Delineating Conformational Variability in Small Protein Structures Using Combinatorial Refinement Strategies

Abstract

As small protein assemblies and even small proteins are becoming more amenable to cryo-Electron Microscopy (EM) structural studies, it is important to consider the complementary dynamic information present in the data. Current computational strategies are limited in their ability to resolve minute differences among low molecular weight entities. Here, we demonstrate a new combinatorial approach to delineate flexible conformations among small proteins using real-space refinement applications. We performed a meta-analysis of structural data for the SARS CoV-2 Nucleocapsid (N) protein using a combination of rigid-body refinement and simulated annealing methods. For the N protein monomer, we determined three new flexible conformers with good stereochemistry and quantitative comparisons provided new evidence of their dynamic properties. A similar analysis performed for the N protein dimer showed only minor structural differences among the flexible models. These results suggested a more stable view of the N protein dimer than the monomer structure. Taken together, the new computational strategies can delineate conformational changes in low molecular weight proteins that may go unnoticed by conventional assessments. The results also suggest that small proteins may be further stabilized for structural studies through the use of solution components that limit the movement of external flexible regions.

1. Introduction

Biomedical research improves our understanding of human health and disease through the development of new technologies. Undeniably, high-resolution imaging has transformed our view of the nanoworld, permitting us to investigate an enormous range of life’s processes. Researchers can now examine in exquisite detail individual protein structures, multi-component machines, and minute pathogens that cause global pandemics [1,2,3]. To fully understand how macromolecules operate in concert, however, their dynamic attributes must be jointly investigated. One caveat with current cryo-Electron Microscopy (EM) practices is that biological entities must be preserved in a frozen environment to acquire sub-nanometer information. Static snapshots of flexible assemblies reveal limited knowledge of their marvelous complexity. Hence, shedding light on the conformational variability among cryo-EM structures can provide a more comprehensive view of their native activities [4].

In recent years, a variety of algorithms have been developed to extract heterogeneous states from large EM data sets as a means to analyze molecular flexibility [5,6,7,8]. For smaller entities, such as individual proteins or RNA structures, identifying subtle differences using standard classification routines can be more challenging. Others have shown that multiple conformers can be modeled within the same EM map to broaden our interpretation of biologically relevant data [7,9]. Molecular models can be calculated through automated routines or by iterative refinement procedures in concert with validation assessments [10,11,12,13,14,15,16,17,18,19,20].

To better determine conformational nuances in small flexible protein structures, we employed our recently determined SARS-CoV-2 Nucleocapsid (N) protein data derived from either a recombinant expression system or from COVID-19 patients [21,22]. These structures were of interest as we aimed to identify changes in flexible regions that may impact their detection in rapid tests for SARS-CoV-2 [23,24]. Importantly, the N protein is the prime target in nearly all of the rapid tests sold in the United States.

Here, we implemented a unique combination of real-space refinement routines coupled with simulated annealing steps to delineate multiple conformations within the confines of the cryo-EM data. Results showed that multiple flexible models could fit well into the EM maps of the N protein monomer and dimer structures. These new interpretations provided a better depiction of the conformational variability present in the protein samples. As each domain of the N protein contains flexible regions, we posit that the dimerization process may reduce the mobility of the N protein in solution. Additionally, the new models suggested that lowering the mobility of disordered loops in the proteins likely favors interactions with binding partners such as immune molecules. Improving knowledge of the mobile parts of virus-related proteins in general can aid in mechanistic drug targeting, rapid detection, or even novel vaccine design and application.

2. Results

2.1. Evaluating Flexible Conformers of the N Protein Monomer

We performed a meta-analysis using EM data (EMD-29002; PDB, 8FD5) for the SARS-CoV-2 N protein monomer (~50 kDa) [22]. Our goal was to test if multiple models were represented in the data and to better define conformational variability in the N protein structure. The EM map was binned to reduce high frequency noise and re-sampled at 1.85 Å/pixel. The re-sampled map was imported into the PHENIX software package (version 1.20.1-4487) [17,19] and auto-boxed as part of the auto-sharpening routine. Auto-sharpening was implemented using a resolution limit of 4.5 Å. The corresponding N protein model was imported into PHENIX at the same origin as the EM map. The model was subjected to iterative rounds of rigid-body refinement with and without simulated annealing routines using standard procedures. Upon convergence, two flexible conformations were identified that fit well within the map (Figure 1 and Figure 2). The new models were further evaluated in terms of stereochemistry and refinement statistics (Table A1).

Flexible conformation 1 was calculated using rigid-body refinement along with simulated annealing and energy minimization routines. The overall fit of the output model within the map showed ~90% occupancy of the c-α backbone and side chain residues, although not all side chains were pristinely defined (Figure 1A,B). Spatial resolution (~4.4 Å) was estimated at the FSC-0.5 value (red line, Figure 1C) for re-sampled half maps employing the Protein Data Bank (PDB) validation server and the Electron Microscopy Data Bank (EMDB) Fourier Shell Correlation (FSC) server. Both FSC curves are shown for comparison in Figure 1C. Map-model resolvability (Q-score) [25] and atom inclusion values were 0.2540 and 0.9020, respectively (Figure 1D). Additional output from PHENIX for the model to map fit cross-correlation included CC_mask (0.5585), CC_volume (0.5593), and CC_peaks (0.4772). The overall MolProbity score [26,27] for flexible conformation 1 was 2.50 and no Ramachandran outliers were identified (Table A1).

Flexible conformation 2 was calculated in PHENIX using rigid-body refinement procedures without simulated annealing and including energy minimization. The second flexible model showed slightly less occupancy (~82%) of the c-α backbone and side chain residues within the same map (Figure 2A,B).

