Allele-Specific Mechanisms of Activation of MEK1 Mutants Determine Their Properties

RAS/RAF/MEK/ERK signaling is activated by upstream signaling pathways, including those initiated by receptor tyrosine kinases, and the strength and duration of the ERK signal is determined by negative feedback (1–3). Genetic lesions that deregulate the ERK pathway are important features of malignant transformation and are often drivers of tumor growth. Gain-of-function mutations of KRAS, NRAS, and BRAF and loss-of-function lesions of NF1 occur commonly in human cancer. Activating RAF family fusion proteins and MEK1, MEK2, ARAF, HRAS, CRAF, and ERK mutations also occur, but more rarely. The overall picture that emerges from these data is that mutations in different components of this pathway occur in many cancers and play a role in their development. What accounts for the differing frequencies of these mutations in cancer as a whole and in specific tumors and how the oncogenic consequences of mutations of the pathway may vary are poorly understood, if at all (4).

MEK1 and MEK2 are dual-specific serine/threonine and tyrosine kinases that become activated when phosphorylated by RAF kinases. They, in turn, phosphorylate and activate ERK1/2. Both MEK1 and MEK2 mutations occur in untreated human tumors (5–9) and have also been associated with the acquired resistance of tumors with BRAF^V600E to RAF/MEK inhibitors (10, 11). MEK1 mutations occur in multiple sites in the protein, without a single dominant hotspot, and the properties of the different mutants have not been well defined.

Here, we have studied the mechanisms of activation of a panel of cancer-associated MEK1 mutants and defined them into three different classes. The first is RAF-dependent and sensitive to ERK-dependent feedback inhibition of RAS/RAF signaling, has low output and transforming activity in untransformed cells, and, in human tumors, coexists with other mutations that activate ERK. The second class of mutant has baseline RAF-independent activity but can be, to a greater or lesser degree, further activated when phosphorylated by RAF. These mutants are also sensitive to ERK-dependent feedback, and they are often but not always associated with other ERK-activating mutants in tumors. The third class of mutant is both RAF and phosphorylation independent, insensitive to feedback, and constitutively active. It induces high levels of ERK output and transformation in untransformed cells and is unassociated with other ERK-activating mutations in tumors. These mutants are hyperactive and, unlike class 1 and 2 mutants, unresponsive to current MEK inhibitors, which bind to the inactive conformation of the enzyme. In contrast, an ATP-competitive MEK inhibitor suppresses the activity of all three classes of mutants. Thus, the work highlights the critical importance of functional characterization of different mutant alleles to understand their mechanism of activation, and their biological effects, and to develop effective targeted therapies.

To characterize the functional consequences of MEK1 mutations, we cloned 17 MEK1 mutants that have been identified as recurrent in human tumors (Fig. 1A) and that, together, have been found in a wide variety of cancer types (cBioPortal; ref. 12). When expressed in 293H cells, each mutant caused higher levels of phosphorylated ERK (pERK) than wild-type (WT) MEK1 does (Fig. 1B; Supplementary Fig. S1). They activated ERK signaling to varying degrees with different MEK/ERK phosphorylation patterns that can separate the mutations into different classes. Some mutants induced pERK more profoundly than pMEK (C121S, L177M, E203K, F53L, ΔF53-Q58, Q56P, K57E, K57N, and ΔE51-Q58; Fig. 1B). In contrast, others (e.g., ΔL98-I103, ΔE102-I103, ΔI99-K104, and ΔI103-K104) induced high levels of both pERK and pMEK. A third group (D67N, P124L, P124S, and L177V) only weakly induced both pERK and pMEK. WT MEK1 is activated by its phosphorylation on S218 and S222 by RAF kinases (13, 14). To test whether phosphorylation of these sites is similarly required for activation of the MEK1 mutants, we introduced both the S218A and S222A mutations into all of them. The double mutation completely abolished ERK activation in WT MEK1 and in four mutants (Fig. 1C), each of which was a weak activator of pERK in 293H cells (Fig. 1B). In contrast, the other 13 mutants all retained some degree of activation when S218 and S222 were replaced by alanine. This suggests some degree of phosphorylation-independent activation of catalysis by these MEK1 mutants. For nine of these, the double mutation clearly decreased, but did not abolish, the induction of pERK by the mutants (Fig. 1D). Thus, phosphorylation of these mutants enhances but is not required for induction of signaling. In four other mutants, those with in-frame deletions in parts of the region between amino acids 98 and 104, signaling was unaffected by mutations of the two phosphorylation sites (Fig. 1E). These were the mutants that, in Fig. 1B, caused hyperphosphorylation of both MEK and ERK. Thus, a set of MEK1 mutants signal in a phosphorylation-independent manner to varying degrees, and the degree to which their activation is phosphorylation independent correlates directly with the level of their phosphorylation in cells.

