The discovery of activated alleles of BRAF in malignant melanoma has changed the therapeutic landscape dramatically (Gray-Schopfer et al., 2007). Data from Plexxikon’s PLX 4032 BRAF inhibitor were a highlight at the 2009 ASCO meeting, and, for the first time, offered hope of significantly improved clinical outcomes for the majority of melanoma sufferers (Flaherty et al., 2009). These data were indeed overdue: earlier attempts to target BRAF or its downstream target MEK, had failed, despite encouraging preclinical data. First or this class to fail was sorafenib. This compound inhibits CRAF quite effectively, but is less active against BRAF, and even less effective against the activated BRAF V600E protein. This is probably because sorafenib is a class II inhibitor: it binds RAF kinase in the inactive configuration. However, BRAF V600E is locked in the active state, and therefore is not the ideal target of this compound. On the other hand, MEK inhibitors potently inhibit growth of cell lines in culture, especially those with V600E BRAF mutations, but have encountered problems with toxicity in the clinic, without showing compelling signs of efficacy. Direct attack on the mutant, activated BRAF kinase therefore seems the most effective strategy and the data presented by Plexxikon support this conclusion. However, complications have arisen, and many questions need to be addressed. For example, how frequently will drug-resistant clones emerge, as they have for drugs targeting BCR-ABL, c-KIT, and EGF-R mutant onco-proteins? What about melanomas with activated NRAS, or with mutant forms of BRAF other than V600E? And what do we make of the emergence of keratoacanthomas and squamous cell carcinoma and related lesions that appear in patients treated with PLX 4032, sorafenib and other RAF kinase inhibitors? (Arnault et al., 2009).
To understand these complexities, we need a better understanding of the RAS-RAF-MEK-ERK pathway. Until recently, this pathway was depicted as a simple linear module, triggered by RAS GTP binding directly to RAF kinase. In this simple model, RAF kinase inhibitors would be indistinguishable from MEK or ERK inhibitors. This indeed was the conventional wisdom when we, and others, began targeting RAF and MEK in the early 1990s (why nobody developed ERK inhibitors remains a mystery). If the pathway were simple and linear, blocking any one step should be sufficient to shut the whole pathway down. Likewise, there was no obvious reason to target CRAF rather than BRAF or ARAF, and no basis for targeting the active configuration versus the inactive configuration. Each of these assumptions (if even considered) turned out to be incorrect. As compounds with selectivity for RAF or MEK were developed, strange effects were observed that suggested that this pathway is a lot more complicated. For example, RAF inhibitors were found to hyper-activate RAF kinase activity, suggesting the presence of a feedback system driven by RAF itself (Hall-Jackson et al., 1999). Later, Adjei and colleagues showed that MEK inhibitors activate RAF kinase, and so increase levels of P-MEK (Friday et al., 2008). These latter effects are due to a feedback loop from ERK to RAF, originally reported by Morrison and co-workers (Dougherty et al., 2005): when ERK is shut down, feedback is lost and RAF kinase becomes hyper-active (Figure 1). This led to the suggestion that combinations of RAF and MEK inhibitors might be needed to shut down the RAF-MEK-ERK pathway effectively, clearly a prediction that is not consistent with a simple linear pathway. More recently, Korn and co-workers (Mirzoeva et al., 2009) found that MEK inhibitors activate EGF-R signaling, again through loss of a feedback loop, and suggested that MEK inhibitors create the need for EGF-R inhibitors to be effective (Mirzoeva et al., 2009). This combination would not have been anticipated based on earlier simplistic logic.
More recently, CRAF has been reported as an inhibitor of BRAF, leading to the suggestion that BRAF mutations are more prevalent in melanoma than other types of tumor because they express lower levels of CRAF (Karreth et al., 2009). This relationship between CRAF may also explain the observation that RAF inhibitors like sorafenib, which are more potent inhibitors of CRAF than BRAF, can activate the MEK-ERK pathway at high doses. These compounds block CRAF effectively, but this leads to BRAF activation and increased signaling downstream.
