Live-cell target engagement of allosteric MEKi on MEK–RAF/KSR–14–3–3 complexes

Abstract

The RAS–mitogen-activated protein kinase (MAPK) pathway includes KSR, RAF, MEK and the phospho-regulatory sensor 14–3–3. Specific assemblies among these components drive various diseases and likely dictate efficacy for numerous targeted therapies, including allosteric MEK inhibitors (MEKi). However, directly measuring drug interactions on physiological RAS–MAPK complexes in live cells has been inherently challenging to query and therefore remains poorly understood. Here we present a series of NanoBRET-based assays to quantify direct target engagement of MEKi on MEK1 and higher-order MEK1-bound complexes with ARAF, BRAF, CRAF, KSR1 and KSR2 in the presence and absence of 14–3–3 in living cells. We find distinct MEKi preferences among these complexes that can be compiled to generate inhibitor binding profiles. Further, these assays can report on the influence of the pathogenic BRAF-V600E mutant on MEKi binding. Taken together, these approaches can be used as a platform to screen for compounds intended to target specific complexes in the RAS–MAPK cascade.

The RAS–mitogen-activated protein kinase (MAPK) pathway is a highly conserved signal transduction pathway important for a variety of biological processes, including growth, differentiation and survival^1–5. The RAS–MAPK pathway has been an intense area of investigation for drug development, as mutations or alterations in the pathway are common in a variety of cancers and inherited disorders⁶. Although many excellent drugs have been developed against specific components of the pathway, such compounds often display limited efficacy or marked susceptibility to resistance^2,7–14. The confounding issue of a narrow therapeutic index in particular is thought to relate to the overlapping function of RAS–MAPK components in disease and normal biology^4,6,15.

Recent structural studies have revealed new insights into the regulation of RAS effector complexes in the MAPK cascade, namely through components BRAF, KSR1/KSR2, MEK1 and 14–3–3 proteins¹. For example, structures of both autoinhibited and active BRAF–MEK1–14–3–3 complexes¹⁶ suggest a model of RAF kinase regulation in which 14–3–3 binds phosphorylated S365 (pS365) and pS729 sites on BRAF to cradle a quiescent state of the BRAF–MEK complex in which MEK is held in an inactive conformation. In the active state, a 14–3–3 dimer binds each C-terminal pS729 site of a BRAF dimer to enable binding to active GTP-bound RAS¹⁷. Distinct structures resolved density for the RAS-binding domain (RBD) of BRAF in the monomeric assembly, indicating its role as a competitor of RAS binding to relieve BRAF autoinhibition¹⁸.

An additional structural state of an asymmetric BRAF–14–3–3 dimer complex was identified that may be relevant to the well-documented paradox whereby inhibitors ‘transactivate’ RAF dimers^19,20. Moreover, it has been demonstrated that 14–3–3 enforces the dimeric arrangement of BRAF, thereby relieving the negative regulatory effect of ATP on BRAF monomers^21,22. More recently, structures and biochemical data have further revealed distinctions in binding of MAPK inhibitors to purified RAF–MEK complexes^23,24. We have solved structures of MEK1 in a similar inactive configuration as seen bound to BRAF and furthermore in complex with several allosteric MEK inhibitors (MEKi). In particular, structures bound to trametinib revealed the ability of compounds to bridge the interface of KSR–MEK1 complexes²⁵.

Given the structural plasticity of MEK, we hypothesized that allosteric MEKi could demonstrate unique pharmacological properties on isolated and bound configurations favored by various KSR and RAF homologs. We recently demonstrated marked differences in potency and residence time for a series of clinical MEKi in free and KSR1-bound forms of the enzyme MEK1 (ref. 25). In contrast to more common biochemical methods with purified enzymes, live-cell target engagement assays capture the dynamic nature of signal transduction pathways in living cells and thereby potential changes in pharmacology due to alterations in target binding pockets that can occur within physiological complexes^26–28. Broadly, live-cell target engagement holds the promise of a complementary approach to more conventional cell line screens and structure-based drug design. Such approaches may be particularly well suited for developing new medicines against specific subcomplexes of pan-essential targets (for example, see ref. 29), where exploiting disease-associated functions can be restricted due to on-target toxicities and apparent limitations to the therapeutic index.

Here, we extend NanoBRET to quantify the binding preferences of several allosteric MEKi on various MEK1–RAF/KSR and MEK1–RAF/KSR–14–3–3 complexes. Our methods reveal the ability to query apparent affinity and selectivity of allosteric MEKi toward specific wild-type (WT) subcomplexes that are favored through coexpression including MEK1–ARAF, MEK1–BRAF, MEK1–CRAF, MEK1–KSR1 and MEK1–KSR2 in the presence and absence of 14–3–3. We additionally explore the role of the highly frequent BRAF-V600E mutant on MEKi binding to MEK1 in the presence and absence of 14–3–3. These studies provide a platform to understand the binding of allosteric MEKi that are currently used therapeutically or that are in advanced clinical development, providing insights into the drug–receptor interactions that occur in living cells on higher-order assemblies of MAPK signaling complexes.

Results

NanoBRET assays of MEKi on MEK–RAF and MEK–KSR complexes

As previously demonstrated, we used a trametinib-based tracer, Tram-bo (1; Fig. 1a), to measure steady-state half-maximal inhibitory concentration (IC₅₀) and residence time of MEKi on a MEK1 construct expressed alone or in complex with mouse KSR1 (ref. 25). Tram-bo is a bifunctional molecule that includes a targeting moiety (trametinib) and Bodipy dye connected by a short PEG-based linker. When added to HEK293T cells transfected with a C-terminal MEK1–NanoLuciferase (MEK1–NL) fusion, Tram-bo binds to MEK1 allosterically and undergoes bioluminescence resonance energy transfer (BRET) through proximity to the NL. In this assay, light is produced via the donor NL substrate furimazine to create BRET with the acceptor Bodipy in Tram-bo (Fig. 1b). The resulting BRET signal reports on all states of MEK1, presumably including the free form of the enzyme and complexes that occur with endogenous binding partners. In competition assays, the BRET signal can be blocked through inhibition of Tram-bo–acceptor binding, which can be observed in a dose-dependent manner using an inhibitor of interest to generate steady-state apparent IC₅₀ values from live cells.

Fig. 1 |. Design of a NanoBRET assay to measure changes of MEKi binding in the presence of KSR and RAF binding partners.

a, Structure of the Tram-bo tracer. b, Schematic of the NanoBRET assay used to measure MEKi binding. The protein denoted by the dotted line indicates MEK binding proteins. c, Cocrystal structure of Tram-bo bound to a KSR2–MEK1–AMP-PNP complex; note that the linker and Bodipy moieties extend into the solvent through a channel at the KSR–MEK1 interface.

