Computational Insights into the Structural Dynamics of the E Protein’s DIII Domain in Brazilian Yellow Fever Virus: Implications of Amino Acid Variations on Protein Plasticity and Potential Impact on Vaccine Efficacy

Abstract

During 2016–2019, Brazil witnessed one of the largest yellow fever virus (YFV) outbreaks. Experimental studies suggested that the epitope recognized by vaccine-specific mAb 411 is strongly associated with the EDIII domain of the E protein. This prompted us to analyze the global protein sequences of the YFV-EDIII domain to explore the conserved mutations that modulate the domain structure. Through comprehensive computational analysis combining sequence alignment, stability prediction, and molecular dynamics simulations, we characterized structural and functional consequences of mutations exclusive to Brazilian strains. Multiple sequence alignments of 223 publicly available YFV protein sequences revealed four prevalent mutations in the EDIII domain: V318A, K331R, I335 M, and I344 V, occurring in more than 65% of the analyzed sequences. Notably, V318A and I335 M mutations were found exclusively in Brazilian strains, while K331R and I344 V showed broader geographic distributions but higher prevalence in South American isolates. Thermodynamic stability predictions using multiple computational approaches revealed that individual mutations, particularly V318A, consistently demonstrated destabilizing effects, with predicted ΔΔG values ranging from −0.43 to −5.40 kcal/mol. Molecular dynamics analysis suggests that the mutations significantly changed the conformational rearrangement and folding of the YFV-EDIII domain at positions 310–312, 315, 321, and 342. In conclusion, our study demonstrates that the four prevalent mutations in Brazilian YFV strains represent a sophisticated example of compensatory evolution, where the virus has optimized its structural stability through a strategic combination of individual substitutions. Therefore, we hypothesized that the significant conformational rearrangement and folding of the YFV-EDIII domain due to mutation would affect the binding of antibodies despite not analyzing virus–antibody interactions.

Introduction

The yellow fever virus (YFV) is an arthropod-transmitted virus classified under the Flaviviridae family. Transmission occurs by the bite of infected Aedes or Haemagogus mosquitoes. YF may be asymptomatic or symptomatic, characterized by fever, myalgia, back pain, and prostration, and in some instances, it may result in multiorgan failure, primarily affecting the liver and kidneys, accompanied by severe jaundice. YF remains a significant threat to public health in Africa and South America, where it is associated with a high case fatality rate of 40% to 60%, especially in South America. The Pan American Health Organization (PAHO) has reported that from 1960 to 2019, the South American nations with the most remarkable recorded occurrences of YF were Brazil (3829 cases), Peru (3189 cases), Bolivia (1546 cases), and Colombia (701 cases).

YFV and its vector, Aedes aegypti, are believed to have been brought to Brazil by slave trafficking vessels from West Africa during the early colonial era. The first YF pandemic in Brazil was documented in 1685. Nevertheless, the last two decades have shown an increase in the geographical distribution of YFV inside the nation. The significant resurgence of YF in Brazil started in late 2016, with the Ministry of Health reporting 2237 human cases and 759 fatalities between December 2016 and June 2019. The aforementioned data suggest that the resurgence of YFV in Brazil may be attributed to one of the following reasons.

First, the recent re-emergence of YFV showed that most of the population affected by YF (82.8% during 2017–2018) were men within the economically active age bracket, residing in rural regions near forested areas. , Consequently, it may be inferred that deforestation may contribute to heightened human exposure to fragmented forest regions, hence elevating the probability of YFV transmission to urban areas via sylvatic cycles of YFV. Furthermore, experimental studies also indicate that YFV may adapt to Aedes albopictus, an opportunistic mosquito species inhabiting urban and peri-urban environments. This might also be a contributing factor to its re-emergence. Third, some studies during the epidemic in Brazil (State of Minas Gerais) have highlighted the absence of neutralizing antibodies against YFV in individuals with proven vaccination records from both rural and urban areas. As far as YFV vaccines are concerned, the live attenuated 17D vaccine for YFV was created in the 1930s, and now, three 17D substrains are in production: 17DD produced in Brazil, 17D-213 produced in Russia, and 17D-204 produced in China, France, Senegal, and the USA. All these vaccines are prequalified and used worldwide in WHO/UNICEF vaccination campaigns and target the E protein of YFV. This may be considered a significant reason for the recent epidemic in the region because, in one of the studies conducted in 2019, it has been observed that the vaccines who had seroconverted following primary vaccination (17DD-YF vaccine) with subdoses (10447 IU; 3013 IU; 587 IU, respectively) in 2009 presented similar levels of neutralizing antibodies with no visible dose-wise increment after eight years of vaccination. This shows that there might be some unexplored reasons behind the absence of neutralizing antibodies in individuals during the Brazilian epidemic.

