Beyond BRAF(V600): clinical mutation panel testing by next-generation sequencing in advanced melanoma.

The management of melanoma has evolved owing to improved understanding of its molecular drivers. To augment the current understanding of the prevalence, patterns, and associations of mutations in this disease, the results of clinical testing of 699 advanced melanoma patients using a pan-cancer next-generation sequencing (NGS) panel of hotspot regions in 46 genes were reviewed. Mutations were identified in 43 of the 46 genes on the panel. The most common mutations were BRAFV600 (36%), NRAS (21%), TP53 (16%), BRAFNon-V600 (6%), and KIT (4%). Approximately one-third of melanomas had >1 mutation detected, and the number of mutations per tumor was associated with melanoma subtype. Concurrent TP53 mutations were the most frequent events in tumors with BRAFV600 and NRAS mutations. Melanomas with BRAFNon-V600mutations frequently harbored concurrent NRAS mutations (18%), which were rare in tumors with BRAFV600 mutations (1.6%). The prevalence of BRAFV600 and KIT mutations were significantly associated with melanoma subtypes, and BRAFV600 and TP53 mutations were significantly associated with cutaneous primary tumor location. Multiple potential therapeutic targets were identified in metastatic unknown primary and cutaneous melanomas that lacked BRAFV600 and NRAS mutations. These results enrich our understanding of the patterns and clinical associations of oncogenic mutations in melanoma.

Introduction

According to the American Cancer Society estimates, 76,100 patients are expected to be diagnosed with melanoma and 9,710 patients are predicted to die from the disease in 2014.(Siegel et al., 2014) Melanoma is a complex, heterogeneous disease with multiple signaling pathways implicated in its molecular pathogenesis. A key advance in the understanding and treatment of this disease was the discovery of frequent recurrent somatic mutations that result in substitutions of the valine at position 600 in the gene encoding the BRAF serine-threonine kinase (BRAF mutations) in the RAS-RAF-MEK-ERK signaling pathway.(Davies et al., 2002) Large single-center studies, meta-analyses, and whole exome-sequencing efforts have subsequently confirmed that BRAF mutations are the most common activating genetic event detected in cutaneous melanomas.(Hocker and Tsao, 2007; Hodis et al., 2012; Jakob et al., 2012; Krauthammer et al., 2012) Studies of clinically annotated specimens have identified significant clinical associations with BRAF mutations, including melanoma subtype, primary tumor location, and prognosis.(Bauer et al., 2011; Curtin et al., 2005; Jakob et al., 2012; Long et al., 2011) More recent studies have also identified significant differences in demographics, primary tumor features, and clinical outcomes between patients with the two most common BRAF substitutions observed in melanoma, BRAF and BRAF.(Bucheit et al., 2013; Menzies et al., 2012) Significant clinical associations have also been identified with oncogenic NRAS and KIT mutations.(Handolias et al., 2010; Jakob et al., 2012) These findings support the concept that studies of mutations may not only help identify therapeutic targets but also provide insights into the molecular pathogenesis and natural history of melanoma. There is also growing evidence that mutations correlate with differential clinical benefit with approved systemic therapies for this disease.(Nathanson et al., 2013; Trunzer et al., 2013) Notably, recent whole-exome sequencing (WES) studies have demonstrated that melanoma has one of the highest rates of somatic mutations among all cancers, and have identified many additional genes that are mutated recurrently in this disease.(Berger et al., 2012; Hodis et al., 2012; Krauthammer et al., 2012)

Due to the approval and clinical testing of treatments that target specific genetic mutations, molecular testing is now performed routinely for patients with advanced melanoma. The expansion of therapeutic options concurrent with technical advances has led to the development of a growing number of clinical molecular testing platforms. Increasingly, panel-based testing approaches are being utilized in order to maximize the efficient use of both patient materials and clinical testing infrastructure and resources. In addition to providing an opportunity to identify therapeutic options for patients, the data generated by such testing provides an opportunity to improve our understanding of the molecular heterogeneity of this disease. Such data can also be used to identify hypotheses for prospective studies that may improve patient testing and/or clinical management.(Curtin et al., 2006; Curtin et al., 2005; Woodman et al., 2012)

In this study, we reviewed our institution's results from clinical molecular testing by next-generation sequencing (NGS) of commonly mutated regions in 46 genes using a pan-cancer panel (AmpliSeq panel, Life Technologies; Table S1) in 699 consecutive patients with advanced melanoma. This data has been analyzed for the prevalence and overlap of mutations, and their concordance in a subset of patients with testing on multiple samples. The molecular data has also been analyzed for associations with clinical subtypes (ie. cutaneous, acral, mucosal, uveal, and unknown primary melanoma) and primary tumor location. The results of this study reinforce the molecular complexity of this disease and identify clinical associations for TP53 and BRAF mutations.

