Metastatic melanoma has traditionally been difficult to treat, and although molecularly based targeted therapies have shown promising results, they have yet to show consistent improvements in overall survival rates. Thus, identifying the key mutation events underlying the etiology of metastatic melanoma will no doubt lead to the improvement of existing therapeutic approaches and the development of new treatment strategies. Significant advances toward understanding the complexity of the melanoma genome have recently been achieved using next-generation sequencing (NGS) technologies. However, identifying those mutations driving tumorigenesis will continue to be a challenge for researchers, in part because of the high rates of mutation compared to other cancers. This article will review the catalog of mutations identified in melanoma through a variety of approaches, including the use of unbiased exome and whole-genome NGS platforms, as well discuss complementary strategies for identifying driver mutations. The promise of personalized medicine afforded by better understanding these mutation events should provide impetus for increased activity and rapid advances in this field.
Melanoma is a malignant skin cancer originating from the unregulated growth of melanocytes and in recent years has been increasing in incidence (Linos et al., 2009). Although not the most common form, melanoma accounts for the majority of skin cancer-related deaths. The high mortality rate associated with melanoma has largely been because of the lack of effective treatment options for late stage or disseminated disease. As such, patients with advanced melanoma have a median life expectancy of 6 months, with only 5% of patients surviving beyond 5 years (AIHW, 2008).
A number of important discoveries within the field of melanoma have been made over the last decade; this includes the approval of two new therapies in 2011 that are showing efficacy within the clinic (see Figure 1 for a timeline of major events). The first of these, Ipilimumab, is an immunotherapeutic approach utilizing anti-CTLA4 antibodies to promote immune responses through T cell proliferation and activation. The use of Ipilimumab has improved overall survival in patients with metastatic melanoma in phase III trials, either used as a monotherapy or in conjunction with dacarbazine (Hodi et al., 2010; Robert et al., 2011).
In contrast to the immunotherapeutic approach of Ipilimumab, the second drug approved in 2011, Vemurafenib (also known as PLX4032 or Zelboraf), is a small molecule inhibitor of constitutive kinase signaling activity caused by a valine to glutamic acid (or lysine) alteration at position 600 in BRAF, which occurs in approximately 50% of melanomas (Davies et al., 2002). Results from a recent phase III randomized clinical trial comparing Vemurafenib to dacarbazine showed an increase in overall survival at 6 months; however, long-term response rates have yet to be demonstrated because of the occurrence of tumor resistance (Chapman et al., 2011). A number of mechanisms of tumor resistance to Vemurafenib have been identified (Emery et al., 2009; Jiang et al., 2011; Nazarian et al., 2010; Poulikakos et al., 2011; Villanueva et al., 2010; Wagle et al., 2011), and it is anticipated that through application of this knowledge, together with continued research, long-term response rates and overall survival will be improved.
The successful development and subsequent approval of these drugs, in particular the latter molecular-based targeted approach of Vemurafenib, can be attributed to the extensive effort toward understanding the genetic etiology of melanoma. Characterization of the multitude of genetic alterations promoting the development and progression of melanoma has led to the identification of frequently mutated genes, of which a large proportion are amenable to therapeutic intervention. Significant progress in discovering and compiling these mutation events has recently been spearheaded with the advent of high-throughput candidate gene sequencing studies, together with whole-genome or exome next-generation sequencing (NGS) strategies.
This review, alongside companion reviews within the same issue of Pigment Cell and Melanoma Research, highlights the diversity of genetic events contributing to the pathogenesis of cutaneous (Walia et al., 2012) and uveal (Harbour, 2012) melanoma and explains how many of these are clinically relevant to the development of new therapies and improving outcomes for patients with metastatic melanoma. Lastly, this review will outline some of the problems associated with the avalanche of data produced by NGS technologies and consider aspects for identifying the therapeutically relevant mutation events that are responsible for driving the growth of melanoma.
Mutations affecting key biological pathways in melanoma
Many somatic mutations have been identified in melanoma. Not surprisingly, a number of these occur in genes within the same biological pathways. Characterizing these pathways will be important for understanding how these mutations contribute to biological effects underlying melanocytic neoplasia, in particular, those that alter signaling events controlling cell cycle regulation and apoptosis.
