Recent clinical data with BRAF and MEK1/2 [MAPK (mitogen-activated protein kinase)/ERK (extracellular-signal-regulated kinase) kinase 1/2] inhibitors have demonstrated the remarkable potential of targeting the RAF–MEK1/2–ERK1/2 signalling cascade for the treatment of certain cancers. Despite these advances, however, only a subset of patients respond to these agents in the first instance, and, of those that do, acquired resistance invariably develops after several months. Studies in vitro have identified various mechanisms that can underpin intrinsic and acquired resistance to MEK1/2 inhibitors, and these frequently recapitulate those observed clinically. In the present article, we review these mechanisms and also discuss recent advances in our understanding of how MEK1/2 inhibitor activity is influenced by pathway feedback.
The RAS–RAF–MEK1/2–ERK1/2 [where ERK is extracellular-signal-regulated kinase and MEK is MAPK (mitogen-activated protein kinase)/ERK kinase] signalling cascade controls the activity and/or expression of a large number of cellular substrates to co-ordinate processes such as cell proliferation, survival, differentiation and motility . In response to growth factors, GTP-loaded RAS proteins (HRAS, KRAS and NRAS) bind to and activate the RAF isoforms (ARAF, BRAF and CRAF). RAFs are serine/threonine protein kinases that phosphorylate and activate the dual-specificity kinases MEK1/2, which in turn phoshorylate and activate ERK1/2 . ERK1/2 are serine/threonine kinases that phosphorylate a plethora of substrates, including nuclear substrates such as the AP1 (activator protein 1) and ETS transcription factors, and also cytoplasmic substrates such as the pro-apoptotic protein BIM (BCL2-interacting mediator of cell death) [2,3]. A consequence of the prominent role that ERK1/2 signalling plays in promoting cellular proliferation and survival is the frequent hyperactivation of ERK1/2 in human cancers, which is the result of mutations in RTKs (receptor tyrosine kinases), RAS or BRAF. This has stimulated the development of numerous inhibitors that target components of the RAF–MEK1/2–ERK1/2 signalling node. The most advanced of these are the BRAF inhibitors vemurafenib and dabrafenib, and the MEK1/2 inhibitor trametinib, which have been approved in the U.S.A. for the treatment of BRAFV600E(/K)-mutant metastatic melanoma. Several other MEK1/2 inhibitors are also in later-stage clinical trials, including selumetinib (AZD6244) and cobimetinib (GDC-0973).
Inhibiting MEK1/2 in tumour cells
The selectivity of BRAF inhibitors for certain BRAF-mutant tumours (including BRAFV600E) affords a broad therapeutic window, which is likely to underpin their clinical success. However, BRAF inhibitors induce ‘paradoxical activation’ of ERK1/2 in BRAFWT (WT is wild-type) tumour cells [4–6] and this is thought to account for the frequent emergence of previously unsuspected RAS-mutant tumours in patients treated with vemurafenib or dabrafenib [7–9]. As a result, BRAF inhibitors are ineffective against tumours driven by alterations in RAS or RTKs. In contrast, MEK1/2 inhibitors inhibit ERK1/2 in cells harbouring mutations in BRAF, RAS or RTKs and do not elicit ‘paradoxical activation’ of ERK1/2. In addition, the majority of MEK1/2 inhibitors in clinical development are not ATP-competitive, but allosteric in action and bind to a unique hydrophobic pocket adjacent to the ATP-binding site, which permits exquisite selectivity [10,11].
