Microchip-based structure determination of low-molecular weight proteins using cryo-electron microscopy

Abstract

Interest in cryo-Electron Microscopy (EM) imaging has skyrocketed in recent years due to its pristine views of macromolecules and materials. As advances in instrumentation and computing algorithms spurred this progress, there is renewed focus to address specimen-related challenges. Here we contribute a microchip-based toolkit to perform complementary structural and biochemical analysis on low-molecular weight proteins. As a model system, we used the SARS-CoV-2 nucleocapsid (N) protein (48 kDa) due to its stability and important role in therapeutic development. Cryo-EM structures of the N protein monomer revealed a flexible N-terminal "top hat" motif and a helical-rich C-terminal domain. To complement our structural findings, we engineered microchip-based immunoprecipitation assays that led to the discovery of the first antibody binding site on the N protein. The data also facilitated molecular modeling of a variety of pandemic and common cold-related coronavirus proteins. Such insights may guide future pandemic-preparedness protocols through immuno-engineering strategies to mitigate viral outbreaks.

Figure 1.. The use of functionalized microchips to prepare low-molecular weight proteins for cryo-EM.

(Step 1) The SARS-CoV-2 N protein in solution was validated by SDS-PAGE and Native gel analysis. N protein migrates at 50 kDa according to SimplyBlue-stained gels and western blots probed against the His-tag (IB: immunoblot). (Step 2) Silicon nitride microchips (2 mm x 2 mm frames) were coated with Ni-NTA layers (yellow), spread over an array of microwells (10 μm x 10 μm each). Etched imaging windows were 20-nm thin and the depth of each microwell was 150 nm. Microchip samples were vitrified in liquid ethane and maintained at −180°C until examined in the TEM. (Step 3) Specimens can be imaged using a variety of high-resolution instruments such as the Talos F200X, Talos F200C, or Titan TEM/STEM.

Figure 2.. Quality assessment of single particle data for N protein specimens.

(A) Images and class averages of frozen-hydrated N protein particles show consistent features from multiple views. Scale bar is 20 nm, box size is 10 nm. (B) The angular distribution plot of particle orientations lacks major limitations. (C) The Fourier shell correlation (FSC) curve and Cref (0.5) evaluation indicate a spatial resolution of 4.5-Å at the 0.143 value using the gold-standard (GS) criteria. (D) A calculated density map of the N protein at 4.5-Å resolution (yellow) is in good agreement with the experimental EM map (gray), scale bar is 10 Å.

Figure 3.. Microchip-enabled cryo-EM structure of the SARS-CoV-2 N protein.

(A) Cryo-EM map resolved to 4.5-Å shows distinctive N- and C-terminal domains with a unique “top hat” motif in the first 50-amino acids of the protein. The map was interpreted with a model for the SARS-CoV-2 N protein (red) calculated using consensus structures in the PHYRE2 server. Rotational views along with a magnified section of the map provide detailed information of the flexible (N-terminal) and rigid (C-terminal) features of the structure. The central helix in the structure from residues D216 to N228 defined a boundary between the two domains. Scale bar is 10 Å. (Movies S1, S2, ESI^†) (B) Surface rendering of the N protein in different views show patches of basic residues in the N-terminus. The C-terminal region contains 3D pockets and clefts for substrates or binding partners. Predicted epitopes mapped onto the N protein surface were evaluated according to their accessibility as high, limited or buried. Highly accessible residues are noted. (C) For nucleotide binding assays, N-protein samples were incubated with SARS-CoV-2 / PCR+ serum in PBS at 37°C for 60 minutes. Reaction mixtures were halted with sample gel loading buffer containing no SDS. Samples were assessed using native gels and immunoblots. N protein migrated at 50 kDa in mixtures lacking viral RNA (RNA−). Control reactions lacking N protein did not show background proteins. Mixtures containing N protein and PCR+ serum (+N/+RNA) showed a shift in the N protein band to a higher molecular weight. Western blots were probed with primary antibodies against the His tag on the N protein.

Figure 4.. Predicted and experimental evidence for N protein-antibody interactions.

(A) N protein samples were incubated with the Ni-NTA coated microchips, followed by serum containing IgG antibodies. Chip contents were assessed using SDS-PAGE analysis. N protein migrated at 50 kDa and was the only band present in −IgG controls. Samples with N protein and antibodies (+N/+IgG) showed bands for the IgG heavy chain (HC), light chain (LC), and the N protein. Controls lacked N protein (−N/+IgG). A separate purified IgG control sample demonstrates the manner in which IgG antibodies migrate on a denaturing gel. (B) Antibody test cassettes contain the His-tagged N protein. Within 10 minutes after applying the reaction mixture serum, a band at the control “C” region indicates a valid test. A positive band at the test region “T” (black arrow, left) indicates the presence of IgG antibodies abound to the N protein analyte. The absence of a band in the test region (right) indicates no detectable IgGs. (C) Image of the N protein incubated with patient antibodies. Class averages of antibody-bound N protein (+Abs) show density (red arrows) not present in controls (−Abs). (D) Particle orientations shows sufficient views without major limitations angular (E) The GS-FSC curve and Cref(0.5) evaluation indicate 14.2-Å resolution.

Figure 5.. Cryo-EM structure of N protein decorated with a Fab fragment from COVID-19 serum.

(A) Cryo-EM map resolved to 14.2-Å shows the placement of the N protein (red) and a corresponding model for a Fab fragment (cyan). The model for the N protein fits in one orientation within the map with the C-terminal domain adjacent to the antigen-binding domain in the Fab model. The flexible loop comprised of residues Q384 – A397 is proximal to the Fab-binding site. Rotational views of the map provide visual clarity of the physical relationship between the two models. Scale bar is 10 Å. (B) Cross-sections through the reconstruction indicate a high-quality fit of the models from side and top views. Sections through the map represent slices produced at ~10-Å increments (Movies S5, S6, ESI^†).

Figure 6.. Structural comparisons of N protein models.

(A) N protein models were calculated using the PHYRE2 server and multiple consensus templates with the highest structural correlational values. The N- and C-terminal domains are unique to SARS-CoV-2 in comparison to other N proteins from human pandemic and common cold strains. Percent sequence identities were generated according to multi-sequence alignment tools and included: Civet, 89.5%; SARS-CoV, 89.7%; MERS 48.6%, OC43, 35.9% HKU1, 35.7%, NL63, 27.9%, and 229E, 25.2%. Proteins were aligned visually to highlight similarities and differences among the predicted structures. Structures are oriented with the N-terminus at the vertical top, the C-terminus at the vertical bottom, and putative antibody epitopes on the right. (B) Structural dendrogram indicates similarities between N protein models, demonstrated in the specified branches and groupings by the DALI protein server. Alpha-coronaviruses (229E and NL63) appear on similar branches and in long range opposition to beta-coronaviruses (OC43 and HKU1). Pandemic beta-coronaviruses, SARS-CoV and SARS-CoV-2, are located in the central branch. This proximity suggests a mixture of features are represented in the structures.