Now, in the case of flaviviruses, including YFV, entry into the target cell is mainly dependent on the E protein interaction with its cognate receptor. Moreover, the E protein of flaviviruses, including YFV, consists of three ectodomains (ED-I, ED-II, and ED-III). ED-III contains important viral entry sites and linear antigenic epitopes that directly interact with potent neutralizing antibodies (Matsui et al., 2010). Therefore, to explore the probable reason behind the re-emergence of YFV in Brazil, we performed the multiple sequence alignment (MSA) of global sequences of YFV-EDIII to look for conserved mutations exclusive to the Brazilian strains. Then, we predict the change in stability (ΔΔG) and vibrational entropy changes (ΔΔS) due to mutations using different tools. Further, to assess the effect of the identified mutations on the structure of the domain and the orientation of the key residues that bind to the monoclonal antibodies, a 300 ns scale molecular dynamics simulation (MDS) in three replicates of the wild-type and mutant variant of the EDIII structure was performed.

Materials and Methods

Sequence Alignment and Mutation Selection

The publicly available 223 sequences of the EDIII domain of YFV E protein belonging to different strains (including the vaccine strains) and distributed worldwide were collected from the NCBI Protein database, followed by MSA using the ClustalW alignment tool. Similar sequences were excluded from the analysis. Those mutations exclusively dominant in all Brazilian strains were selected for further downstream analysis.

Prediction of Stability Change due to Mutations

To study the impact of the selected mutation on the plasticity of domain III of the E protein, we examined the structure-based changes in ΔΔG due to mutations using multiple web servers. The combination of both entropy and energy predictions allows for a more comprehensive understanding of how mutations affect protein stability through different thermodynamic parameters. DynaMut2 was used to predict changes in ΔΔG. Since DynaMut2 does not give any insight into vibrational entropy changes (ΔΔS), DynaMut was also used for the prediction. Further, the ΔΔG predicted in both tools was included in the analysis. , mCSM predicts the impact of mutations by correlating the atomic-distance patterns around an amino acid residue. CUPSAT predicts the impact of mutations by assessing the amino acid environment using prediction model atom potentials and torsion angle distribution. Site Directed Mutator (SDM) employs a statistical potential energy function derived from a large data set of protein structures to predict stability changes upon mutation. DUET combines the predictions from mCSM and SDM using a support vector machine approach to enhance stability change prediction accuracy. Elastic network contact model (ENCoM) utilizes coarse-grained normal-mode analysis to predict thermodynamic stability changes and vibrational entropy variations, providing insights into protein flexibility and conformational dynamics upon mutation.

Wild and Mutant Structure Preparation

The crystal structure of the YFV-rED3 (PDB: 2JQM | https://www.rcsb.org/structure/2JQM; Residue: Met287 to Lys398) was retrieved from the PDB database (https://www.rcsb.org) and considered as a reference and wild-type structure for further downstream analysis. For generating mutant variants, the selected mutations (V318A, K331R, I335 M, and I344 V) from the sequence alignment analysis were introduced into the wild-type structure using the mutagenesis wizard, followed by structure preparation using the pdb 2pqr method in the PyMOL tool.