To better understand the binding mechanism of Tram-bo, we solved a cocrystal structure of this tracer bound to the KSR2–MEK1–AMP-PNP complex to 3.2 Å (Fig. 1c, Extended Data Fig. 1 and Supplementary Table 1). The structure revealed that the trametinib portion of Tram-bo binds in the allosteric site on MEK1 that we previously observed²⁵. However, the linker (with density well resolved) and Bodipy moiety (modeled as unbound in solvent) extended away from the interfacial region of the KSR2–MEK1 complex, with the linker motif extended toward the lower C-lobe of MEK1. In particular, the linker amide oxygen formed a close contact with R189 of MEK1 such that the linker-Bodipy moiety avoids any contacts or clashes with KSR2. The structure supported our original conjecture that Tram-bo may bind within a ternary complex of KSR-bound MEK. Further, based on this structure, and specifically the observed exit vector for the linker-Bodipy motif, we reasoned that Tram-bo may be able to bind within the context of several MEK1-associated complexes, including bound to the catalytic RAFs (ARAF, CRAF and BRAF). To test this hypothesis, we transiently transfected MEK1–NL and coexpressed an excess of each unlabeled binding partner (that is, full-length ARAF/BRAF/CRAF or KSR1/KSR2) to favor specific MEK1–RAF/KSR complexes.

We reasoned that a change in BRET_max (half-maximal effective concentration (EC₅₀) curve height) or EC₅₀ of Tram-bo under coexpression conditions would be indicative of ternary (for example, MEK1–Tram-bo–KSR or MEK1–Tram-bo–RAF) versus binary (for example, MEK1–Tram-bo) complexes (Fig. 1b). The following workflow was used to test the hypothesis of ternary complex formation in living cells: (1) measure Tram-bo binding (EC₅₀ and BRET_max) to each MEK1–NL–RAF/KSR ternary complex at saturation (that is, with each RAF/KSR binding partner expressed in vast excess of MEK1–NL) and compare to assays where MEK1–NL was expressed alone, (2) conduct identical titrations but using RAF–MEK1 and KSR–MEK1 interface mutations as controls for ternary complex formation, (3) move the donor signal source (that is, NL) onto the RAF and KSR partners and coexpress with unlabeled MEK1 such that BRET would be generated only if RAF or KSR come into sufficiently close proximity to form a ternary complex with MEK1-bound Tram-bo and (4) conduct orthogonal assays (for example, coimmunoprecipitation (co-IP) assays) to independently verify ternary complex formation in the presence of Tram-bo.

We first measured the dose-dependent BRET signal buildup of Tram-bo for each binding partner complex to determine a concentration of Tram-bo to use in MEKi steady-state competition assays and found in each case that EC₅₀ values transitioned from free MEK to a saturated complex (Supplementary Fig. 1a–c). We chose a DNA ratio of 0.005 μg ml⁻¹:1 MEK1–NL:RAF/KSR to saturate available MEK1–NL with each binding partner while achieving a strong donor signal (that is, >10⁵ relative light units). Tram-bo appeared to interact most strongly with MEK1–NL alone and the MEK1–NL coexpressed with KSR2 (EC₅₀ = 5.6 ± 0.6 nM for MEK1–NL and 35.2 ± 3.5 nM for MEK1–KSR2) and ~10- to 15-fold weaker for the MEK1–ARAF (EC₅₀ = 167 ± 27 nM), MEK1–BRAF(EC₅₀ = 178 ± 33 nM), MEK1–CRAF(EC₅₀ = 193 ± 28 nM) and MEK1–KSR1(EC₅₀ = 112 ± 1 nM) complexes (Fig. 2a and Supplementary Fig. 1d).

Fig. 2 |. Calibration of the NanoBRET assay used to measure MEKi binding on a ternary MEK1–RAF/KSR–Tram-bo complex.

a, Tram-bo buildup curves on MEK1–NL and MEK1–NL–RAF/KSR complexes show changes in BRET signal; [trametinib], concentration of trametinib. b, A critical tryptophan residue at the interface between MEK1 and KSR2 is shown, and analogous residues for each complex are highlighted below. c, Assay specificity for complexes based on EC₅₀ changes comparing WT MEK1–NL expression alone to coexpression with WT (top) and mutant controls (bottom; ARAF-F520E, BRAF-F667E, CRAF-F559E, KSR1-W831D and KSR2-W884D). Data were analyzed by a one-way analysis of variance (MEK1–NL versus MEK1–NL + ARAF adjusted P = 0.0006; MEK1–NL versus MEK1–NL + BRAF adjusted P = 0.0003; MEK1–NL versus MEK1–NL + CRAF adjusted P = 0.0002; MEK1–NL versus MEK1–NL + KSR1 adjusted P = 0.013); *P < 0.05; ***P < 0.001; NS, not significant. d, Tram-bo EC₅₀ values generated from an assay with NL fused to BRAF or KSR1 indicate a ternary complex. e, BRET signals decrease with an αG mutation on BRAF–NL (F667E) and KSR1–NL (W831D). All experiments were performed in biological triplicate (that is, n = 3) with at least one technical replicate. Error bars represent s.e.m. for all three replicates.

To test if the changes in Tram-bo EC₅₀ after coexpression of RAFs or KSRs with MEK1–NL were due to the formation of helix αG-mediated complex formation, as observed in previous structures of KSR1–MEK1 (Protein Data Bank (PDB) ID 7JUZ), KSR2–MEK1 (PDB ID 7JUR) and BRAF–MEK1 (PDB ID 4MNE), we generated point mutations at a critical and positionally equivalent hydrophobic residue within the RAF–MEK1 and KSR–MEK1 interface, including ARAF-F520E, BRAF-F667E, CRAF-F559E, KSR1-W831D and KSR2-W884D (Fig. 2b). Notably, when the mutants were coexpressed with MEK1–NL, we observed a reversal in the relative Tram-bo EC₅₀ changes compared to coexpression with the WT versions of RAF/KSRl (Fig. 2c and Supplementary Fig. 1e). In a similar but reciprocal experiment, we mutated F311 to arginine on MEK1 to favor the free state; mutation of this residue has been previously shown to disrupt MEK1-bound BRAF, CRAF and KSR1 complexes^30,31. Indeed, after coexpression with WT versions of ARAF, BRAF, CRAF, KSR1 and KSR2 with MEK1-F311R-NL, we observed no substantial change in Tram-bo EC₅₀ (Supplementary Fig. 1f), again supporting that BRET_max and relative EC₅₀ value differences act as sensitive measures of binary and ternary complex formation. These results suggest that the changes in Tram-bo EC₅₀ after our coexpression conditions are due to complex formation via direct reciprocal helix αG-mediated interactions, as observed in X-ray and cryo-electron microscopy structures of KSR–MEK1 and RAF–MEK1 (refs. 16,25,31).