The spatial resolution estimates are again highlighted in Figure 2C. The Q-score (0.1830) and atom inclusion value (0.8200) were lower for the second flexible model than for the first refined model (Figure 2D). The model to map fit cross-correlation output from PHENIX included CC_mask (0.4409), CC_volume (0.4438), and CC_peaks (0.3708). The MolProbity score for flexible conformation 2 was 2.18, which was better than the first conformation. No Ramachandran outliers were identified and the all-atom clashscore for flexible conformation 2 was ~14, also an improvement over flexible conformation 1 (Table A1). Although the map–model correlation for flexible conformation 2 was slightly lower, the overall stereochemistry for the second model was more favorable than the first model.

In an effort to potentially reduce close contacts amongst the output models, the original N monomer structure was then refined into the EM map re-sampled at a pixel size of 2 Å, rather than 1.85 Å (Figure 3). Flexible conformation 3 was calculated in PHENIX using rigid-body refinement, simulated annealing, and energy minimization. The third flexible model had 86% occupancy of the c-α backbone and side chain residues, consistent with the previous two models (Figure 3A,B). Spatial resolution was estimated at the FSC-0.5 value (red line, Figure 3C) using re-sampled half maps and employing the PDB validation server and the EMDB server. The resolution estimate for the new map was ~4.85 Å.

The Q-score (0.1920) and atom inclusion values (0.8640) were higher than flexible conformation 2, but lower than flexible conformation 1 (Figure 3D). The model to map fit cross-correlation values from PHENIX included CC_mask (0.4700), CC_volume (0.4799), and CC_peaks (0.3590). The MolProbity score for flexible conformation 3 was 2.58 and no Ramachandran outliers were identified (Table A1). The all-atom clashscore was ~14, the same value as conformation 2, which was still better than conformation 1 (17). The overall stereochemistry values for the third model were similar to the first one, and the Q-score suggested that the quality of the model fit of flexible conformation 3 was on par with the other two conformations. Therefore, increasing the map sampling by ~10% did relieve some close contacts during refinement. It is also important to note that in each test case, the use of simulating annealing alone without rigid-body refinement protocols produced models with poor stereochemistry and a large clashscore.

As an additional measure, we tested the fit of flexible conformation 1 with the original N protein map (EMD-29002) [22]. The map was evaluated without binning at 0.93 Å/pixel and assessments were performed at a contour level of 0.9000. At this contour value, ~80% of c-α backbone atoms and non-hydrogen atoms were located inside the map. The corresponding Q-score and atom inclusion values were 0.2410 and 0.7980, respectively. Although these numbers were slightly lower than the same model fit into the re-sampled EM map, they meet expectations for interpreting intermediate resolution structures in the range of 4.5–5 Å [10,25]. This check also ensured the new model was consistent with the uncompressed EM data, even though better overall modeling statistics were achieved using the maps calculated with minimal high frequency noise.

To better understand the relationship between the three newly refined models, flexible conformations 1–3, we performed a side-by-side comparison displayed in Figure 4. Models were aligned using the Structure Comparison/MatchMaker tool in the Chimera Software package (version 1.15) [28]. Continuous rainbow rendering was used to highlight the residue progression in the model from the N-terminus (blue) to the C-terminus (red). Previously identified features of the N protein monomer include the RNA-binding region, a dimerization site, and a key antibody epitope [22,29,30,31,32,33,34,35,36,37,38,39].

The three new flexible monomer models showed minor differences throughout each domain with the most variable region located in the N-terminal area. In addition, conformational differences were also noted within the RNA binding site, likely due to its need to accommodate RNA substrates and perform its biological duties of genome packaging and assembly. As there are generally few secondary structural elements in the N protein, some degree of flexibility within the full-length protein is expected. Implications in the methods are that simulated annealing routines augment rigid-body assessments to help reveal newly variable states that are representative of the protein’s natural energy landscape.

To quantify the differences between the two complete models (flexible conformation 1 and 3) we used the Morph Conformations function in Chimera [28] to superimpose the two models (Figure 5A). Flexible conformation 1 (blue) served as a reference model and flexible conformation 3 (yellow) was the test model superimposed upon the reference. Morph parameters included the corkscrew interpolation method and a linear interpolation rate including 20 steps with a core fraction of 0.5, along with 60 minimization steps.

The output frames from the morphed movies were compiled and looped together using Apple iMovie (Apple Inc., Cupertino, CA, USA) (Supplementary Movies S1 and S2). Representative snapshots from the movies show the transition from the superimposed models to the fully separated models (Figure 5, steps 1–6). The root mean square deviation (RMSD) for the fully aligned models was 0.186 Å and the RMSD for the fully separated models was 3.369 Å, with a mid-point RMSD value of 1.778 Å.

An additional comparison between flexible conformation 2 and the original structure (PDB code, 8FD5) yielded a RMSD value of 0.205 Å for the two aligned models and 3.818 Å for the separated models. The mid-point RMSD value was 2.011 Å. Overall, the conformational variability observed for the new models suggested the N protein structure is comprised of dynamic domains needed for proper protein function.

2.2. Understanding Flexible Nuances in the N Protein Dimer

To better understand flexible model interpretations for the N protein dimer, we applied the same meta-analysis approach to previous EM data (EMD-29027; PDB, 8FG2) [22]. Since the molecular weight of the N protein dimer (~100 kDa) is much larger than the monomer (~50 kDa), map features were more clearly delineated from background noise, and it was not necessary to bin the map prior to refinement procedures. The map was imported into PHENIX and auto-boxed while implementing the auto-sharpening routine at a resolution limit of 5.5 Å. The corresponding N protein dimer model was imported into PHENIX at the same origin as the EM map and subjected to iterative real-space refinement routines. Similar to the N protein monomer, two conformations were identified for the dimer structure that fit well within the map (Figure 6 and Figure 7). Model refinement statistics are provided in Table A1.

