Targeting RAF kinases for cancer therapy: BRAF mutated melanoma and beyond

The spectrum of BRAF mutations

Identification of BRAF mutations in cancer ushered in a new era in the treatment of advanced melanomas. BRAF is mutated in ~8% of all cancers, and roughly half of all melanomas harbor a BRAF^T1799A transversion, which encodes the constitutively active BRAF-V600E oncoprotein. In the original description of BRAF mutations in cancer, BRAF-V600E was only one of 14 BRAF alterations identified in cell lines and primary tumor samples¹⁴. Since then, nearly 300¹⁵ distinct missense mutations have been observed in tumor samples and cancer cell lines (Figure 1). These missense mutations encompass 115 of the 766 BRAF codons, yet the majority of mutations are observed in the activation loop (A-loop) near V600, or in the GSGSFG phosphate binding loop (P-loop) at residues 464–469^15,16 (Figure 1). Crystallographic analysis revealed that the inactive conformation of BRAF is stabilized by interactions between the A- and P-loops of the BRAF kinase domain, specifically involving V600 interacting with F468¹⁷. Under normal circumstances, reversible phosphorylation of T599 and S602 in the A-loop regulates the A-loop–P-loop interaction allowing BRAF to convert back and forth from its kinase-active to the kinase-inactive state. Consequently, BRAF mutations that lead to amino acid substitutions in either the A-loop or the P-loop mimic T599 and S602 phosphorylation and, by disrupting the A-loop–P-loop interaction, irreversibly shift the equilibrium of BRAF to the kinase-active conformation.

BRAF V600 point mutations are clearly the most common oncogenic driver in melanoma, but BRAF-V600E melanoma represents only a subset of tumors with BRAF alterations. BRAF point mutations also occur in 60% of thyroid, 10% of colorectal carcinomas and in 6% of lung cancers, as well as nearly all papillary craniopharyngioma¹⁸, classical hairy cell leukemia^19,20, and metanephric kidney adenoma²¹. Unlike other indications where V600 mutations predominate, BRAF alterations in lung cancer often occur in the P-loop at G466 and G469 (Figure 1). While the frequency of BRAF mutation in colon and lung cancer are considerably lower, the relative morbidity for these indications (50,000 and 158,000 deaths respectively in the US²²) may compose an even larger population of patients with BRAF-mutated cancers that are currently ineffectively treated. If 10% of colorectal and 6% of lung cancer patients carry BRAF mutations, that amounts to nearly 16,000 deaths annually due to BRAF-mutated cancers.

The biochemistry of the various altered BRAF proteins varies substantially. While the V600E alteration dramatically elevates kinase activity, several of the less common alterations diminish BRAF kinase activity, yet they promote MEK phosphorylation in a CRAF-dependent manner²³. In one genetically engineered mouse model, conditional melanocyte-specific expression of either KRAS-G12D or BRAF D594A (a kinase impaired P-loop mutation) was insufficient to induce nevi or melanomas, yet co-expression of both alleles promoted cellular dimerization of the catalytically inert BRAF-D594A with catalytically competent CRAF and elicited rapidly growing, pigmented tumors²⁴. These data strongly indicate that kinase impaired BRAF mutations are oncogenic drivers, but require RAS and a catalytically competent RAF isoform to activate downstream signaling. Additionally, the P-loop contains an auto-phosphorylation site that dramatically reduces activity of the wild-type enzyme²⁵. At least some of the BRAF mutant proteins appear to bypass this effect, which may explain why even mutations that impair intrinsic kinase activity can constitutively activate the enzyme by preventing auto-inhibition. Therefore, the various BRAF P-loop and A-loop mutations may have different mechanisms of activation, relying partly on constitutive kinase activity or acting entirely as a scaffold for RAS signal transduction through CRAF.