The latest chapter in this developing story comes from Marais and co-workers (Heidorn et al., 2010), in the current issue of Cell. In this paper, a molecular mechanism is provided to explain how kinase-dead BRAF mutants drive tumor progression. This model (Figure 2) helps to tie together a number of baffling observations, and makes some interesting and testable predictions. When a panel of BRAF mutants from human melanoma was first analyzed biochemically, several appeared to be kinase-impaired, others kinase-dead. The kinase-impaired mutants retain sufficient activity to bind and activate CRAF in a RAS-independent manner (Garnett et al., 2005), but kinase-dead mutants such as D594 must function through a different mechanism. In their current paper, Marais and colleagues show that drugs specific for BRAF provoke association of BRAF with active RAS, resulting in activation of CRAF which is in the same complex. This association occurs when the kinase activity of BRAF is inhibited, pharmacologically or by mutations such as D594. Under these circumstances, a feedback loop from BRAF to itself is interrupted, and the non-phosphorylated BRAF protein joins forces with Ras and CRAF to restore MEK phosphorylation. Oncogenic RAS is the best mediator of this effect, but wild-type RAS can do the same when kicked into the active, GTP-bound states by upstream signals such as EGF-Receptor. This cooperation between kinase-dead RAF mutants and activated RAS was confirmed in an elegant mouse model, and also by the co-existence of RAS and B-RAF mutants of this type in human cancers. In contrast, V600E BRAF never co-exists with mutant RAS, either because mutations in one makes the other redundant, or, as proposed by Tuveson and co-workers (Karreth et al., 2009), because CRAF activation by RAS interferes with V600E signaling.
These new developments underscore many interesting points. They remind us that our understanding of signaling pathways is still superficial, and that even a simple pathway (RAS-RAF-MEK-ERK) is remarkably complex when we drill down to the details. Second, different drugs that are developed against the same target may have very different effects. These differences are not simply a matter of potency and pharmaco-kinetic properties, but of the precise details of drug–protein interaction. An example of this emerged from study of mutations that confer drug-resistance to imatanib (GleVec). This compound, like sorafenib, is a class II inhibitor and stabilizes the inactive form of the BCR-ABL protein. As many as 20 point mutations have been identified that prevent imatinib binding inhibition, yet allow the BCR-ABL protein to function effectively as a kinase. In contrast, very few mutations in BCR-ABL confer resistance to desatanib, a class I ATP-competitive drug that interacts with the active form of the kinase. Clearly, the closer a compound resembles ATP, the fewer mutations can emerge that fail to bind the drug, yet retain ATP-binding and kinase activity. For these reasons, ATP-competitive compounds may be preferable, as drug-resistance is the major downfall of targeted therapy. The new work form Marais and co-workers reveals another reason to consider the molecular details of drug–protein interactions, as the interaction of BRAF with CRAF depends on the activation state of each kinase. Different BRAF and CRAF inhibitors therefore have different effects on this dynamic process. A drug that inhibits BRAF effectively allows CRAF to restore MEK activation and vice versa. A drug, or a combination of drugs, that strongly inhibits both BRAF V600E and CRAF, may therefore be necessary to shut the pathway down completely.
An unexpected consequence, therefore, of the BRAF–CRAF interaction is that RAF inhibitors can lead to activation of the MAPK pathway in normal cells (Figure 2). This is not the case for MEK inhibitors. This difference may explain why PLX 4032 appears to be more effective in treating melanoma than MEK inhibitors. High doses of PLX 4032 and, presumably, other RAF inhibitors, are well tolerated because the MAPK pathway is not shut down in normal cells. MEK inhibitors, on the other hand, shut the pathway down in normal cells and may therefore lead to on-target side effects that limit dosing (Neal Rosen, personal communication). This represents a new paradigm in targeted therapy, one that might form the basis of future target identification and selection. The keratoacanthomas and squamous cell carcinomas that appear in patients treated with RAF inhibitors, but not MEK inhibitors, are likely to be caused by this hyper-activation, though this has yet to be proven. If so, it will be necessary to ascertain whether the benefit of activating MEK and the MAPK pathway in normal cells, thus sparing them from side effects of pathway inhibition, offsets problems associated with induction of these and perhaps other lesions. Today the evidence strongly suggests that the benefits far outweigh the risks, and that BRAF inhibition does indeed represent a tremendous advance in therapy, notwithstanding the unexpected and curious effects that it engenders.