We therefore asked to what degree RAF kinase regulates the phosphorylation and kinase activity of these MEK1 mutants. We purified GST-tagged MEK1-mutant proteins and performed in vitro kinase assays in the absence or presence of active BRAF kinase. Recombinant purified WT MEK1 was not phosphorylated in the absence of RAF kinase, nor were any of the four MEK1-mutant proteins whose kinase activity was abolished by S218A/S222A (Figs. 1C and 2A; Supplementary Fig. S2A). Both the phosphorylation and kinase activities (with ERK2 as a substrate) of WT MEK1 and these mutants were potently induced by activated BRAF kinase (Fig. 2A). These are thus RAF and phosphorylation-dependent MEK1 mutants. Presumably, their elevated activation in cells is due to enhanced activation by phosphorylation compared with WT. In contrast, the 9 mutants whose kinase activities were only partially decreased by S218A/S222A (Fig. 1D) all phosphorylated ERK2 in the absence of BRAF, and their activity was further stimulated, marginally or significantly, by BRAF (Fig. 2A). These are thus RAF-regulated mutants. Phosphorylation of one of these mutants (C121S) was detected in the absence of RAF and was further induced 4-fold by RAF (Fig. 2A). The other 8 mutants had no or barely detectable endogenous phosphorylation of S218 and S222 (Fig. 2A; Supplementary Fig. S2B). These two groups together comprise the mutants that increase pERK to a much greater extent than pMEK in cells (Fig. 1B). In contrast, the four mutants with deletions in the aa 98–104 hyperactivate pMEK and pERK in cells and are completely insensitive to S218 and S222 mutations. In vitro, purified recombinant mutants of this class are phosphorylated in the absence of RAF kinases, and the addition of active BRAF affects neither their phosphorylation nor their kinase activity (Fig. 2A; Supplementary Fig. S2C). These RAF-independent mutants are heavily phosphorylated in vitro and in cells, but their activity is phosphorylation independent.

We asked whether these mutants are phosphorylated by other kinases or whether they undergo autophosphorylation. To answer this question, we introduced a mutation that abrogates MEK kinase activity (K97R) into these MEK1 mutants. The activity of the K97R mutants was assessed with an in vitro kinase assay in which inactive ERK2 was used as the substrate. We found that the kinase dead mutation abolished ERK phosphorylation by the mutant MEK1 as expected. In addition, the phosphorylation of the K97R mutant was also abrogated in the absence of RAF kinases (Fig. 2B), supporting the hypothesis that these mutants autophosphorylate S218/S222. To determine whether this represents true autophosphorylation in cis or whether phosphorylation occurs in trans, we modified MEK1-mutant protein with a FLAG tag and the K97R-mutant protein with a GST tag. The isolated FLAG-tagged mutants and GST-tagged kinase dead mutants were then coincubated in a kinase assay, and the phosphorylation of the tagged proteins was assessed. If phosphorylation occurs in trans, both FLAG-tagged and GST-tagged MEK1 mutants should be phosphorylated; if autophosphorylation is in cis, only the FLAG-tagged protein will be phosphorylated (Fig. 2C). As shown in Fig. 2D, FLAG-tagged activating MEK1 mutants were phosphorylated, but they did not phosphorylate GST-tagged K97R mutants. These data suggest that these MEK1 mutants are autophosphorylated in cis.