Results

Mutation Prevalence

The cohort (n=699) included advanced melanoma patients with known cutaneous (n=484, 69%), acral (n=54, 8%), mucosal (n=43, 6%), and uveal (n=13, 2%) primary melanomas. A subset of patients had metastatic disease without a known primary tumor (referred to as “unknown primary”, n=104, 15%). Next generation sequencing of regions affected recurrently by mutations in cancer identified at least one mutation in 43 of the 46 tested genes in the full cohort of patients (Figure S1 and S2; Table S2). The most prevalent mutations in the entire cohort were BRAF (n = 251; 36% of all patients), NRAS (n=150; 21%), TP53 (n=110; 16%), BRAF (n=39; 6%), and KIT (n=27; 4%) substitutions.

The most common mutations in cutaneous melanomas were BRAF (41%), NRAS (22%), TP53 (17%), and BRAF (7%) (Figure 1a). Metastatic melanomas without a known primary tumor demonstrated a very similar mutation spectrum (39% BRAF, 22% NRAS, 19% TP53, 4% BRAF) (Figure 1b). The most common mutations in acral melanomas were NRAS (24%), BRAF (19%), KIT (11%), and TP53 (6%) (Figure 1c). The most common mutations in mucosal melanomas were NRAS (21%), KIT (16%), TP53 (9%) and BRAF (7%) (Figure 1d). The majority (92%) of the small cohort of uveal melanomas had no mutations detected in the 46 gene panel, which notably did not include BAP1, GNAQ, or GNA11, which are mutated frequently in this melanoma subtype.

Figure 1

Prevalence of detected gene mutations by melanoma subtype. Panels show the rate of gene mutations observed in (a) cutaneous melanomas, (n=484); (b) unknown primary melanomas, (n=104); (c) acral melanomas, (n=54); and (d) mucosal melanoma, (n=43).

Characteristics and Overlap of Detected Mutations

There is evidence that different substitutions in individual oncogenes correlate with distinct molecular and clinical characteristics, including BRAF.(Bucheit et al., 2013; Garnett et al., 2005; Menzies et al., 2012; Wan et al., 2004) In this cohort, the most common BRAFsubstitutions were V600E (76% of all V600 mutations; 28% of all patients), V600K (17%; 6%), and V600R (2.4%; 0.8%) (Figure S3a). Mutations resulting in 20 different substitutions at sites other than V600 in BRAF were also detected. The most frequent BRAF substitutions were G469E (18% of BRAFmutations; 1% of all patients), G469R (13%; 0.7%), and K601E (11%; 0.5%) (Figure S3b and Table S2). For NRAS, mutations affecting Q61 were the most prevalent (77% of all NRAS mutations), followed by G12/13 (20%) (Figure S3c). The most common KIT mutations were L576P (27% of KIT mutations, 1% of all patients), K642E (20%; 0.8%), and N822Y (10%; 0.4%) (Figure S3d). Possible ultraviolet radiation (UVR)-related mutations (C>T or G>A transitions)(Berger et al., 2012) were detected overall in 39 of the 43 mutated genes (Figure S4). Among the genes mutated in >5% of the samples, TP53 displayed the highest frequency of UVR signature mutations (66%). Likely UVR-associated substitutions represented only 11% of detected NRAS mutations, 8% of BRAF mutations, and none of the KIT mutations. CC>TT substitutions, which also provide strong evidence of UVR-induced DNA damage, were detected in 15 tumors, with TP53 (n=5) the most frequently affected gene (Table S2).

