Covalently constrained 'Di-Gembodies' enable parallel structure solutions by cryo-EM

Abstract

Whilst cryo-electron microscopy(cryo-EM) has become a routine methodology in structural biology, obtaining high-resolution cryo-EM structures of small proteins (<100 kDa) and increasing overall throughput remain challenging. One approach to augment protein size and improve particle alignment involves the use of binding proteins or protein-based scaffolds. However, a given imaging scaffold or linking module may prove inadequate for structure solution and availability of such scaffolds remains limited. Here, we describe a strategy that exploits covalent dimerization of nanobodies to trap an engineered, predisposed nanobody-to-nanobody interface, giving Di-Gembodies as modular constructs created in homomeric and heteromeric forms. By exploiting side-chain-to-side-chain assembly, they can simultaneously display two copies of the same or two distinct proteins through a subunit interface that provides sufficient constraint required for cryo-EM structure determination. We validate this method with multiple soluble and membrane structural targets, down to 14 kDa, demonstrating a flexible and scalable platform for expanded protein structure determination.

Extended Data Fig. 1 |. Modular generation of homo Di-Gembodies (homoDiGbs) for cryo-EM.

(a) Gb5–006 (b) GbMBP (c) GbD12 (d) GbLysozyme. Intact protein mass spectra of Gembody monomers and homoDiGb with peaks indicated by circles in grey and green, respectively.

Extended Data Fig. 2 |. Modular generation of hetero Di-Gembodies (heteroDiGbs).

(a) Representative generation process for heteroDiGbs used in the cryo-EM studies GbC4:GbEnhancer and Gb5–006:GbEnhancer (b) demonstrating the modularity. Intact protein mass spectra peaks for Gembody monomers, Gembody-TNB conjugates and heteroDiGbs are indicated by circles in grey, yellow and green, respectively.

Extended Data Fig. 3 |. Cryo-EM structural determination of RECQL5 in complex with a homoDiGb5–006.

(a) Exemplar raw micrographs. Similar images were captured on three separate occasions. (b) 2D classification results of initial particles after Topaz picking. (c) Cryo-EM 3D reconstruction pipeline. (d) Map details of RECQL5’s residues 197–209, 213–221, and 150–158, containing helical, beta sheet, and loop secondary structures, respectively. (e) The Fourier Shell Correlation curves of the overall complex and locally refined RECQL5 after symmetry expansion (SE). (f) Particle orientation distribution of the final locally refined map.

Extended Data Fig. 4 |. Cryo-EM structural determination of SPNS2 in complex with homoDiGbD12.

(a) Exemplar raw micrograph. Similar images were captured on three separate occasions. (b) 2D classification of Topaz-picked particles. (c) Cryo-EM 3D reconstruction pipeline. (d) Cryo-EM map of SPNS2 monomer after symmetry expansion, with map-and-model overlays for TM7 residues 300–346 and central binding site bound with DDM. (e) The Fourier Shell Correlation curves of the overall dimer complex and locally refined SPNS2 after symmetry expansion (SE). (f) Particle orientation distribution of the final locally refined map.

Extended Data Fig. 5 |. Cryo-EM structural determination of MBP in complex with homo Di-Gembody homoDiGbMBP.

(a) Exemplar raw micrograph. Similar images were captured on three separate occasions. (b) 2D classification of Topaz-picked particles. (c) Cryo-EM 3D reconstruction pipeline. (d) Cryo EM map of MBP monomer after symmetry expansion, with map-and-model overlays for residues 145–167 and 259–285. (e) The Fourier Shell Correlation curves of the overall dimer complex and locally refined MBP after symmetry expansion (SE). (f) Particle orientation distribution of the final locally refined map.

Extended Data Fig. 6 |. Cryo-EM structural determination of lysosome in complex with homo Di-Gembody homoDiGbLysozyme.

(a) Exemplar raw micrograph. Similar images were captured on three separate occasions. (b) 2D classification of Topaz-picked particles. (c) Cryo-EM 3D reconstruction pipeline. (d) Cryo-EM map of lysozyme monomer after symmetry expansion, with map-and-model overlays for residues 21–40 and 87–110 and the binding site with the bound chloride. (e) The Fourier Shell Correlation curves of the overall dimer complex and locally refined SPNS2 after lysozyme. (f) Particle orientation distribution of the final locally refined map.

Extended Data Fig. 7 |. Cryo-EM structural determination of the RECQL5:heteroDiGb:sfGFP complex.

(a) Exemplar raw micrograph. Similar images were captured on three separate occasions. (b) 2D classification of particles selected by Topaz picking. (c) Cryo-EM 3D reconstruction pipeline. (d) Overall view of sfGFP after local refinement, and structural details of peripheral regions of the central helix (51–73) and two beta sheets (12–22 and 105–116). (e) Map details of RECQL5 sub regions of the helix (48–68) and the region containing a beta sheet and a loop (214–229). (f) The Fourier Shell Correlation curves of the overall complex, the locally refined map of sfGFP, and the locally refined map of RECQL5. (g) Particle orientation distribution of the overall complex before local refinement of individual targets.

Extended Data Fig. 8 |. Cryo-EM structural determination of the SPNS2:heteroDiGb:sfGFP complex.

(a) Exemplar raw micrograph. Similar images were captured on three separate occasions. (b) 2D classification of particles selected by Topaz picking. (c) Cryo-EM 3D reconstruction pipeline. (d) Map and model of SPNS2’s transmembrane helices of TM2, TM7 and TM11. (e) The Fourier Shell Correlation curves of the overall complex before local refinement, the locally refined map of SPNS2, and the locally refined map of sfGFP. (f) Particle orientation distribution of the dimer complex before local refinement.

