Cryo-EM structure determination of small therapeutic protein targets at 3 Å-resolution using a rigid imaging scaffold

Significance Cryoelectron microscopy (cryo-EM) is emerging as a major method for elucidating the structures of proteins in atomic detail. A key limitation, however, is that cryo-EM is applicable only to sufficiently large macromolecular complexes. This places a great many important proteins of smaller size, especially those of interest for therapeutic drug development, outside the reach of cryo-EM. We describe a protein engineering effort that overcomes the lower mass limit through the development of a modular imaging scaffold able to rigidly bind and display practically any small protein of interest, greatly increasing its effective mass. We show this technology can be used to visualize molecules, such as a key cancer protein, with important implications for drug design and biomedical research.

Comparison of models at the drug binding site after refinement against independent half-maps -i.e.maps obtained by 3-D reconstructions from independent sets of cryo-EM projection images.Model refined with half map A (green), half map B (marine), and the crystal structure 6oim (magenta).The coordinate differences between the two half map-refined models are approximately 1/10 the magnitude of the differences between the crystal structure (6oim) and the scaffolded cryo-EM structure (8g47).

Supplementary text:
Recent cryo-EM scaffolding results: In concurrent developments using antibody/nanobody approaches, the resolutions reported have been better for larger cargo proteins -e.g.2.49 Å for a ~200 kDa receptor complex (2); 2.8 Å for a 64 kDA protein (3), 3.03 Å for a 58 kDa protein (4); ~3.2 Å for a 52 kDa protein (5), 3.47 Å and 3.78 Å for ~50 kDa proteins using NabFabs (6).For proteins smaller than 50 kDa, the finest resolution so far are 3.0−4.0Å for a 11 kDa KIX domain fused to apoferritin (7) and roughly 3.2 Å for a 23 kDa protein bound to a scaffold ensemble (8).An overall resolution of 3.2 Å was reported for the overall complex between the scaffold and KDELR; a resolution range of 3.0 to 3.5 Å was estimated for the just the KDELR protein.These are summarized in Table S3.

Figure S2 .
Figure S2.SDS PAGE gel showing co-elution of the two protein chains, A and B, comprising a scaffold.Subunit A (a component of the cage core) is His-tagged.Subunit B is a fusion between a cage core component and a DARPin that serves to bind diverse cargo proteins for imaging.

Figure S3 .
Figure S3.Negative stained electron micrographs of the rigidified imaging scaffold particles.

Figure S4 .
Figure S4.Local resolution for the focus-refined map of the GFP and DARPin.A. DARPin (salmon) and GFP (green).B and C, two rotated views of the map colored by local resolution.D, a cross section of the map.

Figure S5 .
Figure S5.Cryo-EM densities for the imaging scaffold bound to GFP.The overall reconstruction (left) and a composite map of the focused refinements (right).After symmetry expansion (symmetry T), focused classifications and refinements were performed with a mask encompassing mainly one GFP (green) and one DARPin (salmon).

Figure S6 .
Figure S6.Binding assay of KRAS G13C to rigidified imaging scaffolds.The early SEC fractions show co-elution of the cage components with the KRAS (bound).The later fractions correspond to unbound KRAS.The higher band (~35 kDa) corresponds to the cage subunit B fused to the DARPin, the middle band corresponds to the KRAS protein (19.4 kDa) and the lower band is the cage subunit A (20kDa).Cage subunit A runs slightly smaller than its known size.

Figure S7 .
Figure S7.Local resolution for the focus-refined map of the KRAS G13C and DARPin.A. DARPin (salmon) and KRAS (yellow).B and C, two rotated views of the map colored by local resolution.D, cross section of the map.

Figure S8 .
Figure S8.Cryo-EM densities for the imaging scaffold bound to KRAS G13C.The overall reconstruction (left) and a composite map of the focused refinements (right).After symmetry expansion (symmetry T), focused classifications and refinements were performed with a mask encompassing mainly one KRAS (yellow) and one DARPin (salmon).