Although the tumor initiating potential of the les common BRAF mutations have yet to be demonstrated in vivo, the oncogenicity of BRAF-V600E has been validated in numerous genetically engineered mouse (GEM) models. To study BRAF-V600E driven tumorigenesis, a mouse carrying a Cre recombinase-activated allele of Braf (Braf^CA) was developed, such that normal BRAF is expressed prior to Cre-mediated recombination, and only after exposure to Cre is mutant BRAF-V600E expressed from the endogenous locus.²⁶ Using this model, lungspecific expression of BRAF-V600E caused lung adenomas while concomitant BRAF-V600E expression and homozygous excision of floxed Tp53 alleles caused progression to adenocarcinoma. Expression of BRAF-V600E in melanocyte lineage also cooperated with loss of tumor suppressors (Pten or p16^INK4A) to yield melanoma, with metastatic potential in BRAF^V600E PTEN^null melanocytes^27,28. Targeted expression of BRAF-V600E in thyrocytes produced papillary or anaplastic thyroid cancers^29–32. Expression of BRAF-V600E in the proliferative cells of the mouse gastrointestinal tract has been shown to act as an early driver mutation in the pathogenesis of serrated colorectal cancer^33,34. Finally, recent years have heralded the publication of additional BRAF^V600E-driven GEM models of pancreatic ductal adenocarcinoma³⁵, prostate cancer³⁶, pediatric malignant astrocytoma³⁷, soft tissue sarcoma³⁸, and neoplasms of the histiocyte/monocyte lineage³⁹. Together, these models of BRAF-V600E-driven malignancies highlight the oncogenicity of the mutant protein in a wide array of tissue types, and also emphasize the therapeutic potential of targeting BRAF oncogenes.

CRAF and ARAF mutations

In contrast to BRAF, ARAF and CRAF mutations are exceptionally rare in cancer. Recent data indicate that a small subset (~1%) of patients with adenocarcinoma of the lung carry activating ARAF or CRAF mutations. It has not yet been determined if all ARAF and CRAF mutations constitute oncogenic drivers in all cases, but initial cell culture studies confirmed the transforming potential of ARAF S124C, CRAF S257L and CRAF S259A and as well as the sensitivity of these mutants to RAF inhibition⁴⁰.

Although somatic CRAF point mutations are rare in human cancers, several germ-line CRAF mutations are the cause of Noonan syndrome (NS) (germline mutations in seven other MAP kinase pathway genes are also reported to cause NS; thus the disorder is described as a “RASopathy”)^41,42. NS is an autosomal dominant genetic disorder that is characterized by craniofacial deformations, short stature, cardiac anomalies and a propensity for neurocognitive delay⁴³. One-third of NS patients with CRAF mutations also display multiple nevi, lentigines, or café-au-lait spots, suggesting that germ-line CRAF mutations may predispose patients to cutaneous hyperpigmented lesions, similar to the ability of BRAF-V600E to elicit benign nevi^27,44. However, the majority of CRAF point mutations identified in NS are not observed in human cancers, and fewer than 2% of human cancer derived cell lines harbor mutated CRAF⁴⁵. The oncogenic impotence of CRAF, compared to the potent oncogenicity of BRAF, is thought to arise because CRAF possesses low basal kinase activity compared to that of BRAF⁴⁶. Importantly, one NS mutation, CRAF S259F, has also been identified in lung cancer. Additionally, CRAF S259 mutations have been identified in patients with either colon (S259P) or ovarian cancer (S259A). Phosphorylation of CRAF S259 is associated with reduced activity and promotes direct binding to 14-3-3 proteins, which stabilize the inactive state^41,47. Thus, dephosphorylation or mutation of S259 disrupts the stability of the inactive CRAF conformation and facilitates binding to RAS-GTP at the plasma membrane⁴⁸. The Melan-a cell line, a non-transformed mouse melanocyte-derived cell line that is sensitive to BRAF-V600E induced transformation, is also transformed by expression of CRAF-S259A.⁴⁹ In contrast to the scarcity of CRAF point mutations identified in cancer, increased CRAF expression has been identified in a number of malignancies, the most notable being bladder cancer⁵⁰. The oncogenic significance and therapeutic implications of CRAF amplifications remain unclear.