Tumour cells with mutations in either BRAF or RAS frequently evolve strong dependency on ERK1/2 signalling and consequently exhibit sensitivity to MEK1/2 inhibitors [11–13]. However, recent work has identified previously underappreciated differences in the mechanisms of MEK1/2 inhibitor action. It has been known for several years that certain allosteric MEK1/2 inhibitors, including PD98059, U0126, PD184352, PD0325901, selumetinib and cobimetinib, enhance MEK1/2 activation loop phosphorylation [14–17] (Figure 1A). This is because inhibition of ERK1/2 disables several negative-feedback loops to upstream pathway components such as RAF, RAS and SOS (Son of Sevenless), thereby permitting stronger signalling to MEK1/2  (Figure 1A). However, in cells with an activating BRAF mutation, such as BRAFV600E, RAS and CRAF activity are low and many of these negative-feedback loops are weaker or disabled altogether, and so MEK1/2 inhibition does not result in enhanced phosphorylation of MEK1/2 [15,17,18,19]. A subset of allosteric MEK1/2 inhibitors, such as CH5126766, GDC-0623 and trametinib, can, however, disrupt phosphorylation of the MEK1/2 activation loops and do not promote the phosphorylation of MEK1/2, at least in certain RAS-mutant tumour cells [16,17,20,21] (Figure 1B). This block of phospho-MEK1/2 accumulation can result in stronger inhibition of ERK1/2 phosphorylation and tumour growth [16,17,21]. Structural and genetic analyses suggest that at least some MEK1/2 inhibitors that prevent the induction of MEK1/2 phosphorylation do so by displacing the MEK1/2 activation loops in to conformations that preclude phosphorylation by RAF [16,17]. The structural basis for this is currently debated. In the case of GDC-0623, structural modelling and genetic analyses suggest that a stronger hydrogen-bond interaction between the inhibitor and Ser212 of the MEK1 activation loop accounts for its ability to disrupt MEK1/2 phosphorylation compared with other inhibitors such as cobimetinib . However, Lito et al.  contend that interaction with Ser212 is invariably critical for the binding of MEK1/2 inhibitors to MEK1/2 and is unlikely to be the differentiator that results in the attenuation of MEK1/2 phosphorylation by some drugs, but not others . By solving the ternary crystal structure of CH5126766 in complex with MEK1 and an ATP analogue, Lito et al.  propose that interaction between the inhibitor and Asp221/Ser222 causes lateral displacement of the activation loop and that this is the differentiator that precludes activation loop phosphorylation by RAF compared with other inhibitors, including PD0325901. Different MEK1/2 inhibitors may, however, disrupt MEK1/2 phosphorylation through distinct interactions with MEK1/2, and this will be clarified by further structural analysis.
Mechanisms of ERK1/2 pathway inhibition by two distinct subsets of MEK1/2 inhibitors
Whether dual phosphorylation of the MEK1/2 activation loops is prevented, or just phosphorylation at one site, is unclear. However, another allosteric MEK1/2 inhibitor, trametinib (GSK1120212), was shown to inhibit phosphorylation of MEK1 at Ser218, but not Ser222, resulting in MEK1 with very low activity relative to the dual-phosphory-lated form . The ability of trametinib to prevent accumulation of phospo-MEK1/2 in RAS-mutant cells was cell-type-dependent [20,21], and this perhaps reflects variation in the strength of upstream signals to MEK1/2 and/or ERK1/2-dependent negative-feedback loops.
Interestingly, MEK1/2 inhibitors often promote or disrupt the interaction between RAF and MEK1/2, and this might influence their ability to prevent MEK1/2 phosphorylation by RAF, although no simple correlation is apparent. CH5126766 and GDC-0623, both of which disrupt the induction of MEK1/2 phosphorylation, promote the association of RAF and MEK1 in RAS-mutant cells, as do PD0325901 and selumetinib, which conversely permit the induction of MEK1/2 phosphorylation [16,17,21]. In contrast, cobimetinib and trametinib disrupt RAF–MEK1 complexes in RAS-mutant cells, but, whereas trametinib can prevent increased MEK1/2 phosphorylation, cobimetinib does not. Genetic disruption of the RAF–MEK1 interaction had little effect on basal ERK1/2 phosphorylation, but inhibited the rebound of ERK1/2 phosphorylation following treatment with PD0325901 or selumetinib . This suggests that drug-induced RAF–MEK1 complex formation attenuates drug efficacy in these contexts. The disruption of RAF–MEK1/2 complexes by trametinib has been proposed to at least partly account for its ability to reduce the induction of MEK1/2 phosphorylation, whereas the disruption of RAF–MEK1/2 interactions by cobimetinib was coincident with BRAF–CRAF dimerization and permitted the induction of MEK1/2 phosphorylation [16,17]. CH15126766-induced RAF–MEK1 complexes were disrupted by phospho-mimetic S218E and S222E mutations in MEK1, suggesting that the increased binding of CH15126766-bound MEK1 to RAF is a result of the inability to phosphorylate MEK1/2 . This hypothesis may also apply to GDC-0623, and the induction of RAF–MEK1/2 complexes by GDC-0623 has also been suggested to prevent RAF activation at the plasma membrane . Thus, although incompletely understood, both the intrinsic properties of a particular MEK1/2 inhibitor and cell-type variations in signal wiring are likely to affect the ability of a MEK1/2 inhibitor to robustly block ERK1/2 phosphorylation in scenarios where negative feedback operates.