Our data show that MEK1 mutants may function in one of three different manners: RAF-dependent, RAF-regulated, or RAF-independent activation. To test this idea, we utilized ARAF^lox/lox; BRAF^lox/lox; CRAF^lox/lox; RERT^ert/ert mouse embryonic fibroblast (MEF) cells. The genes encoding ARAF, BRAF, and CRAF can be deleted from this cell with an adenovirus encoding Cre recombinase administered together with 4-hydroxytamoxifen. We used this system to assess the activity of MEK1 mutants in cells that express none of the WT RAF isoforms. As shown in Fig. 2E, MEK1 mutants that function in a completely RAF-independent manner (ΔL98-I103 and ΔE102-I103) are phosphorylated and potently stimulated ERK signaling despite the absence of expression of RAF kinases. In contrast, in cells expressing WT MEK1 or RAF-dependent MEK1 mutants (D67N, P124L, and P124S), both MEK and ERK phosphorylation were decreased to trace levels in RAF-less cells. RAF-regulated MEK1 mutants (F53L, C121S, K57N, and Q56P) have an intermediate effect on signaling when expressed in these cells; pMEK was almost completely eliminated, whereas a variable, but significant, amount of ERK phosphorylation was retained. These data are consistent with our conclusion that MEK1 mutants can be placed into three functional classes based on their dependence on phosphorylation by RAF for activation (Table 1).

The mutants are thus categorized in terms of their dependence on upstream activation of RAF (Fig. 2E): RAF-dependent mutants (blue), which are activated by RAF-dependent phosphorylation; those RAF-regulated mutants (orange) that are partially dependent on RAF but have varying amounts of RAF-independent activity; and those RAF-independent mutants (red) that are completely RAF-independent. These categories, in turn, predict whether signaling output will be sensitive to ERK-dependent feedback inhibition. In NIH3T3 cells, RAF-dependent MEK1 mutants activate pERK and ERK pathway output modestly (Fig. 3A and B; Supplementary Fig. S3A) without affecting levels of RAS-GTP or CRAF phosphorylation (Fig. 3A). Increased ERK output by them is therefore not sufficient to cause feedback inhibition of RAS/RAF. This is as expected if pathway output is self-limited by feedback inhibition of RAS or RAF. In contrast, RAF-independent mutants potently activate pERK while reducing CRAF S338 phosphorylation and RAS activation (Fig. 3A). In these cells, the ERK output was highest (Fig. 3B; Supplementary Fig. S3A) and RAS-GTP and pCRAF were decreased compared with the levels in cells with WT MEK1. RAF-regulated MEK1 mutants have variable intermediate phenotypes: Those with the highest levels of RAF-independent activation of ERK signaling (K57N, Q56P) cause higher levels of pERK and lower levels of RAS-GTP than the other tested mutants in this class (Figs. 2 and 3A and B). This result strengthens the model that the degree of RAF independence of MEK1 mutants is directly associated with their activation of the ERK pathway and inversely associated with their sensitivity to feedback inhibition of RAS/RAF. Degree of RAF independence is similarly associated with the ability of the MEK1 mutants to drive cell proliferation and transformation. The RAF-less cells are viable but do not proliferate (15). The expression of various MEK1 alleles was induced in the RAF-less cells, and their ability to grow in soft agar was assessed (Supplementary Fig. S3B). No growth in soft agar was noted in RAF-less cells in which expression of WT MEK1, RAF-dependent MEK1 mutants, or RAF-regulated MEK1 mutants was induced. In contrast, RAF-independent mutants strongly promoted the anchorage-independent growth of the RAF-less cells. Similar results were obtained when we examined the ability of the MEK1 mutant–expressing RAF-less cells to form tumors in immunosuppressed (nude) mice (Supplementary Fig. S3C). Tumors were observed as early as 14 days after implantation of cells expressing RAF-independent MEK1 mutants (red). In contrast, no palpable tumors were noted as late as 42 days after implantation of cells with RAF-dependent mutants (blue). Tumors did grow from cells in which RAF-regulated mutants (orange) were expressed, but more slowly than those expressing RAF-independent MEK1 mutants. The time of appearance of the former and their rate of growth were directly correlated with the degree to which the catalytic activity of the mutant was RAF independent.

The results show that, unlike the RAF-independent MEK1 mutants, the RAF-dependent or regulated mutants drive transformation only weakly. We speculated that, to significantly activate ERK pathway, they must coexist with other lesions that activate RAF despite ERK-dependent feedback. This turned out to be the case. In Fig. 3C, MEK1 mutants are ordered on the x-axis by their degree of RAF independence. The y-axis represents the fraction of MEK kinase activity that is RAF independent for each of these MEK1 mutants (RAF-independent kinase activity divided by kinase activity when RAF is added to the in vitro kinase assay). Analysis of a large database (MSK-IMPACT) of human tumor–associated mutations showed that the RAF-dependent MEK1 mutants are almost always associated with activating mutations of RAS or RAF or RTK genes or NF1-inactivating lesions (Fig. 3C). In contrast, tumors that express any of the four RAF-independent mutants are unassociated with coexistent mutations that activate RAF signaling. MEK1 mutants with RAF-independent activity that is enhanced by RAF sometimes coexist with RAF-activating mutations and sometimes do not. The data suggest that the frequency of coexistent RAF-activating mutations may be inversely correlated with the degree of RAF independence of the RAF-regulated MEK1 mutants (Fig. 3D).

