Structural basis for the action of the drug trametinib at KSR-bound MEK

Abstract

The MAPK/ERK kinase MEK is a shared effector of the frequent cancer drivers KRAS and BRAF that has long been pursued as a drug target in oncology¹, and more recently in immunotherapy^2,3 and ageing⁴. However, many MEK inhibitors are limited owing to on-target toxicities^5-7 and drug resistance^8-10. Accordingly, a molecular understanding of the structure and function of MEK within physiological complexes could provide a template for the design of safer and more effective therapies. Here we report X-ray crystal structures of MEK bound to the scaffold KSR (kinase suppressor of RAS) with various MEK inhibitors, including the clinical drug trametinib. The structures reveal an unexpected mode of binding in which trametinib directly engages KSR at the MEK interface. In the bound complex, KSR remodels the prototypical allosteric pocket of the MEK inhibitor, thereby affecting binding and kinetics, including the drug-residence time. Moreover, trametinib binds KSR-MEK but disrupts the related RAF-MEK complex through a mechanism that exploits evolutionarily conserved interface residues that distinguish these sub-complexes. On the basis of these insights, we created trametiglue, which limits adaptive resistance to MEK inhibition by enhancing interfacial binding. Our results reveal the plasticity of an interface pocket within MEK sub-complexes and have implications for the design of next-generation drugs that target the RAS pathway.

Extended Data Figure 1:. Summary of ligand bound complexes of KSR1:MEK1 and KSR2:MEK1.

A. Resolution, number of reflections, and ligand omit maps for all described structures. Detailed data collection and refinement statistics are provided in Supplemental Data Table 1. Fo-Fc omit electron density maps are all contoured at 3.0 σ, with a 2.0 Å cutoff, around the ligands and shown as a blue mesh. B. Trametinib bound to KSR2:MEK1:AMP-PNP. C. Trametinib contacts include P878 in the pre-helix aG loop of KSR2. Direct contacts of trametinib with MEK1 also highlighted. D. 2D schematic of the trametinib binding pocket in KSR2:MEK1. E. 2D structures, formulas, and molecular weights of MEK inhibitors (MEKi) used in this study.

Extended Data Figure 2:. Conformational changes in MEK and KSR upon binding to trametinib.

A. Close-up view of the trametinib interactions with KSR1 (left) and KSR2 (right). The terminal acetamide group of trametinib stacks between I216 in MEK1 and A825 in KSR1 or P878 in KSR2. Distances with hydrogens included in the models of trametinib and KSR measure 2.4 Å and 3.5 Å between alpha and beta hydrogens of A825 in KSR1 and the terminal -CH₃ of trametinib. In comparison, the terminal -CH₃ of trametinib measures 2.2 Å and 3.1 Å from beta and gamma hydrogens of P878. Measurements are marked by black arrows. Ser222 at one end of the anti-parallel activation segments between MEK and KSR is highlighted. B. The MEKi allosteric pocket, and activation segment displacement, between the isolated state of MEK1 bound to PD0325901 relative to the KSR1:MEK1 complex bound to trametinib. The displacement in the activation segment was measured based on movement of residue Asn221 in the isolated and KSR1-bound state of MEK1. C. Left: distinct activation loop conformers of isolated MEK1 have been observed in complex with PD0325901 (purple; PDB ID 3VVH), TAK733 (light brown; 3PP1), selumetinib (light blue; 4U7Z), and cobimetinib (light green; 4LMN). Middle and Right: overlay of the KSR1:MEK1 and KSR2:MEK1 structures bound to the indicated MEKi reveal near identical activation segment conformers, with the exception of the trametinib-bound complex of KSR1:MEK1. D. Comparison of activation loop conformations in cobimetinib-bound (left) and trametinib-bound (right) states of the KSR1:MEK1 (top) and KSR2:MEK1 (bottom) complexes. Fo-Fc omit electron density map, contoured at 2.0 σ, with a 3.0 Å cutoff, around the activation loop is shown as a blue mesh. Movement of the MEK activation loop between the two inhibitor-bound states of KSR1:MEK1 is highlighted by a red arrow. Main chain H-bonds between the anti-parallel beta strands in KSR and MEK are shown as dotted lines. E. In the trametinib bound KSR1:MEK1 complex, a four-residue anti-parallel beta strand structure is formed between KSR1 and MEK1. In comparison, the same region forms a three-residue stretch in all other KSR1:MEK1 structures that we determined; the cobimetinib-bound complex is shown as an example for comparison. In contrast, a six-residue long anti-parallel beta strand is formed in the KSR2:MEK1 structures, irrespective of bound MEKi. The three- and four- residue long strands in KSR1:MEK1 include residues 769–771/772 for KSR1 and 222/223–225 for MEK1. The six residue long strands in KSR2:MEK1 include residues 820–825 for KSR2 and 221–226 for MEK1.