Next, to further interrogate binding requirements, we generated mutations of key phosphorylation sites S218 and S222 in MEK1–NL to determine if the changes in Tram-bo EC₅₀ could also be due to MEK1 phosphorylation, which would potentially alter the binding of Tram-bo or free ligand. Indeed, using in vitro assays, trametinib has been shown to prefer the inactive and unphosphorylated state of MEK1 (ref. 32). Tram-bo EC₅₀ values of both WT and S218A/S222A MEK1–NL were similarly elevated after coexpression with WT and untagged BRAF (Supplementary Fig. 1g). The binding affinity of Tram-bo was comparable between WT MEK1–NL (EC₅₀ = 9.9 ± 0.4 nM) and MEK1–NL-S218A/S222A (EC₅₀ = 22 ± 2 nM). Since the S218A/S222A mutant was effectively silent in this assay, whereas point mutations at helix αG of KSR/RAF strongly reduced changes in Tram-bo EC₅₀ values, we conclude that the changes in Tram-bo EC₅₀ after coexpression of MEK1–NL with KSR or RAF occur more significantly through direct interactions between binding partners (for example, MEK1–NL–KSR1 and MEK1–NL–BRAF) and concomitant changes in the MEKi pocket resulting from complexation²⁵ than through alterations in the phosphorylation status of MEK1.

To further evaluate if changes arose from competition on a ternary complex (that is, MEK1–Tram-bo–KSR/RAF) in living cells, we aimed to measure Tram-bo binding but with the NL donor moved onto the C terminus of BRAF or KSR1 as opposed to on MEK1. Both BRAF–NL and KSR1–NL were coexpressed with a tenfold excess of a plasmid encoding full-length unlabeled MEK1 in the presence of Tram-bo. In this assay configuration, we were able to measure similar Tram-bo binding affinities as in the MEK1–NL coexpression assay, supporting that either NL-tagged MEK1 coexpressed with KSR/RAF or, alternatively, NL-tagged KSR1 or BRAF coexpressed with MEK1 treated with Tram-bo results in similar ternary complex formation (Fig. 2d,e and Supplementary Fig. 2).

We next validated the specificity of the ternary complex signals using BRAF and KSR1 αG-helix mutants. Using a constant concentration of Tram-bo, we coexpressed either WT or F667E/W831D mutants with MEK1 and performed a trametinib dose–response assay. We observed a large decrease in signal using the αG-helix mutants relative to the WT versions (Fig. 2e). Although these data support that the formation of this signal for BRAF–MEK1–Tram-bo and KSR–MEK1–Tram-bo are through the helix αG interface, we cannot rule out other more complex models where the proximity of Tram-bo on adjacent complexes is dependent on intact helix αG-mediated binding, for example, similar to a cooperative binding mechanism, as has been eluded to previously³⁰.

In addition to NanoBRET measurements, we performed co-IP experiments where we coexpressed MEK1–NL with FLAG-tagged BRAF or KSR1 and pulled down either MEK1 or FLAG-tagged BRAF/KSR1 in the presence of Tram-bo, the clinical MEKi trametinib and CH5126766 and the new MEKi trametiglue²⁵. In both sets of co-IP experiments, we observed that Tram-bo appeared to reduce the interaction between MEK1 and BRAF but did not completely ablate it (Extended Data Fig. 2a,b). By contrast, the interaction between MEK1 and KSR1 was largely unaffected (Extended Data Fig. 2a,b). Trametiglue and CH5126766 stabilized both MEK1–KSR and MEK1–BRAF complexes, whereas trametinib specifically reduced MEK1–BRAF, as we²⁵ and others³³ have previously shown. Taken together, both RAF and KSR may form sufficiently stable interactions with MEK in the presence of Tram-bo to enable BRET-based assays of specific ternary complexes.

Cellular target engagement of MEKi on free and bound MEK1

To further understand the influence of MEK1–RAF/KSR interactions on MEKi binding, we compared structure–activity relationships among a series of clinical MEKi against free MEK versus KSR- or RAF-bound complexes. Having measured EC₅₀ changes associated with each complex, we then characterized the apparent steady-state IC₅₀ values for the clinical MEKi trametinib, cobimetinib and CH5126766 and the new MEKi trametiglue²⁵ using a constant concentration of 1 μM Tram-bo with MEK–NL and each coexpressed complex (Fig. 3a). A 1 μM Tram-bo concentration represents the highest achievable concentration of tracer (that is below the solubility limit) to enable BRET with MEK1–NL and its complexes. After adjusting the apparent IC₅₀ values to K_i values using the Cheng–Prusoff equation (eq. (1) in Methods), we observed that MEKi prefer MEK1–RAF/KSR complexes (Fig. 3b). Generally, trametinib and trametiglue were more potent binders of all MEK1–RAF/KSR complexes than cobimetinib and CH5126766. Trametinib and trametiglue displayed preferences for MEK1–KSR1 and MEK1–KSR2 complexes relative to the others. Cobimetinib interacted more strongly on MEK1–NL and the MEK1–NL–KSR1 and MEK1–NL–KSR2 complexes than on the MEK1–NL–RAF complexes. CH5126766 showed the largest differences with strong binding to MEK1–ARAF, MEK1–BRAF and MEK1–CRAF complexes over MEK1–KSR1 and MEK1–KSR2 complexes and MEK1–NL alone (Fig. 3b and Supplementary Figs. 3 and 4). Changes in apparent IC₅₀ and thus K_i could be fully reversed using RAF/KSR interfacial mutants (ARAF-F520E, BRAF-F667E, CRAF-F559E, KSR1-W831D and KSR2-W884D; Extended Data Fig. 3b and Supplementary Fig. 4), confirming that the K_i differences of MEKi under coexpression conditions depend on helix αG-mediated complex formation. We also performed reciprocal assays in which NL was moved onto KSR or RAF and specifically queried BRAF–NL–MEK1–Tram-bo and KSR1–NL–MEK1–Tram-bo. Here, we found that the MEKi binding trends were near identical to what we observed using the MEK1–NL assays (Supplementary Fig. 5). Collectively, the differential MEKi profiles reveal distinct binding preferences on MEK1–RAF and MEK1–KSR complexes obtained from living cells.