Our data suggest that the degree of RAF dependence of MEK1 mutants determines their ability to drive ERK pathway output and, consequently, whether they are coselected with other mutations that drive RAF. The recurrence of these mutations in human cancer and their ability to activate ERK output suggested that they are tumor drivers, so we asked whether they were sensitive to MEK inhibitors. Several highly selective inhibitors of MEK are used in the clinic as anticancer agents. These inhibitors bind to an allosteric site adjacent to the ATP pocket (16–19). They do not compete with ATP for binding and preferentially bind to the inactive conformation of the enzyme. Here, we tested the effects of three different allosteric MEK inhibitors in NIH3T3 cells expressing WT or mutant MEK1 (Fig. 4). All three allosteric inhibitors lead to only small shifts in IC₅₀ (∼3- to 10-fold) in some of the RAF-dependent (blue) or RAF-regulated (orange) MEK1 mutant–expressing cells relative to WT MEK1. In contrast, ERK activation in cells expressing RAF-independent (red) MEK1 mutants was inhibited only at approximately 10- to 100-fold higher doses (Fig. 4).

Previous studies suggest that the phosphorylation of MEK1 could affect its sensitivity to allosteric MEK inhibitors (17). Phosphorylation of purified MEK1 has been reported to cause a 20-fold decrease of its sensitivity to trametinib (Supplementary Table S1; ref. 17). We found that prephosphorylation of purified MEK1 kinase significantly reduced its sensitivity to the CH5126766 (100-fold) and PD901 (10-fold; Supplementary Fig. S4A and S4B; Supplementary Table S1). These data suggested that the constitutive RAF-independent hyperphosphorylation of RAF-independent MEK1 mutants (Fig. 2E; Supplementary Fig. S2C) could cause their insensitivity to the allosteric MEK inhibitors. In contrast, the effects of a new selective ATP-competitive MEK inhibitor, MAP855, are not reduced by phosphorylation of the enzyme. As shown in Fig. 5A, the prephosphorylation of MEK1 does not decrease its sensitivity to MAP855. As predicted, this drug effectively inhibited ERK signaling driven by each class of MEK1 mutants, including those with the aa 99-104 deletion (Fig. 5B).

Thus, our work divides mutant MEK1 alleles into three groups, based on the degree of dependence of their catalytic activity on RAF-mediated phosphorylation of S218 and S222. The groups differ in the degree with which they can autonomously activate ERK signaling, their sensitivity to ERK-dependent feedback inhibition of RAS/RAF, whether they require cooperation with other mutant proteins that activate ERK signaling, and in their sensitivity to inhibitors that work by different mechanisms. The first type of mutant is completely RAF dependent, and its activity is hyperstimulated by phosphorylation compared with that of WT. This type of mutant has little or no ability to drive signaling on its own and likely acts as an amplifier of the ERK signaling driven by other mutants with which they coexist. The second group varies in its degree of RAF independence, which varies inversely with their coexistence with other ERK pathway mutants. The third group has unique properties and is comprised of four mutants with deletions in the region bounded by amino acids ⁹⁸LIHLEI¹⁰⁴K. The kinase activity of these mutants is constitutive and RAF and phosphorylation independent, although they are highly phosphorylated in cells due to their autophosphorylation in cis on S218 and S222. A computationally solved structure shows that this region of the protein interacts with BRAF (Fig. 6A and B; Supplementary Fig. S5). This interaction is apparently required for the MEK1/BRAF interaction in cells because, unlike WT MEK1 or RAF-dependent or regulated MEK1 mutants, the RAF-independent mutants do not coimmunoprecipitate with BRAF or CRAF (Fig. 6C). This is not due to hyperphosphorylation of S218/S222 because the S218A/S222A double mutant does not restore binding to BRAF (Fig. 6D). The results are consistent with the structure and suggest that the interaction of 98-104 peptide in the β3-αC loop of MEK1 with BRAF is necessary for the binding of the two proteins. As deletions in this region activate MEK1 in a RAF-independent manner, one might infer that RAF binding to this site inhibits MEK activity and that, for WT MEK1, this inhibition is relieved by RAF-mediated phosphorylation of S218/S222. However, this does not seem to be the case, because in RAF-less cells, WT MEK1 is also inactive. The data are more consistent with the hypothesis that aa 98-104 is a potent negative regulator of MEK1 kinase activity, deletion of which leads to constitutive activation. Deletions in the β3-αC loop have also been identified in EGFR, HER2, and BRAF as activating mutations in patients with cancer (20). Displacement of helix-αC away from the β3-5 core to an inactive αC-out conformation requires a minimum length of the loop for movement, implying that β3-αC deletions work by shifting the equilibrium to the active conformation. Foster and colleagues showed that kinase activity varies as a function of the insert length, with a maximum observed for a five amino acid deletion, also the most prevalent clinical deletion mutation (21). Analysis of all available kinase structures revealed that αC-out conformations preferentially occur in kinases with an intermediate β3-αC insert length (22). The deletions in MEK1 may result in a stabilization of the active conformation by a similar mechanism. Thus, it is likely that activation of WT MEK1 by RAF involves phosphorylation-dependent relief of the negative regulation by this domain. This is supported by the loss of any effect of phosphorylation on the activity of the deletion mutants, whereas activation of all the other mutants is enhanced by phosphorylation.