Extended Data Figure 3:. Structural differences between human KSR1 and KSR2.

A. Comparison of helices αG-αG’ in the KSR1:MEK1 complex (left) and helix αG in KSR2:MEK1 complex. B. 2Fo-Fc omit electron density maps contoured at 1.0 σ, with a 2.0 Å cutoff, around helices αG-αG’ in KSR1 (left) and αG in KSR2 (right). C. 2Fo-Fc omit electron density maps contoured at 1.0 σ, with a 2.0 Å cutoff, around strand β2 in KSR1 (left) and KSR2 (right). D. 2Fo-Fc omit electron density maps contoured at 1.0 σ, with a 2.0 Å cutoff, around the hinge region in KSR1 (left) and KSR2 (right). E. 2Fo-Fc omit electron density maps contoured at 1.0 σ, with a 2.0 Å cutoff, around helix αD in KSR1 (left) and KSR2 (right). F. Positionally equivalent residues H773 in KSR1 and N826 in KSR2 form distinct intra- and inter- molecular contacts, respectively. Specifically, H773 in KSR1 forms a hydrogen bond with the backbone carbonyl of L821 in the αF-αG loop of KSR1 (left). Whereas N826 in KSR2 forms a H-bond across the interfacial region of the KSR2:MEK1 complex via the backbone carbonyl of M219 in MEK1. G. Structure-based sequence alignment of the pseudokinase domains of KSR1 and KSR2 based on structures solved in this study. Boxed regions are highlighted in the upper panels A-F.

Extended Data Figure 4:. Intracellular target engagement on MEK and KSR-bound MEK via bioluminescence resonance energy transfer.

A. Chemical structure of trametinib-bodipy. We refer to this fluorescent probe compound as tram-bo’. B. Legend for schematics used in the lower panels. C. Nano-luciferase tagged fusions of MEK (MEK-luc) and mouse KSR1 (KSR-luc). D. BRET emission signal (red arrow) between MEK-luc and tram-bo is expected to occur within multiple distinct states of MEK, including in the KSR-bound and free states of MEK as depicted. E. BRET emission (red arrow) between KSR-luc and tram-bo is expected to occur exclusively in the KSR-bound state of MEK as depicted. F. Assay design for steady-state competition experiments. G. Assay design for intracellular residence time experiments. H. BRET signals between 1 μM tram-bo and the indicated luciferase tagged fusion proteins expressed in 293T cells. Increasing concentrations of free trametinib were added to these cells to determine IC₅₀ values. Dose-dependent competition for free trametinib was observed on MEK-luc and mouse KSR-luc. However, no discernible dose response for trametinib was observed on controls including RET-luc and SRC-luc using either tram-bo or previously established active site tracers K5 and K4, respectively. I. A helix αG mutant, W781D in mouse KSR1, supports that the BRET signal between wild-type KSR1 and tram-bo depends on intact complex formation between KSR and MEK within cells. In particular, the KSR1-W781D mutant does not produce any dose dependent BRET signal (using 1 μM tram-bo) due to a predicted loss of complexation with MEK1; we previously demonstrated that the W781D mutant (W884D in KSR2 numbering) is a strong loss of function in KSR with respect to ERK pathway activation, and the analogous mutation in BRAF (F667E) prevents direct binding with purified MEK. W781 in mouse KSR1 is equivalent to W831 in human KSR1, W884 in human KSR2, and F667 in human BRAF. Structural depiction of the mouse W781 (ie. W831 in human KSR1) residue at the interface of KSR1:MEK1 complex is shown below.