Molecular Dynamics Simulation

To evaluate the effect of the point mutations on the EDIII’s structural stability, the 300 ns scale MDS in three replicates with the different velocities of the EDIII in the water environment at 300 K using the GROMACS suite v2020.4 with the charmm36-jul2021 force field , were performed. Both variants (wild and mutant) of EDIII were solvated in a cubic periodic box with a Tip3P charm-modified water model. Systems were neutralized by adding NaCl salt of 0.15 molar to replicate the physiological ion concentration, followed by energy minimization of a maximum of 50,000 steps using the steepest descent (SD) method. The systems were then subjected to NVT (constant number of particles, volume, and temperature) and NPT (constant number of particles, pressure, and temperature) equilibrations at pressure and temperature of 1 bar and 300 K, respectively. The LINear Constraint Solver (LINCS) algorithm was used to constrain all the bonds and angles, whereas the Verlet algorithm integrated Newton’s equations of motion. , The temperature and pressure coupling in each equilibration steps were controlled by the V-rescale (modified Berendsen thermostat) and the Parrinello–Rahman method, respectively. However, in NVT equilibrations, no controller was used for pressure coupling. The time evolution of the trajectories was recorded every 10 ps for 300 ns.

Evaluation of Structural Changes

The trajectories of each replicate were evaluated to get root-mean-square deviation (RMSD), root-mean-square fluctuation (RMSF), radius of gyration (RoG), solvent-accessible surface area (SASA), and number of intramolecular hydrogen bond (HBond) formations using the in-built module of GROMACS.

Clustering of Trajectories

The trajectories of each replicate, both wild-type and mutant, were clustered using the TTClust tool. The representative structure (medoids) from the most populated cluster of each replicate was then superimposed. Subsequently, the RMSD between the cluster representatives for the wild-type and mutants was compared.

Essential Dynamics Analysis

To observe the biologically significant motion in the EDIII structure due to mutation in comparison with wild-type, we performed the contact map (cmap)-based principal component analysis (PCA) on trajectories of wild and mutant EDIII structures. Further, the movement and amplitude of the motion in Cα atoms throughout the simulation in the wild and mutant structure, the Porcupine plots of principal components (PCs) 1 and 2 data were generated using the ProDy plugin of the VMD molecular visualization program.

Changes in Secondary Structure Elements

The changes in the secondary structure of each residue for all frames in the wild and mutant structure throughout the simulation, we used the VMD plugin “Timeline” to calculate the secondary structure of each frame.

Free-Energy Landscape

The free-energy landscape (FEL) is a visual representation of the standard changes in Gibbs free energy of a molecular system as a function of its conformational states or reaction coordinates. , It can be deduced by using the following equation:

$$\mathrm{\Delta }G=-k_{\mathrm{B}}T\ln p(\mathrm{C}{\mathrm{V}}_1,\mathrm{C}{\mathrm{V}}_2)$$

where k _B = Boltzmann constant, T = Absolute temperature, and p(CV1, CV2) = probability distribution of the reaction coordinates or collective variables (CVs). In this study, we used RMSD versus RoG and PC1 versus PC2 as the reaction coordinates to analyze the FEL along the MD trajectories of each replicate.

Statistical Analysis

Statistical analysis was conducted throughout the study using suitable tools and methods to derive valuable insights from the trajectory data.

Results

Mutations Analysis Reveals the Four Specific Mutations in Brazilian Strains

MSA of around 223 publicly available protein sequences (52 unique sequences) of the EDIII region belonging to different strains and distributed worldwide highlighted the presence of the amino acid substitutions at various positions (Figure ). However, the mutations V318A, K331R, I335M, and I344 V occur in more than 65% of sequences. Among the Brazilian strains of YFV, the strains H191, RJ258, and BeH526722 contain mutations at five residues (299, 305, 325, 331, and 380). The strain RJ193 includes a mutation at one residue (331). The strains SA39, BeH463676, and BeH622493 contain only one mutation at the 344 position; also, the strain BeH394880 contains only one mutation at the 378 position. However, no mutations were found in the strains H196, H299, and H313 (Figure ). The remaining Brazilian strains contain mutations at positions 318, 331, 335, and 344. Hence, mutations V318A, K331R, I335 M, and I344 V were selected for predictions of ΔΔG and MDS.