MEK1 mutants occur in more than 1% of human tumors with higher frequency in histiocytosis, melanoma, and colorectal, bladder, lung, and other carcinomas. Mutations occur at many sites in the protein, with only three modest hotspots, K57, P124, and E203, the rest distributed at very low frequency at many sites. These frequencies hold for tumors that have not been treated with ERK pathway inhibitors; they may be higher in tumors with acquired resistance to RAF or MEK inhibitors. As is true for most proto-oncogenes, most of the low-frequency mutants have not been studied and, until recently, there has been a general sense that they all function similarly and will respond to the same targeted drugs.

In this work, we have functionally characterized 17 oncogenic MEK1 mutants from human tumors and identified three different functional classes, based on their dependency on phosphorylation by RAF kinase or lack thereof (Table 1). Class 1 mutants are dependent on and hyperactivated by phosphorylation of S218 and S222 by RAF; thus, they are sensitive to ERK-dependent feedback inhibition of upstream signaling. Generation of high ERK output in these cells, therefore, requires other molecular events that reduce the sensitivity of RAF to negative feedback. In support of this model, we found that these mutants always coexist with RAS, RAF, or NF1 alterations in tumors. We believe these mutants serve as amplifiers of ERK signaling rather than autonomous activators. Melanomas with coexistent BRAF^V600E and class 1 MEK1 mutants have been identified in untreated tumors. Although their existence was shown to contribute to the ERK activation in tumors, they are unlikely to cause resistance to the RAF inhibitor treatment (23, 24). In contrast, the class 2 MEK1 mutant kinases have some level of basal, RAF-independent activity but are further activated by RAF. They sometimes coexist with RAS, RAF, or NF1, and the degree to which they do varies inversely with the fraction of their activity that is RAF-independent. This finding suggests that selection of these mutants, alone or together, is based on a requirement for a certain degree of ERK output, in a range that enhances growth but does not cause toxicity. In contrast to class 1, class 2 mutants have been found to cause acquired resistance to RAF inhibitors (24, 25).

Uniquely, class 3 mutants are completely RAF independent. In cells, they drive high levels of ERK output and are insensitive to the ERK-dependent feedback inhibition of RAS and RAF. Thus, unlike class 1 mutants, they autonomously drive signaling. Tumors in which class 3 mutants coexist with RAS, RAF, or NF1 mutants have not been identified so far.