Extended Data Figure 5:. MEKi IC₅₀ measurements and residence time are influenced by protein complex stoichiometry.

A. IC₅₀ values plotted as a function of MEKi for MEK1-luc and KSR1-luc (left); mean and standard error (SEM) plotted from 3 independent experiments, each conducted in technical triplicate. CH5126766 was not plotted due to poor fit. MEK1-luc (middle) and KSR1-luc (right) dose-response curves for plotted IC₅₀ values using 1 μM Tram-bo; mean and SEM plotted for 3 independent experiments, each conducted in technical triplicates. B. Comparison of MEKi IC₅₀ measurements and representative dose response curves of MEK1-luc, KSR1-luc, MEK1-luc co-expressed with KSR1-WT, and MEK1-luc co-expressed with KSR1-W781D. Co-expression of KSR1-WT with MEK1-luc gives rise to dose response curves and IC₅₀ values similar to that of KSR1-luc alone. This effect does not occur for the co-expression of MEK1-luc with KSR1-W781D, implying that IC₅₀ differences between MEK1-luc and KSR1-luc depend on the formation of the KSR-MEK complex mediated by helix αG. Mean and standard errors determined from 3 independent experiments, each conducted in technical duplicate. IC₅₀s derived from KSR1-luc, MEK1-luc co-expressed with KSR1-WT or W781D were compared to those of MEK1-luc for each MEKi using an ANOVA where an asterisk represents a P-value less than 0.05. For trametinib, data were subjected to a Kruskal-Wallis test and Dunn’s multiple comparison post-hoc test (MEK1-luc vs KSR1-luc adjusted P>0.9999, MEK1-luc vs MEK1-luc + KSR1-WT adjusted P>0.9999, MEK1-luc vs MEK1-luc + KSR1-W781D adjusted P=0.4298). All other data were subjected to an ordinary one-way ANOVA and Dunnett’s multiple comparison post-hoc test with a single pooled variance (Cobimetinib: MEK1-luc vs KSR1-luc adjusted P=0.0015, MEK1-luc vs MEK1-luc + KSR1-WT P=0.0021, MEK1-luc vs MEK1-luc + KSR1-W781D P=0.9940; PD0325901: MEK1-luc vs KSR1-luc adjusted P=0.0350, MEK1-luc vs MEK1-luc + KSR1-WT P=0.1524, MEK1-luc vs MEK1-luc + KSR1-W781D P=0.9920; Selumetinib: MEK1-luc vs KSR1-luc adjusted P=0.0578, MEK1-luc vs MEK1-luc + KSR1-WT P=0.0693, MEK1-luc vs MEK1-luc + KSR1-W781D P=0.9994. Cobimetinib displayed the largest difference in IC₅₀ value between MEK1-luc and KSR1-luc or MEK1-luc + KSR1-WT. C. (Left) Schematic for the origin of the BRET signal under co-expression conditions. (Right) Tram-bo build-up curves for MEK1-luc, KSR1-luc, MEK1-luc co-expressed with KSR1-WT, and MEK1-luc co-expressed with KSR1-W781D. Co-expression of MEK1-luc + KSR1-WT resulted in a lower BRET signal and slower tram-bo build-up compared to MEK1-luc alone. Co-expression of MEK1-luc + KSR-W781D gave similar curves to MEK1-luc alone, suggesting that complex formation is disfavored under these conditions.

Extended Data Figure 6:. KSR and RAF share complementary regulatory roles as MEK scaffolds and activators.

