Modulating multi-functional ERK complexes by covalent targeting of a recruitment site in vivo

Recently, the targeting of ERK with ATP-competitive inhibitors has emerged as a potential clinical strategy to overcome acquired resistance to BRAF and MEK inhibitor combination therapies. In this study, we investigate an alternative strategy of targeting the D-recruitment site (DRS) of ERK. The DRS is a conserved region that lies distal to the active site and mediates ERK–protein interactions. We demonstrate that the small molecule BI-78D3 binds to the DRS of ERK2 and forms a covalent adduct with a conserved cysteine residue (C159) within the pocket and disrupts signaling in vivo. BI-78D3 does not covalently modify p38MAPK, JNK or ERK5. BI-78D3 promotes apoptosis in BRAF inhibitor-naive and resistant melanoma cells containing a BRAF V600E mutation. These studies provide the basis for designing modulators of protein–protein interactions involving ERK, with the potential to impact ERK signaling dynamics and to induce cell cycle arrest and apoptosis in ERK-dependent cancers.

a Progressive reduction in the intensities of the resonances corresponding to T108 and C159 in free ERK2 upon addition of an increasing amount of BI-78D3 (also see Fig.   1c). The positions of the peaks that are quenched also experience a small shift as a result of increased presence of DMSO (the solvent for BI-78D3). A similar overall effect is seen for an adjacent yet unassigned peak marked '*'. Also note the appearance of a new peak marked '**' corresponding to T108 in the C159-modified form of ERK2 that appears during the course of the titration (also see Supplementary Fig. 2 above for further examples of the appearance of new peaks due to the "slow exchange" regime). b Addition of 10 mM DTT at the end of the titration course leads to the reappearance of the peaks corresponding to T108, C159 and the peak marked '*' in free ERK2. The slightly shifted new positions of these peaks are consistent with the direction of the shifts seen in (a) due to the presence of DMSO. It is notable that the peak marked '**' (corresponding to T108 in BI-78D3-modified ERK2) disappears completely upon addition of DTT. Samples contained 200 M ERK2. µL Dioxane-d8 containing 4 mg of (1). The reaction was initiated at room temperature by the addition of 300 µL of 50 mM phosphate buffer, pH 7.5 containing an equimolar amount of (2).
The spectrum was acquired in array (every 3 minutes) after initiation of the reaction (30 minutes spectrum is shown). b 13 C NMR spectrum (500 MHz in 1 H frequency, Bruker Avance III equipped with cryogenic probes) of (1) BI78D3, (3) the product of the reaction between (1) and 2methoxyethanethiol (2) with equal molarity (same reaction conditions as above) and an authentic sample of (4) Con-1 (supplementary methods, synthesis). To an NMR tube was added 300 µL Dioxane-d8 containing 4 mg of (1). The reaction was initiated at room temperature by the addition of 300 µL of 50 mM phosphate buffer, pH 7.5 containing an equal molar amount of (2). We started to scan the sample within 2-3 minutes after initiation of the reaction, 512 scans were acquired in a total of 29 minutes. MHz spectrometer (Palo Alto, CA)) of (1), non-phosphorylated His-ERK2 and (6) the product of the reaction between (1) and ERK2. 11 mg of inactive His tagged ERK2 were allowed to react with BI-78D3 in 50 mM phosphate buffer, pH 7.5 and 5% dioxane for 15-20 minutes at room temperature, the labeled protein was desalted using three PD 10 columns, buffer was exchanged for 6 times to phosphate D2O buffer solution of pH 7.5 using Amicon concentrators (Millipore).
Each time the buffer was diluted 1/2 in deuterated buffer without dioxane, and so some water and dioxane was still present (partial exchange of water and dioxane) and the spectrum was acquired immediately. The experiment was repeated three times and the results were consistent. 100 % labeling of ERK2 by BI-78D3 was confirmed by ESI-MS and UV-Vis spectrophotometry. The