Some of the first genes identified to be mutated in melanoma were detected using candidate gene approaches in conjunction with traditional capillary-based Sanger sequencing. What have become known as the most relevant and frequently disrupted biological pathways in the development of melanoma were identified using these approaches. These classical pathways are briefly reviewed later but have also been extensively reviewed elsewhere (de Snoo and Hayward, 2005).
Although the advent of high-throughput technologies such as NGS have enabled researchers to rapidly and comprehensively identify mutations in an unbiased manner, another layer of complexity of analysis has arisen in discerning driver mutation events responsible for the development of the tumor, to those that are merely passenger events. This is of particular significance in understanding the etiology of melanoma, with the observation of significantly higher rates of mutations in melanoma compared to other tumor types (Pleasance et al., 2010; Wei et al., 2011) (and reviewed in more detail in Walia et al., 2012). This aside, a number of important emerging biological pathways and targets have been recently identified in melanoma and are briefly summarized later and in Figure 2.
Classical pathways in melanoma development
One of the most well known and characterized pathways commonly abrogated in melanoma is the mitogen-activating protein kinase (MAPK) signaling cascade, responsible for cell growth regulation and survival. Mutations in NRAS (mutated in 10–20% of melanomas), BRAF (approximately 50%), and less frequently MAP2K1 and MAP2K2 (approximately 8%) result in a high constitutive activity of ERK, leading to an increase in proliferation, angiogenesis, invasiveness, and metastasis (Cohen et al., 2002; Davies et al., 2002; Herlyn and Satyamoorthy, 1996; Murugan et al., 2009; Nikolaev et al., 2012).
In addition to activating the MAPK pathway, mutations in NRAS also constitutively activate the PI3-kinase pathway, leading to increased cell proliferation, apoptosis, and tumor cell chemoresistance (Davies et al., 2008; Omholt et al., 2006). Mutations occur at low frequency in other components of this pathway, including PIK3CA (approximately 3%) and AKT (approximately 1%) (Davies et al., 2008; Omholt et al., 2006). In a similar fashion, activating mutations in KIT (approximately 10%), upstream of PIK3CA and NRAS, can lead to the activation of AKT but are rarely observed in cutaneous melanomas, rather KIT mutation is an event more commonly associated with acral, mucosal, and chronically sun-damaged melanomas (Curtin et al., 2006). The PTEN gene, responsible for negative regulation of AKT signaling, is frequently mutated in melanoma cell lines (approximately 30–50%) through homozygous deletion and a variety of point mutations (Guldberg et al., 1997).
Another important tumor suppressor gene, CDKN2A, is also mutated through homozygous deletion or mutation in approximately 50% of melanomas (Flores et al., 1996). Its two encoded proteins, p16INK4A and p14ARF, generated through alternative splicing and different translation start sites regulate the pRB and p53 pathways, respectively (Pomerantz et al., 1998; Zhang et al., 1998). Additional mutations occur within other components of these important biological pathways controlling cellular senescence, DNA repair, apoptosis, and cell cycle progression, including TP53 (mutated in approximately 20% of melanomas), CDK4 (approximately 3%), and RB1 (approximately 3%) (Bartkova et al., 1996; Soto Martinez et al., 2005; Wolfel et al., 1995). Overall, the vast majority of melanomas have defects somewhere in each of the pRB, p53, MAPK, and PI3K pathways. These thus appear critical ‘gatekeeper’ pathways for melanoma genesis. In time, it may become evident that mutations in components of several other regulatory pathways are also required for melanoma development. The accompanying perspectives by Harbour (2012) and Walia et al. (2012) point to the potential involvement of some of these pathways, in particular, G-protein coupled receptor signaling, the glutamate pathway, and extracellular matrix regulation (summarized in Figure 2). Other pathways for which there is emerging evidence of a crucial role in melanoma are briefly mentioned later.
