing this locus in disease progression (Grafstrom et al., 2005). Activation of the mitrogen-activated protein kinase (MAPK) pathway may also be a compulsory event in CMM pathogenesis, primarily evident as either BRAF (60% of melanomas) or NRAS (30%) point mutations, but never both (Maldonado et al., 2003; Omholt et al., 2003). Known to be early events and present in benign nevi (Pollock et al., 2003), BRAF mutations may influence tumor progression and give rise to distinct gene expression signatures in melanoma cell lines (Pavey et al., 2004). Gain of chromosome 7q is common in CMM suggesting that BRAF, located on 7q34, is a target for gene amplification (Tanami et al., 2004). Moreover, cyclin D1, a down-stream target of the MAPK pathway and p16INK4A antagonist, is amplified in acral-type CMM in which BRAF and NRAS mutations are infrequent (Takata et al., 2005). Finally, activation of the PIK3-pathway is essential and commonly pursued by inactivation of PTEN located on chromosome 10, a frequent deletion target in melanoma (Guldberg et al., 1997b).

A recent genome-wide survey in melanocytic lesions, using array-based comparative genomic hybridization (arrayCGH), revealed patterns of genomic aberrations that could distinguish chronically sun-induced from non-sun-induced melanomas and further classify tumors into clinically relevant groups (Curtin et al., 2005). Here, we used CGH to arrays of 32 433 bacterial artificial chromosome (BAC) clones forming a contiguous and tiling coverage of the human genome with an average resolution of 100 kbp, to precisely map novel DNA copy number gains and losses in 47 melanoma cell lines thoroughly characterized also for mutations in the known major melanoma genes. In addition, the study reveals that mutants BRAF, NRAS, PTEN or TP53 are associated with discrete chromosomal alterations, possibly reflecting cooperative events in melanoma pathogenesis.

Results

DNA copy number profiles were established in cell lines originating from 46 different metastatic CMM and one primary ocular melanoma (Figure 1). The overall genome patterns are complex, involving whole chromosome gains and losses as well as focal alterations such as narrow amplifications and homozygous deletions (Figure 2). Only one (EST97) of the 47 cell lines had no apparent alterations using the current technology. Combining data from all cell lines, alterations were found on all chromosomes at least once. Chromosomal arms with copy number gains in at least 50% of the cell lines included 1q, 7p, 7q, 8q, 17q and 20q. Frequent losses were found on 4p, 4q, 6q, 8p, 9p, 10p, 10q and 11q (Figure 2a).

Figure 1
figure1

Heat-map over DNA copy number aberrations (x-axis) in 47 melanoma cell lines (y-axis). Red represents gains/amplifications and green represents loss/homozygous deletions. Bottom left panel zoom-in on chromosome 3, depicting the frequent DNA copy number gains at the MITF locus at 3p13. Bottom right panel displays chromosome 9 and the common DNA copy number losses on the p-arm, occasionally limited to 9p21.3 and CDKN2A homozygous deletions.

Figure 2
figure2

(a) Genome-wide frequency plot of DNA copy number gains (red) and losses (green) for all 47 melanoma cell lines. (b) Genomic profile of EST73 cells. Homozygous deletions are clearly seen at chromosome 9p21.3 (CDKN2A) and 10p11.21 (PARD3). (c) Genomic profile of FM79 cells. Distinct amplification at chromosome 1p12 (NOTCH2 and other genes). (d) Zoom-in on chromosome 10p11 displaying a homozygous deletion of the PARD3 gene in EST73 cells. (e) Zoom-in on chromosome 1p12 displaying one the recurrently amplified regions including the NOTCH2 gene.

Regions showing frequent copy number losses and gains

The two most common individual copy number losses were detected on chromosomes 10 and 9p24.3–q13, both spanning over large genomic regions harboring known melanoma tumor suppressor genes, that is, PTEN and CDKN2A, which were further pinpointed in some cell lines by homozygous deletions (see below). Other regions frequently affected by large genomic deletions localize to chromosomes 4, 6q and 11q. The smallest regions of overlapping (SRO) losses span 6q23.3–q25.3 and 11q22.3–q24.1, including a large number of candidate genes, whereas complete loss was the most common chromosome 4 aberrations. Restricted deletions were also present on chromosome 1p in 50% of the samples, and the SRO loss was further narrowed down to two loci. The first maps to 1p22.1 (92029228–94254061 bp; UCSC May 2004 build hg17) and includes 20 genes, for instance TGFBR3 and CDC18. The second region maps to 1p21.3 (95472183–98455030 bp) and includes only two known coding genes, PTBP2 and DPYD, and a putative microRNA, miR-137. Moreover, chromosome 9q was deleted in 40% of the cell lines with an SRO on 9q21.13–q21.2 (72394581–78138360 bp).