Extended Data Fig. 9 |. 3D variability analyses of Di-Gembody structures.

Sixty models of each structure are overlaid and aligned using the static Gembody. Static and moving Gembodies are shown in light blue and teal, respectively. The left insets indicate the reference points (yellow spheres), vectors (${\stackrel{\rightharpoonup }{D}}_i$ shown as a yellow arrow) and angle $\beta$ for the moving Gembody used in the analyses. ${\stackrel{\rightharpoonup }{D}}_a$ shown as a black arrow is the average vector of ${\stackrel{\rightharpoonup }{D}}_i$. The right insets show the Di-Gembody interfaces with the local density shown in mesh, interacting residues shown in stick, and hydrogen bonds shown with blue dashes. Distances are shown in Å. The analyses include (a) SPNS2:homoDiGb (b) RECQL5:homoDiGb (whole map and wobbling angle shown in Fig. 3a) (c) RECQL5:heteroDiGb:sfGFP (whole map and wobbling angle shown in Fig. 3c) and (d) SPNS2:heteroDiGb:sfGFP.

Extended Data Fig. 10 |. A penta-hetero-DiGb workflow that allows assembly of higher-order complexes.

(a) Construction pipeline for pentameric complexes via hetero DiGembody. 2D classification results for (b) Stx2aB alone, (c) Stx2aB in complex with newly constructed anti-Stx2aB, Gb113, (d) Stx2aB and lysozyme in complex via penta-hetero DiGb113:GbLysozyme. (e) Stx2aB and SARS-CoV2-RBD in complex via penta-hetero-DiGb113:GbRBD1. Data collected from 300 kV Krios microscope.

Fig. 1 |. Application of homoDiGbs enables high-resolution cryo-EM structure determination for dual copies of small proteins.

a, Designed Gb core substitutions (S7N;L12C;Q14K;T125M in gold) on the nanobody backbone enable chemically driven, covalent DiGb generation. The sequences are aligned using the IMGT scheme developed for immunoglobulin folds. The sequence difference between anti-GFP nanobody NbEnhancer (PDB 3K1K) and its Gb equivalent (Supplementary Fig. 5) is shown as an example. b, Schematic cartoon of the anti-GFP nanobody NbEnhancer showing the Gb substitution sites (gold) away from CDRs (pink). c, Construction pipeline for homoDiGb. BioRender.com license for the SEC column item: XM289VST7B.

Fig. 2 |. Application of homoDiGbs enables high-resolution cryo-EM structure determination for dual copies of small proteins.

a–g, High-resolution cryo-EM reconstruction maps of RECQL5 (a), SPNS2 (c), lysozyme (e) and MBP (g) in complex with their respective homoDiGbs. Local resolution comparisons before and after local refinement with or without C₂ symmetry expansion for RECQL5 (b), SPNS2 (d), lysozyme (f) and MBP (h).

Fig. 3 |. Synthesis of heteroDiGbs enables simultaneous high-resolution structure solution of two different small proteins.

Construction pipeline for heteroDiGb, through a trapped intermediate. BioRender.com license for the SEC column item: XM289VST7B. RT, room temperature.

Fig. 4 |. Synthesis of heteroDiGbs enables simultaneous high-resolution structure solution of two different small proteins.

a, High-resolution cryo-EM map of RECQL5 and sfGFP in complex with heteroDiGb Gb5–006:GbEnhancer. b,c, Individual local refinement resulted in higher resolutions of sfGFP (b) and RECQL5 (c). d, High-resolution cryo-EM map of SPNS2 and sfGFP in complex with heteroDiGb GbC4:GbEnhancer. e,f, Individual local refinement resulted in higher resolutions of SPNS2 (e) and sfGFP (f).

Fig. 5 |. Kinetic trapping allows use of even modest underpinning affinities for protein targets.

BLI measurements for RECQL5 nanobody, Gb and homoDiGb and sfGFP Gb show affinity reductions after Gb mutagenesis and generation of homoDiGbs. a–d, BLI measurement plots with fitted lines are shown in the sequence of anti-RECQL5 wild-type nanobody (a), anti-RECQL5 monomeric Gb5–006 (b), anti-RECQL5 homoDiGb5–006 (c) and anti-sfGFP monomeric GbEnhancer (d). The kinetic dissociation constants are indicated above the individual plot titles. One replicate is shown in each plot.

Fig. 6 |. Side-chain-to-side-chain-linked DiGb interfaces are tilted and asymmetric.

a–d, A 3DVA revealing the flexibility around the DiGb interfaces of RECQL5:homoDiGb (a) and RECQL5:sfGFP:heteroDiGb (c) with the respective insets (b,d) showing the degree of flexibility by wobbling angles. e, Comparison of homoDiGb to the G5–006 crystallographic pattern. The pairs are aligned to one copy of Gb5–006. The angle of deviation is indicated. f, Comparison of heteroDiGb Gb5–006:GbEnhancer to the G5–006 crystallographic pattern. The pairs are aligned to one copy of Gb5–006. The angle of deviation is indicated. g–i,The interacting residues of the cysteine–cysteine interfaces in crystallo (g), in homoDiGb5–006 (h) and heteroDiGb Gb5–006:GbEnhancer (i), with blue dashes showing hydrogen bonds and red dashes showing the minimum distances (not counting hydrogen atoms) between the P46 or G47 and Q120 pairs. j, Top-view cartoon of the cysteine–cysteine interface for all three types. k–m, Side-view cartoons of RECQL5:homoDiGb (k), in crystallo RECQL5:Gb5–006 (l) and RECQL5:heteroDiGb:sfGFP (m), with key residues annotated and the minimum distances (not counting hydrogen atoms) between the P46 or G47 and Q120 pairs in red dashes, identical to g–i.