The role of fibroblast growth factor receptors in melanoma progression
The majority of molecular-based therapies target constitutive activity of mutated kinases, a family of >500 proteins involved in cell signal transduction which are highly amenable to therapeutic intervention. As such, it is not surprising that several research efforts have investigated the role of kinases in the development of cancer (Futreal et al., 2005; Stephens et al., 2005). The high rate of mutations of BRAF and less frequently, mutation of KIT in specific subtypes of melanoma, would suggest that other kinases may also be mutated and activated in this tumor type. Analysis of the four members of the fibroblast growth factor receptor (FGFR) family of tyrosine kinases implicated these as having a role in melanoma progression. Sequence analysis of FGFR1-4 in an initial cohort of 47 melanoma cell lines, followed by additional sequencing of FGFR2 in 66 samples, revealed a total of 15 different mutations in FGFR2 (mutated in approximately 10% of all samples tested). Additional analysis revealed mutations in three of 28 metastases and five of 72 primary tumors. However, unlike ERBB4, bioinformatic analysis and in-vitro functional assays indicated that the majority of mutations in FGFR2 result in a loss of receptor activity, rather than constitutive activation (Gartside et al., 2009).
The role of MAP3Ks in apoptosis
Recent exome sequencing has revealed inactivating mutations in another set of kinases (the MAP3 kinases) that, in contrast to promoting cell proliferation, are responsible for apoptosis (Stark et al., 2012). Targeted exome capture of eight metastatic melanoma cell lines and their matched normal DNA revealed non-synonymous mutations in MAP3K5 and MAP3K9. Prevalence screening for mutations in a larger series of samples revealed mutually exclusive mutations in MAP3K5 and MAP3K9 in 9 and 15% of melanoma cell lines, respectively. These inactivating mutations affected downstream signaling of the MAPK and JNK pathways, while mutations in MAP3K9 also contributed to resistance to the chemotherapeutic agent Temozolomide.
Genes mutated in melanoma requiring further functional analysis
Plexins are a family of transmembrane receptors involved in cell guidance regulation, motility, and invasion. Sequencing of this family of nine proteins in addition to copy number analysis through comparative genomic hybridization (CGH) identified copy gains of PLXNA4 as well as somatic mutations in PLXNA4, PLXNB3, and PLXNC1 in melanoma (Balakrishnan et al., 2009). A total of four mutations in 24 samples were identified (approximately 16%). Functional analysis revealed a number of these mutations resulted in impaired plexin function, suggesting a possible important role for plexins in tumor progression of a subset of melanomas.
Candidate gene sequencing of focal homozygous deletions involving four or less genes has recently identified a putative tumor suppressor, Trk-fused gene (TFG), mutated in approximately 5% of melanomas (Dutton-Regester et al., 2012, in preparation). Although no functional analysis was performed in this study, TFG has been implicated in numerous cancer types through gene fusions (Greco et al., 1995; Hernandez et al., 1999; Hisaoka et al., 2004). Non-synonymous mutations in TFG were mutually exclusive of BRAF/NRAS mutations, suggesting the possibility of an alternative mechanism to MAPK pathway activation. Indeed, a previous study indicated that TFG can activate both the MAPK and NF-KB pathways in a yeast hybrid screening assay (Matsuda et al., 2003). Further investigation into the functional significance of TFG in melanoma is warranted as tumors with mutations in TFG may be amenable to therapeutic approaches targeting the MAPK pathway.
An alternative immunological approach to Ipilimumab in the treatment of metastatic melanoma is through adjuvant therapy using vaccines containing highly expressed tumor-specific antigens (Rosenberg et al., 2011). Commonly upregulated genes in melanoma are those that are highly expressed in testis but undetectable in somatic tissues. One family that has not been extensively functionally characterized but has been implicated in melanoma genesis is the MAGE family. Sequencing revealed 37% of cell lines and 32% of fresh melanoma tumors contained at least one mutated MAGE family member; however, the biological significance of these mutations has yet to be determined (Caballero et al., 2010).