Intriguingly, MEK1/2 inhibitors that permit accumulation of phospho-MEK1/2, such as cobimetinib and selumetinib, may exhibit better relative efficacy against BRAFV600E-positive tumours [16,20]. Given that MEK1/2 phosphorylation is often constitutively high in BRAFV600E-mutant cells [16,18], this may reflect stronger binding of these inhibitors to the active dual-phosphorylated form of MEK1/2 than those inhibitors that robustly contact the MEK1/2 activation loops. Indeed, trametinib exhibited ~20-fold greater potency towards unphosphorylated MEK1 when combined before CRAF activation than pre-activated dual-phosphorylated MEK1, suggesting that unphosphorylated MEK1/2 exhibits higher affinity for trametinib . In addition, binding of GDC-0623 to constitutively active MEK1 (S218D S222D) was 10-fold weaker than to MEK1WT, whereas binding of cobimetinib was unaffected, and this difference was reflected in the ability of these inhibitors to block ERK1/2 phosphorylation in cells . Similar properties have been observed for other MEK1/2 inhibitors . Thus the interaction between inhibitor (such as GDC-0623 or trametinib) and the MEK1/2 activation loop can attenuate activation loop phosphorylation or inhibitor binding, and the balance between these extremes may be determined by the relative proportion of phospho-MEK1/2 before treatment. Clearly, rational selection of a particular MEK1/2 inhibitor will be important to optimize clinical responses, and these results may motivate the structural refinement or discovery of new MEK1/2 inhibitors that further exploit these recent findings.
Intrinsic resistance to MEK1/2 inhibitors
Reactivation of MEK1/2 following loss of negative-feedback loops is one mechanism that can promote rebound of ERK1/2 activity and confer intrinsic resistance to MEK1/2 inhibitors. A variety of ERK1/2-independent mechanisms that attenuate the short-term responses of tumour cells to MEK1/2 inhibitors have also been identified. BRAF- and KRAS-mutant CRC (colorectal cancer) tumour cells can vary greatly in their sensitivity to the MEK1/2 inhibitor selumetinib [13,24]. In some cases, insensitivity correlated with high phospho-PKB (protein kinase B; also known as Akt) and/or low phospho-ERK1/2, and inhibition of PI3K (phosphoinositide 3-kinase) could sensitize these cells to MEK1/2 inhibitors, demonstrating the importance of PI3K signalling in mediating intrinsic resistance in some contexts [13,24,25]. In contrast with CRC cells, BRAFV600E-mutant melanoma cells are generally highly sensitive to MEK1/2 inhibitors, whereas the sensitivity of cells with NRAS mutations can vary widely . This highlights the strong dependency that BRAFV600E-positive melanoma cells develop on this oncogene and ERK1/2 signal-ling. Intrinsic resistance of melanoma cells to MEK1/2 inhibitors has not been as thoroughly investigated as for BRAF inhibitors, but it is likely that many of the mechanisms elucidated for BRAF inhibitors will also apply to MEK1/2 inhibitors. These include the absence or low expression of PTEN (phosphatase and tensin homologue deleted on chromosome 10) , CCND1 (cyclin D1) amplification  and A1 (BCL2A1, BFL1) or MITF (microphthalmia-associated transcription factor) amplification .
Other reported mechanisms of intrinsic resistance to MEK1/2 inhibitors include: EGFR (epidermal growth factor receptor)-dependent feedback activation of PI3K signalling in PDAC (pancreatic ductal adenocarcinoma) ; high PKA (protein kinase A) activity in CRC and NSCLC (non-small-cell lung cancer) ; elevated expression of Wnt signalling pathway components in CRC ; activation of STAT3 (signal transducer and activator of transcription 3) in lung cancer ; ERBB3 overexpression in thyroid cancer ; FOXD3 (forkhead box D3) and ERBB2/ERBB3 overexpression in melanoma [34,35]; and high ERK1/2-independent mTORC1 (mammalian target of rapamycin complex 1) activation in melanoma .