First, the N dimer structure was subjected to multiple rounds of rigid-body refinement along with simulated annealing and energy minimization (Figure 6A,B).Spatial resolution (~5.4 Å) was determined at the FSC-0.5 value (red line, Figure 6C) using the EMDB FSC server. The PDB validation server slightly over-estimated the resolution value. The overall Q-score and atom inclusion values were 0.1250 and 0.6950, respectively (Figure 6D). The model to map fit cross-correlation output from PHENIX included CC_mask (0.5719), CC_volume (0.5866), and CC_peaks (0.4644). The overall MolProbity score for flexible conformation 1 of the N protein dimer was 2.51 and no Ramachandran outliers were identified. In addition, the all-atom clashscore was ~14 (Table A1).

A second flexible conformation was calculated in PHENIX using only rigid-body refinement procedures and energy minimization. Simulated annealing was not employed for this test case. Flexible conformation 2 was analyzed using 95% of Chain A and 86% of Chain B, including the c-α backbone and side chain residues (Figure 7A,B). The resolution estimate of the map is again highlighted in Figure 7C. The Q-score (0.1210) and atom inclusion values (0.6920) were slightly lower for the model calculated without simulated annealing (Figure 7D). The model to map fit cross-correlation values from PHENIX included CC_mask (0.4930), CC_volume (0.4981), and CC_peaks (0.4157). These values were also lower for the second model than for flexible conformation 1. The MolProbity score for flexible conformation 2 was 2.50, very similar to flexible conformation 1 (2.51), and no Ramachandran outliers were identified (Table A1). The all-atom clashscore was ~14, the same for both conformers. Since both models had an equivalent number of close contacts, we did not further test re-sampling of the EM map at a different pixel size. As noted for the N protein monomer, the use of simulating annealing alone without rigid-body refinement protocols yielded output models with poor stereochemistry and unreasonably large clashscores.

A side-by-side comparison of the two new conformers is shown in Figure 8. The models were aligned in Chimera using the same Structure Comparison/MatchMaker tools implemented for the N protein monomer. Continuous rainbow rendering highlights the transition from the N-terminus to the C-terminus. Features of the dimer structure are labeled, including the RNA-binding site, dimerization site, and an antibody epitope. The labeled antibody site was still freely available to interact with immune molecules in both flexible dimer structures (Figure 8).

Superimposition procedures for flexible conformations 1 and 2 were performed using the Morph Conformations function in Chimera (Figure 9, top panel). The reference model (flexible conformation 1, blue) was held in place while the test model, flexible conformation 2 (yellow), was superimposed upon the reference model. The morphing routine used the corkscrew interpolation method with a linear interpolation rate of 20 steps and a core fraction of 0.5, along with 60 minimization steps.

The output frames from the morphing procedure were compiled and looped using Apple iMovie (Supplementary Movies S3 and S4). Snapshots of the process showed only subtle differences between the aligned and separated states, highlighted in the magnified region of Figure 9. The RMSD value for the aligned models was 0.077 Å and for the separated models was 0.576 Å. The mid-point RMSD was 0.326 Å, indicating that the dimer models had little variability, especially compared to the monomer structures.

3. Discussion

Here we sought to understand the dynamic properties of small protein structures that are difficult to discern via standard classification methods. A meta-analysis was performed on the protein models using different real space refinement procedures executed in the PHENIX software package (version 1.20.1-4487) [17]. For the N protein monomer, we identified three closely related flexible conformers that each fit well within the EM maps and had good stereochemistry. Better map-to-model agreement was achieved using rigid-body refinement routines that incorporated simulated annealing and energy minimization steps. One caveat with this approach is that additional close contacts between amino acids were introduced into the models when simulated annealing was utilized as indicated by slightly higher clashscore values in MolProbity assessments. These contacts could be reduced to some extent by increasing the map sampling by ~10%. Other strategies such as molecular dynamics-based flexible fitting may also help alleviate close contacts during refinement [5,12,14,15,16,20].

A direct comparison of N dimer models showed only minor differences when implementing simulated annealing steps during refinement. Unlike the monomer models, the dimer models had an equivalent number of close contacts. It was, however, still beneficial to incorporate simulated annealing protocols into the refinement routines for the N protein dimer as it led to improved model statistics. Although two conformations were identified for the N protein dimer, RMSD values suggested minimal differences among the models (0.576 Å).

By contrast, RMSD values for the two complete monomer structures varied by a greater degree (3.369 Å). These results were somewhat unexpected, considering the N protein dimer was isolated from the serum of COVID-19 patients and the N protein monomer was recombinantly expressed in bacteria. The natively sourced N protein from patients likely contains a heterogenous distribution of post-translational modifications. Other teams have identified modifications to the N protein such as phosphorylation, ADP-ribosylation, and glycosylation [40,41,42]. The presence of these modifications may influence the accuracy of model fitting and account for the lower Q-score values observed for the dimer structures derived from patients, compared to the recombinant monomer structures. Mass spectrometry experiments are underway to better define these modifications and their impact on protein structure and function.

In general, the limited variability noted in the dimer structures painted a more stable picture of the N protein dimer than the monomer. Results from this work also suggest that solution additives that stabilize flexible external loops among protein structures may better preserve their native architectures. Equally important, the use of these collective modeling procedures can reveal meaningful conformational changes that go unnoticed by conventional structural determination strategies.