Approximately one-third (n=213) of the melanomas had ≥2 mutations. Concomitant mutations were present in 35% of tumors with BRAFmutations, 55% of tumors with BRAFmutations, and 50% of tumors with NRAS mutations (Fig. 2a-c). TP53 mutations were the most common overlapping mutation in tumors with NRAS (17%) and BRAF (12%) mutations (12% of V600E; 7% of V600K; Figure S5). The most frequent overlapping mutation in melanomas with BRAF mutations was NRAS, which was mutated in 18% of these tumors. In contrast, concurrent NRAS mutations were detected in only 1.6% of tumors with BRAFmutations. Mutations in TP53 (13%) and KRAS (10%) were also relatively common in tumors with BRAFmutations. ATM (11%), NRAS (7%), and CTNNB1 (7%) were the most prevalent concomitant mutations in tumors with KIT mutations (Figure 2d.) The rate of co-occurring BRAF and NRAS mutations in tumors with rare but potentially targetable mutations (i.e. PI3K-AKT pathway, EGFR, MET) are presented in Table S3.

Figure 2

Prevalence of concurrent mutations among melanomas with common gene mutations. Panels show the rate of concurrent mutations present in melanomas with (a) BRAF; (b) BRAF; (c) NRAS; and (d) KIT.

Associations with Melanoma Subtype and Primary Tumor Location

The overall rate of mutations varied significantly by melanoma subtype. Mucosal (44%) and acral (33%) melanomas were more likely to have no mutations detected than cutaneous (15%) and unknown primary (20%) melanomas (p<0.0001) (Figure S6). As mentioned previously, no mutations were detected in the majority (92%) of the small cohort of uveal melanomas.

BRAF, and KIT mutations, which were the most frequent events overall in the cohort, were assessed for associations with clinically-defined melanoma subtypes (Figure 3a). This analysis identified significant associations for BRAF (p<0.001) and KIT (p<0.001) mutations. BRAF mutations were more frequent in cutaneous and unknown primary melanomas, whereas KIT mutations were more prevalent in acral and mucosal melanomas. TP53 mutations were also more common in cutaneous and unknown primary melanomas, but this differential distribution did not reach statistical significance (p=0.059). The prevalence of NRAS mutations varied very little by subtype.

Figure 3

Associations of mutations with clinical subtypes and primary tumor location. (a) Gene mutation rates in different melanoma subtypes (Black, cutaneous; Striped, acral; White, mucosal, Spotted, unknown primary). * BRAF mutations (p<0.001) were significantly associated with cutaneous and unknown primary melanomas, whereas KIT mutations (p<0.001) were significantly associated with acral and mucosal melanomas. (b) Primary tumor location of prevalent gene mutations in cutaneous melanomas (Black, head/neck; Striped, trunk; White, extremity). * BRAF mutations were significantly associated with the trunk compared to the head/neck (p=0.0004), while TP53 mutations were significantly associated with the head/neck compared to the trunk (p=0.03) or extremities (p<0.0001).

Among non-acral cutaneous melanomas, the prevalence of BRAF (p=0.001) and TP53 (p=0.0002) mutations were significantly associated with primary tumor location (Figure 3b). The rate of BRAFmutations was higher in the primary tumors of the trunk (49%) compared to the head/neck (30%, p=0.0004). TP53 mutations were more frequent in primary tumors of the head/neck (26%) compared to the trunk (16%, p=0.03) or extremities (8%, p<0.0001). BRAF mutations trended towards an association with primary tumor location (p=0.055), and were more common in primary tumors of the head/neck (10%) compared to the extremities (4%) on pairwise comparison (p=0.025). NRAS mutations were not significantly associated with primary tumor site.

Among the cutaneous and unknown primary melanomas, 113 tumors (19%) were wild-type for both BRAFV600 and NRAS mutations. The most common mutations in this cohort were TP53 (n=51; 45%), BRAF (n=24; 19%) and KIT (n=13; 10%) (Figure 4). Other potentially targetable genes in which mutations were detected in this cohort included EGFR (n=7; 6%), ERBB4 (n=7; 6%); CDKN2A (n=5; 4%); PIK3CA (n=4; 3%), PDGFRA (n=3; 2%) and PTEN (n=2; 2%). Mutations in KRAS (n=7; 7%) and HRAS (n=5; 4%) were also identified in the cohort.

Figure 4

Mutations detected in cutaneous and unknown primary melanomas without a BRAF or NRAS mutation (n=113).