1.

Multiple sequence alignment of publicly available protein sequences (52 unique sequences) of the YFV-EDIII region belonging to different strains and distributed worldwide highlighted the presence of the amino acid substitutions at various positions. The YFV-EDIII sequences of Brazilian strains are shown bold and dark red color.

Predicted Stability Changes Reveal Variable Mutational Effects on Protein Thermodynamic Stability

We employed multiple computational approaches, including structure-based methods (DynaMut, DynaMut2, mCSM, and DUET) and machine learning algorithms (CUPSAT, SDM, and ENCoM), to comprehensively assess the thermodynamic consequences of mutations in terms of ΔΔG and their impact on protein stability (Table ). The V318A mutation consistently demonstrated destabilizing effects across all computational tools, with predicted ΔΔG values ranging from −0.43 kcal/mol (ENCoM) to −5.40 kcal/mol (CUPSAT). Notably, CUPSAT and SDM indicated this mutation as highly destabilizing (−5.40 and −2.66 kcal/mol, respectively), while mCSM and DUET showed moderate destabilization (−2.12 and −2.55 kcal/mol). The consensus among all methods suggests that the V318A substitution significantly compromises protein stability. The K331R mutation exhibited the most variable predictions among the tested variants. While DynaMut, DynaMut2, CUPSAT, and SDM predicted stabilizing effects (ΔΔG values of 0.13, 0.02, 0.60, and 0.77 kcal/mol, respectively), mCSM, DUET, and ENCoM indicated destabilizing outcomes (−1.33, −0.99, and −0.12 kcal/mol). This discrepancy likely reflects the conservative nature of the lysine-to-arginine substitution, where both residues maintain a positive charge but differ in side-chain geometry and hydrogen bonding capacity. The I335M mutation showed predominantly destabilizing effects across most computational approaches. DynaMut2, mCSM, CUPSAT, SDM, and DUET predicted significant destabilization (ΔΔG values of −1.14, −1.26, −4.67, −1.91, and −1.59 kcal/mol, respectively), with CUPSAT again showing the most severe prediction. Only DynaMut and ENCoM suggested stabilizing effects (1.14 and 0.50 kcal/mol), indicating that the isoleucine-to-methionine substitution generally compromises structural integrity. For the I344V mutation, predictions were moderately variable but typically indicated destabilizing effects. While DynaMut, DynaMut2, and ENCoM predicted slight stabilization (0.09, 0.07, and 0.12 kcal/mol), mCSM, CUPSAT, SDM, and DUET indicated destabilization ranging from −0.18 to −1.68 kcal/mol. The relatively small magnitude of predicted changes suggests that this conservative hydrophobic substitution has modest thermodynamic consequences.

1. Predicted Stability Change (ΔΔG, Kcal/mol) and Vibrational Entropy Energy (ΔΔS, kcal·mol^–1·K^–1) due to Mutations and Their Outcome Using Different Computational Algorithms in the Structure of YFV EDIII Domain.

prediction type and tools	V318A	K331R	I335M	I344V
ΔΔG (DynaMut)	–2.21 (destabilizing)	0.13 (stabilizing)	1.14 (stabilizing)	0.09 (stabilizing)
ΔΔG (DynaMut2)	–1.89 (destabilizing)	0.02 (stabilizing)	–1.14 (destabilizing)	0.07 (stabilizing)
ΔΔG (mCSM)	–2.12 (destabilizing)	–1.33 (destabilizing)	–1.26 (destabilizing)	–0.52 (destabilizing)
ΔΔG (CUPSAT)	–5.4 (destabilizing)	0.60 (stabilizing)	–4.67 (destabilizing)	–1.68 (destabilizing)
ΔΔG (SDM)	–2.66 (destabilizing)	0.77 (stabilizing)	–1.91 (destabilizing)	–0.18 (destabilizing)
ΔΔG (DUET)	–2.55 (destabilizing)	–0.99 (destabilizing)	–1.59 (destabilizing)	–0.23 (destabilizing)
ΔΔG (ENCoM)	–0.43 (destabilizing)	–0.12 (destabilizing)	0.5 (stabilizing)	0.12 (stabilizing)
ΔΔS (ENCoM)	0.54 (increase of molecule flexibility)	0.15 (increase of molecule flexibility)	–0.62 (decrease of molecule flexibility)	–0.15 (decrease of molecule flexibility)

