One of the hallmarks of cancer treatment is the heterogeneity of clinical outcomes. Such heterogeneity has been observed with both classical cytotoxics and newer targeted approaches that inhibit the pathways dysregulated by alterations in the oncogenes or tumor suppressors responsible for cancer initiation and progression. With inhibitors of activated kinases such as BRAF, variability in treatment response has been hypothesized to be the result of differences in the complement of genetic and epigenetic changes present within individual tumors.
In contrast to many solid tumors, cell line models can be readily derived from the majority of patients with invasive melanomas. Furthermore, a single missense mutation in BRAF (V600E) is found in over 90% of BRAF-mutant melanomas. These two factors allowed the generation of a large number of V600E
BRAF-mutant melanoma models for use in studying the genomic diversity of this disease. To identify the V600E
BRAF-mutant cohort, we first screened 149 melanoma cell lines and short-term cultures for mutations in BRAF. Genome-wide copy-number profiling was then performed to identify genomic alterations that co-occur with and potential diminish dependence on V600E
BRAF. In our analysis, we noted a large number of both broad and focal copy-number alterations in cells expressing V600E
BRAF. This high mutational burden in melanoma raised concern that inhibition of a single oncogenic driver such as BRAF would be unlikely to result in meaningful anti-tumor activity. Despite such concerns, we observed that V600E
BRAF melanoma cell lines were dependent upon MAP kinase pathway activation for cell proliferation with only rare exception. Thus, the existence of multiple genetic lesions in melanoma does not, in general, eliminate dependence on the V600E
BRAF mutation. This conclusion is consistent with a recent clinical trial of the RAF inhibitor PLX4032 in which an 81% partial response rate was observed in patients with advanced melanoma whose tumor harbored V600E
BRAF (Flaherty et al., 2010
However, our data and that of others do suggest that the complement of mutations that are found to co-occur with mutant BRAF does impact the response of such cells to selective BRAF and more generally ERK pathway inhibition (Gopal et al., 2010
; Sondergaard et al., 2010
). This contention is consistent with the clinical data, in which the responses to PLX4032 varied widely in both durability and the extent of tumor regression. In the current study, we identify PTEN loss as one genetic event that abrogates the effects of ERK pathway inhibition in V600E
BRAF tumors. Loss of PTEN expression was observed in approximately one quarter (9/40) of the V600E
BRAF melanoma cell lines examined. We observed that cells with concurrent BRAF mutation and PTEN loss were typically dependent upon ERK signaling, but the nature of the dependence was altered by concurrent loss of PTEN. Although, ERK was required for the proliferation of 17 of the 19 V600E
BRAF cell lines examined in detail, it only promoted survival in the subset wild type for PTEN. It thus appears that secondary mutation of PTEN in a tumor with mutant BRAF renders the survival function of the latter redundant. These data suggest that tumor regressions are likely to be less frequent and of decreased magnitude in such tumors. Furthermore, our data provide a rationale for combined targeting of the RAF/MEK and PI3-kinase pathways in the cohort of patients with concurrent BRAF mutation and PTEN loss.
Resistance to the anti-proliferative effects of RAF and MEK inhibition was observed in two of the 19 V600E BRAF-mutant melanoma cell lines, a fraction consistent with the ~20% of patients who fail to derive any clinical benefit from the RAF inhibitor PLX4032 (Flaherty et al., 2010
). In each case, both PTEN and RB1-loss coexisted with V600E
BRAF. Furthermore, inactivation of RB1 in MEK-dependent, V600E
BRAF-mutant, PTEN-null, RB1 wild-type cell lines upon infection with viral E7 conferred high-level resistance to MEK inhibition. These results suggest that loss of RB1 function in the setting of PTEN loss is sufficient to render the proliferation of V600E
BRAF-mutant cell lines independent of oncogenic BRAF.