Mutations in melanoma have also been identified in the mitochondrial genome indicating that this might be an alternative mechanism driving growth compared to genomic mutation (Deichmann et al., 2004). Comprehensive analysis of the entire mitochondrial genome, using mitochondria sequencing arrays, identified 100 somatic mutations in 12 of 16 (75%) melanomas, of which nine had non-synonymous mutations in functional mitochondrial genes (Mithani et al., 2008). The role of these mutations in tumorigenesis is controversial as it is unclear whether they are driver mutations or are instead passenger events resulting as a consequence of oxidative damage during tumor development.
Complementary strategies for discerning between passenger and driver mutations in the development of melanoma
The genetic landscape of the melanoma genome is emerging as a complex picture. The high rate of mutation in melanoma caused by carcinogenic UV exposure is a major issue for discerning between driver and passenger mutation events. It is becoming apparent that rapid and comprehensive identification of driver mutations will require a variety of bioinformatic and experimental approaches (summarized in Figure 3). In addition, there are a number of biological considerations, including comparison of different histological subtypes of melanoma; the time of occurrence of the mutations in respect to the stage of tumor development (primary versus local versus distant metastasis); and if the mutations occur in nevi and other benign lesions. This section describes a number of complementary approaches that can be used in conjunction with data generated using NGS technologies.
RNA sequencing and transcriptomic analysis
RNA sequencing (RNA-seq) is a powerful sequencing approach that can be used to elucidate a number of mechanisms contributing to the development of cancer. Benefits of RNA-seq compared to DNA sequencing strategies include the identification of transcribed mutations that are translated at biologically relevant levels and the detection of transcribed in-frame fusion events. The significance of gene fusions, such as the BCR-ABL fusion in chronic myeloid leukemia, has extensively been documented in blood cancers; however, recent RNA sequencing studies indicate an emerging role for oncogenic fusions in a variety of solid epithelial cancers (Berger et al., 2010; Edgren et al., 2011; Ha et al., 2011; Nacu et al., 2011; Pflueger et al., 2011).
Gene fusions can structurally occur in a variety of ways, and those fusions that are translated into functional proteins are clinically relevant because they may be amenable to therapeutic approaches. In an attempt to identify druggable fusion events, RNA sequencing of ETS rearrangement-negative prostate cancers revealed RAF-associated fusion events mutually exclusive to BRAF V600 mutations. These were present in a variety of cancers, including approximately 2% (2/131) of melanomas (Palanisamy et al., 2010). Functional analysis indicated SLC45A3-BRAF and ESRP1-RAF1 (or CRAF) fusions promoted cell proliferation and colony foci formation and were susceptible to RAF and MEK inhibitor treatment.
In the first comprehensive survey of the melanoma transcriptome, Berger et al. (2010) used RNA-seq combined with high-resolution SNP arrays on 10 melanoma short-term cultures. This study revealed a number of findings that included the identification of 11 novel gene fusions, seven of which were translated in-frame. Interestingly, all of these fusion events were private or patient-specific events and were not detected in an additional set of 90 melanomas. As recurrent mutations are indicative of a causal biological role in tumorigenesis, determining the significance of private fusion events will likely require extensive functional experimental validation.
In addition to gene fusions, 12 novel read-through transcripts joining neighboring genes transcribed in the same orientation were identified. One of these read-through transcripts, involving 264 of 294 amino acids of CDK2 with an additional two residues from RAB5B, was detected in seven melanoma samples. Although the significance was not functionally determined, the observation of increased expression levels of CDK2 in CDK2-RAB5B read-through melanomas suggested a putative contribution to transcript stability influencing cellular proliferation. Evidence for the significance of read-through transcripts in cancer is accumulating, such as the recurrent SLC45A3-ELK4 fusion in prostate cancer (Maher et al., 2009). More importantly, there is currently no association of genomic aberrations with read-through transcripts, highlighting the benefits of complementary RNA sequencing strategies.
Lastly, RNA-seq analysis performed by Berger et al. (2010) also informed on point mutations in melanoma. A total of 721 novel non-synonymous mutations were identified; however, as matching germline DNA from the 10 melanoma specimens was not sequenced, it was estimated that approximately 30% of these represented true somatic mutations, while the remaining 70% comprised novel germline variants. Although RNA-seq analysis alone would not be sufficient to capture the entire complexity of mutation events occurring in melanoma, pairing this technique with exome or whole-genome sequencing would be an effective, versatile, and complementary approach for identifying clinically relevant mutations.
