Over the past year, there has been considerable excitement following the positive therapeutic effects of the compound PLX4032 in early clinical trials in patients who have metastatic melanoma1. The compound is thought to selectively inhibit BRAFV600E, a mutant, cancer-associated form of the kinase enzyme BRAF. The positive clinical effects of PLX4032 contrast with clinical responses to inhibitors of MEK, another kinase directly downstream of BRAF in the RAS–RAF–MEK–ERK signalling cascade. MEK inhibitors have been less effective than PLX4032 in clinical trials2, despite the fact that they — like the BRAF inhibitor — selectively killed melanoma cells that carry BRAFV600E mutations in preclinical studies3.

But despite the antitumour effects of PLX4032 on melanomas, about 25% of patients treated with the drug unexpectedly developed hyperproliferative skin lesions (keratoacanthomas) and, in some cases, squamous-cell carcinomas, which disappeared following drug withdrawal1. Adding to this paradox, although most cancer-causing (oncogenic) BRAF mutations — such as BRAFV600E — enhance the kinase's activity, some patients with melanoma harbour BRAF-inactivating mutations4. Three studies, two in this issue5,6 and another in Cell7, begin to provide an explanation for these unexpected findings.

All three studies5,6,7 started with the same intriguing observation: whereas selective BRAF inhibitors potently suppress RAF–MEK–ERK signalling in BRAF-mutant melanoma cells, the same inhibitors unexpectedly activate this pathway in a different type of tumour cell — those that carry an oncogenic mutation in the gene KRAS, but which possess a wild-type BRAF gene (Fig. 1). Heidorn et al.7 report that the selective BRAF inhibitor 885-A binds to wild-type BRAF in KRAS-mutant cells and promotes the formation of BRAF–CRAF complexes, in which CRAF is a RAF kinase related to BRAF in this signalling pathway. Formation of the complex activates CRAF, and consequently the MEK–ERK pathway.

Figure 1: The effects of RAF inhibitors in mutant and wild-type cells.
figure 1

Activation of the RAS–RAF–MEK–ERK signalling pathway promotes cell growth, but excessive activation is associated with cancer. RAS enzymes activate signalling when bound to the nucleotide GTP. b, BRAFV600E is a mutant, oncogenic version of a RAF kinase enzyme found in the pathway. In tumour cells that harbour BRAFV600E, RAS–GTP activation is low, and ERK signalling is predominantly activated by dimers of the mutant kinase. Phosphate groups (blue) are present on the activated kinase. c, In the presence of a BRAF inhibitor that blocks phosphorylation (green), MEK–ERK signalling in BRAFV600E mutant cells is suppressed, causing cell death and tumour shrinkage. d, In RAS mutant cells that contain wild-type BRAF, RAF inhibitors block phosphorylation of RAF monomers (BRAF or CRAF), which go on to form dimeric complexes5,6,7 such as BRAF–CRAF or CRAF–CRAF. Similarly, a 'kinase-dead' BRAF that cannot phosphorylate (grey block) forms a complex with CRAF7. The formation of these RAF complexes causes excessive signalling, stimulated by the mutant RAS enzyme. This might explain the formation of skin lesions (keratoacanthomas) and squamous-cell carcinomas observed in melanoma patients given RAF inhibitors during clinical trials1.

Full size image

Heidorn et al. also demonstrate that a 'kinase-dead' mutant of BRAF (which behaves like an inhibited BRAF enzyme) similarly binds to CRAF and cooperates with a mutant form of the RAS gene to promote MEK–ERK activation, thus inducing melanomas in genetically engineered mice (Fig. 1). The authors therefore provide a unifying explanation for the oncogenic effect of kinase-dead BRAF proteins, and for the pro-tumorigenic effects of BRAF inhibitors in some settings (such as in cells that lack BRAF-activating mutations).

Hatzivassiliou and colleagues' studies6 (page 431) are thematically similar, but their focus is on working out the molecular mechanisms by which two selective, structurally unrelated BRAF inhibitors — GDC-0879 and PLX4720 — activate the RAF–MEK–ERK pathway in RAS-mutant cells. The authors' findings suggest that GDC-0879 promotes the formation of dimeric complexes of RAF-type kinases, specifically BRAF–CRAF, BRAF–ARAF and CRAF–CRAF complexes. The authors also obtained a crystal structure of a close analogue of GDC-0879 bound to CRAF. This revealed that drug binding causes a conformational change in the kinase that promotes the formation of dimeric RAF complexes.

Although PLX4720 also induces activation of the RAF–MEK–ERK pathway, Hatzivassiliou et al. found that it actually prevents the formation of BRAF–CRAF complexes. Their models suggest that this is because the inhibitor induces a different conformational change in BRAF, which specifically impairs BRAF binding to other RAF kinases. It therefore seems that PLX4720 mediates its effects by promoting the formation of CRAF–CRAF dimers (Fig. 1).