Overall, we found the outcomes to be compelling and exciting for interpreting small protein structures < 100 Å in length at intermediate resolutions in the range of 4–6 Å. Due to the intrinsic dynamics of the SARS-CoV-2 N protein, high-resolution experimental maps below 3 Å are not currently available, hence the presented approach aims to extend the analysis and interpretation of cryo-EM data within this intermediate resolution regime. Until more high-resolution cryo-EM structures are available for low molecular weight entities between 50 and 100 kDa, the community is left dealing with less resolved maps not readily determined at 3 Å or better.

An additional factor is that high-resolution structures show less variability due to more precisely determined features; therefore, flexible regions are naturally absent in well-resolved EM maps. As such, we envision the development of new computational strategies for interpreting molecular nuances and flexible states at intermediate resolutions (4–6 Å range) to be a positive step forward, where our approach is most useful. Collectively, we found that simulated annealing methods incorporated into real-space refinement procedures elevated model interpretations of the data and provided new information regarding the conformational variability for Nucleocapsid proteins of interest.

4. Materials and Methods

Protein sources. The recombinant N protein monomer sample is commercially available from RayBiotech Inc. (Peachtree Corners, USA) (230-01104-100). The N protein dimer was isolated from serum samples of COVID-19 patients that tested PCR+ and they are commercially available from RayBiotech Inc. (Peachtree Corners, USA) (CoV-PosPCR-S-100). Procedures to isolate the native N dimer protein have been demonstrated in prior work [22].

EM data and initial modeling. Specimen preparation methods, EM data collection parameters, and initial modeling procedures have been previously described [22]. A summary of map parameters is included in Table A1.

Real-space refinement routines. EM maps and model coordinates for the N protein monomer (EMD-29002; PDB, 8FD5) and dimer structures (EMD-29072; PDB 8FG2) were used for this analysis. The N monomer map was binned and re-sampled at 1.85 and 2.0 Å/pixel using PROC3D operating in the EMAN2 software package (version 1.9) [43]. The re-sampled maps were imported into PHENIX [19] and auto-boxed during import as part of the auto-sharpening procedure, implementing a resolution limit of 4.5 Å. A resolution cutoff of 4 Å was applied to the maps used for model refinement. The N monomer model was fit into the re-sampled maps in PHENIX and subjected to 5 macrocycles of rigid-body refinement with and without simulated annealing. Iterative rounds of refinement and energy minimization were implemented until convergence. Post refinement, a circular mask was applied to the maps using a radius of 27 pixels. The N dimer map was directly imported into PHENIX at a sampling value of 1.4 Å/pixel. The map was auto-boxed during import as part of the auto-sharpening procedure, which was implemented at a resolution limit of 5.5 Å. A resolution cutoff of 5 Å was applied to the maps that were used for subsequent model refinement. The N dimer model was subjected to 5 macrocycles of rigid-body refinement with and without simulated annealing. Iterative rounds of rigid-body refinement and energy minimization were implemented until convergence. Q-scores and atom inclusion values were calculated using the PDB validation server.

Model alignment and superimposition procedures. Flexible models of the N protein monomer were aligned using the Structure Comparison/MatchMaker tool in the Chimera software package [28]. Root mean square deviation (RMSD) measurements for the flexible models were calculated using the Morph Conformations function in Chimera. The reference model was held in place while the test model was superimposed upon the reference. Morphing parameters included the corkscrew interpolation method with a linear interpolation rate of 20 steps and a core fraction of 0.5, along with 60 minimization steps. Output frames from the morphed movies were compiled and looped together using Apple iMovie (Apple Inc., Cupertino, California). RMSD values were output via the Chimera program per movie frame. The same alignment and superimposition procedures were implemented in the Chimera program to compare the N protein dimer models.

Author Contributions

Conceptualization, D.F.K., G.M.J., M.J.S., S.B. and L.-A.D.; methodology, D.F.K., W.D., L.K. and M.C.; software, D.F.K., G.M.J. and W.D.; validation, D.F.K. and M.C.; investigation, G.M.J., L.K. and M.C.; writing: D.F.K., L.-A.D. and M.J.S.; supervision, D.F.K.; funding acquisition, D.F.K. All authors have read and agreed to the published version of the manuscript.

Data Availability Statement

Molecular structures related to this work have been deposited in the Protein Data Bank and will be released upon publication. Additional data may be made available upon request at the discretion of the principal investigator.

Conflicts of Interest

The authors declare no conflict of interest.

Footnotes

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Figures

Figure 1. Evaluating flexible conformation 1 determined for the N protein monomer. (A) The refined model had ~90% occupancy of the c-α backbone and side chain residues within the EM map sampled at 1.85 Å/pixel during model fitting. Side, top, and bottom views of the fit model are displayed. Scale bar is ~10 Å. (B) Slices through the map and model (1–3) highlight some of the fit residues. (C) Spatial resolution (~4.4 Å) was estimated at the FSC-0.5 value (red line) for half map comparisons employing the PDB validation server and the EMDB server. Both curves are shown for comparison. Half-bit criterion is designated by the purple line, and the green line is the 0.143 cutoff. (D) The Q-score and atom inclusion values were 0.2540 and 0.9020 and are shown mapped to the model with scale values ranging from 0.0 (red) to 1.0 (cyan).