The role of mutations in non-coding regions
Cancer is a complex process involving numerous diverse mutation events that occur throughout the entire genome. Currently, there has been a strong focus on mutations residing within coding regions; understanding the functional effects of altered protein structures can be easily assessed using bioinformatics and can be subjected to robust experimental validation. However, as indicated by the first melanoma genome sequenced (Pleasance et al., 2010), large numbers of mutations are not only synonymous but also occur outside of coding regions. Although the majority of these are likely to be passenger mutations, some may have putative driving roles in tumorigenesis, affecting transcriptional control, regulation, splicing, and mRNA stability.
Of emerging significance in cancer is the role of small non-coding microRNAs (miRNAs) in the transcriptional control of gene expression. Expression of miRNAs is frequently deregulated in melanoma, and a number of miRNAs have been associated with pro-metastatic, proliferative, and invasive functions (Boyle et al., 2011; Felicetti et al., 2008; Segura et al., 2009; Stark et al., 2010). Although the mechanisms responsible for the deregulation of miRNA expression are incompletely understood, genomic aberrations may be important, as array CGH (aCGH) analysis has revealed significant alterations of chromosomal regions containing miRNAs in a variety of cancers (Zhang et al., 2006). Of these cancers, melanoma had the highest rate of copy number changes, with 86% of samples exhibiting alterations compared to 37% in ovarian cancer and 73% in breast cancer.
Building on these findings, a study comparing data from published cancer genomes recently analyzed the contribution of somatic mutations affecting miRNA binding in untranslated regions (UTR) of genes (Greenberg et al., 2011). As transcriptional control of genes is regulated through binding of miRNAs to specific motifs in the 3′UTRs, the finding that 50% of somatic mutations occur in UTRs prompted a computational approach to assess binding efficiency of miRNAs in mutated UTRs. Bioinformatic analysis revealed a statistically significant preference for miRNA binding to wild-type sequence in 207 mutated 3′UTR sites tested (P = 2.5 × 10−6) and that somatic mutation led to a general decrease of putative miRNA sites relative to the establishment of new sites.
Evidence suggests that through direct or indirect mechanisms, epigenetic regulation can also affect miRNA expression levels in cancer (Han et al., 2007). Studies utilizing methylation arrays, expression arrays, and bisulfite sequencing strategies have begun to elucidate the role of miRNA in melanoma (Bonazzi et al., 2011; Koga et al., 2009; Mazar et al., 2011); however, combining these data sets to explore the contribution of somatic mutation to this mechanism has not yet been comprehensively explored. Heritable germline mutations occurring within promoter regions affecting methylation and associated transcript expression, or ‘epimutations’, have been identified in the DNA mismatch repair genes MSH2 and MLH1 in colorectal cancer (Chan et al., 2006; Hitchins et al., 2007, 2011). It is interesting to speculate whether the occurrence of somatically acquired epimutations can occur and adds another layer of complexity to the melanoma genome, already complicated by high rates of mutation.
Lastly, the role of mutations in epigenetic regulation through chromatin and histone modification in melanoma will need to be established. Numerous genes involved in chromatin modification frequently exhibit mutations in a variety of cancers, including non-Hodgkin lymphoma, hepatocellular, gastric, and bladder cancer (Gui et al., 2011; Li et al., 2011; Morin et al., 2011; Wang et al., 2011). With regard to melanoma, recurrent genomic amplifications and upregulated expression of the histone methyl-transferase SETDB1 accelerated tumor formation in BRAF V600E mutated melanomas in zebrafish (Ceol et al., 2011), while loss of histone variant macroH2A contributes to CDK8 up-regulation and increased cellular proliferation (Kapoor et al., 2010). Furthermore, recurrent homozygous deletions of histone deacetylase 4 (HDAC4), nonsense mutations in HDAC3, and recurrent S722F mutations in transformation/transcription domain-associated protein (TRRAP) may indicate an important functional role of chromatin regulation in human melanoma development (Nikolaev et al., 2012; Stark and Hayward, 2007; Wei et al., 2011).
