Towards a Unified Model of RAF Inhibitor Resistance

RAF inhibitors (vemurafenib and dabrafenib) have profound clinical activity in patients with BRAF-mutant melanoma, but their therapeutic effects are limited by the emergence of drug resistance. Initial responses to RAF inhibitor therapy are also highly variable and complete remissions are rare. This has spurred extensive laboratory and clinical research efforts to understand the molecular basis of acquisition of resistance to RAF inhibitors. The goals of this work include the development of therapeutic strategies that are useful for treating resistant tumors or, preferably, that delay or prevent the emergence of drug-resistant clones.

The unique feature of selective, ATP-competitive RAF inhibitors is that they inhibit extracellular signal-regulated kinase (ERK) signaling and thus cellular proliferation only in cells that harbor BRAF mutations. In other tumor cells, particularly those with RAS activation, these drugs cause a paradoxical activation of ERK signaling that can be associated with accelerated cellular proliferation and cancer progression. Physiologic activation of RAS causes the dimerization and activation of RAF family protein kinases. Binding of the RAF inhibitor to one member of the RAF dimer causes the allosteric transactivation of the other, unbound RAF protein (1). In contrast, BRAF^V600E signals as a monomer that is inhibited by drug binding (2).

The mechanism of action of RAF inhibitors suggests that molecular lesions that cause BRAF^V600E to dimerize in tumor cells would cause resistance to vemurafenib. Laboratory and clinical studies now support this model. Expression of an alternatively spliced variant of BRAF^V600E that constitutively dimerizes causes resistance to RAF inhibitors in a model system and has been shown to be expressed in a significant fraction of melanomas from patients with acquired resistance (2). Moreover, mutations in NRAS, loss of NF1 function, and activation of upstream receptor tyrosine kinases, lesions expected to induce RAS-dependent RAF dimerization, have all been shown to confer acquired resistance to RAF inhibitors in model systems, patients, or both (2–4). Additional molecular events, particularly amplification of BRAF^V600E and mutations in the MAP2K1 gene, which encodes MEK1, have also been shown to occur in tumors with acquired resistance and functionally validated as causal (5, 6).

The common feature of each of these mechanisms of resistance is that they result in activation of ERK signaling that is insensitive to the RAF inhibitor. Thus, RAF inhibitor resistance is often associated with maintenance of activation of the oncogene-driven pathway. Previously only a small number of samples had been analyzed for each of the above resistance lesions, and thus their prevalence and whether they coexist with other lesions that affect the phenotype was unknown. Two studies in the current issue of Cancer Discovery provide further data as to the prevalence of RAF inhibitor resistance mechanisms and identify novel resistance mechanisms, several of which are consistent with prior laboratory studies.

In the study by Shi and colleagues (7), the investigators analyzed 100 tumor samples collected from 44 patients using targeted methods to identify known mechanisms of RAF inhibitor resistance and, in parallel, unbiased whole exome sequencing (WES), to attempt to identify novel resistance mechanisms. In their cohort, approximately 60% of disease progression biopsies harbored mutations within the mitogen-activated protein kinase (MAPK) pathway identified in prior work as responsible for RAF inhibitor resistance. These included NRAS mutations (18%), BRAF amplification (19%), BRAF splice variants (13%), and MAP2K1 mutations (3%). CDKN2A deletion was observed in 7% of biopsies, and was also grouped by the investigators with the ERK pathway alterations, as cyclin D expression is dependent on the ERK pathway in BRAF-mutant cells (8). However, it is not clear that loss of the CDKN2A product p16^INK4a is sufficient to cause resistance, as, in such tumors, cyclin D1 is still ERK dependent and CDKN2A-null models often retain sensitivity to RAF and MAP-ERK kinase (MEK) inhibitors (8, 9).

Consistent with prior preclinical studies, mutations in the phosphoinositide 3-kinase (PI3K)–PTEN–AKT pathway were also identified by WES as potential mediators of RAF inhibitor resistance (9). Mutations in AKT1/AKT3 (Q79K and E17K) and putative functional alterations in positive regulators of the pathway (PIK3CA, PIK3CG) and in negative regulatory genes (PIK3R2, PTEN, and PHLPP1) were detected in 22% of resistant samples. The functional consequences of these latter lesions and their role in conferring resistance have in most cases not yet been fully defined and require further laboratory study.

In the study by Shi and colleagues (7), multiple biopsies were obtained at different times or from disparate locations from several patients, and more than a single lesion in the ERK pathway was identified in multiple patients typically within different tumor biopsies. A detailed phylogenetic analysis of multiple progressive lesions from a subset of these patients suggested branching evolution of tumors in which the development of genetic diversity was not linearly associated with time. In one patient, for whom nine samples from progressing tumors were obtained over 726 days of BRAF inhibitor therapy, five distinct drivers of acquired BRAF inhibitor resistance were detected, including KRAS mutation, BRAF alternative splicing, and BRAF amplification. These results are consistent with a prior case report, which suggested that distinct mechanisms of BRAF inhibitor resistance were present in two different progressing lesions from a single patient (10). This finding of heterogeneity of resistance drivers within individual patients suggests that biopsy of individual lesions may have limited utility in guiding subsequent therapy selection. Novel methods for the analysis of circulating tumor cells or tumor-derived DNA in plasma may provide a better picture of the spectrum of resistance mechanisms within individual patients and may allow for the detection of resistant clones and therapy modification before evidence of clinical progression.