A. KSR and RAF family members appear to have co-evolved. Phylogenetic tree diagrams for the indicated species were generated from reported kinome sequence data that can be found at http://kinase.com/web/current/kinbase/. All species that we analysed include at least one RAF and one KSR homolog. B. Structures of MEK1 in complex with KSR1 and KSR2 determined here, and previously determined structures of MEK1:BRAF-active conformation (PDB ID: 4MNE), and MEK1:BRAF-inactive conformation (PDB ID: 6U2G). C. Structural overlay of MEK1-associated complexes highlights variations in the quaternary arrangements of KSR-bound MEK and RAF-bound MEK. Shown are overlays of MEK1:KSR1 with MEK1:KSR2 (left); MEK1:BRAF (PDB ID: 4MNE) with MEK1:BRAF (PDB ID: 6U2G) (center); and MEK1:KSR1 with MEK1:BRAF (PDB ID: 4MNE). In particular, the N-lobe, including helix αC, in KSR and RAF proteins are significantly displaced between distinct complexes. However, in contrast, the lower C lobe, including helix αG, appears relatively fixed in all sets of complexes. D. Overlay of all structures, using MEK1 C-lobe as an anchor (center), demonstrates helix αG as a common docking site for reciprocal kinase domain interactions between MEK and BRAF or KSR (left inset). Further, the pre-helix αG loop regions within BRAF and KSR proteins occupy a relatively fixed location relative to MEK (right inset).

Extended Data Figure 7:. Variance in the pre-helix αG loops of KSR and RAF proteins determines selectivity for trametinib.

A. The pre-helix αG loop in BRAF (left; N660-N661-R662) includes an insertion and larger amino acid side chains compared to KSR1 (middle; GAP-A825-A826) and KSR2 (right; GAP-P878-A879), creating a clash with trametinib. B. Sequence alignment highlighting conserved variations between RAF kinases and KSR pseudokinases at the trametinib-binding site. Native sequences and mutants in mouse KSR1 and human BRAF used for functional studies in main Figure 3C,D are listed. Mouse KSR1 mutants include K1 (KSR1_P775N), K2 (KSR1_A776R), K3 (KSR1_P775N/A776R), and K4 (KSR1_insertionN/P775N/A776R). Human BRAF mutants include B1 (BRAF_N661A), B2 (BRAF_R662A), B3 (BRAF_N661A/R662A), and B4 (BRAF_N660deletion/N661A/R662A). C. IP/WB of endogenous MEK1 from lysates of HCT116 cells transfected with (left) wild-type KSR1 and mutant K1 (P775N, mouse KSR1 numbering); (middle) wild-type BRAF and mutant B2 (R662A); (right) untransfected controls. Cells were treated with DMSO (D), 200 nM trametinib (T), or 200 nM cobimetinib (C) for 1 hour prior to harvesting cells. IgG was used as a control for non-specific binding of proteins during IPs. Transfected KSR1 or BRAF were detected using an anti-FLAG antibody. All other western blot signals were detected using specific antibodies against endogenous proteins. Blots are representative of three independent experiments. We conducted side-by-side analysis of cobimetinib as a control compound that does not generate direct interfacial contacts like trametinib but displays a similar IC₅₀ on the KSR:MEK complex. Note; compare the effects of cobimetinib addition on complex stability to the effects of trametinib in Figure 3C,D. Unlike trametinib, cobimetinib does not impact the KSR1 or BRAF mutants in terms of pulldown through endogenous MEK similar to trametinib. This data supports that the ‘bump-and-hole’ model for trametinib selectivity between KSR-bound MEK and RAF-bound MEK. Further note from Figure 3C,D that all of the tested KSR1 alleles, and also the full swaps of the pre-helix αG loops between RAF and KSR proteins, resulted in partial or complete loss of pulldown via MEK (Figure 3C,D; lanes 2 vs 10 for mutants K4 and B4), which suggests that the length and composition of interfacial residues within both KSR and RAF proteins are critical and unique determinants of binding towards MEK. D. Overlay of four clinical MEKi highlights the phenyl acetamide group of trametinib as a unique ‘bump’ not found in the other compounds including cobimetinib. E. BRET buildup curves with increasing concentrations of tram-bo on the indicated luciferase-tagged versions of human KSR1, KSR2, ARAF, BRAF, and CRAF/RAF1. KSR1-luc and KSR2-luc both show higher BRET ratios, and also ~10-fold tighter binding, with tram-bo relative to ARAF-luc, BRAF-luc, and CRAF-luc. Lower inset is a y-axis magnification of the top inset. Data points represent the average of two technical replicates; experiments were conducted at least three independent times with similar results.

Extended Data Figure 8:. In vitro binding of purified MEK, KSR:MEK, and RAF:MEK to trametinib.