Chromosomal regions displaying frequent DNA copy number gains include 1q, 7, 8q, 17q and 20q. The SRO gain on 1q23.3–q25.3 (157822335–181823355 bp) spans 24 Mbp and contains a large number of genes. The SRO gains on chromosome 7 were divided into two distinct loci. The first (found in 59%) maps to 7q21.13–q31.1 (87891721–107605147 bp) and includes CDK6, whereas the second (65%) is more distal (7q32.1–q34; 126600073–142561955 bp) and includes BRAF. Furthermore, two SRO gains were detected on chromosome 8, the first maps to 8q22.1–q22.3 (96297981–104661726 bp) and the second to 8q24.11–q24.22 (118528414–134896260 bp) and includes MYC. Copy number gains on chromosome 17 were observed in 50% of the cell lines and two SRO gains map to 17q23.2–q24.1 (55386720–60690169 bp) and 17q25.1–q25.3 (69684458–74597283 bp), respectively. Finally, two SRO gains on chromosome 20 map to 20q12–q13.12 (38554072–45485732 bp) and 20q13.31–q13.33 (54727799–62434349 bp), respectively.

High-level amplifications

More narrow peaks corresponding to focal amplifications (defined as log2 ratio>1.5) were detected at 25 loci in 15 different cell lines (Table 1). Recurrent amplifications were located on 11q13, 3p14 and 1p12. The 11q13 amplicon includes CCND1 and was seen in three cell lines. Chromosome 3p14, including the melanoma oncogene microphthalmia-associated transcription factor (MITF), was amplified in three cell lines and MITF copy number gain was found in additional cell lines, in total 36%. Chromosome 1p12 was amplified in three cell lines, one (FM79) of which harbored a narrow peak including only NOTCH2. Other known amplified oncogenes include CCNE1 on 19q13, MDM2 on 12q15 and BRAF on 7q34. Moreover, 13 additional amplicons without obvious target genes were identified (Table 1).

Table 1 Gene alterations in cell lines derived from 46 CMM and one ocular melanoma (EST128)

Homozygous deletions

Previously recognized homozygous deletions on chromosomes 9p21.3 and 10q23.31 were confirmed and shown to target CDKN2A and PTEN, respectively, often as the single-affected gene (Table 1). Moreover, two novel homozygous deletions including single genes, 10p11.22 (PARD3) and 11q14.1 (RAB38), were detected. A recurrent homozygous deletion on 4q34.3 including no known genes was found in three cell lines.

Mutation screening

BRAF, NRAS, PTEN, TP53, CDKN2A, CDK4 and CTNNB1 were analysed for mutations. Two-thirds (n=32) of the 47 cell lines contained a BRAF mutation, in all but one the V600E mutation. Ten (21%) cell lines had NRAS mutation, all at codon 61. Six cell lines harbored neither BRAF nor NRAS mutations, including the ocular melanoma EST128 and EST97, the latter devoid of CGH alterations. TP53 mutations were present in 31%, whereas PTEN mutations were identified in 27% of the cell lines. CDKN2A was inactivated by point mutations, homozygous deletions or methylation in 10, 57 and 10%, respectively, any form of inactivation being found in 77% of the cell lines. In addition, one cell line (FM56) harbored a mutation in CDK4 and another (FM9) contained a CTNNB1 mutation (Table 1). Accordingly, the mutation frequency (below in parentheses) seen here in 47 melanoma cell lines is similar to the Cancer Genome Project database (http://www.sanger.ac.uk/genetics/CGP/) on 46 (other) malignant melanoma cell lines: BRAF 60% (63%), CDKN2A 54% (77%), TP53 32% (31%), NRAS 21% (21%) and PTEN 17% (27%).

