Figure 9. NMR chemical shift perturbation and CHESCA analysis for FGFR2K mutations.

(A) ¹H/¹⁵N correlation spectra are shown in the middle four panels for residues A567, G646, G542, and I623 located near the molecular brake, within the DFG latch, near the αC tether, and within the catalytic loop, respectively. The colors of this spectral overlay match to those of the seven mutants listed in the top three and bottom four panels. Perturbations are mapped onto the active WT FGFR2K structure (PDB ID: 2PVF [Chen et al., 2007]) with the magnitude of changes reflecting the difference between unphosphorylated WT FGFR2K and the respective activating mutation (red indicates the maximum perturbation, while yellow corresponds with no perturbation). Blue colored regions correspond to residues whose chemical shifts disappeared or shifted beyond detection for the given mutant. The mutated residue in each structure is colored green. (B) Ile, Leu, and Val chemical shift perturbations of K659E (left) and E565A (right) relative to those of unphosphorylated WT FGFR2K. The methyl perturbation sites shown in spheres are mapped onto the autoinhibited WT FGFR1K (PDB ID: 3KY2 [Bae et al., 2010]); note that the residue numbering convention corresponds to that of FGFR2K. White spheres correspond to residues unassigned or overlapped in the mutants. (C) Residues within the allosteric network identified using CHESCA are mapped onto the FGFR1K structure with a sphere at the backbone nitrogen position (PDB ID: 3KY2 [Bae et al., 2010]). The chemical shifts from a series of mutations at K659 (T, N, Q, M, E) and WT phosphorylated and unphosphorylated FGFR2K were used for the analysis. (D) Phylogenic tree showing the three separate clusters of residues with correlation coefficients |r_ij| = 0.97. Based on the similar functional network, these three clusters comprise the same allosteric network (Figure 9—figure supplement 2).

DOI: http://dx.doi.org/10.7554/eLife.21137.015

Figure 9—figure supplement 1. Chemical shift perturbations for pathogenic mutants of FGFR2K.

(A) The combined chemical shift perturbation (Δδ) as a function of residue was calculated according to equation 1 between the unphosphorylated WT FGFR2K and that of the indicated mutant. Green circles show residues that are within 10 Å of the mutation site and therefore may have chemical shift perturbations resulting from electrostatic changes of the mutant. The grey bars in the background correspond to the combined chemical shift perturbation between phosphorylated and unphosphorylated FGFR2K (i.e., active vs. autoinhibited). The red vertical lines are residues that were missing in the spectrum relative to unphosphorylated FGFR2K. (B) ¹H/¹⁵N TROSY spectra showing molecular brake residues N549 and E565 and how they are perturbed in the three brake mutants (N549T, E565A, K641R) relative to wild-type FGFR2K.

DOI: http://dx.doi.org/10.7554/eLife.21137.016

Figure 9—figure supplement 2. CHESCA functional network and single-value decomposition analysis.

(A) The three dendrograms shown in Figure 9D were obtained using the complete linkage CHESCA approach (Boulton et al., 2014) where each cluster corresponds to residues having correlation coefficients |r_ij| > 0.97. (B) State-based dendrograms showing the results of agglomerative clustering using residues corresponding to the individual clusters I-III. (C and D) Score and loading plots indicating that the residues identified as part of the allosteric network (in red) align along the same principle component (PC1), which further validates the conclusion that clusters I-III belong to the same allosteric network. The analysis was performed as described by Melacini and co-workers (Selvaratnam et al., 2011; Boulton et al., 2014).

DOI: http://dx.doi.org/10.7554/eLife.21137.017