Use of functional screens and systems biology approaches
Identifying the catalog of mutations in cancer is no longer a rate-limiting step because of recent advances in NGS technologies. It is now apparent that discerning mutations driving tumorigenesis from bystander or passenger mutations represents a significant hurdle in the analysis of NGS data. Although a number of bioinformatic techniques and experimental design strategies can be utilized to prioritize putative driving candidates [briefly reviewed in the accompanying Perspective by Walia et al. (2012)], all mutations will ultimately require functional analysis to determine their respective significance. Coupling NGS data sets with high-throughput functional validation screens will be a powerful tool for the identification of novel genes contributing to melanoma development.
High-throughput functional screening approaches often link a variety of phenotypic responses from perturbed functions of singular genes on a genome-wide scale. It is hypothesized that abrogated gene function leading to a range of phenotypes associated with tumorigenesis, such as cell cycle control, proliferation, and invasiveness, will recapitulate similar effects for somatically mutated genes. The efficacy of RNA interference strategies using siRNA or shRNA knockdown of individual gene transcripts in the identification of cancer driver genes has been experimentally validated. An example of RNAi-assisted protein target identification (RAPID) screening in primary leukemia has demonstrated the versatility of such a high-throughput functional screening approach (Tyner et al., 2009). This study utilized a siRNA knockdown screen to identify members of the tyrosine kinase family critical to the survival of malignant melanoma cells. As a validation of the approach, the screen identified oncogenic addiction to JAK2 (V617F) and KRAS (G13D) in samples harboring these mutations, in addition to a number of novel tyrosine kinases including the identification of a previously undocumented activating insertion mutation in MPL. Similar RNAi analyses have recently identified a number of therapeutically relevant genes in other cancers including breast (Brough et al., 2011), colorectal (Eskiocak et al., 2011), and pancreatic cancer (Henderson et al., 2011).
Functional RNAi screens have begun to elucidate a number of regulatory proteins in melanomagenesis and pigment regulation and may give insight into mechanisms of melanoma development. This includes siRNA genome-wide synthetic library analysis to identify 92 novel genes responsible for pigment production in melanocytes (Ganesan et al., 2008) and an RNAi-based computational approach to identify components clustering to pathways of the known pigment regulator endothelin receptor type B (Ho et al., 2010). Utilizing RNAi-based screening strategies in a similar manner to target phenotypic changes associated with proliferative or metastatic potential in melanoma cells should prove useful for identifying driver genes or mutations.
Combining high-throughput data sets from multiple analytical approaches
Analysis of the first melanoma genome and subsequent exome reports has indicated that the genetic events contributing to melanoma development are complex. It is apparent that to understand the heterogeneous nature of the melanoma genome, a multifaceted approach using a variety of analytical and technological methods will be required. A number of studies have combined multiple data sets of aCGH and expression array data to identify causal drivers of melanoma development, progression, and metastasis. Chromosomal copy number gains and losses can occur as small focal events but commonly involve large stretches of genomic regions. This can present challenges in identifying causal genes involved in these regions because of the presence of large numbers of genes for which expression may be altered. In order to apply a methodological approach to identify driver events, Lin et al. (2008) combined CGH data with expression array data for 101 melanoma short-term cultures and cell lines. Essentially, genes involved in regions of chromosomal loss exhibiting a decrease in expression were deemed likely to represent candidate tumor suppressor genes, while those in regions of gain associated with increased expression as putative oncogenes. Results from this study indicated, although not functionally or experimentally confirmed, that such an analytical approach could identify driver genes involved in the development of melanoma.
Another study building on this analytical approach through use of a Bayesian network-based algorithm, postulated that driving mutation events are associated with a specific gene expression signature (Akavia et al., 2010). Analysis of combined copy number and expression array data with respect to pathway and modular connections identified a number of candidate driver genes in melanoma development. As proof of principle, this approach identified 10 known oncogenes and tumor suppressor genes within the top 30 hits, including the highest ranked gene MITF, a gene of high importance to melanocyte and melanoma biology. Functional analysis of two other candidates, TBC1D16 and RAB27A, indicated these genes were essential to tumor dependency and proliferation, further supporting this approach in identifying drivers of melanoma.