The currently available MEK inhibitors are all allosteric and preferentially bind to the inactive form of the enzyme. Class 1 and 2 mutants were also sensitive to these drugs, suggesting that a significant fraction of these enzymes is in the unphosphorylated, inactive state. Our data suggests that the aa 98–104 region of MEK1 is autoinhibitory, and its deletion causes class 3 enzymes to permanently adopt the active conformation. In support of this idea, class 3 mutants are resistant to allosteric MEK inhibitors. The other option to inhibit the activity of this class of mutant MEK1 would be using an ATP-competitive MEK inhibitor. Although there is one such compound, E6201, which is currently under clinical investigation, it also has many non-MEK targets (26–28). Here, we examined a newly developed selective MEK inhibitor, MAP855, that functions by competing ATP binding to MEK protein. In contrast to the cases with allosteric MEK inhibitors, all three classes of MEK1 mutants are sensitive to this new ATP-competitive inhibitor. Such drugs may be useful in treating tumors dependent on any classes of MEK1 mutants and in preventing class 2 or class 3 MEK1 mutant–dependent acquired resistance.

Taken together, our work suggests that detailed functional analysis of multiple oncogenic alleles in a given protein is a powerful means for understanding different biochemical strategies for oncogenic activation and is necessary if we are to successfully use the genomics of the tumor to develop effective therapies.

The 293H, NIH-3T3, and Phoenix AMPHO cells were purchased from the ATCC between 2013 and 2015. ARAF^lox/lox; BRAF^lox/lox; CRAF^lox/lox; RERT^ert/ert MEF cells were provided by M. Barbacid in February 2016 and were confirmed by MSK-IMPACT. These cell lines and all other cells with inducible expression of MEK1 mutants were tested negative for Mycoplasma contamination. All the cells were grown in DMEM with glutamine, antibiotics, and 10% FBS. Cells engineered to inducibly express a protein were maintained in standard medium with 100 μg/mL hygromycin and 2 μg/mL puromycin.

NIH-3T3 cells were used to construct the stable lines with inducible expression of mutant MEK1s to study MEK1 mutant–driven signaling output and their response to different types of MEK inhibitors. Conditional RAF knockout MEF cells were employed to determine the RAF dependency of mutant MEK1–driven ERK pathway activation. Coexpression and coimmunoprecipitation experiments were performed with the 293H cells, in which more than two exogenous proteins were transiently expressed. Cell lines were used after less than 3 months of passages after receipt for the above experiments.

Western blot analysis, immunoprecipitation, and in vitro kinase assays were performed as described previously. The following antibodies were used: anti-p217/p221-MEK1/2 (pMEK1/2; #9154), anti-p202/p204-ERK1/2 (pERK1/2; #4370), anti-MEK1/2 (#4694), anti-ERK1/2 (#4696) from Cell Signaling Technology; anti-V5 (R960-25) from Thermo Fisher Scientific; anti-ARAF (sc-408) and anti-BRAF (sc-5284) from Santa Cruz Biotechnology; anti-FLAG (F3165) from Sigma; anti-CRAF (610152) from BD Transduction Laboratories; anti-pCRAF S338 (05-538) from Millipore; and anti-RAS antibody from the Active RAS Pull-Down and Detection Kit (Thermo Fisher Scientific, #16117). For immunoprecipitations of tagged proteins, the following reagents were used: anti-V5 agarose affinity gel (Invitrogen), anti-FLAG M2 affinity gel (Sigma), protein G agarose gel (Invitrogen).

The pcDNA3-FLAG-MEK1 was obtained from Biomyx. The pcDNA3-BRAF-V5 was constructed as described previously (29). The pGEX6P1 was obtained from Addgene. The MEK1 gene was subcloned into pGEX-6P-1 for in vitro protein purification. Plasmids TTIGFP-MLUEX and pMSCV-rtTA3-PGK-Hygro for inducible gene expression were provided by Scott Lowe's lab at Memorial Sloan Kettering Cancer Center (MSKCC; New York, NY). The MEK1 gene was subcloned into TTIGFP-MLUEX vector harboring the Tet-responsive promoter. Mutations were introduced using the Site-Directed Mutagenesis Kit (Stratagene).

Trametinib and CH5126766 were obtained from GlaxoSmithKline and Chugai Pharmaceuticals, respectively. PD0325901 was synthesized in the MSKCC Organic Synthesis Core Facility by O. Ouerfelli. MAP855((1-((3S,4S)-4-(8-(2-chloro-4-(pyrimidin-2-yloxy)phenyl)-7-fluoro-2-methyl-1H-imidazo[4,5-c]quinolin-1-yl)-3-fluoropiperidin-1-yl)-2-hydroxyethanone)) was obtained from Novartis (compound No. 1, WO2015022662). The referenced patent provides detailed structure and methods for synthesis. Doxycycline and 4-hydroxytamoxifen were from Sigma Aldrich; puromycin and hygromycin stock solution were from Invitrogen. 4-Hydroxytamoxifen was dissolved in methanol to yield 5 mmol/L stock; all other drugs were dissolved in DMSO to yield 10 mmol/L stock and stored at −20°C.