Concordance of Mutations

Thirty-seven patients had NGS data available for more than one tumor, including 23 patients with paired primary tumors and metastases. Highly concordant results were observed for BRAF (100%) and NRAS (97%) among the matched primaries and metastases (Figure S7). One patient with two primaries had discordant BRAF testing results. However, this was not unexpected as the patient had a primary mucosal melanoma (BRAF wild-type) and a primary cutaneous melanoma (BRAF). Three patients (8%) had discordant results for TP53. Two patients had a TP53 mutation in the primary lesion that was not detected in the metastasis (H179Y → WT; L194F → WT), while the third patient had a wild-type TP53 in the primary tumor and mutation in the metastasis (WT → R306*). One patient each demonstrated discordance in PTEN, KRAS, CTNNB1, APC, and FBXW7.

Mutation prevalence between primary (n=248) and metastatic (n=486) samples overall demonstrated similar distribution. For both primary and metastatic samples, BRAF (32% primary; 37% metastatic), NRAS (18%; 23%), and TP53 (17%; 15%) were the most common gene mutations in descending order. KIT (6%) was the next most common in primary tumors followed by BRAF (5%) whereas the reverse was true for metastatic tumors (BRAF 6%; KIT 4%). Only STK11 showed a significant difference in prevalence between metastatic samples (2%) and primary tumors (0%, p=0.033). A total of 68 patients (10%) received chemotherapy before the removal of the lesion that was used for sequencing. Only 2 genes showed significant increases in mutation rates in the post-chemotherapy tumors compared to the chemotherapy-naïve tumors (MLH1, 3% vs 0.3%, p=0.05; RB1, 4% vs 1%, p=0.03).

Discussion

Molecular testing is now performed routinely for patients with advanced melanoma. In addition to guiding therapeutic decision making, the information from such testing can also provide insights into the molecular basis and heterogeneity of this disease. In this study we have reviewed the results of clinical next generation sequencing (NGS) of regions of 46 genes in a pan-cancer panel in a cohort of 699 melanoma patients. The results represent the largest cohort of melanomas to date analyzed by multiplexed NGS and add insights to some of the discoveries from recent whole exome sequencing efforts, including to our knowledge previously unreported molecular and clinical associations for TP53 and BRAFmutations.(Hodis et al., 2012; Krauthammer et al., 2012) While this pan-cancer panel does not examine certain melanoma-specific genes of interest, it does provide an opportunity to assess other genes not commonly tested by focused, single gene approaches in this tumor type.

Mutations affecting the V600 site of BRAF and hotspots in NRAS were the most frequent mutations observed in our cohort of 699 patients that underwent next generation sequencing for regions of 46 genes. Despite the possible bias that could have occurred due to patients being selected for molecular testing in the clinical setting, the mutation rates for both of these genes, particularly in the cutaneous melanomas, are similar to other large series and meta-analyses of melanoma patients tested for these mutations.(Hocker and Tsao, 2007; Jakob et al., 2012) Hotspot mutations in BRAF and NRAS were also the most common mutations identified in two recent whole-exome sequencing (WES) studies of cohorts of 121 and 147 melanomas (Hodis et al., 2012; Krauthammer et al., 2012), and in the preliminary publicly available results reported for the melanoma component of the Cancer Genome Atlas (TCGA) effort.[1] While the rates of BRAF mutations in the cutaneous melanomas in the published WES studies were slightly higher than observed here, one of those studies included cell lines, and both studies were limited by the requirement for frozen tumors with sufficient DNA for WES. In addition, patients with known BRAF mutations detected by outside testing before patients were seen at our institution may not have undergone CMS46 analysis, thus potentially contributing to the lower percentage of BRAF mutations in this cohort.

Consistent with previous studies by ourselves and others, BRAFmutations were more frequent in cutaneous melanomas than in acral and mucosal melanomas.(Hocker and Tsao, 2007; Jakob et al., 2012) The rates of BRAF, and TP53 mutations in melanomas with an unknown primary tumor were nearly identical to the rates observed in cutaneous melanomas. This result is also consistent with similar rates observed between unknown primary melanomas and melanomas with a known cutaneous primary in our previous analysis of patients at our center who underwent DNA pyrosequencing for BRAFmutations and NRAS hotspot mutations only, and also consistent with a more recent study which analyzed the mutational status of BRAF, NRAS, and KIT in 44 patients with unknown primary melanoma.(Egberts et al., 2014; Jakob et al., 2012) The additional data for the similar rates of BRAF and TP53 mutations provides further support for the hypothesis that the majority of melanomas with an unknown primary tumor likely had an occult cutaneous primary.