Candidate drivers of melanoma metastasis have also been identified through genomic and transcriptomic profile analysis combined with functional genetic screens for enhancers of cell invasion (Scott et al., 2011). This approach gathered 1597 differentially expressed genes between melanoma mouse models exhibiting polar differences in metastatic capabilities and by comparison to copy number and expression array data in human primary melanoma, narrowed the candidate list to 360 genes. Focusing on 295 genes with upregulated expression potentially amenable to therapeutic intervention, high-throughput functional screening invasive assays identified six pro-invasive metastatic candidate drivers. Further analysis revealed that one of these genes, encoding acid phosphatase 5 (ACP5), is a prognostic marker in human primary melanomas and involved in metastasis in vivo.
The aforementioned studies are excellent examples of multidimensional analytical approaches for identifying drivers in melanoma development; however, what remains is the collective analysis of copy number and expression data combined with somatic mutation data generated through NGS. Combining high-throughput data from a variety of methodological and technological platforms in a large number of samples will be essential for key identifying drivers in melanoma development.
A similar vision of overlaying high-throughput data sets to identify cancer mutations is currently being performed by large-scale collaborative efforts of the International Cancer Genome Consortium and The Cancer Genome Atlas, which aim to extensively characterize 50 tumor types (Hudson et al., 2010; Zhang et al., 2011). Using strict criteria, protocols, and sample requirements, each tumor will be interrogated for somatic mutation and germline variant analysis, structural rearrangements, copy number alteration, gene expression, and miRNA analysis. This approach has already led to significant success in understanding the genetic landscape of glioblastoma multiforme (CGARN, 2008) and ovarian cancer (CGARN, 2011). Although currently there is little data publically available, analysis of the melanoma genome is in progress and will be a valuable resource and reference in the future.
Utilizing mutation data in a personalized therapy approach to treat melanoma
The ultimate purpose of understanding the genetic mechanisms contributing to the development and progression of melanoma is to improve therapeutic outcomes for patients with the disease. Recent success with molecularly targeted therapies in melanoma has begun to impact patient survival and is quickly becoming a viable avenue for positive treatment outcomes. However, effective use of these approaches in the clinic will require the understanding of complex interactions between driver and passenger mutation events, the identification of prognostic indicators, and determining mechanisms of acquired resistance. To achieve this goal, a number of important experimental and analytical considerations will be necessary.
The use of BRAF inhibitors in melanoma has resulted in dramatic improvements on progression-free survival and tumor regression; however, acquired drug resistance is an important issue that needs to be addressed in order to improve rates of overall survival. It is hoped that by understanding mechanisms of drug resistance, a crucial weakness or ‘Achilles heel’ of the tumor will be exposed and concurrently be amenable to alternative therapeutic avenues. A number of acquired resistance mechanisms have already been identified, including secondary mutation events upstream and downstream of BRAF occurring in NRAS (Nazarian et al., 2010) and MAP2K1 (MEK1) (Emery et al., 2009), respectively. The detection of these resistance mutations was aided by the collection and analysis of patient tumor biopsies or established cell lines prior and post-drug treatment. Collections of these samples are crucial for understanding mechanisms of drug resistance with BRAF inhibitors and other emerging molecularly targeted therapies.
It is interesting to note at this point that a number of alternative mechanisms to BRAF inhibitor drug resistance have been identified that are currently not associated with nucleotide sequence mutations. Such acquired resistance mechanisms include up-regulation of protein tyrosine kinases PDGFRB (Nazarian et al., 2010) and COT (Johannessen et al., 2010) and more recently expression of aberrantly spliced BRAF V600E transcripts (Poulikakos et al., 2011). Although NGS has been extremely valuable in comprehensive mutation analysis, this technology will need to be paired with complementary analytical approaches to assess the entire complexity of resistance.