Poulikakos and colleagues5 (page 427) also investigated the biochemical effects of RAF inhibitors. They found that these agents activate MEK and ERK not only in tumour cells that harbour a mutant version of RAS, but also in cells in which the RAS–RAF–MEK–ERK pathway is activated by other oncogenes, such as HER2. Consistent with Hatzivassiliou and colleagues' report6, Poulikakos et al. suggest that PLX4720 functions by promoting CRAF–CRAF formation (Fig. 1). In addition, they make the intriguing observation that, at low concentrations, compounds that are thought to inhibit all RAFs (pan-RAF inhibitors) stimulate the RAF–MEK–ERK pathway. Presumably this is because, at low concentrations, the inhibitors bind to only a fraction of the available RAF kinases, promoting the formation of complexes of the drug-bound kinases with non-drug-bound RAFs. Because only one kinase within a given complex is drug-bound (inhibited), such complexation induces the activation of the non-drug-bound kinase, thus promoting MEK–ERK signalling.

Although the precise mechanisms by which different RAF inhibitors activate the RAF–MEK–ERK pathway may differ, all three studies5,6,7 demonstrate that inhibitors thought to be selective for mutant BRAF can activate CRAF through the formation of dimeric RAF complexes — a process that is enhanced by the presence of an oncogenic RAS mutation. Consequently, these studies support the same fundamental conclusion: mutant-specific BRAF inhibitors should be used for treating cancers caused by BRAF mutants (such as BRAFV600E-associated melanomas), but should be avoided in cancers caused by RAS mutants (including other kinds of melanoma). This underscores the need to select patients who have BRAF-mutant cancers for current and future clinical studies of BRAF inhibitors.

These findings5,6,7 may also help to explain the side effects of RAF inhibitors in patients with melanoma, and they raise additional therapeutic considerations. Although it remains to be formally proved, it is tempting to speculate that the keratoacanthomas and squamous-cell carcinomas observed in clinical trials8 of RAF inhibitors are attributable to the activation of ERK signalling by one or more of the proposed mechanisms5,6,7 (Fig. 1). Consistent with this possibility, Hatzivassiliou et al.6 report that GDC-0879 causes hyperactivation of ERK and hyperproliferation of skin cells in mice. Moreover, RAS mutations are found in up to 20% of squamous-cell carcinomas. RAS mutations are also present7,9 in a subset of actinic keratoses — thick, scaly patches of skin that sometimes precede keratoacanthomas — supporting the notion that RAS mutations may enhance the pro-tumorigenic side effects of RAF inhibitors in patients. Although this particular lesion is easily removed, it could be that other, less visible pre-malignant lesions associated with RAS mutations might be induced by RAF inhibitors, such as those in the airways of smokers. This suggests that patients treated with RAF inhibitors should be monitored carefully for side effects. Nevertheless, the success of BRAF inhibitors in patients with metastatic melanoma1, a group of patients who until recently had a poor prognosis, suggests that clinical development of these compounds should continue.

The trio of studies5,6,7 might also help us to understand the clinical differences observed when melanomas are treated with MEK or BRAF inhibitors, and may provide insight into which drugs should be selected to treat genetically distinct tumours. The reports suggest that BRAF inhibitors do not inhibit (and may even potentiate) ERK signalling in normal tissues, such as skin. This effect might actually provide a greater therapeutic window for anticancer treatment, allowing BRAF inhibitors to be administered at sufficient doses to fully inhibit ERK signalling in BRAF-mutant tumours, while sparing normal tissue. In contrast, MEK inhibitors block ERK signalling in both tumour and normal tissue, so that the therapeutically acceptable doses of these compounds are limited by the side effects associated with ERK inhibition in normal tissue, such as intolerable skin rashes2. It is currently not known whether the greater success of PLX4032 in clinical trials, compared with MEK inhibitors, is solely the result of a greater therapeutic window, or whether this success also (or instead) stems from the selectivity of PLX4032 for BRAFV600E. The answer to this question is important for future drug development, as BRAF inhibitors that do not activate CRAF might ultimately cause side effects similar to those caused by MEK inhibitors. Thus, a compound that inhibits BRAFV600E alone, if it is possible to develop such an agent, might ultimately be the best drug for treating BRAFV600E-mutant melanomas.

Finally, the observation5 that pan-RAF inhibitors might also activate the RAS–RAF–MEK–ERK pathway in RAS-mutant cells at low concentrations may suggest that only those inhibitors that potently inhibit CRAF at clinically achievable drug concentrations should be considered as treatments for RAS-mutant cancers. Alternatively, MEK inhibitors may be a more appropriate choice of therapeutic agent. Regardless of the agent chosen, it will probably need to be used in combination with another drug to effectively treat RAS-mutant tumours.

More fundamentally, these reports5,6,7 suggest that, despite 25 years of intensive study, we still do not fully appreciate the complexities of the RAS–RAF–MEK–ERK pathway. The unexpected effects (and, in some cases, lack of effects) of cancer therapies that target it underscore the value of studies aimed at garnering detailed mechanistic insight into both normal and pathogenic signalling mediated by this pathway. Such insight might ultimately lead to the development of effective new anticancer therapies.