In addition to position 600, our pan-cancer panel included sequencing of multiple other residues in exons 11 and 15 of BRAF, where most mutations in this gene have been identified. (Davies et al., 2002) Mutations that affected sites in the BRAF protein other than V600 (BRAF) were detected overall in 5% of the patients in this study. BRAF mutations were detected in 7% of cutaneous, 4% of unknown primary, 2% of acral, and 2% of mucosal melanomas. Among the cutaneous and unknown primary tumors that did not have BRAFor NRAS mutations (n=113), BRAF mutations (19%) were the second most common mutation detected. Dahlman et al previously reported an 8% prevalence of BRAF mutations in a smaller cohort (n=49) of patients without BRAFor NRAS mutations.(Dahlman et al., 2012) A review of publicly available preliminary data from the melanoma TCGA effort, which is restricted to melanomas with a primary tumor arising on non-glabrous skin but includes sequencing of all exons of BRAF, identified 21 BRAF mutations in 266 melanomas (8%). Among the TCGA melanoma cases that were wild-type for BRAF and NRAS mutations (n=72) the rate of BRAF mutations was 21%. As 10 of the 21 BRAF mutations detected in the TCGA were in exons other than 11 or 15, it is possible that the actual prevalence of BRAF mutations in our cohort was slightly higher, although the functional significance of most of those mutations is unknown. A portion of these mutations also co-occurred with classic activating mutations in exon 15 and thus are not favored to be oncogenic. Preclinical studies have demonstrated that BRAF mutations do not respond to the FDA-approved BRAF inhibitors vemurafenib and dabrafenib, which were selected for development based on their specificity for BRAF proteins with V600E substitutions. However, other agents may be active in melanomas with BRAF mutations, including MEK inhibitors and pan-RAF inhibitor sorafenib.(Dahlman et al., 2012; Garnett et al., 2005; Smalley et al., 2009; Wan et al., 2004) Dramatic and durable clinical responses have been reported in isolated metastatic melanoma patients with BRAF mutations affecting the L597 residue in early phase clinical trials of the MEK inhibitors trametinib and TAK-733.(Dahlman et al., 2012; Falchook et al., 2012; Kim et al., 2013) Clinical trials have been planned to systematically evaluate the activity of the MEK inhibitor trametinib in patients with BRAF mutations.

BRAF mutations demonstrated significant clinical and molecular differences compared to BRAFmutations. While BRAFmutations were detected at higher rates in melanoma patients with cutaneous and unknown primary tumors, the prevalence of BRAF mutations was not significantly associated with melanoma subtype. However, the power to detect significant differences was limited by the comparatively lower prevalence of the BRAF mutations. More strikingly, the pattern of co-mutations was distinct for BRAF and BRAF mutations. In this large cohort, only 1.6% of the melanomas with detected BRAFmutations had a concurrent NRAS mutation. This prevalence is similar to the rate observed in our previous pyrosequencing study.(Jakob et al., 2012) In contrast, concurrent NRAS (18%) and KRAS (11%) mutations were frequent events in melanomas with BRAF mutations. Previous in vitro characterization of 15 BRAF mutations demonstrated heterogeneous effects on the serine-threonine catalytic activity of the BRAF protein.(Garnett et al., 2005; Wan et al., 2004) While some mutations increase the activity of BRAF to a level comparable to that observed with V600 mutations, other mutations are only partially activating, and some cause the kinase activity to be decreased compared to the wild-type protein. Experimental data supports that the non-activating BRAF mutations still increase the activity of MAPK pathway signaling through increased formation of multiprotein complexes with CRAF and active RAS proteins.(Wan et al., 2004) The co-occurrence of BRAF mutations with activating RAS mutations may therefore cooperate to activate the MAPK pathway to a greater degree than is achieved with either event alone.