Figure 2. Assessing flexible conformation 2 determined for the N protein monomer. (A) The second refined model was fitted into the EM map sampled at 1.85 Å/pixel with ~82% occupancy of the c-α backbone and side chain residues. Side, top, and bottom views are displayed. Scale bar is ~10 Å. (B) Slices through the map and model (1–3) highlight some of the fit residues. (C) Spatial resolution (~4.4 Å) was estimated according to the FSC-0.5 value (red line) calculated using the PDB validation server and the EMDB server. Half-bit criterion is designated by the purple line, and the green line is the 0.143 cutoff. (D) The Q-score and atom inclusion values were 0.1830 and 0.8200 and are shown mapped to the model with scale values ranging from 0.0 (red) to 1.0 (cyan).

Figure 3. Model evaluation for flexible conformation 3 determined for the N protein monomer. (A) The third refined model was fitted into the EM map sampled at 2.0 Å/pixel with ~86% occupancy of the c-α backbone and side chain residues. Side, top, and bottom views are displayed. Scale bar is ~10 Å. (B) Slices through the map and model (1–3) highlight some of the fit residues in the map. (C) Spatial resolution (~4.85 Å) was estimated at the FSC-0.5 value (red line) calculated using the PDB validation server and the EMDB server. Half-bit criterion is designated by the purple line, and the green line is the 0.143 cutoff. (D) Q-score and atom inclusion values were 0.1920 and 0.8640 and are shown mapped to the model with scale values ranging from 0.0 (red) to 1.0 (cyan).

Figure 4. Side-by-side comparison of the three new flexible N monomer models. The newly refined output models were aligned using the Structure Comparison/MatchMaker tool in the Chimera software package. Models are displayed in their side, top, and bottom views. Rainbow rendering was used to highlight residue progression in each model starting from the N-terminus (blue) to the C-terminus (red). Features of the N protein monomer have been previously identified and include an RNA-binding region, a dimerization site, and a key antibody epitope.

Figure 5. Comparing flexible conformations among N monomer models. (A) Flexible conformation 1 (blue) was used as the reference model and flexible conformation 3 (yellow) was used as the test model. The test model was superimposed upon the reference model using the Morph Conformations function in Chimera. (B) Representative snapshots of the process show the transition from the aligned models to the fully separated models during the procedure (steps 1–6). The RMSD for the fully aligned models was 0.186 Å and the RMSD for the fully separated models was 3.369 Å, with a mid-point RMSD of 1.778 Å.

Figure 6. Evaluating flexible conformation 1 determined for the N protein dimer. (A) The refined model of the N protein dimer was fitted into the EM map sampled at 1.4 Å/pixel. Side, top, and bottom views of the new model are shown. Scale bar is ~10 Å. (B) Slices through the map and model (1–3) highlight some of the fit residues within the map. (C) Spatial resolution (~5.4 Å) was estimated at the FSC-0.5 value (red line) for half map comparisons employing the EMDB FSC server. The PDB server slightly over-estimated the resolution. Half-bit criterion is designated by the purple line, and the green line is the 0.143 cutoff. (D) Overall Q-score (0.1250) and atom inclusion values (0.6950) are shown mapped to the model. Scale ranges from 0.0 (red) to 1.0 (cyan).

Figure 7. Assessing flexible conformation 2 determined for the N protein dimer. (A) The new model of the N dimer is shown within the EM map sampled at 1.4 Å/pixel during model fitting and displays side, top, and bottom views. Scale bar is ~10 Å. (B) Slices through the fit model (1–3) highlight some residues within the map. (C) Spatial resolution (~5.4 Å) was estimated at the FSC-0.5 value (red line) employing the EMDB FSC server. The PDB server slightly over-estimated the resolution. Half-bit criterion is designated by the purple line, and the green line is the 0.143 cutoff. (D) Overall Q-score (0.1210) and atom inclusion values (0.6920) were mapped to the model with scale values ranging from 0.0 (red) to 1.0 (cyan).

Figure 8. Side-by-side comparison of the refined flexible N dimer models. The new flexible models were aligned using the Chimera software package and the same tools employed for the N protein monomer comparisons. Different views of the models are displayed. Rainbow rendering was used to highlight residue progression in each model starting from the N-terminus (blue) to the C-terminus (red).

Figure 9. Comparing flexible N dimer models and their RMSD values. The same alignment and superimposition routines were used to evaluate the N dimer models. Flexible conformation 1 (blue) served as the reference model and flexible conformation 2 (yellow) served as the test model. The test model was superimposed upon the reference model using the Morph Conformations function in Chimera. The separated models showed very little variation. Some minor differences were noted in the magnified region (black arrows). The RMSD for the aligned models was 0.077 Å and the RMSD for the fully separated models was 0.576Å, with a mid-point RMSD value of 0.326 Å.

Supplementary Materials

The following supporting information can be downloaded at https://www.mdpi.com/article/10.3390/mi14101869/s1, Title: “Supplemental reports”, Movie S1: superimposition of the N protein monomer models. Movie S2: conformational flexibility in the N protein monomer. Movie S3: superimposition of the N protein dimer models. Movie S4: magnified view of a flexible region in the N protein dimer.