Vibrational Entropy Changes Reveal Altered Protein Flexibility and Conformational Dynamics

To understand the dynamic consequences of mutations beyond thermodynamic stability, we analyzed vibrational entropy changes (ΔΔS) using ENCoM integrated in DynaMut, which provides insights into protein flexibility and conformational dynamics (Table and Figure S1). The vibrational entropy analysis revealed distinct patterns of flexibility modulation across mutation sites. The V318A mutation exhibited a substantial increase in molecular flexibility (ΔΔS = 0.54 kcal·mol^–1·K^–1), indicating that the removal of the bulky valine side chain creates a more dynamic local environment. This increased flexibility correlates with the destabilizing thermodynamic predictions and is visualized in Figure S1A, where red regions indicate enhanced conformational freedom extending beyond the immediate mutation site. Similarly, the K331R mutation showed increased molecular flexibility (ΔΔS = 0.15 kcal·mol^–1·K^–1), though to a lesser extent than V318A. The structural visualization (Figure S1B) reveals localized flexibility changes with some rigidification in distant regions (blue areas), suggesting complex allosteric effects, despite the conservative nature of the substitution.

In contrast, both I335M and I344V mutations resulted in decreased molecular flexibility (ΔΔS = −0.62 and −0.15 kcal·mol^–1·K^–1, respectively), indicating an increased structural rigidity. The I335M mutation showed the most pronounced rigidification effect, potentially due to the introduction of the sulfur-containing methionine side chain, creating new intramolecular interactions. Figure S1C,D illustrates these rigidification patterns, with blue regions predominating around the mutation sites and extending to structurally connected regions. Integrating thermodynamic stability predictions with vibrational entropy analysis reveals that mutations in the EDIII domain can simultaneously affect protein stability and conformational dynamics through distinct mechanisms. The observed inconsistency among computational predictions highlights the complex nature of mutational effects and the importance of using multiple computational approaches to systematically assess the structural and functional impacts of amino acid substitutions in the YFV EDIII domain.

Comparative Structural Parameters Analysis Highlights the Mutant’s Greater Structural Stability

The wild (reference) and the mutant (Brazilian) structures of YFV-EDIII were subjected to 300 ns MD simulation in three replicates to observe the possible impact of four substitutions (V318A, K331R, I335M, and I344V) on structural and conformational changes of the domain. Time-evolution analysis of structural parameters (RMSD, RoG, SASA, and HBonds) for individual simulation replicates of wild-type and mutant EDIII domains across 300 ns MDS is represented in Figure S2. The RMSD range for the Cα atoms of the wild-type structure is notably broader, spanning from 0.94 to 8.69 Å. In comparison, the mutant variants display a narrower range between 1.0 Å and 7.0 Å (Figure A). On average, the mutant exhibits a relatively lower mean RMSD (4.85 ± 0.76 Å) compared to the wild-type’s mean RMSD (5.94 ± 1.06 Å) (Figure B). In the final 10 ns of the simulation (from 291 to 300 ns), the mutant structure maintains a significantly lower RMSD of 5.41 ± 0.34 Å. In contrast, the wild-type averages 6.17 ± 0.21 Å with a difference of 0.76 Å (Figure A). This comparative analysis highlights the mutant’s greater structural stability and reduced conformational variability relative to the wild-type.

2.

Comparative molecular dynamics simulation analysis of wild-type and mutant YFV EDIII domain structures over 300 ns. The figure presents four key structural parameters comparing the dynamic behavior of wild-type (blue) and mutant (orange) variants of the YFV envelope protein domain III (EDIII). (A) Root mean square deviation (RMSD). (B) Kernel density estimation (KDE) distribution of RMSD values. (C) Radius of gyration (RoG). (D) KDE distribution of RoG values. (E) Solvent-accessible surface area (SASA). (F) KDE distribution of SASA values. (G) Hydrogen bond formation patterns. (H) KDE distribution of hydrogen bond formation. The kernel density estimation plots (Panels B,D,F,H) provide statistical insights into the distribution patterns of each structural parameter, with dashed vertical lines indicating mean values for each variant.

