An Unholy Alliance: Cooperation between BRAF and NF1 in Melanoma Development and BRAF Inhibitor Resistance

Ultraviolet radiation (UVR) exposure is the major causative factor of cutaneous melanoma. The central role of UVR in melanoma etiology has been convincingly shown through epidemiologic studies as well as whole genome and exome sequencing, which have shown the overwhelming majority of mutations in melanoma to be “UV-signature” C>T transitions. Despite the wealth of genetic data now available, the exact series of mutational hits required to transform a melanocyte into a melanoma cell is not fully understood. Most research to date has focused upon the role of nevi— common, benign melanocytic proliferations—as the precursor of melanoma development. The link between nevi and melanoma is suggested by the frequent occurrence of activating mutations in BRAF, a serine/threonine kinase known to drive the development and progression of 50% of all cutaneous melanomas (1). Despite harboring this deleterious oncogene, most nevi never undergo malignant transformation and remain growth arrested throughout the lifetime of the individual. The explanation for this apparent paradox was suggested by the observation that the introduction of oncogenic BRAF into primary melanocytes led to oncogene-induced senescence (OIS) and not cell proliferation (ref. 2; Fig. 1A). Histopathologic examination also showed the majority of human nevi samples to stain positively for senescence-associated β-galactosidase (SA-β-Gal; ref. 2). The mechanism by which some melanocytes harboring oncogenic BRAF apparently avoid OIS and undergo malignant transformation remains to be elucidated. Although mutant BRAF plays a critical role in melanoma development, inputs from other signaling cascades are also required. Of these, the best-studied pathway is phosphoinositide 3-kinase (PI3K)/AKT, which is constitutively activated in melanoma through the loss of the tumor suppressor PTEN, increased expression of AKT3, and rarely as the result of mutations in either PI3K or AKT3 (3). Both animal modeling and cell culture studies have shown a cooperating role for the PI3K/AKT pathway in BRAF-mediated melanoma development, in which its activity seems to prevent the entry of melanocytes into OIS (ref. 4; Fig. 1A).

In this issue of Cancer Discovery, 2 studies provide new data implicating the loss of expression or function of neurofibromin (NF1) in melanoma pathogenesis (5, 6). The first, from Maertens and colleagues from the Cichowski laboratory, describes the development of a transgenic mouse model in which mutant BRAF and NF1 loss cooperate to drive melanoma initiation and progression (5). The second, from Whittaker and colleagues from the Garraway laboratory, details a high-throughput short hairpin RNA (shRNA) screen that identified NF1 loss as a mediator of acquired and intrinsic BRAF inhibitor resistance (6). Both of these studies follow earlier work from the 1990s showing loss of neurofibromin expression to be a frequent event in both melanoma cell lines and tissue specimens (7).

NF1 is a known tumor suppressor and a negative regulator of RAS signaling. Under physiologic conditions, it acts to stimulate the GTPase activity of RAS, leading to its accumulation in the GDP-bound (inactive) state (ref. 8; Fig. 1B). Loss of NF1 function releases the negative regulation of RAS, resulting in increased signaling through downstream pathways, including the PI3K/AKT and the mitogen-activated protein kinase (MAPK) signal transduction cascades (Fig. 1B). The NF1 gene was first described in the context of the familial cancer syndrome neurofibromatosis type I (occurrence: 1 in 3,500), a disorder characterized by inactivating mutations in NF1 leading to multiple neurofibromas, hyperpigmented macules of the skin (café-au-lait macules), freckling, and iris hamartomas (Lisch nodules; ref. 8). Patients with NF1 typically show an increased cancer incidence and a reduced lifespan, with tumor development occurring following the acquisition of secondary somatic mutations in NF1 (leading to its homozygous inactivation). Despite pigmentation defects being an important clinical characteristic of NF1 patients, increases in melanoma incidence have not been reported and the pigmented lesions that arise are not accompanied by increased melanocyte proliferation (9, 10). Café-au-lait macules, which present as large areas of skin pigmentation, result from the increased accumulation of melanosomes in skin melanocytes and keratinocytes. Similarly, freckles are areas of increased skin pigmentation that also occur independently of melanocyte proliferation. In a transgenic mouse model of Nf1 haploinsufficiency (+/−), neurofibromin regulates the KIT-MITF signaling axis during melanocyte development (9). Melanocytes derived from Nf1^+/− mice exhibit higher basal MAPK pathway activity than their wild-type counterparts and show increased expression of melanin synthesis genes (9). Likewise, melanocytes from NF1 patients have higher melanin levels and a higher level of tyrosine hydroxylase activity than do those derived from normal controls.