Fig. 3 |. Characterization of MEKi binding on MEK1 and MEK1–RAF family complexes.

a, Structures of the MEKi trametinib, trametiglue, cobimetinib and CH5126766. b, Calculated K_i values of MEKi using a 0.005:1 μg ml⁻¹ MEK1–NL:RAF/KSR DNA ratio for all coexpressions. K_i values were calculated from the Cheng–Prusoff equation using mean IC₅₀ values (as shown in Supplementary Fig. 3), mean EC₅₀ values of Tram-bo (values from Supplementary Fig. 1) and a concentration of 1 μM Tram-bo. All data points represent individual biological replicates. Means and s.e.m. of six biological replicates are shown (n = 6). Error bars represent the s.e.m.

We next sought to compare MEKi K_i values and binding profiles using a distinct tracer. Using the MEKi PD0325901, we designed and synthesized a tracer named PD-bo (6; Extended Data Fig. 4a and Supplementary Information). This tracer consists of a PEG-3 spacer between PD0325901 and a Bodipy dye. Tracer buildup curves of PD-bo showed a ~45-fold weaker binding affinity to MEK1–NL than Tram-bo (253 nM versus 6 nM; Extended Data Fig. 4b and Supplementary Fig. 6a) and, further, several orders of magnitude weaker and highly non-reproducible binding for MEK1–RAF/KSR complexes. Moreover, as a result of its lowered affinity, apparent IC₅₀ values measured for MEKi on MEK1–NL were substantially lower for PD-bo than for Tram-bo but followed the same rank order as Tram-bo (Extended Data Fig. 4c and Supplementary Fig. 6b). This comparison suggests that the apparent IC₅₀ values are a direct function of tracer affinity as would be expected. Further, the poor binding affinity of PD-bo to MEK1–RAF/KSR complexes made it impossible to measure steady-state IC₅₀ values for the complexes. Thus, analogous MEKi-based tracers, such as Tram-bo and PD-bo, can be interchangeably used to profile compounds against MEK1–NL. However for MEK1–RAF/KSR complexes, a higher-affinity tracer with an appropriate exit vector from the interfacial space between binding partners, as is achieved with Tram-bo, is required.

Target engagement of MEKi in 14–3–3-bound complexes

We next sought to assay MEKi binding in the context of larger physiological complexes involving 14–3–3. Based on PDB 6Q0J and 6NYB in ref. 16, we reasoned that an NL placed on the C terminus of 14–3–3 would be within ~100 Å of Tram-bo bound to MEK1 within macromolecular MEK1–RAF/KSR–14–3–3 complexes (Fig. 4a,b). To test this, we transfected 14–3–3–NL into HEK293T cells treated with Tram-bo and performed a trametinib dose–response assay. Single expression of 14–3–3–NL was not sufficient to generate BRET above baseline or a trametinib dose response (Fig. 4c). However, coexpression of MEK1 and BRAF generated a strong BRET_max signal through 14–3–3–NL–Tram-bo above baseline (Fig. 4c), which could also be outcompeted using free trametinib to generate a dose–response curve and K_i value of 1.19 ± 0.09 nM (Fig. 4d). Mutation of F667E on the αG-helix of BRAF largely decreased the overall BRET signal, suggesting that Tram-bo-bound MEK1 engages RAF–14–3–3 in living cells in a manner observed in cryo-electron microscopy and crystal structures (Fig. 4c and Extended Data Fig. 5). Coexpression using other RAF family kinases, including ARAF, CRAF, KSR1 and KSR2, also yielded dose responses that gave similar K_i values as those in the MEK1–NL assay configuration (Fig. 4d and Supplementary Fig. 7). These results highlight the utility of the 14–3–3–NL donor with Tram-bo-based acceptor combinations to measure binding in the context of MEK1–RAF–14–3–3 and MEK1–KSR–14–3–3 complexes.

Fig. 4 |. Development of a NanoBRET assay to measure MEKi binding within larger complexes involving 14–3–3.

a, Structures of autoinhibited (PDB 6NYB) and active (PDB 6Q0J) MEK1–BRAF–14–3–3 signaling complexes. Stars indicate the position of Tram-bo on MEK1 (orange) and the C terminus of 14–3–3 where NL is attached (cyan). Dashed lines indicate the shortest potential interactions between Tram-bo and NL. b, Assay design to measure MEKi binding using C-terminal placement of NL on 14–3–3. The assay can report on the autoinhibited and active arrangements; h14–3–3-ζδ, human 14–3–3-ζδ; CRD, cysteine-rich domain. c, BRET signal between 14–3–3–NL and Tram-bo requires coexpression of MEK1 and BRAF. Each curve represents the average of four biological replicates from Extended Data Fig. 5. d, Characterization of MEKi on MEK1–RAF–14–3–3–Tram-bo and MEK1–KSR–14–3–3–Tram-bo complexes. K_i values were calculated using the Cheng–Prusoff equation and mean apparent IC₅₀ values (as shown in Supplementary Fig. 7), mean EC₅₀ values of Tram-bo (values from Supplementary Fig. 7) and a concentration of 1 μM Tram-bo. e, RAF dimer-influencing mutations BRAF-R509H (cobimetinib), CRAF-R401H (trametiglue) and KSR1-R665H (trametinib, trametiglue and cobimetinib) significantly and differentially influence MEKi binding. The asterisk indicates the mutation site on the cartoon model (trametinib: 14–3–3–NL + MEK1 + KSR1 WT versus 14–3–3–NL + MEK1 + KSR1-R665H P = 0.0343, t = 3.156, d = 4, two-tailed unpaired t-test; trametiglue: 14–3–3–NL + MEK1 + CRAF WT versus 14–3–3–NL + MEK1 + CRAF-R401H, P = 0.019, t = 3.803, d = 4, two-tailed unpaired t-test; 14–3–3–NL + MEK1 + KSR1 WT versus 14–3–3–NL + MEK1 + KSR1-R665H P = 0.0442, t = 2.899, d = 4, two-tailed unpaired t-test; cobimetinib: 14–3–3–NL + MEK1 + BRAF WT versus 14–3–3–NL + MEK1 + BRAF-R509H, P = 0.0006, U = 0, two-sided Mann–Whitney U-test; MEK1 + KSR1 WT versus 14–3–3–NL + MEK1 + KSR1-R665H P = 0.02, t = 3.748, d = 4, two-tailed unpaired t-test); *P < 0.05; ***P < 0.001. Note that a lack of an asterisk for any comparison implies that there are no statistical differences. All data points for d and e represent individual biological replicates. Mean and s.e.m. values of at least three biological replicates are shown (n = 3 or n = 7). Each biological replicate has at least one technical replicate.