Acquired resistance to MEK1/2 inhibitors
Acquired resistance to targeted agents, including protein kinase inhibitors, is a major obstacle to the successful treatment of cancer. Our understanding of the mechanisms that can drive resistance to agents that target particular oncogenic signalling cascades has expanded enormously in recent years. A particularly valuable approach has been to model acquired resistance in vitro, for instance by chronically exposing tumour cell lines to a targeted agent; this has frequently recapitulated and even predicted acquired resistance mechanisms that occur in patients [37–40]. Several studies have modelled acquired resistance to MEK1/2 inhibitors or analysed samples from patients following disease progression on MEK1/2 inhibitors, and these are discussed below (Figure 2).
Mechanisms of acquired resistance to MEK1/2 inhibitors
To generate CRC tumour cells with acquired resistance to selumetinib, two independent studies chronically exposed four BRAFV600E-positive CRC tumour cell lines to selumetinib [41,42]. These cell lines exhibited high sensitivity to the anti-proliferative effects of selumetinib initially, but, following ~2–3 months of continual culture in drug, evolved 50–100-fold resistance. In all cases, resistance arose through an intrachromosomal amplification of BRAF, with no evidence of acquired mutations in the drug targets MEK1 or MEK2. Sequencing confirmed the selective amplification of BRAFT1779A (encoding BRAFV600E) in cells with acquired resistance. RNAi-mediated knockdown or inhibition of BRAF, and converse experiments overexpressing BRAFV600E, demonstrated that elevated BRAF expression was necessary and sufficient for resistance [41,42]. Up-regulation of BRAFV600E resulted in a larger pool of phospho-MEK1/2 with enough residual activity in the presence of selumetinib to reinstate ERK1/2 phosphorylation and cell proliferation (Figure 2).
Another study chronically exposed HT29 cells, a BRAFV600E-mutant CRC cell line, to the allosteric MEK1/2 inhibitor RO4927350 . Similar to the findings above, these cells exhibited increased phospho-MEK1/2 levels and possible BRAF protein up-regulation. However, these cells also harboured a MEK1F129L mutation, which was shown to increase the intrinsic kinase activity of MEK1, enhance ERK1/2 phosphorylation and confer resistance to MEK1/2 inhibitors when expressed in drug-naïve A375 melanoma cells . MEK1F129L mutation had been identified previously as a potential mechanism of acquired resistance in a random mutagenesis screen of MEK1 mutations that confer resistance to selumetinib . Various other potential activating MEK1 mutations were identified in this screen, such as MEK1Q56P and MEK1P124S, as were several MEK1 mutations that were predicted to abrogate drug binding, such as MEK1I111N, MEK1L115P and MEK1V211D. Indeed, a MEK1V211D mutation was identified in RKO (BRAFV600E CRC) cells with acquired resistance to the MEK1/2 inhibitor trametinib . The clinical importance of MEK1 mutations as a mechanism of acquired resistance to MEK1/2 inhibitors was demonstrated by the identification of a MEK1P124L mutation in the metastastic focus of a patient post-selumetinib relapse that was undetectable in pre-treatment samples .