References

1. Walls, A.C.; Park, Y.-J.; Tortorici, M.A.; Wall, A.; McGuire, A.T.; Veesler, D. Structure, Function, and Antigenicity of the SARS-CoV-2 Spike Glycoprotein. Cell; 2020; 181, pp. 281-292.e6. [DOI: https://dx.doi.org/10.1016/j.cell.2020.02.058] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/32155444]

2. Barnes, C.O.; West, A.P., Jr.; Huey-Tubman, K.E.; Hoffmann, M.A.G.; Sharaf, N.G.; Hoffman, P.R.; Koranda, N.; Gristick, H.B.; Gaebler, C.; Muecksch, F. et al. Structures of Human Antibodies Bound to SARS-CoV-2 Spike Reveal Common Epitopes and Recurrent Features of Antibodies. Cell; 2020; 182, pp. 828-842.e16. [DOI: https://dx.doi.org/10.1016/j.cell.2020.06.025] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/32645326]

3. Lan, J.; Ge, J.; Yu, J.; Shan, S.; Zhou, H.; Fan, S.; Zhang, Q.; Shi, X.; Wang, Q.; Zhang, L. et al. Structure of the SARS-CoV-2 Spike Receptor-Binding Domain Bound to the ACE2 Receptor. Nature; 2020; 581, pp. 215-220. [DOI: https://dx.doi.org/10.1038/s41586-020-2180-5]

4. Murata, K.; Wolf, M. Cryo-Electron Microscopy for Structural Analysis of Dynamic Biological Macromolecules. Biochim. Biophys. Acta (BBA) Gen. Subj.; 2018; 1862, pp. 324-334. [DOI: https://dx.doi.org/10.1016/j.bbagen.2017.07.020]

5. Tjioe, E.; Lasker, K.; Webb, B.; Wolfson, H.J.; Sali, A. MultiFit: A Web Server for Fitting Multiple Protein Structures into Their Electron Microscopy Density Map. Nucleic Acids Res.; 2011; 39, pp. W167-W170. [DOI: https://dx.doi.org/10.1093/nar/gkr490] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/21715383]

6. DiMaio, F.; Song, Y.; Li, X.; Brunner, M.J.; Xu, C.; Conticello, V.; Egelman, E.; Marlovits, T.C.; Cheng, Y.; Baker, D. Atomic-Accuracy Models from 4.5-Å Cryo-Electron Microscopy Data with Density-Guided Iterative Local Refinement. Nat. Methods; 2015; 12, pp. 361-365. [DOI: https://dx.doi.org/10.1038/nmeth.3286] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/25707030]

7. Pintilie, G.; Chen, D.-H.; Haase-Pettingell, C.A.; King, J.A.; Chiu, W. Resolution and Probabilistic Models of Components in CryoEM Maps of Mature P22 Bacteriophage. Biophys. J.; 2015; 110, pp. 827-839. [DOI: https://dx.doi.org/10.1016/j.bpj.2015.11.3522]

8. Herzik, M.A.; Fraser, J.S.; Lander, G.C. A Multi-Model Approach to Assessing Local and Global Cryo-EM Map Quality. Structure; 2018; 27, pp. 344-358.e3. [DOI: https://dx.doi.org/10.1016/j.str.2018.10.003]

9. Topf, M.; Baker, M.L.; Marti-Renom, M.A.; Chiu, W.; Sali, A. Refinement of Protein Structures by Iterative Comparative Modeling and CryoEM Density Fitting. J. Mol. Biol.; 2006; 357, pp. 1655-1668. [DOI: https://dx.doi.org/10.1016/j.jmb.2006.01.062]

10. Pintilie, G.; Chiu, W. Validation, Analysis and Annotation of Cryo-EM Structures. Acta Crystallogr. Sect. D Struct. Biol.; 2021; 77, pp. 1142-1152. [DOI: https://dx.doi.org/10.1107/S2059798321006069]

11. Rosenthal, P.B.; Rubinstein, J.L. Validating Maps from Single Particle Electron Cryomicroscopy. Curr. Opin. Struct. Biol.; 2015; 34, pp. 135-144. [DOI: https://dx.doi.org/10.1016/j.sbi.2015.07.002] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/26605834]

12. Trabuco, L.G.; Villa, E.; Schreiner, E.; Harrison, C.B.; Schulten, K. Molecular Dynamics Flexible Fitting: A Practical Guide to Combine Cryo-Electron Microscopy and X-ray Crystallography. Methods; 2009; 49, pp. 174-180. [DOI: https://dx.doi.org/10.1016/j.ymeth.2009.04.005] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/19398010]

13. Afonine, P.V.; Klaholz, B.P.; Moriarty, N.W.; Poon, B.K.; Sobolev, O.V.; Terwilliger, T.C.; Adams, P.D.; Urzhumtsev, A. New Tools for the Analysis and Validation of Cryo-EM Maps and Atomic Models. Acta Crystallogr. Sect. D Struct. Biol.; 2018; 74, pp. 814-840. [DOI: https://dx.doi.org/10.1107/S2059798318009324] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/30198894]

14. Trabuco, L.G.; Schreiner, E.; Gumbart, J.; Hsin, J.; Villa, E.; Schulten, K. Applications of the Molecular Dynamics Flexible Fitting Method. J. Struct. Biol.; 2011; 173, pp. 420-427. [DOI: https://dx.doi.org/10.1016/j.jsb.2010.09.024] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/20932910]

15. DiMaio, F.; Tyka, M.D.; Baker, M.L.; Chiu, W.; Baker, D. Refinement of Protein Structures into Low-Resolution Density Maps Using Rosetta. J. Mol. Biol.; 2009; 392, pp. 181-190. [DOI: https://dx.doi.org/10.1016/j.jmb.2009.07.008] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/19596339]

16. Ian Croll, T. ISOLDE: A Physically Realistic Environment for Model Building into Low-resolution Electron-density Maps. Acta Crystallogr. Sect. D Struct. Biol.; 2018; 74, pp. 519-530. [DOI: https://dx.doi.org/10.1107/S2059798318002425] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/29872003]