In this assay configuration, both monomeric and dimeric versions of RAF or KSR could be compatible with long-range 14–3–3–NL–Tram-bo interactions according to models that we generated based on recently determined structures^16,18. We, therefore, sought to test the influence of RAF/KSR dimerization on MEKi binding using our 14–3–3–NL–Tram-bo assay configuration. Residue R509 in BRAF, for example, is critical for the side-to-side dimer arrangement^34,35, and this analogous residue in ARAF, CRAF, KSR1 and KSR2 when mutated to histidine has also been shown to play a role in dimerization (ARAF-R362H³⁶, BRAF-R509H, CRAF-R401H³⁷, KSR1-R665H³⁰ and KSR2-R718H³⁸). Characterization of MEKi on these mutants revealed a small but significant decrease in potency between KSR1 and KSR1-R665H for trametinib, a decrease in potency between CRAF and CRAF-R401H and KSR1 and KSR1-R665H for trametiglue, an increase in potency for BRAF and BRAF-R509H and a decrease in potency for KSR1 and KSR1-R665H for cobimetinib (Fig. 4e and Supplementary Fig. 7). Strikingly, the sensitivity of MEKi to the arginine-to-histidine mutation was paralog and inhibitor specific; for example, trametinib, trametiglue and CH5126766 were insensitive to the arginine-to-histidine mutation in the context of the MEK1–BRAF–14–3–3 assay, whereas cobimetinib demonstrated a significant increase in potency, suggesting that cobimetinib has a unique preference for binding toward BRAF monomers over dimers. Moreover, structure–activity relationships were highly sensitive. For example, trametinib differs from trametiglue by virtue of an acetamide versus sulfamide moiety. In ARAF and CRAF assays, trametiglue demonstrated a significant preference for RAF dimers or monomers, whereas trametinib did not show any specificity toward either state. These results imply that dimerization of BRAF, CRAF and KSR1 plays a measurable role in the potency and binding of specific MEKi, such as trametinib, trametiglue and cobimetinib in our assays.

These results prompted us to evaluate the contribution of endogenous 14–3–3 in our MEK1–NL coexpression assays. Because each RAF and KSR is thought to form key interactions with 14–3–3 via analogous C-terminal phospho-regulatory residues (ARAF-S582, BRAF-S729, CRAF-S621, KSR1-S888 and KSR2-S939), we envisioned that those individual requirements for 14–3–3 binding at these sites could be interrogated by measuring potential changes in Tram-bo and MEKi binding after mutation of these residues to alanine. We tested the 14–3–3 C-terminal site mutants first in the context of our 14–3–3–NL assay and observed a decrease in BRET signal in all cases based on trametinib dose responses, indicating the importance of functional interactions between RAF/KSR and 14–3–3 (Extended Data Fig. 6). Tram-bo binding affinity was next measured for each of the mutants using the MEK1–NL assay. Interestingly, we observed a significant change in Tram-bo binding for RAF mutants, but not for KSR mutants, relative to the WT proteins, implying that endogenous 14–3–3 plays a specific role in complexes containing RAFs more so than KSRs (Supplementary Fig. 8a). Moreover, we noticed small but significant changes in the calculated K_i values for specific inhibitor (that is, trametinib, trametiglue, cobimetinib or CH5126766) and various KSR- or RAF-based paralogs (Supplementary Fig. 8b–e). Taken together, these data suggest that 14–3–3 participates in our MEK–NL assays, and the impact on specific MEKi binding profiles is small and can be assessed using C-terminal phospho-regulatory site mutants in RAFs and KSRs.

Influence of the BRAF-V600E mutant on MEKi binding

We further sought to test if we could gain insight into how the highly frequent BRAF-V600E mutant influences MEKi binding preferences in our MEK1–NL and 14–3–3–NL assays (Fig. 5a). The BRAF-V600E mutant has high intrinsic activity, is not RAS or dimer dependent and can signal as a monomer^34,39–41. Using our MEK1–NL coexpression assay configuration, we observed that BRAF-V600E showed similar Tram-bo binding affinity to WT BRAF but displayed weakened binding affinity (calculated K_i) for all MEKi, with the largest decrease in potency for CH5126766 (~100-fold; Fig. 5b, Extended Data Fig. 7 and Supplementary Fig. 9a). BRET measurements and K_i values could not be generated for BRAF-V600E using the 14–3–3–NL assay configuration, as coexpression of the V600E variant eliminated any detectable BRET signal between 14–3–3–NL and Tram-bo. This result suggests that the BRAF-V600E mutant disfavors assembly of the MEK1–BRAF(V600E)–14–3–3–NL–Tram-bo complex (Fig. 5c and Supplementary Fig. 10).

Fig. 5 |. Characterization of MEKi binding in the presence of BRAF-V600E.

a, Location of V600E on a MEK1–BRAF complex, as highlighted by a red surface. b, Characterization of MEKi within MEK1–BRAF complexes using the MEK1–NL–RAF–Tram-bo coexpression assay. Data were analyzed by one-way ANOVA (trametinib: MEK1–NL + BRAF WT versus MEK1–NL + BRAF-V600E adjusted P = 0.0013; trametiglue: MEK1–NL + BRAF WT versus MEK1–NL + BRAF-D576N adjusted P < 0.0001, MEK1–NL + BRAF WT versus MEK1–NL + BRAF-V600E adjusted P = 0.0279; cobimetinib: MEK1–NL + BRAF WT versus MEK1–NL + BRAF-V600E adjusted P < 0.0001). Note that each number (1–8) under each graph acts as an identifier for each condition; *P < 0.05; **P < 0.01; ****P < 0.0001. c, Characterization of MEKi within MEK1–BRAF complexes using the MEK1–RAF–14–3–3–NL coexpression assay. Note that each number (1–3) under each graph acts as an identifier for each condition, and NB refers to No BRET, meaning that BRET could not be observed for the indicated condition. K_i values were calculated from the Cheng–Prusoff equation using mean apparent IC₅₀ values (as shown in Supplementary Figs. 9 and 10), mean EC₅₀ values of Tram-bo (values from Supplementary Figs. 9 and 10) and a concentration of 1 μM Tram-bo. All data are shown as means and s.e.m., with each individual point representing one biological replicate (n = 3). Each biological replicate has at least one technical replicate.