A recent study in BRAF-mutant melanoma  has investigated acquired resistance to the MEK1/2 inhibitor trametinib, which, unlike selumetinib, at least partially inhibits the phosphorylation of MEK1/2 by RAF . Chronic exposure to trametinib led to the emergence of resistant derivatives harbouring a MEK2Q60P mutation in two of five subcell lines. This mutation occurs within helix A of the N-terminal negative regulatory region of MEK2, and mutations to proline may enhance intrinsic kinase activity . Indeed, ectopic expression of MEK2Q60P, but not MEK2WT, enhanced phosphorylation of both MEK1/2 and ERK1/2, and knockdown of MEK2, but not MEK1, partially restored trametinib sensitivity. These cells also exhibited a 20-fold amplification of the BRAF locus that selectively targeted the mutant BRAF allele, and knockdown of BRAF to parental levels partially restored trametinib sensitivity. The authors did not simultaneously knockdown MEK2 and BRAF to confirm that these alterations fully accounted for acquired resistance to trametinib in this context, but no other secondary mutations or known mechanisms of acquired resistance to MEK1/2 or RAF inhibitors were detected. In addition co-expression of MEK2Q60P and BRAFV600E in parental cells recapitulated trametinib resistance. Thus concurrent MEK2Q60P mutation and BRAFV600E amplification may account for resistance to trametinib in this context. In addition, a resistant post-progression biopsy from a patient treated with trametinib harboured a MEK2Q60P mutation that was not detected before treatment.
Several studies have reported increased BRAF copy number in MEK1/2 inhibitor-naïve tumour cells and pre-treatment samples from patients with BRAF-mutant CRC or melanoma [40,41]. Rare (~3–4%) COLO201 and COLO206F CRC tumour cells were found to exhibit a pre-existing increase in BRAF copy number, and one of 11 BRAFV600E-positive primary human CRC specimens was found to have a substantial pre-existing BRAF amplification . Enhanced BRAF copy number was also detected in pre-treatment biopsies from a BRAFV600E metastatic melanoma patient . These results suggest that acquired resistance may arise through the Darwinian selection of pre-existing cells with BRAFT1799A amplification, both in vitro and in patients. Consistent with this, mutational activation of oncogenes is often accompanied by local copy number gain of the mutant allele [45,46].
Related mechanisms of acquired resistance have been observed in CRC cells harbouring KRAS mutations. At least three studies have independently generated HCT116 cells (KRASG13D) with acquired resistance to MEK1/2 inhibitors. In the first, selumetinib resistance was driven by a substantial KRAS amplification; no evidence of MEK1 or MEK2 mutations was found  (Figure 2). As with BRAF up-regulation, KRAS up-regulation resulted in an increased pool of phospho-MEK1/2 that reinstated ERK1/2 phosphorylation and ERK1/2 pathway targets, such as CCND1, in the presence of selumetinib. Hatzivassiliou et al.  generated HCT116 cells with acquired resistance to PD0325901, and these also harboured KRAS amplification, but in addition acquired an activating MEK1F129L mutation . Single-cell cloning revealed substantial heterogeneity within this pool of PD0325901-resistant HCT116 cells. KRAS copy number between eight individual clones varied greatly from 3.5 to 89.5 copies, and, in the majority of cases, KRAS amplification was coincident with MEK1F129L mutation. In addition to MEK1F129L mutation and KRAS amplification, two of the clones analysed exhibited a MEK1I111N mutation, which is predicted to preclude drug binding [39,47]. In a separate study that chronically exposed HCT116 cells to trametinib, acquired resistance was associated with MEK1G128D/L215P mutations, consistent with previously identified mutations that are predicted to confer selumetinib resistance by disrupting the allosteric drug-binding pocket [39,44]. It is not clear why in two of these studies, MEK1 mutations emerged as a mechanism of acquired resistance, but not in the study by Little et al. , although this probably reflects heterogeneity in the drug-naïve HCT116 cell populations and/or the differing MEK1/2 inhibitors and drug doses employed [44,47]. Furthermore, chronic exposure of KRAS-mutant LoVo (CRC) and MDA-MB-231 (breast) tumour cells to PD0325901 resulted in acquired mutations in MEK2 (MEK2V215E) and MEK1 (MEK1L115P) respectively . These mutations map to the drug-binding allosteric site within MEK1 or MEK2 and are predicted to abrogate MEK1/2 inhibitor binding [39,40]. Interestingly, KRAS amplification in HCT116 cells also promoted PI3K-dependent phosphorylation of PKB . Strong PI3K/PKB signalling is a recognized mechanism of intrinsic resistance to MEK1/2 inhibitors [13,24], but the role, if any, of this enhanced PI3K activity in driving acquired resistance to selumetinib in HCT116 cells is unclear. Inhibition of PI3K did not resensitize these cells to the anti-proliferative effects of selumetinib, and cyclin D1 levels remained high even when both MEK1/2 and PI3K were inhibited, suggesting that other, as yet undefined, KRAS effector pathways contribute to resistance .