Members of the INK4 protein family inhibit cdk4 (cyclin-dependent kinase 4) and cdk6-mediated phosphorylation of the retinoblastoma susceptibility gene product (RB1). p16INK4A
, a prototypic INK4 protein, functions as a tumor suppressor in many human cancers and is mutated or deleted in most but not all melanomas. Patients with germline mutations in CDK4
and those with hereditary retinoblastoma demonstrate an increased risk of developing melanoma, and in patients with CDK4 mutations, somatic alterations in the INK4A gene locus are not observed (Draper et al., 1986
; Eng et al., 1993
; Ohta et al., 1994
). Consistent with these reports, the SKMEL-1 and SKMEL-28 cell lines, which harbor cdk4 amplification and mutation, respectively, express p16INK4A
(Supplementary Figure 5
). We observed that loss of RB1 and p16INK4A
were also mutually exclusive events in our cohort of V600E
BRAF melanoma cell lines. Detailed characterization of the two V600E
BRAF, RB1-null models, demonstrated that in both cases, p16INK4A
protein was expressed (). Furthermore, direct sequencing indicated that both RB1-null models were CDKN2A
wild type within the p16 coding region of the gene (Supplementary Table 3
). In the case of A2058, a small focal deletion was detected within the CDKN2A
locus as previously reported (Kumar et al., 1998
). Close inspection of this deletion confirmed that it encompassed only exon 1β of the ARF
gene, which encodes the N-terminal portion of p14ARF
but is outside of the coding region of INK4A
(). These findings are consistent with the notion that p14ARF
have distinct tumor suppressive functions in melanoma. They also suggest that somatic loss of RB1 function may represent an alternative pathogenic event in melanomagenesis, which is mutually exclusive with INK4A inactivation.
Recent data demonstrating that RB1, like p16INK4A
, can abrogate oncogene-induced senescence, provides a potential basis for the co-selection for RB1-loss in V600E
BRAF tumors and its mutual exclusivity with CDKN2A
(Chicas et al., 2010
). However, whereas loss of both p16INK4A
and RB1 may have overlapping roles in promoting tumor formation by preventing oncogene-induced senescence, our data suggest that loss of RB1, in contrast to loss of p16INK4A
, leads to a diminished dependence upon BRAF.
Our work has translational implications for the ongoing clinical trials of RAF and MEK-selective inhibitors. Specifically, the data suggest that despite the high level of genomic complexity in melanoma, the vast majority of V600EBRAF melanoma models exhibit BRAF dependence and therefore inhibitors of BRAF or MEK may be effective in this setting. Early clinical trials of the MEK inhibitors PD0325901 and AZD6244 demonstrated low, but reproducible response rates (Adjei et al., 2008
; Dummer et al., 2008
). Although these early trials were not stratified for BRAF mutational status, the majority of responders were patients with melanoma. Additionally, in those cases where tissue was available for study, most patients with objective responses had tumors with BRAF mutations (Dummer et al., 2008
). Nevertheless, the low response rate with MEK inhibitors even in patients whose tumors harbored BRAF mutation, raised the possibility that the high level of genomic complexity observed in melanoma was the basis for the limited efficacy of this approach in patients. The high response rate of PLX4032 along with our current data, however, imply that the lower response rate of MEK inhibitors in BRAF-mutant patients was unlikely to be due to a lack of BRAF dependence in the majority of patients harboring the oncogene.
One explanation for the significantly greater clinical efficacy of the RAF versus the MEK inhibitors in patients with BRAF-mutant melanoma is that oncogenic BRAF may promote tumor progression in part through MEK-independent effectors. The data we present here suggest that this possibility is unlikely to be the major determinant of individual variability, as BRAF-mutant tumors with concurrent PTEN and RB1 loss that exhibited resistance to MEK inhibition were also resistant to RAF inhibition. Rather, the greater efficacy of the RAF inhibitor observed in clinical trials is more likely attributable to the more potent inhibition of MAP kinase pathway activity achievable with PLX4032 in patients (Bollag et al., 2010
). The MEK inhibitor downregulates MAP kinase activity in all cells including normal tissues whereas the RAF inhibitor PLX4032 inhibits MAP kinase activity only in cells expressing mutant BRAF (; Joseph et al., 2010
). This appears to confer a broader therapeutic index, which allows for greater pathway inhibition with PLX4032 and, thus, increased anti-tumor efficacy in patients with BRAF-mutant tumors.
In summary, our data suggest that the complement of mutations associated with the evolution of mutant BRAF melanoma condition the biologic function of ERK signaling and thus sensitivity to selective MAP kinase pathway inhibitors. The prevalence of functional inactivation of RB1 in melanoma remains unknown. Given the higher frequency of CDKN2A
deletions in cultured melanoma cells versus tumors, it is possible that the prevalence of RB1 alterations will also prove to be lower in primary and metastatic melanomas than that observed in our cell line panel (Gonzalgo et al., 1997
). It is anticipated that the ongoing Tumor Cancer Genome Atlas project in melanoma will define the frequency of RB1 alterations in previously untreated melanomas and their co-occurrence with BRAF mutation and PTEN loss. The data presented here support screening for RB1 and PTEN inactivation in patients with advanced metastatic melanoma to determine whether RB1 and PTEN loss are associated with diminished clinical benefit with selective inhibitors of RAF and MEK.