ARAF^lox/lox; BRAF^lox/lox; CRAF^lox/lox; RERT^ert/ert MEF cells were generously provided by M. Barbacid. MEF cells were isolated from embryonic day 13.5 embryos and immortalized as reported previously (15). For generation of RAF-less cells, ARAF^lox/lox; BRAF^lox/lox; CRAF^lox/lox; RERT^ert/ert MEF cells were infected with Adeno-Cre particles (multiplicity of infection = 100) for 24 hours; then, cells were cultured in medium with 1 μmol/L 4-hydroxytamoxifen (Sigma) for a week.

Retroviruses encoding rtTA or MEK1 genes were packaged in Phoenix-AMPHO cells obtained from the ATCC. The supernatant-containing virus was filtered with 0.45 μm PVDF membrane. The target cells were infected with virus for 8 hours. Forty-eight hours later, cells were selected in medium containing puromycin (2 μg/mL) or hygromycin (100 μg/mL) for 3 days. The positive infected cell populations were further sorted using GFP as a marker after overnight exposure to 1 μg/mL doxycycline. GFP-positive cells were then cultured and expanded in medium with doxycycline along with antibiotics. For the conditional RAF knockout cells, the inducible expression cells were created in the absence of 4-hydroxytamoxifen.

Cells were seeded in 60 or 100 mm plates and transfected the following day using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. A ratio of DNA to lipofectamine of 1 μg DNA/3 μL lipofectamine was employed.

Total RNA was collected from cultured cells using the SV Total RNA Isolation System (Promega). cDNA was synthesized with ImProm-II Reverse Transcription System using Oligo(dT) Primer (Promega), and RT-PCR was performed on a ViiA 7 Real-Time PCR instrument with SYBR Green PCR Master Mix (Applied Biosystems), using the following program: holding at 50°C for 2 minutes and polymerase activation at 95°C for 10 minutes, followed by 40 cycles of amplification (95°C for 15 seconds, 60°C for 1 minute). GAPDH expression was used as an internal reference to normalize input cDNA. Fold changes of the gene expression between the mutants and WT were averaged from the three replicates and logarithm transformed before clustering analysis. The heat map was plotted by R package, with genes in the row clustered on the basis of similarity of Pearson correlation coefficient. One-tailed t tests were performed using R package for comparisons among three different groups.

Cells were cultured in 10-cm dishes until 70% to 80% confluence. GTP-bound RAS was quantitated using the RAF1 RAS-binding domain pulldown from Detection Kit (Thermo Fisher Scientific), as instructed by the manufacturers.

RAF-less cells were seeded in 0.4% agar on 6-well plates, incubated at 37°C for 4 weeks, and stained with 0.005% crystal violet (Sigma-Aldrich) for 1 hour at room temperature. Soft-agar colonies were quantified using Gen5 (BioTek). Captured images were analyzed using uniform size and shape parameters.

Cells were collected on ice and lysed in lysis buffer (50 mmol/L Tris, pH 7.5, 150 mmol/L NaCl, 0.1% NP40, 1 mmol/L EDTA) supplemented with protease and phosphatase inhibitor cocktail tablets (Roche). Immunoprecipitations were carried out at 4°C for 1 hour, followed by five washes with lysis buffer. The pulldown complexes were eluted with 1× SDS loading buffer and assayed by Western blotting.

Human WT MEK1, as well as all the mutants used in this study, were subcloned into pGEX-6P-1, expressed as glutathione-S-transferase fusions, and purified by Pierce Glutathione Agarose (Thermo Fisher Scientific).

In vitro kinase assays were conducted in the presence of 200 μmol/L ATP, at 30°C for 15 minutes. Briefly, GST-MEK1 or mutants were incubated in the absence or presence of active BRAF^V600E protein (Upstate). Changes in MEK1 phosphorylation were estimated by immunoblotting for pMEK. To test the kinase activity of WT or mutant MEK1 protein, recombinant inactive ERK2 protein (GenWay Biotech) was used as a substrate, and the reaction was terminated with the addition of 1× SDS loading buffer and boiling. Kinase activity was estimated by immunoblotting for pERK. When the subsequent in vitro kinase assay was performed after immunoprecipitation, 0.02% SDS was added to the wash buffer and one extra wash with kinase buffer was required.