Although the link between NF1 loss and melanocyte dysregulation is well established, a role for NF1 inactivation in melanoma development has never been shown. In their study, Maertens and colleagues (5) began investigating the influence of NF1 loss upon signaling pathways implicated in the escape from RAF-induced OIS. In cells that retained NF1 function, RAF activation suppressed RAS and AKT signaling, an effect associated with growth inhibition. In contrast, induction of RAF in Nf1-null cells did not inhibit RAS and AKT signaling and proliferation were both maintained (ref. 5; Fig. 1B). The authors next asked whether mutant BRAF and NF1 loss cooperated to drive melanoma development in vivo. To explore this hypothesis, a transgenic model was developed in which mice carrying a conditional inactivating mutation in Nf1 were crossed with those harboring a conditional activating mutation in Braf. Significant increases in the incidence of melanoma development were noted in mice harboring mutant Braf and Nf1 loss, compared with those with mutant Braf alone (melanoma incidence: 57% vs. 22%, respectively). A link between melanoma development and escape from BRAF-mediated OIS was suggested by the increased PI3K/AKT signaling and reduced levels of SA-β-Gal staining observed in tumors lacking Nf1 (5). These data confirm the growing body of evidence implicating PI3K/AKT signaling in BRAF-driven melanomagenesis and suggest the existence of alternate routes to malignant transformation for melanocytes (4).

Recent years have seen major progress in the therapeutic management of disseminated melanoma, with BRAF inhibitors (such as dabrafenib and vemurafenib) showing good levels of tumor regression and increased progression-free survival in patients whose melanomas harbor V600 position mutations in BRAF (11, 12). Despite these successes, only half of these patients showed objective responses to therapy, and the median progression-free survival was less than 7 months (11). This heterogeneity in response and the ultimate development of resistance in nearly all patients have been attributed to multiple underlying mechanisms, including BRAF splice forms, PTEN loss, increased receptor tyrosine kinase (RTK) signaling, COT activity, and the acquisition of NRAS and MAP–ERK kinase (MEK ) mutations (13).

Despite reports of a number of potential BRAF inhibitor resistance mechanisms, relatively few unbiased screens have been conducted to identify novel mechanisms of therapeutic escape. To address this, Whittaker and colleagues (6) conducted a high-throughput assay in which a BRAF inhibitor–sensitive melanoma cell line was targeted with a library of 90,000 shRNAs specific for 16,600 genes. Among 31 candidates genes, NF1 was identified as the top-ranking hit whose knockdown abrogated the growth inhibitory effects of the BRAF inhibitor PLX4720. As expected, the effects of NF1 loss upon PLX4720 resistance were mediated through the derepression of RAS signaling leading to CRAF-mediated activation of the MAPK pathway (5, 6). These observations agree with previous reports implicating RAS and CRAF in acquired BRAF inhibitor resistance and confirm the central role of the MAPK signaling pathway in the escape from BRAF inhibitor therapy (14). Although clinical trials have already shown the use of combining BRAF and MEK inhibitors in patients with melanoma, these may not be effective for every BRAF-mutant melanoma genotype (15). In melanoma cell lines harboring mutant BRAF/NF1, the BRAF + MEK (PLX4720 + AZD6244) inhibitor combination had little effect upon either MAPK signaling or cell growth, with more impressive results seen following treatment with either a pan-RAF inhibitor (AZ628) or an extracellular signal–regulated kinase (ERK) inhibitor (VTX-11e; ref. 6; Fig. 1C). The lack of MEK inhibitor efficacy in this context is not easily explained, but could be a consequence of either limited drug potency (other MEK inhibitors could be more effective) or differences in feedback inhibition seen following MAPK pathway targeting at different nodes. In the transgenic Braf/Nf1 mouse model, limited antitumor responses were observed following single-agent MEK inhibition (PD0325901), with enhanced effects seen following combination with the mTOR inhibitor rapamycin (ref. 5; Fig. 1C). Although more work is needed to establish the best combination therapeutic strategy for BRAF/NF1-mutant melanomas, the data so far suggest that this genetic subgroup may require similar approaches to those under investigation for RAS-mutant melanoma.