The V600E allele could decrease MEKi K_i values through several mechanisms, including alterations in RAF side-to-side dimerization, alterations in BRAF-V600E kinase activity that would influence MEK1–NL phosphorylation and loss of MEK–RAF complexation with 14–3–3. We therefore conducted several experiments including (1) assessing the impact of altering back-to-back kinase dimerization using the R509H mutant and (2) testing the impact of a BRAF kinase-inactivating mutant D576N.

To test the impact of dimerization, we coexpressed a R509H/V600E double mutant with MEK1–NL and calculated MEKi K_i values. In this context, we found that the apparent IC₅₀ and K_i values for all MEKi shifted back to values that were similar to WT BRAF, highlighting that disrupting BRAF dimerization can reverse the impact of the V600E mutant on MEKi binding (Fig. 5b (condition 1 versus 5), Extended Data Fig. 7 and Supplementary Fig. 9b,c). The R509H mutant by itself imparted no changes on apparent MEKi IC₅₀ and K_i values relative to WT BRAF coexpressed with MEK1–NL (Fig. 5b (condition 1 versus 3), Extended Data Fig. 7 and Supplementary Fig. 9b,c).

We next tested the requirement for catalytic activity in the context of the BRAF-V600E mutant by creating the kinase-dead D576N/V600E double mutant. D576 serves as the catalytic base in BRAF and is critical to lower the activation energy of substrate phosphorylation^42,43. Much like R509H, the D576N mutant was also able to rescue MEKi binding, with the effects of the V600E mutant reverted to similar apparent IC₅₀ and K_i values as MEK–NL + WT BRAF for all MEKi (Fig. 5b (see condition 5 versus 6 relative to condition 1 versus 2), Extended Data Fig. 7 and Supplementary Fig. 9b,c). We interpret these data to suggest that the BRAF-V600E mutant creates a population of phosphorylated MEK that does not readily form a MEK1–BRAF complex, thereby disfavoring Tram-bo binding and further greatly diminishing formation of MEKi–MEK1–BRAF-V600E ternary complexes (Fig. 5b, Extended Data Fig. 7 and Supplementary Fig. 9b,c). In addition, the D576N/V600E double mutant was able to restore the BRET signal in the 14–3–3–NL configuration, generating similar apparent IC₅₀ and K_i values as observed following coexpression of 14–3–3–NL with WT BRAF (Fig. 5c (see condition 2 versus 3) and Supplementary Fig. 10). Thus, these experiments collectively reveal that BRAF-V600E negatively impacts the interaction between MEKi and MEK1, mostly likely through phosphorylation on the S218/S222 sites, resulting in reduced MEK1–RAF and/or MEK1–KSR complex formation.

Discussion

Here, we developed assays for the comparison of binary and ternary complexes in live cells and reveal differences in MEKi binding preferences among various KSR and RAF paralogs in complex with MEK1. We further extended the approach to query MEKi binding using 14–3–3 as part of our reporter system; these constructs ensure that observed binding signals occur through higher-order assemblies minimally including MEK1, RAF/KSR, 14–3–3 and Tram-bo in live cells.

Our first set of assays favor complexes through coexpression of various MEK1–RAF and MEK1–KSR complexes that are mediated through reciprocal helix αG interactions, which form ternary binding interactions as measured by relative changes in both Tram-bo EC₅₀ and the corresponding calculated K_i values of MEKi (for example, as shown in Fig. 2a,b and Supplementary Figs. 1 and 4). Based on these experiments, we conclude that changes in Tram-bo EC₅₀ and MEKi IC₅₀ values that occur from coexpression of WT KSR or RAF with MEK1–NL relative to MEK1–NL alone reflect a change in ternary complex formation and thereby affinity of Tram-bo or free MEKi to the KSR- and RAF-bound states of MEK. Notably, in all cases, any differences that we observed between MEKi through coexpression experiments could be fully abolished using helix αG mutants on RAF or KSR, demonstrating that MEK1–KSR and MEK1–RAF form the key subcomplexes that we are assaying in living cells.

We first determined a working concentration of Tram-bo below its solubility limit of ~ 2 μM that would produce high BRET signal on all MEK1–NL and MEK1–RAF/KSR complexes and generated subsequent steady-state IC₅₀ values through competition experiments. The K_i values that we calculated for trametinib, trametiglue, cobimetinib and CH5126766 revealed that MEKi can have binding preferences for certain complexes (Fig. 3b), which could relate to the established differences in mechanism and efficiency of these drugs^25,33,44. The MEKi trends and preferences for KSR or RAF homologs were also similar using our 14–3–3–NL reporter assays. Overall, our 14–3–3–NL assay provides a screenable platform to discover MEKi that are influenced by 14–3–3-mediated long-range allosteric interactions.

Further, we show that the pathological BRAF mutation V600E negatively affects the formation of MEK1–BRAF–14–3–3 complexes, and the binding of MEKi to MEK1–BRAF complexes relative to WT BRAF as a result of increased BRAF kinase activity and likely reduced inactive state complex formation between MEK1 and BRAF. This result is counterintuitive given that the BRAF-V600E mutation predicts sensitivity to MEKi^45,46. Our data imply that the sensitivity of BRAF-V600E cell lines to MEKi could result predominantly from binding to non-BRAF-V600E–MEK1 complexes in living cells, such as isolated MEK or WT KSR-bound MEK. Consistent with this idea, loss of KSR1 in a CRISPR screen using a BRAF-V600E cell line (SK-MEL-239) provided marked resistance to a clinical MEKi-based combination (trametinib + dabrafenib)^25,47.

We previously demonstrated decreased potency of CH5126766 relative to trametinib and trametiglue in clonogenic assays of RAS-driven (HCT116 and A549) and BRAF-V600E-driven (SKMEL239 and A375) cell lines²⁵. These data are consistent with the structure–activity relationships and profiles obtained here (Fig. 3). For example, poor target engagement of CH5216766 on MEK1 and/or MEK1–KSR1/KSR2 complexes could explain the activity of this analog on RAS- and RAF-mutant cell lines. Overall, our Tram-bo-based reporter systems suggest unique pharmacological properties of allosteric MEKi on distinct complexes. The approaches and design for profiling clinical drugs could (1) enable specific applications of various drugs to inhibit unique states of MEK1 and (2) define new design parameters and binding pockets for targeting MEK1–RAF/KSR–14–3–3.