One of the important insights from recent whole exome sequencing studies of melanoma was the identification of frequent TP53 mutations. While previous studies had suggested that TP53 mutations were quite rare in melanoma (Albino et al., 1994; Castresana et al., 1993; Lubbe et al., 1994), a recent WES study identified a rate of 19%.(Hodis et al., 2012) While we cover most exons but did not fully sequence the TP53 gene, and thus may underestimate their prevalence, we found a very similar rate of TP53 mutations (16%) in this cohort. TP53 mutations were the most common mutations identified after BRAF and NRAS mutations in both this study and the WES study, underscoring their frequency and potential significance. TP53 mutations strongly trended (p=0.06) toward an association with melanoma subtype, with lower prevalence in acral and mucosal melanomas compared to cutaneous and unknown primary melanomas. TP53 mutations were also associated with primary tumor location, with higher prevalence observed in melanomas with head/neck primary tumor location. This result could reflect an etiological role for UVR in TP53 mutations. Consistent with this hypothesis, 66% of the observed TP53 mutations were associated with typical UVR-induced changes. TP53 mutations were frequent in melanomas with concurrent BRAF, BRAF, and NRAS mutations. As TP53 mutations have been significantly associated with clinical outcomes and therapeutic resistance in other cancers,(Hoffmann et al., 2008; Lindenbergh-van der Plas et al., 2011; Poeta et al., 2007; Temam et al., 2000) future studies will test the predictive and prognostic significance of these events in melanoma. Notably, we did observe discordant results for TP53 mutation status in three patients (13%) who had molecular testing data for both primary tumors and metastases. This suggests that testing of archival material alone may not be adequate to accurately determine the significance of TP53 mutations as a predictive marker of response to systemic therapies in patients with metastatic disease. However, the analysis of the paired specimens did overall demonstrate highly concordant results for this panel of genes mutated recurrently in cancer.

While BRAFand NRAS were the most frequent mutations observed in our study, 19% of patients with cutaneous and unknown primary melanomas had neither of these mutations. The identification of therapeutic targets in these patients is a key challenge and clinical need. The most common genes in this cohort in which mutations were detected by our panel were TP53 (45%), BRAF (19%), and KIT (10%). As described above, clinical responses have been observed in early-phase clinical trials of the MEK inhibitors trametinib and TAK-733 in metastatic melanoma patients with BRAF mutations.(Dahlman et al., 2012; Falchook et al., 2012; Kim et al., 2013) A number of non-randomized clinical trials with the KIT inhibitor imatinib have been conducted in advanced melanoma patients with KIT mutations. The clinical response rates in these trials have ranged from 16% to 29%. (Carvajal et al., 2011; Guo et al., 2011; Hodi et al., 2013) One imatinib study identified the presence of a concurrent NRAS mutation, which we observed in 7% of the KIT-mutant patients in this cohort, as a predictor of resistance. Other genes that are potentially actionable in which rare mutations were detected include EGFR (6%), ERBB4 (6%), PIK3CA (3%), and PDGFRA (2%). However, the functional and clinical significance of the majority of the mutations detected in these genes is currently unknown.

As described above, the AmpliSeq 46-gene panel can provide important clinical information. Notably the panel allows for the consolidated evaluation of multiple important cancer genes and oncogenic mutations in one assay, including a number of clinically actionable aberrations. However, we recognize that this pan-cancer panel has significant limitations for this study. The panel includes a number of genes that currently have unknown relevance to melanoma. The panel also fails to include certain mutations that have been detected in cutaneous melanomas in recent WES studies, such as TERT, NF1, and RAC1, as well as mutations detected in other melanoma subtypes (i.e. BAP1, GNAQ, GNA11). The high frequency of acral, mucosal, and uveal melanomas with no mutations detected supports the need to consider other molecular testing panels for those subtypes, or the augmentation of the panel with genes relevant for those subtypes. In addition, the sequencing of only certain regions of many of the genes may not be adequate to annotate certain genes, particularly tumor suppressors that can be affected by frameshift mutations at many loci. For example, the observed mutation rates for CDKN2A (2.4%) and PTEN (1.6%) in our cohort are lower than those observed in recent WES studies. The inclusion of only 46 genes also makes it technically challenging to accurately quantify significant copy number variation (CNV), which could have affected these and other genes in the panel. This may further explain why tumor suppressors that show frequent losses through deletions that are not adequately detected by this platform, such as CDKN2A and PTEN, were less prevalent in our study than those that are more frequently affected by missense and frameshift mutations, such as TP53. Exploratory studies are underway to determine what CNVs can be detected and validated reliably using the AmpliSeq platform, and will be reported in the future. Such CNV information may also be forthcoming from panels that include sequencing of all exons for genes of interest, and larger numbers of genes. Currently, our pan-cancer panel is expanding to include more comprehensive coverage of prevalent cancer genes, including those that may prove clinically relevant to melanoma. However, gene translocations, such as those recently reported for BRAF (Botton et al., 2013; Hutchinson et al., 2013), are not assessable using this panel. Ultimately, the integration of multiple types of molecular data that will be available in the near future from the melanoma TCGA effort will likely provide additional molecular insights into the mutations observed in this study. Subsequent studies in which molecular data can be integrated with clinical characteristics and outcomes will be critical to personalizing and optimizing patient management.