RAF fusion proteins as oncogenic drivers

Analyses of prostate cancer and pediatric astrocytoma or glioma have revealed that point mutation is not the only mechanism that can reveal the oncogenic potential of RAF protein kinases. The presence of the Philadelphia chromosome in chronic myelogenous leukemia is the archetypal example of an oncogenic protein kinase (BCR-ABL) that is generated via a chromosome translocation⁵¹. Similarly, RAF fusion transcripts have been identified in melanoma, prostate, gastric, thyroid and breast cancers and gliomas^52–57. Early studies demonstrated the constitutive activity and transforming potential of amino-terminally truncated RAF proteins^58–60. In every case where chromosomal break points have been identified within the BRAF locus in human cancer, the C-terminal portion of the enzyme is fused in-frame using the start codon of another gene, resulting in a RAF kinase fusion protein lacking the aminoterminal domain. Several 5’ fusion partners have been identified for BRAF, including angiotensin II receptor-associated protein (AGTRAP) and solute carrier family 45, member 3 (SLC45A3) in prostate cancer⁵²; A kinase anchor protein (AKAP9) in thyroid cancer⁵³; and MARVEL domain containing 1 (MALD1), acylglycerol kinase (AGK)⁶¹ and heat shock 27kDa associated protein 1 (PASS1) in melanoma⁶². CRAF fusions with epithelial splicing regulatory protein 1 (ESRP1) have also been described in prostate cancers⁵², although most RAF kinase fusion events seem to occur in pediatric gliomas. Despite the low frequency of BRAF point mutations in low-grade pediatric astrocytomas (~1%), BRAF gene fusions are observed in the majority (70%) of such cases. To date, the KIAA1549-BRAF fusion is most commonly observed^63,64, while FAM131B-BRAF⁵⁴ and SLIT-ROBO Rho GTPase activating protein 3 (SRGAP3)-CRAF fusions appear less frequently (~2–8% of cases)⁵⁷.

The biochemistry of RAF fusion proteins remains poorly characterized, but canonical RAF signaling is thought to rely heavily on RAS-mediated dimerization of BRAF and CRAF protomers. Deletion of the amino-terminal RAS binding domain (RBD) has long been known to constitutively activate RAF kinases^59,65 and is thought to promote dimerization, suggesting that the amino-terminus of RAF contains an auto-inhibitory domain that prevents dimerization. Dimerization is associated with increased MEK phosphorylation, and disruption of the dimerization interface identified by the crystal structures^17,23 nearly abolishes catalytic activity^66,67. Indeed, disrupted dimerization also prevents KIAA1549-BRAF-induced oncogenic transformation in vitro⁶⁸, indicating that loss of the RAF RBD through fusion of the kinase domain to KIAA1549 may constitutively activate RAF proteins through a similar mechanism.

The RAF inhibitor paradox

Use of RAF inhibitors in RAS-mutated cancers has unveiled the complexity of RAF signal transduction. In 2009 three groups showed that catalytic inhibition of RAF kinases resulted in paradoxical activation of CRAF, increased proliferation of RAS mutated cells in vitro, and in some cases induction of tumorigenesis in vivo^24,76,77. While each group proposed a distinct mechanism to explain the effect, several themes overlapped (Figure 2B). RAF inhibitors promoted co-immunoprecipitation of RAF isoforms as well as association with RAS-GTP at cell membranes^25,76,78, suggesting that the inhibitors promote CRAF homo- and heterodimerization and RAS-mediated activation. Moreover, site-directed mutagenesis of the putative dimer interface rendered CRAF resistant to paradoxical activation⁷⁷. The drug-induced association of RAF with RAS-GTP as well as increased RAF dimerization likely both contribute to the activation of ERK signaling, but it remains unclear if either event alone is sufficient to induce paradoxical activation.

Interestingly, paradoxical activation by RAF inhibitors can be recapitulated genetically using catalytically-impaired BRAF²⁴. In this model, BRAF catalytically suppresses CRAF activity in RAS-mutated cells such that disrupting BRAF function, either genetically or with a BRAF inhibitor, alleviates CRAF suppression and activates it in a RAS-dependent manner. This mechanism is further supported by the discovery of a RAF auto-inhibitory phosphorylation site that is required for paradoxical activation²⁵ (Figure 2B), indicating that paradoxical activation may be intrinsic to all catalytic RAF inhibitors.