Binding interactions in the RAS–MAPK pathway are dependent on the ordered and cooperative assembly of specific subcomplexes. For example, the formation of MEK–KSR complexes has been suggested to serve as a primer for the formation of KSR–RAF dimers that enable higher-order RAS–RAF–KSR–MEK supercomplexes³⁰. Moreover, the binding of RAS to RAF can serve to nucleate RAF dimers and may enable the formation of large supramolecular assemblies (reviewed in ref. 1). Additionally, 14–3–3 dimers can regulate the monomer-to-dimer transition of RAF kinases through interactions that are modulated at various phospho-regulatory sites; for example, pS365 in BRAF is dephosphorylated to favor release of 14–3–3 from the kinase domain dimer interface, thereby allowing for BRAF dimerization and phosphorylation of MEK (reviewed in ref. 5). These examples highlight the complexity of allostery and cooperativity in the regulation of RAS–MAPK signaling through various subcomplexes and binding interactions; our data highlight the potential for these interactions to also impact MEKi binding specificity and profiles in living cells.

Online content

Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41589-023-01454-8.

Methods

Expression and purification of KSR2–MEK1

A heterotetrameric complex of KSR2–MEK1 was expressed and purified as previously described²⁵.

Crystallization and structure determination of the Tram-bo-bound KSR2–MEK1 complex

Crystallization and structure determination were performed as previously described²⁵. Crystals were soaked using a concentration of 1 mM Tram-bo. Data were collected at the National Synchrotron Light Source II (17-ID-1-AMX, wavelength 0.92 Å, −196 °C), a US Department of Energy Office of Science User Facility operated for the Department of Energy Office of Science by Brookhaven National Laboratory under contract number DE-AC02–98CH10886. Following refinement using PHENIX, Ramachandran outliers were 0%.

Immunoprecipitation of endogenous MEK1 or FLAG-tagged BRAF/KSR1 to evaluate the effect of different MEKi on MAPK complex assembly

HEK293T cells (350,000) were seeded in a six-well, flat-bottom tissue culture-treated plate. After 24 h, cells were transfected with either FLAG-tagged KSR1 or FLAG-tagged BRAF constructs at a concentration of 1 μg ml^–1 in combination with 9 μg ml^–1 carrier DNA for the FLAG pulldowns or 0.5 μg ml^–1 in combination with 9 μg ml^–1 carrier DNA for the endogenous MEK1 pulldown. Fugene HD was added to both conditions. Twenty-four hours after transfection, cells were treated with trametinib, Tram-bo, trametiglue or CH5126766, each at 1 μM, at a final DMSO concentration of 0.1% in the medium. After 24 h, cells were washed with sterile 1× PBS and collected in Pierce IP lysis buffer (Thermo Fisher, 87787) supplemented with HALT phosphatase and protease inhibitor cocktail and quantified and normalized using an RC DC Protein Assay kit (Bio-Rad, 5000122). Dynabeads M-280 sheep anti-rabbit IgG (Invitrogen, 11203D) were preconjugated to 5 μg per sample of anti-MEK1 (EMD Millipore, 07–641), anti-FLAG (Cell Signaling Technologies, 14793) or normal rabbit IgG antibody (Cell Signaling Technologies, 2729) by rotating for 1 h at room temperature. Next, the conjugated mixtures were washed thoroughly with PBS (0.1% Tween 20) three times using a magnetic rack. For endogenous MEK1 IP, 100 μg of total cell lysate was mixed with preimmobilized anti-MEK1 conjugated to anti-rabbit magnetic Dynabeads. For FLAG IP, 100 μg of total cell lysate was mixed with preimmobilized anti-FLAG conjugated to anti-rabbit magnetic Dynabeads. As a control for both conditions, 100 μg of total cell lysate was mixed with preimmobilized normal rabbit IgG conjugated to anti-rabbit Dynabeads (5 μg per 100 μg of cell lysate). Samples were incubated for 1 h at 4 °C on a rotator and were subsequently washed three times with PBS (0.1% Tween 20). Protein lysates were eluted off beads by adding 70 μl of 2× Laemmli buffer to each sample and heating to 80 °C for 5 min. Eluted proteins were analyzed via western blotting. The following primary antibodies were used for western blotting: anti-BRAF (Santa Cruz Biotechnology, sc-5284; 1:1,000), anti-MEK1 (Cell Signaling Technologies, 2352; 1:1,000) and anti-FLAG (Cell Signaling Technologies, 8146; 1:1,000). The secondary antibody used was anti-mouse horseradish peroxidase (Cell Signaling Technology, 7076S; 1:10,000)

Steady-state IC₅₀ measurements using Tram-bo and PD-bo

Apparent IC₅₀ values for trametinib, trametiglue, cobimetinib and CH5126766 were measured according to the standard protocol provided by Promega for the in-cell kinase assay with minor modifications. For human MEK1–NL binary assays, HEK293T cells were transfected with Fugene HD and human MEK1–NL (UniProt Q02750) DNA at a concentration of 0.005 μg ml⁻¹ in combination with 9.995 μg ml⁻¹ carrier DNA (part of the Promega kit) at a density of 200,000 cells per ml. A DNA ratio of 0.005:1:8.995 μg ml⁻¹ MEK1–NL:RAF/KSR:carrier DNA was used for all MEK1–NL coexpressions (ARAF UniProt P10398, BRAF UniProt P15056, CRAF UniProt P04049, KSR1 UniProt Q8IVT5 and KSR2 UniProt Q6VAB6). Note that these DNA ratios were chosen from a DNA ratio titration where steady-state apparent IC₅₀ values were measured across multiple DNA ratios (Supplementary Fig. 3). The 0.005 μg ml⁻¹ transfection concentration for MEK was the lowest concentration where sufficient NL signal (>50,000 relative light units) and apparent IC₅₀ saturation was observed for all coexpressions and MEKi. For MEK1–ARAF–14–3–3–NL, MEK1–BRAF–14–3–3–NL, MEK1–CRAF–14–3–3–NL, MEK1–KSR1–14–3–3–NL and MEK1–KSR2–14–3–3–NL assays, a DNA ratio of 0.1:1:1:7.9 μg ml⁻¹ 14–3–3–NL:MEK1:RAF/KSR:carrier DNA was used (14–3–3-ζ/δ UniProt P63104). For BRAF–NL and KSR1–NL coexpression assays with unlabeled MEK1, a 0.1:1:8.9 μg ml⁻¹ BRAF/KSR1–NL:MEK1:carrier DNA ratio was used. All mutations were introduced using QuikChange mutagenesis. Transfected cells were incubated overnight at 37 °C. Cells were trypsinized and plated into white, low-adherence 96-well plates in Opti-MEM at a density of 20,000 cells per well. All compounds were dissolved in DMSO and further in a 96-well plate using Opti-MEM to a 10× concentration of each dose used to make the dose–response curves. Following the addition of the drugs to cells, a 20× Tram-bo or PD-bo solution (final concentration of 1 μM) in a mixture of DMSO/Tracer dilution buffer (Promega) was immediately added to each well, and the plate was incubated at 37 °C for 2 h. The order of addition of drug first and then tracer was very important to produce dose–response curves that gave the most BRET ratio separation between the highest and lowest doses (that it, BRET_max and BRET_min). To generate dose–response curves, a 3× solution of NL inhibitor and Nano-Glo substrate in Opti-MEM was added to each well, and plates were read on a GloMax plate reader using the standard protocol on the GloMax software. All data were analyzed in Prism 9 (GraphPad). Conversion of apparent IC₅₀ values to calculated K_i was performed using eq. (1):