In summary, our study presents the results for the largest cohort to date of melanoma patients to be analyzed by clinical multiplexed NGS. Consistent with previous studies, we found that BRAFand NRAS hotspot mutations were the most common molecular aberrations detected. We also observed frequent TP53 mutations, consistent with recent data from WES studies, with information about their associations with melanoma subtypes and primary tumor location. In addition, our study adds to the growing understanding of the prevalence and molecular patterns of BRAF mutations, which have emerged as a therapeutic target. These results provide a basis for future focused studies of these molecular events, and further supports the rationale for integrated analyses of clinical molecular testing data from other centers.

Materials and Methods

Mutation Testing

Molecular testing was performed on DNA extracted from formalin-fixed, paraffin-embedded (FFPE) tissues from melanoma primary tumors or metastases for which molecular testing was clinically indicated. DNA was extracted using standard methods, and was analyzed for mutations using the AmpliSeq sequencing panel (Life Technologies) as previously described.(Singh et al., 2013) Detailed methodology is provided in Supplementary Materials. The regions analyzed for mutations in each of the 46 genes in the panel are listed in Table S1.

Data Analysis

This study was conducted according to the Declaration of Helsinki Principles, and all analyses were performed under an Institutional Review Board approved protocol. In accordance with this protocol, all samples used were obtained from patients who gave their written informed consent for the use of their archival tissue for research purposes or from deceased patients who have samples stored in the Department of Pathology at MD Anderson Cancer Center. Clinical NGS data, patient demographics, and disease characteristics were obtained from institutional pathology and clinical databases. Review of the publicly available TCGA melanoma samples (batches 180, 198, 206, and 240) was performed through the online TCGA data matrix (https://tcga-data.nci.nih.gov/tcga/dataAccessMatrix.htm; accessed on 04/01/2014). Associations were evaluated using either Fisher's exact tests or Freeman-Halton tests using SAS v9.3 for Windows. P-values less than 0.05 were considered statistically significant.

Table S1. Genes and associated codons tested by AmpliSeq 46-gene panel, Life Technologies.

Table S3. Co-mutations with less common mutated genes.

Figure S1. Gene mutatons identfied in cohort of melanoma patents (n=699). Each column reflects the results for one patent. The first row indicates clinical melanoma subtype (Purple, cutaneous; Green, mucosal; Blue, acral; Yellow, uveal; White, unknown primary). Bars in each subsequent row show the presence of a mutaton in the indicated genes. Specific BRAF mutatons are indicated by color (Red, V600E; Pink, V600-other; Gray, non-V600; Blue, >1 BRAF mutaton type) in the second row.

Figure S2. Prevalence of mutations in each gene in the full cohort of patients. The prevalence of BRAF mutations and BRAF mutations are shown separately.

Figure S3. Molecular characteristics of detected mutations. Panels show the rate of the indicated substitutions observed among melanomas with (a) BRAF (n=251); (b) BRAF (n=39); (c) NRAS (n=150); and (d) KIT (n=27) mutations.

Figure S4. Percentage of mutations detected in each gene that reflect likely UVR damage (White, G>A substitutions; Black, C>T substitutions).

Figure S5. Prevalence of concurrent mutations among melanomas with specific BRAF gene mutations. Panels show the rate of concurrent mutations present in melanomas with (a) BRAF; and (b) BRAF

Figure S6. Percent of patents with 0/1/2/3+ mutations by melanoma subtype (Black, 0; Striped, 1; White, 2; Spotted, 3 or more mutations).

Figure S7. Concordance of driver genes in patents with tumor pairs (n=37). Each bar represents the percent concordance of the specific driver gene mutation (Black, BRAF; Striped, NRAS; White, TP53; Spotted, Other) among tumor pairs tested. * 1 case of BRAF discordance occurred in a patient with primary mucosal melanoma (BRAF wild-type) and primary cutaneous melanoma (BRAF)