Correlation of cancer gene mutation status and genetic alterations

Two-sided t-test was performed to determine whether PTEN, TP53, BRAF or NRAS mutation status was associated with specific DNA copy number aberrations. Indeed, mutation in PTEN was correlated (P<0.05) to loss of 1p22.1–p21.1, a region spanning 10 Mbp and >50 known genes. Other altered regions overrepresented in PTEN-mutated cell lines included 10q loss, 13q14–q33 loss, 14q21–q23 loss and 19q13 gain (P<0.05). Mutation in TP53 was inversely correlated (P<0.05) to 1p34.3–p13.2 loss, in fact, TP53-mutated cell lines were instead associated with gains at 1p34.3–p13.2. Moreover, 5q35 gain, 12q14 loss and 12q15–q21 loss distinguished TP53 mutated from wild-type cells (P<0.05) (Figure 4). Cells with BRAF mutation had a higher frequency of chromosome 7 gains, often comprising large regions and always including BRAF on 7q34 (P<0.05). Chromosome 7 gains were also seen in cell lines with wild-type BRAF and NRAS; however, less commonly in NRAS-mutated cells. Also, 2p11–q13 gain, 6p22 loss, 10 loss, 11q loss, 14q21–q23 loss and 20 gain were more frequently observed in BRAF mutant cell lines (P<0.05). Alterations specific for NRAS-mutated cells included only 3q13.12–q13.31 gain (P<0.05), although regions commonly altered in BRAF-mutated and wild-type cells (chromosome 7 and 20) were less frequently affected in NRAS-mutant cells (Figure 3). Furthermore, BRAF and NRAS mutations were found to be mutually exclusive (P<0.00001), PTEN and NRAS mutations were rarely mutual (P=0.038), whereas combined BRAF and PTEN mutations were common (P=0.0023). No correlation was found between TP53 mutations and mutations in other genes.

Figure 4
figure3

(a) DNA copy number frequency plots for PTEN-mutated (upper panel) and wild-type cell lines (lower panel). Regions altered significantly (P<0.05) more frequent in PTEN mutant compared to wild-type cell lines are indicated (*) in the upper panel. Chromosomes 1p, 9p, 10 and 13q losses are significantly more common in PTEN mutated cell lines. (b) DNA copy number frequency plots for TP53 mutated (upper panel) and wild-type cell lines (lower panel). Regions altered significantly (P<0.05) more frequent in TP53 mutant compared to wild-type cell lines are indicated (*) in the upper panel. Chromosome 1p and 5q35 gains and losses at 12q14 and 12q15–q21 are significantly more common in TP53-mutated cell lines.

Figure 3
figure4

(a) DNA copy number frequency plots for BRAF-mutated (upper panel) and wild-type cell lines (lower panel). Regions altered significantly (P<0.05) more frequent in BRAF mutant compared to wild-type cell lines are indicated (*) in the upper panel. (b) DNA copy number frequency plots for NRAS-mutated (upper panel) and wild-type cell lines (lower panel). The region (3q13.12–q13.31) altered (gained) significantly (P<0.05) more frequent in NRAS mutant compared to wild-type cell lines is indicated (*) in the upper panel.

Concomitant copy number alterations

Regions that are affected by DNA copy number changes in at least 40% of the cell lines, including eight gains and eight losses (as defined in Materials and methods), were analysed to detect patterns in copy number aberrations. Twelve pairs of aberrations were seen significantly more often than expected under the null hypothesis of independent genetic events (P<0.05) (Table 2). The highest correlation between alterations not affecting the same chromosomal arm was seen for 8p23.3–p23.1 loss and 8q12.1–q23.1 gain. Significant correlation was also seen for gain of chromosome 7 (where BRAF resides) and loss of chromosome 10 (PTEN).

Table 2 Altered regions that co-exist more often than expected by chance (P<0.05)

Discussion

CMM is an aggressive, heterogeneous disease where new markers for diagnosis, prognosis and treatment effect are needed. Genomic and gene expression profiling are powerful tools in this respect, for example, revealing gene sets that allow discrimination of vertical and radial growing CMM and that can be used for class discovery (Bittner et al., 2000; Haqq et al., 2005). Moreover, arrayCGH unraveled genomic aberrations specific for chronic sun-induced melanomas, and indicated that alteration in the CDKN2A and PI3K pathways are independent but complementary events in melanoma pathogenesis (Curtin et al., 2005).

Here, in a screen of 47 melanoma cell lines using tiling BAC-arrayCGH, we confirm the high frequency of chromosome 9p24.3–q13 and 10 loss, encompassing CDKN2A (9p21.3) and PTEN (10q23.31). Of interest is the novel melanoma susceptibility locus on 1p identified by linkage analysis and mapped by critical recombinants in linked families to a 15 Mbp region between D1S430 and D1S2664 at 1p31.1–1p21.3 (Gillanders et al., 2003). It was shown that 80% of familial tumors exhibited loss of heterozygosity at this region, with an SRO deletions of 9 Mbp between D1S207 and D1S435 at 1p31.1–1p22.2 (Walker et al., 2004), strongly suggestive of a classical tumor suppressor gene locus. Our data indicate two SRO losses in this region. The first maps to 1p22.1 and includes candidate genes such as TGFBR3 and CDC18, whereas the second SRO loss maps to 1p21.3 and includes only two known coding genes, PTBP2 and DPYD, and miR-137, a putative microRNA predicted (TargetScan; PicTar) to target the MITF, among others, and obviously warrants further analysis.