In the second study by Van Allen and colleagues (11), WES was performed on paired pretreatment and progression samples collected from 45 patients, of whom 14 developed resistance soon after initiation of therapy (within 12 weeks). They also detected several resistance mechanisms that had been previously identified to confer RAF inhibitor resistance, including mutations in NRAS, MAP2K1, and NF1 and BRAF amplification. They did not seek to assess the expression of alternatively spliced BRAF^V600E isoforms. Mutations in the MAP2K2 gene (which encodes the MEK2 kinase) were also identified in four patients, and focal amplification of MITF, a master lineage transcription factor, was identified in one. In vitro studies confirmed that the resistance mutations in MAP2K1 and MAP2K2 detected exclusively in the resistant samples were associated with resistance to both RAF and MEK inhibitors, albeit to varying degrees. Expression of these resistance alleles did not, however, confer resistance to a selective ERK inhibitor. In contrast, forced overexpression of MITF was sufficient to induce resistance to the BRAF, MEK, and ERK inhibitors studied. Furthermore, consistent with the results of Shi and colleagues (7), multiple resistance alterations within the MAPK pathway were found to co-occur in several patients for whom multiple disease progression samples were available.

In sum, the data support prior studies of BRAF inhibitor resistance in which alterations that cause reactivation of ERK signaling were found to occur in most patients. These studies provide support for the testing of RAF and MEK inhibitors in combination in patients with BRAF-mutant melanoma. Early results from a randomized phase II trial comparing combined RAF and MEK inhibition to RAF inhibitor monotherapy suggest a delay in median time to progression with the combination (9.4 months with the combination of RAF and MEK inhibitors versus 5.8 months with RAF monotherapy; ref. 12). In contrast, MEK inhibitors have shown little activity if initiated following disease progression on RAF inhibitors. A third study by Wagle and colleagues (13) in this issue of Cancer Discovery reports on the first in depth study of tumor samples collected from patients treated with the RAF and MEK inhibitor combination. Consistent with the preclinical studies highlighted above demonstrating that MEK1 and MEK2 mutations can confer RAF and MEK inhibitor resistance, a MEK2^Q60P mutation was identified in 1 of 5 patients studied. Of greater surprise to the investigators, one patient had a BRAF splice variant lacking exons 2-10 and a second patient had BRAF amplification. It should be noted that amplification of BRAF has previously been identified as a mechanism of MEK inhibitor resistance in cells with mutant BRAF (14). One hypothesis to explain this result is that increased abundance of the oncogenic driver (in this case BRAF) in response to prolonged drug treatment results in increased flux through the ERK pathway and restoration of ERK activity above the threshold required for inhibition of cell proliferation. As the antitumor activity of RAF inhibitors requires near-complete inhibition of ERK activation, minimal reactivation of ERK signaling may be sufficient for clinical progression (15). A similar mechanism of increased basal ERK activation in cells expressing RAF splice variants may also explain the selection for this alteration in patients treated with the RAF and MEK inhibitor combination. These early results from Wagle and colleagues (13) suggest that clinical strategies that achieve more maximal inhibition of ERK signaling may have greater clinical efficacy. These include novel RAF and MEK inhibitors with alternative mechanisms of kinase inhibition or greater potency and the use of intermittent treatment schedules (16). Alternatively, combined RAF and ERK inhibitor therapy may result in more durable treatment responses or may simply result in the selection of alterations downstream of ERK such as RB1 loss with similar kinetics. Finally, it should be noted that early trials of combined inhibition of RAF and MEK did not escalate the doses of each drug beyond their recommended single-agent doses. As the combination of a RAF inhibitor (which activates ERK in normal cells) and a MEK inhibitor (which inhibits ERK in normal cells) has less ERK-dependent toxicities than the use of either drug alone, further dose escalation of these agents may be achievable and may result in greater antitumor effects by inducing more maximal ERK pathway inhibition in BRAF-mutant tumors.

Finally, recent laboratory studies have shown that in BRAF-mutant tumors, high levels of ERK-dependent negative feedback potently suppress ligand-dependent mitogenic signaling (17). In the context of BRAF-mutant cells, relief of this ERK-dependent feedback results in the rapid adaptation of the cancer cell that attenuates the antitumor effects of the RAF inhibitor. One consequence of this adaptive response is RAS activation, the formation of RAF dimers, and a partial rebound in ERK activation (Fig. 1). The term “adaptive resistance” has been proposed to describe this blunting of the therapeutic response caused by rapid, adaptive changes to the signaling network following treatment with selective inhibitors of oncoprotein-activated pathways. In a fourth study by Shi and colleagues (18) in this issue of Cancer Discovery, the investigators demonstrate that BRAF (and BRAF + MEK) inhibitor therapy results in an increase in PI3K/AKT signaling that is associated with enhanced clonogenic growth dependence on PI3K and AKT. This adaptive response was attenuated by PTEN expression but could be rescued by the introduction of a mutant AKT, in this case AKT1^Q79K, a resistance allele identified in their study of disease progression samples (7). Taken in sum, the results suggest that the early adaptive response of BRAF-mutant cells to ERK pathway inhibition may promote the selection of resistant clones that harbor additional genomic events that confer higher levels of RAF inhibitor resistance. The data also support combinatorial approaches that attenuate the adaptive response, including the addition of a PI3K or AKT inhibitor to the RAF and MEK (or ERK) inhibitor combination. As previous attempts to combine MAPK and PI3K pathway inhibitors have been limited by overlapping toxicities, upfront testing of intermittent treatment schedules should be considered. As cyclin D1 expression is coregulated by both ERK and PI3K/AKT activation, evaluation of a CDK4 inhibitor in combination with the RAF/MEK inhibitor combination should also be pursued. In summary, the four studies reviewed above add significant new insights to our understanding of resistance to RAF inhibitor therapy and provide support for rational combinatorial treatment approaches that delay or prevent the emergence of drug-resistant clones.

D.B. Solit is a consultant/advisory board member of Pfizer. No potential conflicts of interest were disclosed by the other author.