Cutaneous lesions induced by BRAF inhibitors

The ability of RAF inhibitors to paradoxically activate ERK MAP kinase signaling carries critical clinical implications. Results from the vemurafenib phase III clinical trial indicate that 18% of patients experienced squamous cell carcinomas (SCC) and/or keratoacanthomas (KAs)⁸¹. The list of secondary cutaneous lesions induced by BRAF inhibitors has since expanded to include papillomas, cystic lesions, milia, eruptive nevi, and basal cell carcinomas⁸⁷. While all drug-induced lesions are BRAF wild-type, multiple other genetic mutations have been identified. Importantly, samples of secondary KAs display a high frequency of HRAS mutations^88–90. This suggests that the ability of BRAF inhibitors to elicit paradoxical MAP kinase pathway activation activates a proliferative program in dormant keratinocytes harboring a latent RAS mutation. Indeed, studies using the K5-SOS-F transgenic mouse model, in which mice are predisposed to epidermal tumors through keratinocyte-specific expression of a dominant active, farnesylated Son of Sevenless (SOS-F), have shown that RAF inhibitor treatment promotes RAF dimerization, activates ERK signaling, and suppresses Rho-Associated, Coiled-Coil Containing Protein Kinase 2 (Rokα) to potentiate Ras-driven epidermal tumorigenesis.⁷⁸ In addition to RAS mutations, oncoviruses have been implicated as initiating factors in a subset of cutaneous lesions. Papilloma virus and Merkel cell polyomavirus DNA were identified in BRAF inhibitor treated patient samples and vemurafenib promoted tumorigenesis in an HPV driven mouse model of SCC⁹¹. Thus far the secondary cutaneous malignancies arising in patients receiving BRAF inhibitor therapy have been well-differentiated and amenable to local resection with no reports of metastasis⁹².

More worrisome than the cutaneous toxicities is the possibility that BRAF inhibitors may induce the proliferation of pre-malignant cells harboring RAS mutations in non-cutaneous tissues, such as the lung, colon, or pancreas. The ability of BRAF inhibitors to promote the growth of pre-malignant precursor lesions along the path to adenocarcinoma is of considerable concern. Indeed, a recent report described a patient with metastatic BRAF^V600K melanoma who developed NRAS-mutant chronic myelomonocytic leukemia during the course of vemurafenib treatment^93,94. Upon withdrawal of vemurafenib, the patient’s leukemic cell counts dropped, thus necessitating an empirically determined intermittent dosing schedule of vemurafenib and cobimetinib (a MEK inhibitor) to simultaneously control the patient’s metastatic melanoma and also to prevent the outgrowth of leukemia⁹⁵. A second case report described a patient presenting with BRAF-mutated metastatic melanoma, who had undergone a prior successful surgical resection of a KRAS-G12D colorectal cancer⁹⁶. While the combination of dabrafenib and trametinib caused a PR of the patient’s melanoma, at 12 weeks the patient presented with a brain metastasis that, upon resection, was characterized as originating from a KRAS-mutated colon cancer. After progressive metastatic colon cancer necessitated cessation of dabrafenib and trametinib combination therapy, the patient received trametinib monotherapy and experienced a brief improvement in performance status. In vitro studies performed using tissue derived from the aforementioned patients suggested that MEK inhibitors suppress the growth of BRAF-V600E melanoma and also prevent expansion of the RAS-mutated malignancies. These case reports not only emphasize the caution that must be exercised when considering the use of RAF inhibitors as an adjuvant therapy, but also demonstrate the importance of determining which node of the MAP kinase pathway is most appropriate to target therapeutically in patients with BRAF-mutated melanoma.

Improving responses to RAF inhibition with other therapeutic targets

It is currently unclear if survival benefit alone will cause RAF inhibitors to be used as single agents, but preclinical and early clinical data strongly suggest that some combination of RAF and either MEK or conceivably an ERK inhibitor will have the greatest efficacy. FDA approval was recently announced for the combination of dabrafenib and trametinib in advanced melanoma, and detailed results of the phase III trial are forthcoming. Phase I/II results demonstrated that the combination increased PFS from 5.8 months for single agent treatment to 9.4 months in the combination group, and OR increased from 54% to 76%¹²⁵. Perhaps unexpectedly, tolerability of either agent improved in the combination group. Acneiform dermatitis, the most common dose limiting toxicity associated with trametinib (8%), was not observed in the combination group. Likewise, the frequency of hyperkeratosis, cutaneous SCCs and papillomas, all RAF inhibitor class effects, each decreased when dabrafenib was co-administered with trametinib. Mechanistically, this most likely occurs because the paradoxical activation of the MEK1/2→ERK1/2 MAP kinase pathway induced by RAF inhibitors is suppressed by MEK inhibition.