The potential clinical significance of these findings was confirmed in a panel of melanoma cell lines and tumors (tissue microarray), in which NF1 expression was found to be reduced or absent in 36% and 18% of samples, respectively (5). However, protein expression assays may underestimate the prevalence of NF1 deficiency, as loss of NF1 function may also result from acquired mutations and altered posttranslational processing without lowering total protein levels. With this in mind, the percentage of patients whose melanomas show impaired NF1 function could be higher than what the above numbers suggest. Although defects in NF1 sometimes overlapped with BRAF mutation status, this was not universally exclusive and some co-occurrences with NRAS mutations were noted (6). Interestingly, loss of PTEN expression and impaired NF1 function were not mutually exclusive either, suggesting some nonoverlapping functions between the 2 tumor suppressors. An additional analysis of a cohort of 121 melanoma specimens and cell lines identified potentially pathogenic NF1 mutations in 25% (5/21) of samples that were wild-type for both BRAF and NRAS (6).

The study of samples from patients with melanoma who received BRAF inhibitor therapy suggested a role for NF1 loss in both intrinsic and acquired drug resistance. Whole-exome sequencing of pre- and posttreatment specimens conducted by Whittaker and colleagues (6) identified 4 patients with NF1 mutations, with one individual harboring a deleterious nonsense mutation (R2450*) in NF1 in both pre– and post–drug treatment specimens. A second series of 5 matched pre- and posttreatment melanoma samples analyzed for neurofibromin expression by immunohistochemistry showed that 2 of 5 expressed little or no protein before BRAF inhibitor treatment (5). In the remaining 2 of 3 with initial neurofibromin expression, vemurafenib treatment was associated with loss of protein expression.

As further discoveries in melanoma genomics and proteomics are made, a complex and heterogeneous field of personalized medicine is evolving. It is important to keep in mind that not all mutations and activated pathways identified in the laboratory or in whole-genome studies will actually be causative in patients' disease. The identification of NF1 as a modifier of BRAF inhibitor response seems supported in both laboratory and patient specimen data but still requires prospective clinical evaluation. Whether patients with BRAF-mutant/NF1-deficient melanoma will derive more benefit with a pan-RAF, ERK, or MEK + mTOR inhibitor combination remains to be answered. The concept of BRAF^V600 mutational status in melanoma as the sole biomarker for BRAF-targeted therapy is quickly becoming obsolete. Although the phase III studies underway to determine whether BRAF/MEK inhibitor combination therapy is better than BRAF inhibitor monotherapy (NCT01584648; NCT01689519) are important for drug approval, they do not directly address this issue of the complex heterogeneity of patients with BRAF-mutant melanoma. An adaptive strategy is needed to integrate real-time genomic and proteomic data for an individualized approach to optimize targeted therapy. It is hoped that greater strides will be achieved by prospectively selecting combination therapies based on broader molecular profiles to optimize a personalized approach in patients with melanoma.

G.T. Gibney is a consultant/advisory board member of Genentech/Roche. No potential conflicts of interest were disclosed by the other author.

The authors thank Inna Fedorenko for assistance with the figure and Dr. Katherine Nathanson (The University of Pennsylvania, Philadelphia, PA) for helpful discussions about the role of NF1 in cancer development.

The work in K.S.M. Smalley's laboratory is supported by grants U54 CA143970-01 and R01 CA161107-01 from the NIH.