High frequency of copy number gains were found on 7q21.13–q31.1 and includes CDK6, encoding a p16INK4a antagonist, reported in increased gene copy numbers in 35% of ultraviolet-induced murine melanomas (Kannan et al., 2003). A second and more distal SRO gain maps to 7q32.1–q34 and includes BRAF, earlier found to be involved in complex alteration patterns (Daniotti et al., 2004). CCND1 amplification has been associated with the acral subtype of melanomas (Sauter et al., 2002) and here, 11q13 amplification was found in only three cell lines corroborating suggestions that CCND1 is not a general CMM oncogene. A gene that requires further analysis in this respect is CDK3 on 17q25.1, found to have an increased gene copy number in 57% of the cell lines. In contrast to the well-characterized cyclin D/cdk4/6-mediated inactivation of retinoblastoma protein (pRB) at the G1/S transition, cyclin C/cdk3 is implicated in promoting G0 cell-cycle exit through pRB phosphorylation (Ren and Rollins, 2004).

An interesting adjacent gene on 17q23.2 is TBX2, highly expressed in melanoma cells where it plays an important role in maintaining proliferation and suppression of senescence (Vance et al., 2005), possibly by downregulation of p21CIP1 and CDKN2A/p19ARF (Jacobs et al., 2000; Vance et al., 2005). TBX2 is one of the known targets for MITF in melanocytes, and a strong candidate melanoma oncogene (Carreira et al., 2000). Here, we showed that TBX2 has an increased gene copy number in 43% of the cell lines. Moreover, MITF at 3p13, was amplified in three cell lines and showed increased gene copy number in 36% of the cell lines, supporting a role in progressive and metastatic melanoma. Additionally, recurrent amplification was found on 1p12, the smallest amplicon including only NOTCH2 (Figure 2e). Intriguingly, a recent study has shown that Notch2 protein is significantly up regulated in dysplastic nevi and melanomas but not in common melanocytic nevi (Massi et al., 2006). Notch proteins are transmembrane receptors that are activated by specific ligands and increase signaling via the MAPK and PI3K pathways in melanoma cells (Liu et al., 2006). Well-known oncogenes such as MDM2, CCNE1 and BRAF were each found amplified in single samples. High-level amplification with unknown target genes were also identified, for example, EST69 melanoma cells had two distinct and narrow peaks on 18q11.2, spanning 950 and 870 kbp, respectively. The first peak contained six known genes including a laminin alpha 3 gene thought to be involved in cell adhesion, signal transduction and differentiation of keratinocytes (McLean et al., 2003), and the second peak included only ZNF521. A complex and intriguing amplification pattern was found on chromosome 5p in FM72 cells, including a large number (>20) of narrow peaks, each with only a few or a single gene.

CDKN2A is the major melanoma susceptibility gene and also a major tumor suppressor gene in nonfamilial melanoma (Grafstrom et al., 2005). Homozygous CDKN2A deletions were identified in 27/47 cell lines emphasizing their importance in melanoma cells and in vitro establishment. Although frequent in cultured cells, PTEN homozygous deletions or somatic mutations are less common in primary melanomas, suggesting that this is a late genetic event or that epigenetic mechanisms are responsible for PTEN silencing in primary tumors (Pollock et al., 2002). Moreover, PTEN is likely not the only target gene on chromosome 10. Interestingly, one cell line harbored a homozygous deletion on 10p11.21 spanning a single gene (PARD3), encoding an evolutionarily conserved protein and key regulator of epithelial and neural cell polarity and migration (Macara, 2004). Another novel homozygous deletion at 4q34 encompasses no known genes and was found in three samples. Whether this deletion is of biological importance in melanoma development or merely a copy number polymorphism is unknown. Finally, a homozygous deletion on 11q14 included the RAB38 gene, a homolog to a 9q21 melanoma susceptibility gene candidate (Jönsson et al., 2005a), and previously characterized as a melanocyte differentiating antigen, but also abundantly expressed in melanoma tissue (Zippelius et al., 2006).