Tumors that have acquired on-pathway mechanisms of resistance to BRAF inhibitors compose an important group in which to test the efficacy of second-line MEK or ERK inhibitors. Unfortunately, early clinical data suggest limited efficacy of MEK inhibitor monotherapy for BRAF-mutated melanoma patients with drug resistant disease. Within a cohort of 40 patients who had progressed on either vemurafenib or dabrafenib monotherapy, subsequent treatment with trametinib offered only a 1.8 month median PFS and a 5.8 month median OS¹²⁶. Only two patients within the cohort had a complete or partial response and both of these patients had discontinued BRAF inhibitor therapy due to adverse events rather than the onset of drugresistant disease. Thus second-line MEK inhibitor monotherapy to treat BRAF inhibitor resistant disease seems limited. These results may be partially explained by a recent report demonstrating that oncogenic BRAF over-expression or alternative splicing – previously known to confer resistance to BRAF inhibitor monotherapy – also emerges as a resistance mechanism among patients who receive first-line BRAF and MEK combination therapy¹²⁷. Another tactic to target tumors that have acquired MAP kinase pathway-dependent resistance mechanisms is the use of ERK inhibitors. While ERK inhibitors have only recently entered clinical trials, preclinical data suggest that cells that have developed resistance to RAF and MEK inhibitors remain sensitive to a selective ERK1 and ERK2 inhibitor¹²⁸. The acquisition of MAP kinase pathway-independent mechanisms of resistance also may sensitize tumors to other pathway-targeted therapies. For example, elevated PDGFRβ or IGF1R signaling might be combated with receptor tyrosine kinase or PI3’-K inhibitors¹¹⁸.

BRAF-V600E oncogene overdose and intermittent drug dosing

Although BRAF-V600E constitutively activates the MEK→ERK pathway, many cancer cells remain sensitive to the quantity of MAP kinase pathway signaling (as measured by ERK phosphorylation or transcription of target genes). More than ten years ago, it was demonstrated that mouse or human fibroblasts display a peak proliferative response with an intermediate amount of RAF→MEK→ERK pathway activation^129,130. Indeed, in nonimmortalized human IMR-90 fibroblasts, sustained RAF activation induced an irreversible cell cycle arrest with features of cellular senescence¹³¹. This further substantiated the hypothesis that RAF oncogene induced transformation is determined not only by activation of the MAPK pathway, but the quantity of pathway activation, as reviewed by Marshall¹³².

The notion that malignant cells might remain sensitive to the strength of MAP kinase pathway signaling was recently substantiated using a patient-derived melanoma xenograft model. Upon transplantation into immunocompromised mice, a chemonaïve, BRAF mutated melanoma displayed striking regression in response to vemurafenib treatment¹³⁰. However, sustained drug treatment led to the emergence of lethal drug-resistant disease due to elevated BRAF-V600E expression. Remarkably, the drug resistant tumors and cell cultures were not merely drug-resistant but also drug-dependent for their peak proliferation (Figure 5A). Moreover, when vemurafenib was removed from drug resistant cells or tumors, striking antiproliferative effects were observed (Figure 5B). These data indicate that BRAF-V600E ‘addicted’ melanoma cells can remain sensitive to the magnitude of BRAFV600E→MEK→ERK signal pathway activation such that too much of the oncoprotein to which they are addicted can lead to “oncogene overdose”. The demonstration that vemurafenibresistant melanoma cells can have a fitness deficit in the absence of drug has led to the design of clinical trials to test the efficacy of intermittent dosing to forestall the onset of drug resistant disease and therefore enhance the durability of patient responses.