The mechanisms of acquired MEK1/2 inhibitor resistance described above share the common theme of ERK1/2 pathway reactivation in the presence of drug (Figure 2), highlighting the strong dependency that these tumour cells have on ERK1/2 signalling. Although some ERK1/2-independent resistance mechanisms have been described for BRAF inhibitors, the vast majority also involve the reactivation of ERK1/2 signalling . How tumour cells from other tissue types, such as the lung, pancreas and thyroid, which frequently exhibit mutations in BRAF or KRAS and dependency on ERK1/2 signalling, evolve resistance to MEK1/2 inhibitors is largely unexplored. In addition, further studies are required to verify whether tumour cells evolve the same or distinct mechanisms of resistance to MEK1/2 inhibitors with different properties, such as those that permit the accumulation of phospho-MEK1/2 compared with those that do not.
The inevitable development of acquired resistance to targeted monotherapies, including ERK1/2 pathway inhibitors, is driving the search for rational drug combinations that can improve primary efficacy and stall the development of acquired resistance. One approach is to inhibit the RAF–MEK1/2–ERK1/2 cascade at two or more levels, thereby ensuring that the robust inhibition of ERK1/2 required for tumour regressions is achieved . Co-targeting BRAF and MEK1/2 is a strategy applicable to BRAF-mutant tumours, and, in a recent Phase I/II clinical trial of patients with BRAFV600E/K-positive melanoma, combination of a BRAF inhibitor (dabrafenib) and a MEK1/2 inhibitor (trametinib) improved progression-free survival, the objective response rate and response duration relative to dabrafenib monotherapy . This combination was recently approved by the FDA (U.S. Food and Drug Administration) for the treatment of BRAFV600E/K unresectable or metastatic melanoma; however, acquired resistance still developed after an average 9.4 months treatment, with MEK2Q60P mutation, BRAF amplification or alternative BRAF splice variants being identified as potential resistance mechanisms .
Another promising approach is to harness the apoptotic potential of ERK1/2 pathway inhibitors using BH3 (BCL2 homology domain 3) mimetics . Despite invariably enhancing the activity and/or expression of pro-apoptotic BH3-only proteins, including BIM, BMF (BCL2-modifying factor) and PUMA (p53 up-regulated modulator of apoptosis), MEK1/2 and BRAF inhibitors frequently only elicit a cytostatic response in BRAF- or RAS-mutant CRC and melanoma tumour cells . However, recent studies have demonstrated that combining the MEK1/2 inhibitor selumetinib with navitoclax (ABT-263), a small-molecule inhibitor of the pro-survival BCL2 family proteins BCL2, BCL-w and BCL-XL, can elicit strong synergistic apoptosis in a variety of ERK1/2-addicted tumour cells [53,54]. Importantly, this combination delayed the development of acquired resistance in vitro and induced sustained tumour regressions in murine models [53,54]. A Phase I clinical trial combining dabrafenib, trametinib and navitoclax is currently recruiting patients (http://clinicaltrials.gov identifier NCT01989585). The results of these and trials evaluating combinations of MEK1/2 inhibitors with other targeted therapies will be eagerly anticipated as the search for more durable treatments continues.
Signalling and Acquired Resistance to Targeted Cancer Therapeutics: A Biochemical Society Focused Meeting held at Robinson College, Cambridge, U.K., 5–7 January 2014. Organized and Edited by Jim Caunt (University of Bath, U.K.), Simon Cook (Babraham Institute, Cambridge, U.K.) and David MacEwan (University of Liverpool, U.K.).
We thank members of the Cook group for helpful discussions and advice. We also thank Dr Paul Smith (AstraZeneca) for provision of selumetinib and many useful discussions.
This work was supported by a Biotechnology and Biological Sciences Research Council Ph.D. studentship (to M.J.S.), a Biotechnology and Biological Sciences Research Council Institute Strategic Programme Grant (to S.J.C.) and a research collaboration agreement between AstraZeneca and The Babraham Institute (to S.J.C.). Neither S.J.C. nor M.J.S. received any personal remuneration from AstraZeneca.