CDKN2A was affected by homozygous deletion, mutation or methylation in the majority of cell lines. As expected, the single-cell line that harbored a CDK4 mutation was wild type for CDKN2A, but did show high-level CCNE1 amplification. In addition, we also confirmed that BRAF and NRAS mutations are mutually exclusive (Maldonado et al., 2003). However, our data also suggest that as there is no selection for coupled PTEN and NRAS mutations, both PTEN and BRAF mutation commonly occur in the same cells, emphasizing that BRAF is downstream of NRAS and does not affect the PI3K/PTEN pathway (Tsao et al., 2004). By searching for similar patterns of copy number changes, 11 concomitantly altered genomic pairs appeared (Table 2). The two top pairs included changes located on the same chromosome arm, possibly reflecting mechanistically related events. However, chromosome 7 gain was significantly correlated with chromosome 10 loss, functionally in accordance with coupled BRAF and PTEN mutation. Moreover, gain of chromosome 20 was correlated to gain of 15q22.2–q26.3, loss of 11q14.1–q25 and loss of chromosome 4. Another combination of interest was chromosome 10 loss and 17q21.31–qter gain, both regions including potential genes such as PTEN and CDK3/TBX2, respectively.

These patterns suggest that mutations in key genes, such as BRAF, NRAS, PTEN or TP53, can direct pathogenesis and the genomic pathways leading to specific melanoma phenotypes. Accordingly, global gene expression analysis was recently used to identify a set of 80 genes that separated BRAF-mutated and wild-type cell lines (Pavey et al., 2004). We used two-sided t-test to show that BRAF mutated cells have a high frequency (84%) of copy number gain on large regions of chromosome 7, which includes the BRAF gene on 7q34. A recent study found that both the mutant BRAF and the wild-type allele are targeted in copy number gains (Christensen and Guldberg, 2005). Chromosome 7 gain was less common in NRAS-mutated cells, the only aberration occurring at a significantly different frequency in NRAS mutant (40%) and wild-type cells (14%) was 3q13 gain. This indicates that BRAF-mutated cells are more homogenous in their genomic profiles than NRAS-mutated samples, corroborating the idea that BRAF mutation is an early genetic event. We further revealed that PTEN-mutated samples displayed a high frequency of 1p22.1–p21.1 loss, affecting a 10 Mb region and >50 genes, among them an elusive melanoma susceptibility gene (Gillanders et al., 2003).

In summary, we observed a considerable variety in genomic aberrations reflecting the heterogeneous nature of CMM. Careful molecular characterization of well-defined melanoma cell lines discloses the major genetic events that determine pathogenesis. Further analysis on clinical tumor samples will decipher more complex and hierarchical patterns and their relation to clinical outcome.

Materials and methods

Cell lines and culture

FM melanoma cell lines were established as described previously (Guldberg et al., 1997b). SK-MEL-3, SK-MEL-28, SK-MEL-19 and SK-MEL-31 cells were obtained from the American-type culture collection and EST cell lines from the ESTDAB Cell Bank (http://www.ebi.ac.uk/ipd/estdab/). All cells were cultured and maintained in RPMI 1640 medium with 10% fetal bovine serum. All were derived from CMM patients except EST128, which was derived from a primary ocular melanoma (Table 1).

DNA isolation and mutation analysis

Genomic DNA for arrayCGH and mutation analysis was isolated using the nucleospin tissue DNA extraction kit (Machery–Nagel, Easton, PA, USA) or the purescript DNA isolation kit (Gentra Systems, Minneapolis, MN, USA). BRAF, NRAS, TP53, PTEN, CDKN2A, CTNNB1 and CDK4 were screened using a combination of PCR and denaturing gradient gel electrophoresis (Guldberg et al., 1997a, 1997b; Christensen and Guldberg, 2005). Analysis of CDKN2A for deletions, mutations and promoter methylation was carried out as previously described (Gronbaek et al., 2000).