17. Adams, P.D.; Afonine, P.V.; Bunkóczi, G.; Chen, V.B.; Davis, I.W.; Echols, N.; Headd, J.J.; Hung, L.-W.; Kapral, G.J.; Grosse-Kunstleve, R.W. et al. PHENIX: A Comprehensive Python-Based System for Macromolecular Structure Solution. Acta Crystallogr. Sect. D Biol. Crystallogr.; 2010; 66, pp. 213-221. [DOI: https://dx.doi.org/10.1107/S0907444909052925]

18. DiMaio, F.; Echols, N.; Headd, J.J.; Terwilliger, T.C.; Adams, P.D.; Baker, D. Improved Low-Resolution Crystallographic Refinement with Phenix and Rosetta. Nat. Methods; 2013; 10, pp. 1102-1104. [DOI: https://dx.doi.org/10.1038/nmeth.2648]

19. Adams, P.D.; Afonine, P.V.; Bunkóczi, G.; Chen, V.B.; Davis, I.W.; Echols, N.; Headd, J.J.; Hung, L.-W.; Kapral, G.J.; Grosse-Kunstleve, R.W. et al. PHENIX: A Comprehensive Python-Based System for Macromolecular Structure Solution. Int. Tables Crystallogr.; 2012; 66, Pt 2, pp. 539-547. [DOI: https://dx.doi.org/10.1107/97809553602060000865]

20. Trabuco, L.G.; Villa, E.; Mitra, K.; Frank, J.; Schulten, K. Flexible Fitting of Atomic Structures into Electron Microscopy Maps Using Molecular Dynamics. Structure; 2008; 16, pp. 673-683. [DOI: https://dx.doi.org/10.1016/j.str.2008.03.005]

21. Casasanta, M.A.; Jonaid, G.M.; Kaylor, L.; Luqiu, W.Y.; Solares, M.J.; Schroen, M.L.; Dearnaley, W.J.; Wilson, J.; Dukes, M.J.; Kelly, D.F. Microchip-Based Structure Determination of Low-Molecular Weight Proteins Using Cryo-Electron Microscopy. Nanoscale; 2021; 13, pp. 7285-7293. [DOI: https://dx.doi.org/10.1039/D1NR00388G] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/33889923]

22. Casasanta, M.A.; Jonaid, G.M.; Kaylor, L.; Luqiu, W.Y.; DiCecco, L.-A.; Solares, M.J.; Berry, S.; Dearnaley, W.J.; Kelly, D.F. Structural Insights of the SARS-CoV-2 Nucleocapsid Protein: Implications for the Inner-Workings of Rapid Antigen Tests. Microsc. Microanal.; 2022; 29, pp. 649-657. [DOI: https://dx.doi.org/10.1093/micmic/ozac036] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/37749713]

23. Mahmoudinobar, F.; Britton, D.; Kim Montclare, J. Protein-Based Lateral Flow Assays for COVID-19 Detection. Protein Eng. Des. Sel.; 2021; 34, gzab010. [DOI: https://dx.doi.org/10.1093/protein/gzab010] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/33991088]

24. Burbelo, P.D.; Riedo, F.X.; Morishima, C.; Rawlings, S.; Smith, D.; Das, S.; Strich, J.R.; Chertow, D.S.; Davey, R.T.; I Cohen, J. Sensitivity in Detection of Antibodies to Nucleocapsid and Spike Proteins of Severe Acute Respiratory Syndrome Coronavirus 2 in Patients with Coronavirus Disease 2019. J. Infect. Dis.; 2020; 222, pp. 206-213. [DOI: https://dx.doi.org/10.1093/infdis/jiaa273]

25. Pintilie, G.; Zhang, K.; Su, Z.; Li, S.; Schmid, M.F.; Chiu, W. Measurement of Atom Resolvability in Cryo-EM Maps with Q-Scores. Nat. Methods; 2020; 17, pp. 328-334. [DOI: https://dx.doi.org/10.1038/s41592-020-0731-1]

26. Chen, V.B.; Arendall, W.B., III; Headd, J.J.; Keedy, D.A.; Immormino, R.M.; Kapral, G.J.; Murray, L.W.; Richardson, J.S.; Richardson, D.C. MolProbity: All-Atom Structure Validation for Macromolecular Crystallography. Acta Crystallogr. D Biol. Crystallogr.; 2010; 66, pp. 12-21. [DOI: https://dx.doi.org/10.1107/S0907444909042073]

27. Williams, C.J.; Headd, J.J.; Moriarty, N.W.; Prisant, M.G.; Videau, L.L.; Deis, L.N.; Verma, V.; Keedy, D.A.; Hintze, B.J.; Chen, V.B. et al. MolProbity: More and Better Reference Data for Improved All-atom Structure Validation. Protein Sci.; 2018; 27, pp. 293-315. [DOI: https://dx.doi.org/10.1002/pro.3330]

28. Pettersen, E.F.; Goddard, T.D.; Huang, C.C.; Couch, G.S.; Greenblatt, D.M.; Meng, E.C.; Ferrin, T.E. UCSF Chimera—A Visualization System for Exploratory Research and Analysis. J. Comput. Chem.; 2004; 25, pp. 1605-1612. [DOI: https://dx.doi.org/10.1002/jcc.20084]

29. Ye, Q.; West, A.M.V.; Silletti, S.; Corbett, K.D. Architecture and Self-assembly of the SARS-CoV-2 Nucleocapsid Protein. Protein Sci.; 2020; 29, pp. 1890-1901. [DOI: https://dx.doi.org/10.1002/pro.3909]

30. Peng, Y.; Du, N.; Lei, Y.; Dorje, S.; Qi, J.; Luo, T.; F Gao, G.; Song, H. Structures of the SARS-CoV-2 Nucleocapsid and Their Perspectives for Drug Design. EMBO J.; 2020; 39, e105938. [DOI: https://dx.doi.org/10.15252/embj.2020105938]

