The Ras-Raf protein association initiates the downstream mitogen-activated protein kinase (MAPK) signaling cascade. This interaction depends on the Ras-nucleotide ligand, whereby Raf preferentially binds to the active guanosine triphosphate (GTP)-loaded state over the guanosine diphosphate (GDP)-loaded state of Ras. Whether GDP-bound Ras can also form specific protein-protein interactions with Raf, however, remains unclear. Here, we characterize the interaction between human K-Ras in both nucleotide states and the Ras-binding domain (RBD) of B-Raf by solution NMR. Active K-Ras•GTP forms a tight, conformationally restricted complex with RBD, causing pronounced chemical shift perturbations at the binding interface consistent with existing cryo-EM and X-ray crystallography structures. Surprisingly, we detect specific binding between “inactive” wild-type K-Ras•GDP with RBD in the millimolar affinity range, whereby the oncogenic G12D mutant of K-Ras further strengthens this interaction. NMR relaxation dispersion and chemical exchange saturation transfer (CEST) experiments allow the detailed structural characterization of the transient K-Ras•GDP•RBD complex, along with the determination of the interaction affinity and kinetics. The results demonstrate that the transient K-Ras•GDP•RBD bound state closely resembles the active K-Ras•GTP•RBD complex, providing a quantitative understanding at the backbone ¹⁵N-level of the nucleotide-specific K-Ras-Raf interaction. These findings underscore the prospect of the highly adaptable GDP-bound state of K-Ras for both signaling and as a drug target.

All NMR experiments were performed at 298 K on a Bruker magnet with an AVANCE III console operating at 850 MHz (19.97 T) equipped with a 5 mm TCI cryoprobe. Backbone amide ¹⁵N-(STCW-)CPMG NMR relaxation dispersion and ¹⁵N-CEST experiments were performed for ¹⁵N-labeled K-Ras•GDP in the presence of unlabeled RBD and ¹⁵N-labeled RBD in the presence of unlabeled K-Ras•GDP, for both WT K-Ras and G12D mutant. Amide HSQC/HMQC experiments were measured to assist the sign determination of the chemical shift differences Δϖ during the exchange fittings. All dynamics experiments were measured with recovery delay d1 of 2 s. The constant relaxation time T_CPMG for the CPMG block was set to 30 or 40 ms. The CPMG ¹⁵N-refocusing pulses were applied with a B₁-field of approximately 6.1 kHz, and the CPMG pulsing frequency ν_CPMG was varied from 25 Hz to 2 kHz. For CEST experiments, mixing times ranged from 150 ms to 250 ms with the saturation field strength B₁ set around 20 Hz.

The dynamics NMR experiments were processed with NMRPipe and visualized with NMRFAM-SPARKY or POKY. The cross-peak intensities were extracted with either autoFit in NMRPipe or a Voigt fitter. CPMG and CEST profiles were analyzed collectively using ChemEx. The profiles were fitted to a two-state exchange model using the two-step procedure. Briefly, in the first step, residue-specific ¹⁵N R_ex values were estimated from the CPMG profiles as the difference between the effective R₂ values at the minimal and maximal ν_CPMG frequencies (R_ex = R_2,eff(min(ν_CPMG)) – R_2,eff(max(ν_CPMG)). CEST and CPMG profiles of residues with R_ex > 5 s^-1 were then selected to determine the global exchange parameters with ChemEx. ChemEx allows numerical simulation of the CEST and CPMG pulse sequences with a two-state exchange model. The output includes the global parameters exchange rate k_ex = k_AB + k_BA and populations of each state p_A and p_B = 1 - p_A (subscripts A and B stand for state A and B), as well as the site-specific chemical shift differences Δϖ. In the second step, CEST and CPMG profiles of all residues were analyzed with the global parameters k_ex and p determined in step 1 constrained to obtain Δϖ of each residue. The sign of Δϖ was obtained from CEST and/or HSQC/HMQC experiments. The errors in peak intensities were estimated from replicate planes in CPMG or CEST experiments, which were then propagated through the analysis. Errors in the global exchange parameters and Δϖ were determined from 500 bootstrap replicate simulations.