ArrayCGH

Microarrays were produced from the 32 K BAC clone library (CHORI BACPAC Resources, http://bacpac.chori.org/genomicRearrays.php) at the SWEGENE DNA Microarray Resource Center. Mapping data for each BAC clone was based on build hg17 (UCSC May 2004). DOP–PCR products were obtained from BAC DNA template and purified using filter–based 96-wells plates (PALL, East Hills, NY, USA), dried and resuspended in 50% dimethyl sulfoxide to a concentration of 500–1000 ng/μl. Arrays were printed on UltraGAPS slides (Corning, Corning, NY, USA) using a MicroGrid II spotter (Biorobotics, Cambridge, UK) (Jönsson et al., 2005a). For all samples, 2 μg of genomic DNA was labeled using a random labeling kit (Invitrogen Life Technologies, Carlsbad, CA, USA) (Jönsson et al., 2005b). Test DNA and male commercial reference DNA was differentially labeled, pooled, mixed with human COT-1 DNA, dried and resuspended in a formamide-based buffer. The hybridization reactions were applied to arrays, which were incubated under cover slips for 48–72 h at 37°C. Slides were washed (Snijders et al., 2001) and scanned using an Agilent Microarray scanner (Agilent Technologies, Santa Clara, CA, USA). Identification of individual spots on scanned arrays was performed with Gene Pix Pro 4.0 (Axon Instruments, Wheatherford, TX, USA), and the quantified data matrix was loaded into BioArray Software Environment (BASE) (Saal et al., 2002). Background-correction of Cy3 and Cy5 intensities was calculated using the median-feature and median-local background intensities. Within arrays, intensity ratios for individual probes were calculated as background-corrected intensity of sample divided by background-corrected intensity of reference sample. A signal-to-noise filter of 5 for the sample and reference channels was applied, and spots that failed to pass these criteria were excluded. The filtered data was, for each array separately, centralized to a median ratio of unity excluding X and Y chromosome clones. All filtering, normalization and analysis were performed in BASE. Subsequently a moving average of 150 kbp was applied and a BASE implementation of CGH Plotter was used to determine deletion/amplicon boundaries (Autio et al., 2003). The noise constant was set to 15 and the amplification/deletion limits were set to ±0.2. The noise constant is used to modulate the amount of segmental breakpoints on each chromosome. A high-noise constant corresponds to a large number of breakpoints per chromosome. CGH Plotter returned a sample data set transformed to a ternary scale (−1, 0 and 1) corresponding to (loss, normal and gain) for 27 834 BACs as described (Jönsson et al., 2005b). Amplification was defined as gains exceeding a log2 (ratio)>1.5 and homozygous deletions were considered when a log2 (ratio)−2 was observed.

Statistical analysis

Identification of regions associated with specific-mutated genes was performed using a standard two-sided two-sample t-test for individual BAC clones. The analysis of pair-wise genomic regions was performed as follows. Genomic regions, defined as two or more sequential gained or lost clones, altered in at least 40% of the cell lines were identified and recoded as binary variables (1 for gain/loss and 0 for no alteration). A 10% in the frequency variation was accepted inside a region, and consequently the end of an amplified region was determined by two or more consecutive clones with a frequency inferior to 36%. Associations between copy number changes, that is, analysis of whether certain alterations occurred more (or less) often than expected under the null hypothesis of no association, were evaluated using a permutation test. For each pair of two alterations, the observed number of co-events was compared to the null distribution constructed by simulating 10 000 replicates conditional on the observed marginal totals. All associations with P<0.05 were listed in Table 2. It should be noted that only those with P<0.0001 were considered significant after Bonferroni correction for multiple comparisons, most probably a too conservative correction. The same permutation-based association analysis was used when searching for excess co-occurrence of gene-specific point mutations. Data from EST128 were not included.

References

  1. Autio R, Hautaniemi S, Kauraniemi P, Yli-Harja O, Astola J, Wolf M et al. (2003). CGH-Plotter: MATLAB toolbox for CGH-data analysis. Bioinformatics 19: 1714–1715.

  2. Bittner M, Meltzer P, Chen Y, Jiang Y, Seftor E, Hendrix M et al. (2000). Molecular classification of cutaneous malignant melanoma by gene expression profiling. Nature 406: 536–540.

  3. Cachia AR, Indsto JO, McLaren KM, Mann GJ, Arends MJ . (2000). CDKN 2A mutation and deletion status in thin and thick primary melanoma. Clin Cancer Res 6: 3511–3515.

  4. Carreira S, Liu B, Goding CR . (2000). The gene encoding the T-box factor Tbx2 is a target for the microphthalmia-associated transcription factor in melanocytes. J Biol Chem 275: 21920–21927.

  5. Christensen C, Guldberg P . (2005). Growth factors rescue cutaneous melanoma cells from apoptosis induced by knockdown of mutated (V 600 E) B-RAF. Oncogene 24: 6292–6302.

  6. Curtin JA, Fridlyand J, Kageshita T, Patel HN, Busam KJ, Kutzner H et al. (2005). Distinct sets of genetic alterations in melanoma. N Engl J Med 353: 2135–2147.