The majority of research investigating resistance to BRAF inhibitors has currently remained within the realm of post-treatment tumor analysis/acquired resistance as compared to inherent or de novo resistance (except for the COT study by Johannessen et al., 2010). Further research into this area is important as a proportion of BRAF V600E mutated patients do not respond to BRAF inhibitors (Chapman et al., 2011). Understanding de novo resistance may allude to a variety of unknown mechanisms or alternatively may be explained by resistant mutations already existing within the tumor. An example is the identification of a MAP2K1 (MEK1) P124L mutation in a treatment naïve early passage cell line which also carried a BRAF V600K mutation (Dutton-Regester et al., 2012, in preparation). However, from an experimental standpoint, understanding mechanisms of de novo resistance will require more complex analysis requiring intricate experimental design or the analysis of large collections of clinical specimens to observe trends of molecular alterations.
Effective utilization of mutation data within the clinic will require the development of methodologies that can rapidly and accurately assess the scope of mutations present within an individual’s tumor. A number of approaches have already been established including a panel for screening recurrent ‘hot spot’ mutations across large numbers of tumors using a single base pair extension reaction and detection via mass spectrometry (MacConaill et al., 2009; Thomas et al., 2007). An extension of this analysis was the development of a melanoma-specific mutation panel targeting 39 clinically relevant mutations in 20 genes mutated in melanoma, the majority of which are amenable to therapeutic intervention (Dutton-Regester et al., in press, in preparation).
One limitation to screening panels interrogating recurrent oncogenic events is that a large proportion of clinically relevant mutations driving tumor growth may not be assessed. Mutations that occur throughout the coding region of genes in a typical ‘tumor suppressor’ fashion are unable to be assessed using the aforementioned approaches. Ideally, use of NGS within the clinic is highly desirable, as the method can capture a comprehensive snapshot of mutations present in the tumor, and patient management can then be tailored based on this knowledge. It seems very likely that while NGS technology will eventually become part of routine clinical practice, there remains a variety of issues to debate and discuss, including overall costs, clinical utility and relevance, bioinformatic analysis, and timeline requirements.
Recently, a proof of principle report has begun to outline the key ethical issues, technological hurdles, and methodological outline of personalized oncology through integrative high-throughput sequencing (Roychowdhury et al., 2011). In this study, data from whole-genome sequencing, exome sequencing, and RNA-seq analysis were integrated to produce a comprehensive list of mutations within a timeframe of 3–4 weeks. These data were then analyzed by a board consisting of experienced clinicians and researchers so that optimal therapeutic avenues could be decided based on the mutation profile of the patient’s tumor. Using this process, a number of clinically relevant mutations were identified including mutations in NRAS, TP53, and amplification of CDK8 in a patient with malignant colorectal cancer, and mutations in MAP2K1 (MEK1), PI3K, and HRAS in a patient with an ulcerated spitzoid metastatic melanoma. Financially, the costs associated with this study were estimated at $6137 per patient, covering sequencing, bioinformatics and labor, thus indicating the high feasibility of such an approach within the clinic.
Regardless of the method of data generation, combining data from frequently observed genetic alterations in melanoma has also led to the development of a melanoma molecular disease model (Vidwans et al., 2011). This approach uses mutation data to separate melanoma development into five principal molecular subtypes and three additional secondary classes with potentially relevant therapeutics, respectively. This model could act as a scaffold to be continually added to over time as new mutations or clinically relevant mutation events are identified through NGS approaches.
The advent of NGS technology has rapidly changed the approach to understanding cancer genomics. While significant progress has been made in identifying key mutation events underlying melanoma development, the first forays into whole-genome and exome sequencing have shown we are only just beginning to understand the complex landscape of the melanoma genome. It has become apparent that the genetic architecture of melanoma is very heterogeneous and complex, having a very high mutation rate, which poses a challenge to identify genes driving tumorigenesis. Their discovery will require comprehensive bioinformatic strategies in conjunction with a multifaceted analytical approach by overlaying multiple complementary data sets. In addition, further improvements in functional screening assays should allow driver mutations to be more quickly distinguished from passenger mutations. Lastly, combining the above high-throughput approaches to clinical samples, both pre and post-treatment, should allow for improved patient management through molecularly based targeted treatment strategies. This ultimate vision of personalized therapy should lead to significantly better outcomes and improved rates of overall survival from melanoma.