31. Jia, Z.; Liu, C.; Chen, Y.; Jiang, H.; Wang, Z.; Yao, J.; Yang, J.; Zhu, J.; Zhang, B.; Yuchi, Z. Crystal Structures of the SARS-CoV-2 Nucleocapsid Protein C-terminal Domain and Development of Nucleocapsid-targeting Nanobodies. FEBS J.; 2021; 289, pp. 3813-3825. [DOI: https://dx.doi.org/10.1111/febs.16239] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/34665939]

32. Matsuo, T. Viewing SARS-CoV-2 Nucleocapsid Protein in Terms of Molecular Flexibility. Biology; 2021; 10, 454. [DOI: https://dx.doi.org/10.3390/biology10060454] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/34064163]

33. Chandrasekaran Dinesh, D.; Chalupska, D.; Silhan, J.; Koutna, E.; Nencka, R.; Veverka, V.; Boura, E. Structural Basis of RNA Recognition by the SARS-CoV-2 Nucleocapsid Phosphoprotein. PLoS Pathog.; 2020; 16, e1009100. [DOI: https://dx.doi.org/10.1371/journal.ppat.1009100]

34. Zhou, R.; Zeng, R.; von Brunn, A.; Lei, J. Structural Characterization of the C-Terminal Domain of SARS-CoV-2 Nucleocapsid Protein. Mol. Biomed.; 2020; 1, 2. [DOI: https://dx.doi.org/10.1186/s43556-020-00001-4]

35. Ye, Q.; Lu, S.; Corbett, K.D. SStructural Basis for SARS-CoV-2 Nucleocapsid Protein Recognition by Single-Domain Antibodies. Front. Immunol.; 2021; 12, 719037. [DOI: https://dx.doi.org/10.3389/fimmu.2021.719037] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/34381460]

36. Sarkar, S.; Runge, B.; Russell, R.W.; Movellan, K.T.; Calero, D.; Zeinalilathori, S.; Quinn, C.M.; Lu, M.; Calero, G.; Gronenborn, A.M. et al. Atomic-Resolution Structure of SARS-CoV-2 Nucleocapsid Protein N-Terminal Domain. J. Am. Chem. Soc.; 2022; 144, pp. 10543-10555. [DOI: https://dx.doi.org/10.1021/jacs.2c03320]

37. Redzic, J.S.; Lee, E.; Born, A.; Issaian, A.; Henen, M.A.; Nichols, P.J.; Blue, A.; Hansen, K.C.; D’Alessandro, A.; Vögeli, B. et al. The Inherent Dynamics and Interaction Sites of the SARS-CoV-2 Nucleocapsid N-Terminal Region. J. Mol. Biol.; 2021; 433, 167108. [DOI: https://dx.doi.org/10.1016/j.jmb.2021.167108]

38. Kang, S.; Yang, M.; Hong, Z.; Zhang, L.; Huang, Z.; Chen, X.; He, S.; Zhou, Z.; Zhou, Z.; Chen, Q. et al. Crystal Structure of SARS-CoV-2 Nucleocapsid Protein RNA Binding Domain Reveals Potential Unique Drug Targeting Sites. Acta Pharm. Sin. B; 2020; 10, pp. 1228-1238. [DOI: https://dx.doi.org/10.1016/j.apsb.2020.04.009]

39. Khan, A.; Tahir Khan, M.; Saleem, S.; Junaid, M.; Ali, A.; Shujait Ali, S.; Khan, M.; Wei, D.-Q. Structural Insights into the Mechanism of RNA Recognition by the N-Terminal RNA-Binding Domain of the SARS-CoV-2 Nucleocapsid Phosphoprotein. Comput. Struct. Biotechnol. J.; 2020; 18, pp. 2174-2184. [DOI: https://dx.doi.org/10.1016/j.csbj.2020.08.006]

40. Shajahan, A.; Pepi, L.E.; Rouhani, D.S.; Heiss, C.; Azadi, P. Glycosylation of SARS-CoV-2: Structural and Functional Insights. Anal. Bioanal. Chem.; 2021; 413, pp. 7179-7193. [DOI: https://dx.doi.org/10.1007/s00216-021-03499-x]

41. Cubuk, J.; Alston, J.J.; Incicco, J.J.; Singh, S.; Stuchell-Brereton, M.D.; Ward, M.D.; Zimmerman, M.I.; Vithani, N.; Griffith, D.; Wagoner, J.A. et al. The SARS-CoV-2 Nucleocapsid Protein Is Dynamic, Disordered, and Phase Separates with RNA. Nat. Commun.; 2021; 12, 1936. [DOI: https://dx.doi.org/10.1038/s41467-021-21953-3] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/33782395]

42. Sun, Z.; Zheng, X.; Ji, F.; Zhou, M.; Su, X.; Ren, K.; Li, L. Mass Spectrometry Analysis of SARS-CoV-2 Nucleocapsid Protein Reveals Camouflaging Glycans and Unique Post-Translational Modifications. Infect. Microbes Dis.; 2021; 3, pp. 149-157. [DOI: https://dx.doi.org/10.1097/IM9.0000000000000071]

43. Tang, G.; Peng, L.; Baldwin, P.R.; Mann, D.S.; Jiang, W.; Rees, I.; Ludtke, S.J. EMAN2: An Extensible Image Processing Suite for Electron Microscopy. J. Struct. Biol.; 2007; 157, pp. 38-46. [DOI: https://dx.doi.org/10.1016/j.jsb.2006.05.009] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/16859925]