  7. Daniotti M, Oggionni M, Ranzani T, Vallacchi V, Campi V, Di Stasi D et al. (2004). BRAF alterations are associated with complex mutational profiles in malignant melanoma. Oncogene 23: 5968–5977.

  8. Gillanders E, Juo SH, Holland EA, Jones M, Nancarrow D, Freas-Lutz D et al. (2003). Localization of a novel melanoma susceptibility locus to 1p22. Am J Hum Genet 73: 301–313.

  9. Grafstrom E, Egyhazi S, Ringborg U, Hansson J, Platz A . (2005). Biallelic deletions in INK4 in cutaneous melanoma are common and associated with decreased survival. Clin Cancer Res 11: 2991–2997.

  10. Gronbaek K, de Nully Brown P, Moller MB, Nedergaard T, Ralfkiaer E, Holler P et al. (2000). Concurrent disruption of p16INK4a and the ARF-p53 pathway predicts poor prognosis in aggressive non-Hodgkin's lymphoma. Leukemia 14: 1727–1735.

  11. Guldberg P, Nedergaard T, Nielsen HJ, Olsen AC, Ahrenkiel V, Zeuthen J . (1997a). Single-step DGGE-based mutation scanning of the p53 gene: application to genetic diagnosis of colorectal cancer. Hum Mutat 9: 348–355.

  12. Guldberg P, thor Straten P, Birck A, Ahrenkiel V, Kirkin AF, Zeuthen J . (1997b). Disruption of the MMAC1/PTEN gene by deletion or mutation is a frequent event in malignant melanoma. Cancer Res 57: 3660–3663.

  13. Haqq C, Nosrati M, Sudilovsky D, Crothers J, Khodabakhsh D, Pulliam BL et al. (2005). The gene expression signatures of melanoma progression. Proc Natl Acad Sci USA 102: 6092–6997.

  14. Hayward NK . (2003). Genetics of melanoma predisposition. Oncogene 22: 3053–3062.

  15. Jacobs JJ, Keblusek P, Robanus-Maandag E, Kristel P, Lingbeek M, Nederlof PM et al. (2000). Senescence bypass screen identifies TBX2, which represses Cdkn2a (p19(ARF)) and is amplified in a subset of human breast cancers. Nat Genet 26: 291–299.

  16. Jönsson G, Bendahl PO, Sandberg T, Kurbasic A, Staaf J, Sunde L et al. (2005a). Mapping of a novel ocular and cutaneous malignant melanoma susceptibility locus to chromosome 9q21.32. J Natl Cancer Inst 97: 1377–1382.

  17. Jönsson G, Naylor TL, Vallon-Christersson J, Staaf J, Huang J, Ward MR et al. (2005b). Distinct genomic profiles in hereditary breast tumors identified by array-based comparative genomic hybridization. Cancer Res 65: 7612–7621.

  18. Kannan K, Sharpless NE, Xu J, O'Hagan RC, Bosenberg M, Chin L . (2003). Components of the Rb pathway are critical targets of UV mutagenesis in a murine melanoma model. Proc Natl Acad Sci USA 100: 1221–1225.

  19. Liu ZJ, Xiao M, Balint K, Smalley KS, Brafford P, Qiu R et al. (2006). Notch1 signaling promotes primary melanoma progression by activating mitogen-activated protein kinase/phosphatidylinositol 3-kinase-Akt pathways and up-regulating N-cadherin expression. Cancer Res 66: 4182–4190.

  20. Macara IG . (2004). Parsing the polarity code. Nat Rev Mol Cell Biol 5: 220–231.

  21. Maldonado JL, Fridlyand J, Patel H, Jain AN, Busam K, Kageshita T et al. (2003). Determinants of BRAF mutations in primary melanomas. J Natl Cancer Inst 95: 1878–1890.

  22. Massi D, Tarantini F, Franchi A, Paglierani M, Di Serio C, Pellerito S et al. (2006). Evidence for differential expression of Notch receptors and their ligands in melanocytic nevi and cutaneous malignant melanoma. Modern Pathol 19: 616.

  23. McLean WH, Irvine AD, Hamill KJ, Whittock NV, Coleman-Campbell CM, Mellerio JE et al. (2003). An unusual N-terminal deletion of the laminin alpha3a isoform leads to the chronic granulation tissue disorder laryngo-onycho-cutaneous syndrome. Hum Mol Genet 12: 2395–2409.