A. Representative binding sensograms for 500 nM each of isolated MEK1 or the indicated KSR:MEK and BRAF:MEK complexes on a biosensor immobilized with biotin-conjugated trametinib. Fitting of association and dissociation phases based on one-to-one binding is provided in Source Data Extended Data Fig. 8. B. K_D (M), k_on (1/Ms), and k_dis (1/s) values for MEK1 (M), KSR1:MEK1 (K1M1), KSR2:MEK1 (K2M2), BRAF:MEK1 (BRM1) on biotin-linked trametinib. Individual data points from independent binding experiments (n=29, 14, 22, and 9 for MEK1, KSR1:MEK1, KSR2:MEK1, and BRAF:MEK1, respectively) were used for statistical comparisons (**** for p ≤ 0.0001, respectively; Source Data Extended Data Fig. 8). Note, trametinib likely favours dissociation of BRAF from MEK1 for binding. For example, whereas the association and K_D data between BRAF:MEK1 and isolated MEK1 markedly differ, the off rate and residence time calculations are similar. This data would be consistent with a model in which the equilibrium of BRAF:MEK1 shifts so as to populate the dissociated state under the conditions of the BLI assays. C. Residence time values plotted as a function of protein concentration. MEK1 and BRAF:MEK1 display small variations in residence time over the concentrations tested. Whereas KSR2:MEK1 and KSR1:MEK1 demonstrate concentration-dependent changes in residence time. In particular, at low concentrations of KSR-MEK, where the complexes would be expected to more readily dissociate, the kinetic values of purified KSR1:MEK1 and KSR2:MEK1 approached isolated MEK1 and BRAF:MEK1. D. Full binding curve experiment including loading of biotin-conjugated trametinib for 10 mins, followed by a wash step, and subsequently treating a low-density streptavidin (SA) sensor with a blocking agent, biocytin for 3 min. The sensors were washed extensively to acquire a zero baseline prior to binding analysis. Following, sensors were dipped in wells containing 500 nM of each protein for 15 min, followed by a dissociation in running buffer for 15 min. E. A biotin conjugated version of trametinib was immobilized on sensor-heads and binding to MEK1, KSR1:MEK1, KSR2:MEK1 or BRAF:MEK1 was monitored using bio-layer interferometry. Increasing concentrations in 2-fold increments of proteins from 31.25 nM to 500 nM for MEK1, KSR1:MEK1, and KSR2:MEK1 and 500 nM to 2000 nM for BRAF:MEK1 were tested. A blank sensor head without immobilized trametinib was used as a control for non-specific binding. K_D (M), k_on (1/Ms), and k_dis (1/s) values were derived from fitting each binding curve.

Extended Data Figure 9:. KSR as a co-receptor for binding to trametinib.

A. Literature data on CRISPR depletion screens highlight strong functional interactions between trametinib and KSR. For example, in a Drosophila cellular fitness model (left) and a human BRAF V600E mutant cell line (right), sgRNAs towards KSR generated relative outlier sensitivity or resistance to trametinib or a trametinib+dabrafenib combination, respectively. Raw data from Viswanatha et al. was plotted based on the authors determination of a Z-score for log₂-fold change in sgRNA reads for S2 cells treated with trametinib versus a no treatment control (left). Raw data from Strub et al. was plotted based on the authors determination of log₂-fold change in sgRNA reads for SKMEL-239 cells treated with a trametinib+dabrafenib combination relative to a no treatment control (right). sgRNAs towards KSR are highlighted as a red dot; all other sgRNAs analysed in the respective studies are shown as grey dots. KSR emerged as a strong outlier beyond the mean plus standard deviation (black cross hairs) of all genes analysed in each respective study. These screens could be re-investigated based on the model that KSR functions as a direct co-receptor for binding to trametinib and MEK. B. Model for the action of trametinib on KSR:MEK and RAF:MEK complexes. In the absence of drug, MEK activation depends on heterodimerization of both RAF and KSR, with phosphorylation on the sites S218/S222 occurring through active RAF kinases. This model is adapted from structural and biochemical studies in^,,,. Trametinib could down-regulate ERK signaling by impeding direct binding of MEK towards RAF in favor of KSR. In the KSR-bound state of MEK, trametinib would be expected to reside on target for extended periods of time.