Supplementary Note 1 -Mutational analysis of recombinant ERK2
We mutated a number of residues that exhibited significant perturbations in the NMR spectra of ERK2 upon adduct formation, to alanine. These included L155, N156, and C159 of loop 11, T108 of the inter-lobe linker, H123 of helix E, D316 and D319 of the chg pocket and C164 of the active site. Each mutant was expressed, purified, and activated (see the Methods section) and shown to possess Michaelis-Menten parameters using a peptide substrate, indistinguishable from that of the wild type enzyme (Supplementary Table 1). As expected, mutations at the chg pocket (D316A and D319A) and in the active site (C164A) had no effect on the susceptibility of ERK2 to BI-78D3 (determined as an apparent IC50). Similarly, L155A, H123A were not significantly distinguishable from the wild type protein. However, mutation of C159S completely abrogated the ability of BI-78D3 to inhibit and the T108A and N156A mutants showed a 3-to-4-fold increase in the apparent IC50 (Supplementary Table 1). ** Effect of ERK2 mutations on its Michaelis-Menten kinetic parameters using Ets-1 as a protein substrate.

Supplementary Note 2 -The reactivity of BI-78D3 towards other proteins
Upon incubation of BI-78D3 with recombinant ERK2, p38- MAPK, ERK5, JNK1, JNK2, JNK3, and JNK2 C163A (C163 corresponds to C159 in ERK2) which were expressed, purified and stored in 50 mM phosphate buffer (pH 7.5) containing 10% glycerol. ERK2 was the only protein that showed a characteristic change in the absorption spectrum, consistent with thiol addition ( Supplementary Fig. 12). On the other hand, reaction of DNTB with each tested protein revealed one or more surface accessible cysteines (Supplementary Table 2). Interestingly, mutation of C163 of JNK2 to alanine did not impact the reactivity of the protein with DNTB, suggesting that C163 is probably not accessible to DNTB. Additionally, we used LC-MS to assess whether other MAPKs could form an adduct following incubation with BI-78D3 (10 µM) for 60 minutes. We could not detect the labeling of either His-JNK2 ( Supplementary Fig. 13a), p38- MAPK ( Supplementary Fig. 13c) or ERK5 (Supplementary Fig. 13b).  ** UV spectra of 50 µM enzyme reacting with 10 µM BI-78D3 in 50 mM phosphate buffer (pH 7.5) containing 2% dioxane were recorded every 10 seconds over 600 seconds on an Agilent 8453 diode-array spectrophotometer.

78D3 in melanoma cells
To investigate whether the vulnerability of A375 cells to BI-78D3 can be attributed to the suppression of ERK docking through the DRS, we employed a chemical genetics approach.
Transient expression of ERK2 C159A mutant (an inhibitor-resistant form of ERK2) in A375 cells rescued the ability of BI-78D3 to suppress ERK signaling ( Supplementary Fig. 19), ERK nuclear localization ( Supplementary Fig. 20), and both anchorage-independent ( Supplementary Fig. 21a) and dependent ( Supplementary Fig. 21b) colonies formation. The transient over-expression of wild type ERK also blunted the effect of BI-78D3, but the effect was smaller. The same observations were seen in HEK 293 cells ( Supplementary Fig. 22). This suggests that BI-78D3 suppresses cell proliferation and survival through covalent binding to C159 of ERK.
37 Supplementary Fig. 20 Rendering transiently transfected ERK2 refractive to BI-78D3 in A375 cells suppresses inhibition of EGF-stimulated nuclear translocation by BI-78D3. Transfected A375 cells were serum starved overnight, then treated with 25 µM of BI-78D3 for 1 hour and induced with EGF. Cells were fixed, immune-stained and imaged using a confocal microscope.

Molecular Dynamics
Simulations. An initial complex was generated by docking the compound into crystal structure 4ERK using GOLD. 9 A distance constraint of between 1.5 and 2.5 angstroms was placed between the sulfur of C159 and the appropriate carbon atom of the ligand. All simulations were carried out with a 3 fs time step using the AMOEBA force field, 10 runs on GPUs using Tinker-OpenMM. 11 Protein parameters have defaulted ameobapro13. A distant constraint of 5 kcal A -2 and distance between 1.5 and 2.0 A was used to sample the pre-reaction complex between the sulfur of C159 and the carbon atom. AMOEBA parameters of the compound were generated using poltype. 12 A solvated model (100 Å on each side) of the complex was generated using the structure generated by docking. The resulting complex was heated for from 25 K to 298 K (increasing 25 degrees every 100 ps). After reaching 298 K, the complex was simulated for 100 ns, and the final frame was analyzed.