All structure models of MEK1 or MEK1/BRAF complex were generated from 10ns AMBER15 constant pressure simulations using the PARM99 (30) force field for the proteins and PARM@FROSST force field for the small molecules at 1 atm and 300 K. Conformationally and topologically averaged AM1-BCC (31) charges were generated for the small molecules. WATMD (32) was used to generate time-averaged structures for evaluation of the PPIs.

Mutational data were obtained from three sources: (i) The Cancer Genome Atlas; (ii) various published studies in which mutational data were publicly available; and (iii) an ongoing prospective clinical sequencing initiative at MSKCC that utilizes a targeted capture-based sequencing of protein-coding exons and selected introns of 314 or 410 cancer-associated genes (MSK-IMPACT). Because mutation calling algorithms and mutation filtering and reporting practices varied from study to study, somatic mutational data review and correction were undertaken where possible as described previously. Mutation calls were uniformly reannotated to gene transcripts in Ensembl release 75 (Gencode release 19), and a single canonical effect per mutation was reported using Variant Effect Predictor version 81 and vcf2maf version 1.6.2. To exclude putative germline variants misattributed as somatic mutations, we excluded any variant identified in ExAC r0.3 with a minor population allele frequency greater than 0.06%. We sought to determine the enrichment of co-occurring RTK/RAS/RAF/NF1 lesions in tumors possessing MEK1 mutants in all cancer types.

Nu/nu athymic mice were obtained from Harlan Laboratories and maintained in compliance with Institutional Animal Care and Use Committee (IACUC) guidelines. RAF-less cells with WT or mutant MEK1 expression were implanted subcutaneously into the left and right posterior flanks of 4- to 6-week-old Nu/nu athymic female mice (n = 5 mice/group). Tumor formation was measured twice a week. All studies were performed in compliance with institutional guidelines under an IACUC-approved protocol. Investigators were not blinded when assessing the outcome of the in vivo experiments.

Results are mean values ± SD. Two-tailed t tests were performed using GraphPad Prism 7.01 for comparisons among different groups. Investigators were not blinded when assessing the outcome of the in vivo experiments. All cellular experiments were repeated at least three times.

R. Yaeger is a consultant/advisory board member for Glaxo-SmithKline. D. Stuart reports receiving a commercial research grant from Novartis and has ownership interest (including patents) in the same. H. Moebitz has ownership interest in Novartis stock. N. Rosen reports receiving a commercial research grant from Chugai and is a consultant/advisory board member for AstraZeneca, Beigene, Chugai, and Novartis. No potential conflicts of interest were disclosed by the other authors.

Conception and design: Y. Gao, D. Stuart, Z. Yao, N. Rosen

Development of methodology: Y. Gao, K. Muniz, M. Drosten, G. Caponigro, Z. Yao

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y. Gao, M.T. Chang, N. Na, M. Drosten, M. Barbacid, D.B. Solit, O. Abdel-Wahab, Z. Yao

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y. Gao, M.T. Chang, D. McKay, N. Na, B. Zhou, R. Yaeger, D. Stuart, B.S. Taylor, Z. Yao, N. Rosen

Writing, review, and/or revision of the manuscript: Y. Gao, R. Yaeger, G. Caponigro, D. Stuart, H. Moebitz, D.B. Solit, O. Abdel-Wahab, B.S. Taylor, Z. Yao, N. Rosen

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Y. Gao, N. Na, N.M. Torres, K. Muniz, H. Moebitz, D.B. Solit, Z. Yao, N. Rosen

Study supervision: Z. Yao, N. Rosen

We are grateful to Piro Lito and Vanessa S. Rodrik-Outmezguine for useful discussions. This research is supported by grants from the NIH (R01 CA169351, P01 CA129243, R35 CA21008, and U54 OD202355), William H. Goodwin and Alice Goodwin, the Commonwealth Foundation for Cancer Research, and The Center for Experimental Therapeutics at Memorial Sloan Kettering Cancer Center. This work is also supported by the Functional Genomics Initiative of MSKCC (to O. Abdel-Wahab) and NIH/NCI Cancer Center Support Grant P30 CA008748.