Extended Data Figure 10:. Trametiglue provides durable inhibition of RAS/ERK signaling in models of mutant KRAS and BRAF.

A. (Left) Immunoblot of stable HCT-116 (KRAS G13D) cancer cells including parental, scramble control (shSCR), and KSR1 knockdown (shKSR1). Cells were treated with 10 nM trametinib for the indicated time points and harvested for analysis on the indicated markers. (Right) Quantitative PCR was used to confirm specific knockdown of KSR1 in the shKSR1 cells. KSR1 knockdown slows the rebound of activated RAS-MAPK signaling in the presence of trametinib as measured by recovered phosphorylated-ERK1/2 over time (lanes 1–5 and 6–10 versus 11–15). This data supports that KSR1 plays a positive role in the adaptive resistance of HCT-116 cells to trametinib, suggesting that knockdown or trapping of the KSR-bound MEK complex could mitigate this intrinsic drug resistance mechanism. Experiment was conducted twice with similar results. B. EC₅₀ values for cell viability assays for the indicated compounds against a series of human cancer cell lines. Mean and standard deviation determined from three independent experiments, each conducted in technical triplicate. Raw data is included in Source Data Extended Data Fig. 10. C. X-ray crystal structure of trametinib bound to the KSR2:MEK1:AMP-PNP complex. MEK1 and KSR2 are colored pink and green respectively, with several key residues highlighted. Trametinib is shown in stick representation. A Fo-Fc omit electron density map, contoured at 3.0 σ with a 2.0 Å cutoff around ligand, is shown as a blue mesh. Left panel shows the entire inhibitor binding pocket; right panel highlights contacts around the phenyl acetamide group of trametinib. D. Bar graph plot of mean EC₅₀ values from B. E. Clonogenic assay of KRAS-mutant and BRAF-mutant cancer cell lines treated with 10 nM trametinib or 10 nM trametiglue, and 10 nM or 50 nM CH5126766 for 10 days. Experiment was conducted twice with similar results. F. Immunoblot analysis of the indicated cell lines treated for 1 hour with increasing concentrations of trametiglue and trametinib. This data supports that trametiglue, relative to trametinib, is a higher potency inhibitor of RAS-MAPK signaling as measured by phosphorylated ERK1/2 at residues T202 and Y204 (pERK). Experiment was conducted three times with similar results. G. Immunoblot of KRAS-mutant and BRAF-mutant cancer cell lines treated with 10 nM trametinib or trametiglue for various times. Experiment was conducted twice with similar results.

Figure 1.. The trametinib binding pocket in MEK extends to the KSR interaction interface.

A. Trametinib bound to KSR1:MEK1:AMP-PNP. See Extended Data Figure 1 for trametinib bound to KSR2:MEK1:AMP-PNP. B. Trametinib contacts include A825 in the pre-helix αG loop of KSR1. Direct contacts of trametinib with MEK1 also highlighted. C. 2D schematic of the trametinib binding pocket.

Figure 2:. Binding of KSR to MEK creates an enlarged allosteric binding pocket for inhibitors.

A. Binding comparison of MEKi in KSR1:MEK1. Of the analysed MEKi, only trametinib directly engages KSR1. B. Structures of isolated MEK1 bound to cobimetinib (PDB ID 4LMN; activation segment colored green), selumetinib (4U7Z; blue) and PD0325901 (3VVH; purple) compared to the KSR1:MEK1 complex for the indicated MEKi. Binding of KSR1 to MEK1 stabilizes an outward orientation of the MEK1 activation segment. See Extended Data Figure 2 for additional analysis and electron density maps. C. Comparison of trametinib IC₅₀ values vs other MEKi on MEK1-luc and KSR1-luc. Mean and standard errors determined from 3 independent experiments. Each drug was compared to trametinib using an unpaired two-tailed t test where an asterisk represents a P-value less than 0.05. MEK-luc: T(trametinib) vs. C(cobimetinib) p<0.0001, t=16.85, df=4; T vs P(PD0325901) P=0.0004, t=11.26, df=4; T vs S(selumetinib) P=0.0081, t=4.985, df=4; KSR-luc: T vs C P=0.223, t=1.425, df=4; T vs P P=0.0046, t=5.735, df=4, T vs S P=0.0036, df=4. D. Representative intracellular residence time plots for cells pre-treated with a range of sub-saturating levels of trametinib and cobimetinib that were transfected with MEK1-luc (left) and mouse KSR1-luc (right). The build-up signal (height and rate) is proportional to the dissociation of the indicated compounds on MEK1 or KSR1-bound MEK complex following the addition of Tram-bo (1 μM) on cells that were pre-treated and then washed of the MEKi. Note the DMSO curves are the same in both the trametinib and cobimetinib plots due to the experimental design where all data was collected at once on one plate. See replicate data in Supplementary Figure 2.