  24. Omholt K, Platz A, Kanter L, Ringborg U, Hansson J . (2003). NRAS and BRAF mutations arise early during melanoma pathogenesis and are preserved throughout tumor progression. Clin Cancer Res 9: 6483–6488.

  25. Pavey S, Johansson P, Packer L, Taylor J, Stark M, Pollock PM et al. (2004). Microarray expression profiling in melanoma reveals a BRAF mutation signature. Oncogene 23: 4060–4067.

  26. Pollock PM, Harper UL, Hansen KS, Yudt LM, Stark M, Robbins CM et al. (2003). High frequency of BRAF mutations in nevi. Nat Genet 33: 19–20.

  27. Pollock PM, Walker GJ, Glendening JM, Que Noy T, Bloch NC, Fountain JW et al. (2002). PTEN inactivation is rare in melanoma tumours but occurs frequently in melanoma cell lines. Melanoma Res 12: 565–575.

  28. Ren S, Rollins BJ . (2004). Cyclin C/cdk3 promotes Rb-dependent G0 exit. Cell 117: 239–251.

  29. Saal LH, Troein C, Vallon-Christersson J, Gruvberger S, Borg A, Peterson C . (2002). BioArray Software Environment (BASE): a platform for comprehensive management and analysis of microarray data. Genome Biol 3: software0003.

  30. Sauter ER, Yeo UC, von Stemm A, Zhu W, Litwin S, Tichansky DS et al. (2002). Cyclin D1 is a candidate oncogene in cutaneous melanoma. Cancer Res 62: 3200–3206.

  31. Snijders AM, Nowak N, Segraves R, Blackwood S, Brown N, Conroy J et al. (2001). Assembly of microarrays for genome-wide measurement of DNA copy number. Nat Genet 29: 263–264.

  32. Takata M, Goto Y, Ichii N, Yamaura M, Murata H, Koga H et al. (2005). Constitutive activation of the mitogen-activated protein kinase signaling pathway in acral melanomas. J Invest Dermatol 125: 318–322.

  33. Tanami H, Imoto I, Hirasawa A, Yuki Y, Sonoda I, Inoue J et al. (2004). Involvement of overexpressed wild-type BRAF in the growth of malignant melanoma cell lines. Oncogene 23: 8796–8804.

  34. Tsao H, Goel V, Wu H, Yang G, Haluska FG . (2004). Genetic interaction between NRAS and BRAF mutations and PTEN/MMAC1 inactivation in melanoma. J Invest Dermatol 122: 337–341.

  35. Tucker MA, Goldstein AM . (2003). Melanoma etiology: where are we? Oncogene 22: 3042–3052.

  36. Walker GJ, Indsto JO, Sood R, Faruque MU, Hu P, Pollock PM et al. (2004). Deletion mapping suggests that the 1p22 melanoma susceptibility gene is a tumor suppressor localized to a 9-Mb interval. Genes Chromosomes Cancer 41: 56–64.

  37. Vance KW, Carreira S, Brosch G, Goding CR . (2005). Tbx2 is overexpressed and plays an important role in maintaining proliferation and suppression of senescence in melanomas. Cancer Res 65: 2260–2268.

  38. Zippelius A, Gati A, Bartnick T, Walton S, Odermatt B, Jaeger E et al. (2006). Melanocyte differentiation antigen RAB38/NY-MEL-1 induces frequent antibody responses exclusively in melanoma patients. Cancer Immunol Immunother Epub ahead of print.

Download references

Acknowledgements

This work was supported by grants from the Swedish Cancer Society, the Swedish Research Council, the Mrs Berta Kamprad Foundation, the Gunnar Nilsson Cancer Foundation, the Franke & Margareta Bergqvist Foundation, the American Cancer Society, the Lund University Hospital Foundations, the King Gustav Vs Jubilee Foundation, the Ingabritt and Arne Lundberg Foundation, the Swedish Foundation for Strategic Research, the Marianne and Marcus Wallenberg Foundation and by the Knut and Alice Wallenberg Foundation via the SWEGENE program. The support from Pieter de Jong and Kazutoyo Osoegawa, BACPAC Resource Center at CHORI, is acknowledged.

Author information

Correspondence to Å Borg.

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Jönsson, G., Dahl, C., Staaf, J. et al. Genomic profiling of malignant melanoma using tiling-resolution arrayCGH. Oncogene 26, 4738–4748 (2007). https://doi.org/10.1038/sj.onc.1210252

Download citation

Keywords

  • melanoma
  • arrayCGH
  • mutation

Further reading