$$K_{\mathrm{i}}=\frac{{\mathrm{I}\mathrm{C}}_{50}}{(1+\frac{[\mathrm{T}\mathrm{r}\mathrm{a}\mathrm{m}-\mathrm{b}\mathrm{o}]}{{\mathrm{E}\mathrm{C}}_{50,\text{Tram-bo}}})}$$ (1)

Tram-bo and PD-bo buildup curves

HEK293T cells were transfected as described above with MEK1–NL at a concentration of 0.005 μg ml⁻¹ in combination with 9.995 μg ml⁻¹ carrier DNA. Following incubation at 37 °C overnight, cells were trypsinized and plated into white, low-adherence 96-well plates in Opti-MEM at a density of 20,000 cells per well. A 20× solution of either PD-bo or Tram-bo was added to each well to give final concentrations of 2, 1, 0.5, 0.1, 0.05, 0.01, 0.001 and 0.0001 μM for PD-bo and 1, 0.5, 0.1, 0.05, 0.01, 0.005, 0.001 and 0.0001 μM for Tram-bo. The plate was incubated for 2 h at 37 °C, and buildup curves were generated using the standard GloMax protocol after the addition of a 3× solution of NL inhibitor and Nano-Glo substrate in Opti-MEM. All data were analyzed using Prism 9 (GraphPad).

Synthesis of Tram-bo and PD-bo

Please see Supplementary Information. NMR data were collected and processed using TopSpin v3.2.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Extended Data

Extended Data Fig. 1 |. Omit Maps and structural description of Tram-bo’s interactions at the KSR2-MEK1 interface.

Dotted lines in the zoomed in interface images indicate a distance between side-chains of 3.5 angstroms.

Extended Data Fig. 2 |. Tram-bo can form a ternary complex with MEK1 and BRAF/KSR1.

a, Co-IP experiment pulling down MEK-NL reveals that BRAF-FLAG and KSR-FLAG co-associate in the presence of Tram-bo. b, Co-IP experiment pulling down BRAF-FLAG or KSR1-FLAG reveals that MEK-NL co-associates in the presence of Tram-bo.

Extended Data Fig. 3 |. αG RAF/KSR interface mutants highlight assay specificity by reversing changes in MEKi apparent IC₅₀ values.

a, Dose-response curves for MEK1-NL in the absence and presence of WT or αG RAF/KSR interface mutants using a DNA ratio of 0.1:1 μg/mL MEK-NL:RAF/KSR DNA. Each curve is the average of six biological replicates (n = 6) for WT co-expressions and three biological replicates (n = 3) for αG RAF/KSR interface mutants. Each biological replicate has at least one technical replicate. Fitted values are plotted in graphs below the dose-reponse curves. Error bars are the standard error of the mean. b, Same as a using a DNA ratio of 0.005:1 μg/mL MEK-NL:RAF/KSR DNA.

Extended Data Fig. 4 |. PD-Bo as a tracer to measure MEKi binding on MEK1.

a, Structures of PD-Bo and Tram-bo. b, PD-Bo binds worse to MEK1-NL and its ternary RAF/KSR complexes than Tram-bo. Each point represents one biological replicate (n = 3) with at least one technical replicate. Error bars are the standard error of the mean. c, Apparent IC₅₀ values for MEKi appear more potent using PD-Bo than Tram-bo. Each point represents one biological replicate (n = 6 for Tram-bo experiments, and n = 3 for PD-Bo experiments) with at least one technical replicate. Error bars are the standard error of the mean. 0.005:1 μg/mL refers to the DNA transfection ratio of MEK-NL:RAF/KSR DNA.

Extended Data Fig. 5 |. Co-expression of all components of each 14–3–3-NL-MEK1-RAF/KSR complex are necessary to generate a strong BRET signal.

This signal can be disfavored via αG-helix mutations on each RAF/KSR. Each point of each curve represents the average of three biological replicates (n = 3; the BRAF condition had n = 4) with one technical replicate. Error bars represent the standard error of the mean.

Extended Data Fig. 6 |. Mutation of the 14–3–3 interacting residue on RAF/KSR affects its ability to associate with 14–3–3.

Each point of each curve represents the average of four biological replicates (n = 4) with one technical replicate. Error bars represent the standard error of the mean.

Extended Data Fig. 7 |. BRAF-V600E decreases the potency of MEKi binding through its enhanced catalytic activity.

a, Tram-bo binding affinity to MEK1-BRAF mutant ternary complexes. Tram-bo: 1-way ANOVA, MEK1-NL + BRAF-WT vs. MEK1-NL + BRAF-S365A/S729A adjusted P = 0.0013; 1-way ANOVA, MEK1-NL + BRAF-WT vs. MEK1-NL + BRAF-V600E/D576N adjusted P = 0.0015; 1-way ANOVA, MEK1-NL + BRAF-WT vs. MEK1-NL + BRAF-V600E/R509H adjusted P < 0.001;1-way ANOVA, MEK1-NL + BRAF-WT vs. MEK1-NL + BRAF-S365A/V600E/S729A adjusted P < 0.001. b, Apparent binding affinity values and dose-response curves for MEK1-BRAF mutant ternary complexes. For all experiments, dose-response curves are the average of three biological replicates (n = 3). Each curve was fitted separately to generate EC₅₀/apparent IC₅₀ values. Error bars represent the standard error of the mean. Asterisk meanings are as follows: *=p < 0.05, **=p < 0.01, ***=p < 0.001.

Data availability

Atomic coordinates and structure factors have been deposited in the PDB under accession code 7UMB. Source data are provided with this paper.