Figure S9 .
Figure S9.Binding position of the AMG510 drug to KRAS G12C in the cryo-EM structure reported here, compared to an x-ray crystal structure of the same complex.The cryo-EM density is shown with the refined cryo-EM model in blue.An earlier, high resolution x-ray crystal structure (pdb 6oim) is shown in salmon.[The comparison is shown after overlapping the protein backbone structures from the two coordinate sets.]The Q-score (i.e.agreement between atomic model and density map) for the terminal part of the drug molecule refined in the cryo-EM map is 0.59.For comparison, the Q-score for the drug molecule in the X-ray position compared to the cryo-EM map is considerably lower, 0.45.The refined conformation for the cryo-EM model is a better fit compared to the conformation see in the X-ray crystal structure.

Figure S10 .
Figure S10.Imaging scaffold binding to KRAS G12C-AMG510.AMG510 shows maximum absorption at λmax=354 nm, allowing for validation of complex formation between imaging scaffold and AMG510-bound KRAS.Covalent attachment of AMG510 to KRAS G12C increases the molecular weight by 560 Da which can be resolved on SDS-PAGE (inset).The SEC profile shows preferential binding of AMG510-bound KRAS G12C over inhibitor-free KRAS G12C to the imaging scaffold.

Figure S11 .
Figure S11.Cryo-EM maps summary.Cryo-EM maps reported in this manuscript with corresponding plots reporting the resolutions via gold standard FSC at a cutoff of 0.143 and the angular distribution of particles.The corrected FSC from EMDB:29700 is the FSC curve calculated using the tight mask with correction by noise substitution, provided when using the Homogeneous Refinement tool in cryoSPARC.Particle orientation density plots are shown in the right column.

Figure S12 .
Figure S12.Improvements in Q-score values (i.e.correlation between atomic model and density map) are observed for the new rigidified imaging scaffolds (left) when compared with the old imaging scaffolds (right).The old scaffold exhibits a distribution of Q-scores that increases dramatically as a function of distance from the hinge point.The new scaffold largely eliminates this dependence, showing that elimination of the hinge motion is largely responsible for the improved quality of the resulting density map.

Figure S13 .
Figure S13.Comparison of automatic atomic model-building trials using density maps from the present scaffold vs a prior scaffold prior to engineering for rigidification.For the new scaffold, 95% of the GFP scaffold could be correctly built automatically, compared to 28% for the map obtained from the prior scaffold.The models automatically built by the program ModelAngelo are shown in green, overlaid with the final structure in gray.

Figure S14.
Figure S14.Comparison of models at the drug binding site after refinement against independent half-maps -i.e.maps obtained by 3-D reconstructions from independent sets of cryo-EM projection images.Model refined with half map A (green), half map B (marine), and the crystal structure 6oim (magenta).The coordinate differences between the two half map-refined models are approximately 1/10 the magnitude of the differences between the crystal structure (6oim) and the scaffolded cryo-EM structure (8g47).

Figure S15 .Figure S16 .
Figure S15.Potential influence of protein packing forces on the conformation of the AMG510 drug molecule and the neighboring binding pocket on KRAS.(A) Crystal structure of KRAS-AMG510 (6oim).Gray colors indicate residues structurally conserved with the new scaffolded cryo-EM KRAS structure (8g47).Red indicates structurally variable residues.Pink indicates a crystallographically related symmetry copy of the KRAS molecule.The neighboring molecule impinges on a helix (red) that contacts AMG510.(B) Scaffolded cryo-EM structure of KRAS-AMG510 (8g47).Residues that differ in structure from 6oim are shown in deep blue color.Light blue indicates a segment of DARPin used to display KRAS.(C) Superposition with arrows showing the different angles at which packing forces influence the AMG510 binding site in the two structures.

Table S1 .
Cryo-EM data collection, image analysis, modeling, refinement, and validation statistics

Table S3 .
Comparison of size and resolution for contemporary Cryo-EM protein fiducial/scaffolds including their binders and targets.