Figure 3:. The trametinib binding site distinguishes KSR from RAF.

A. Structural overlay of the BRAF:MEK1, KSR1:MEK1, and KSR2:MEK1 complexes predicts a clash between trametinib and the pre-helix αG loop of BRAF. B. Sequence alignment of RAF kinases and KSR pseudokinases at the trametinib-binding site. Numbering for human KSR1, KSR2, and BRAF highlighted. C. And D. IP/WB of endogenous MEK1 from lysates of HCT116 cells transfected with FLAG-tagged wild-type and mutant versions of full-length mouse KSR1 (panel C) or human BRAF (panel D). Cells were treated with DMSO or 200 nM trametinib for 1 hour prior to harvesting cells. IgG was used as a control for non-specific binding of proteins during IPs (lanes 1 vs 2). Transfected KSR1 and BRAF were detected using anti-FLAG antibody. All other western blot signals were detected using specific antibodies against endogenous proteins; note, the antibody against BRAF detects both endogenous and transfected FLAG-tagged BRAF. Blots are representative of three independent experiments with similar results, uncropped blots in Supplementary Figure 1.

Figure 4:. Trametiglue targets both KSR-MEK and RAF-MEK with unprecedented potency and selectivity via unique interfacial binding interactions.

A. Chemical structures of trametinib and trametiglue. B. X-ray crystal structure of trametiglue bound to KSR2:MEK1:AMP-PNP. Fo-Fc omit electron density map, contoured at 3.0 σ with a 2.0 Å cutoff around ligand. Left panel shows the entire inhibitor binding pocket; right panel highlights contacts around the sulfamide group of trametiglue. C. Overlay of trametinib (yellow) and trametiglue (orange). The sulfamide group of trametiglue, relative to the acetamide in trametinib, generates unique contacts at the interfacial binding region of KSR-bound MEK. In particular, the -NHSO₂NHCH₃ module of trametiglue facilitates unique space-filling via M230 and the peptide backbone of R189 in MEK1 and a water-mediated H-bond towards the backbone of T876 in KSR2. D. Trametiglue retains the strong binding potency and residence time of trametinib on KSR-bound MEK as determined under steady-state conditions (left) and intracellular residence (right panel; all compounds tested at 6.25 nM) formats. Each point and error bars represent the mean and SEM of three independent experiments. Data points for the intracellular residence time experiments represent the average of two technical replicates, each repeated three independent times. Additional data in Supplementary Figure 2. E. Trametiglue, unlike trametinib but similar to CH5126766, enhances interactions between endogenous BRAF and MEK1. IP/WB of endogenous MEK1 from HCT116 cells. Lanes 1–4 are cells transfected with FLAG-KSR1, and lanes 5–8 are untransfected samples. Cells were treated with DMSO, 200 nM CH5126766, 200 nM trametinib, or 200 nM trametiglue for 1 hour prior to harvesting cells and IPs. Blots are representative of three independent experiments with similar results. F. In vitro profiling of 1 μM trametiglue demonstrates high selectivity towards MEK1 and MEK2 in direct binding assays (top). Trametiglue also displays high selectivity in a panel of active kinases measured for inhibition of MEK1 and MEK2 substrate phosphorylation or direct MEK1 phosphorylation by the upstream kinases as indicated (bottom). See Source Data Fig. 4. G. Cell viability dose-responses on K-RAS and BRAF mutant lines. Assays conducted under low-adherence conditions and representative of three independent experiments, each conducted in technical triplicate. Mean and standard deviations in Extended Data Figure 10B.