The RoG measures the compactness of the molecular structure over the simulation time. The RoG analysis suggested that the mutant structure of EDIII is slightly more compact than that of the wild-type. The RoG for wild-type is distributed from 13.63 to 15.98 Å with an average of 14.90 ± 0.31 Å. In contrast, for the mutant, it was distributed from 13.68 Å to 15.89 Å with an average of 14.69 ± 0.36 Å (Figure C,D). This difference in average RoG (ΔR _oG = 0.21 Å) indicates that the mutation modestly affects the overall compactness of the EDIII structure, with the mutant adopting a slightly more compact conformation throughout the simulation.

The SASA quantifies the molecular surface area accessible to solvent molecules, indicating the protein compactness and structural stability. Our analysis identified clear differences between the wild-type and mutant variants. The mutant structure showed a lower average SASA value (74.29 ± 2.00 Ų) compared to the wild-type (75.78 ± 1.73 Ų), which indicates a more compact molecular architecture (Figure E). This decrease in the SASA suggests that the mutation induces structural rearrangements that result in enhanced compactness, increased structural stability, and altered surface topology. The density distribution plots (Figure F) further illustrate this distinction, with the mutant showing a narrower, more defined peak centered at lower SASA values, indicating more consistent structural compactness throughout the simulation trajectory. Therefore, the mutation alters the protein’s surface, possibly by influencing binding sites, catalytic regions, or interaction interfaces.

Intramolecular HBonds play an essential role in maintaining the structural integrity and stability of proteins. These noncovalent interactions contribute significantly to the overall conformational stability of the protein throughout the MDS. Our analysis of HBond formation patterns revealed notable differences between the wild-type and mutant variants. The mutant structure demonstrated a relatively higher average number of HBonds per frame (54.41 ± 2.59) compared to the wild-type (53.42 ± 2.61), indicating enhanced intramolecular connectivity (Figure G,H). This increase in HBond formation suggests that the mutation facilitates more favorable geometric arrangements between amino acid residues, allowing for additional stabilizing interactions within the protein structure. Since more intramolecular HBonds contribute to enhanced structural stability, the increased HBond formation in the mutant EDIII suggests improved protein stability over the simulation time. These additional stabilizing interactions likely complement the structural compactness observed in the SASA analysis, collectively contributing to the overall enhanced stability profile of the mutant variant.

Solvent Accessibility and Fluctuations Unveil Structurally Important Residues

The RMSF analysis of the trajectories measures the amplitude of plasticity for individual atoms or groups of atoms in a molecular system. The RMSF of Cα atoms shows overall stability, except for the N-terminal (residues 287–295) and C-terminal (residues 390–398) regions (Figure A,B). Upon detailed examination of the mutant variant compared to the wild-type structure, the most statistically significant changes were observed at positions 310–312, where the mutant demonstrated markedly reduced fluctuations with differences of −0.28 Å (p = 0.02), −0.31 Å (p = 0.01), and −0.30 Å (p = 0.01), respectively, indicating a substantial increase in local structural compactness and reduced conformational sampling in this region (Table S1).

3.

Per-residue structural flexibility (RMSF) and solvent exposure (SASA) analysis of yellow fever virus envelope protein domain III (EDIII). (A) RMSF profile for the wild-type protein. (B) RMSF profiles for the mutant protein. (C) SASA profile for the wild-type protein. (D) SASA profiles for the mutant protein. Line colors represent independent replicates for each measurement. These analyses highlight differences in structural flexibility and solvent exposure between wild-type and mutant forms, focusing on the central region of the protein, excluding the N-terminal (residues 287–295) and C-terminal (residues 390–398) regions.

This stabilization effect extends to residue 315, which also displayed a significantly lower fluctuation (−0.40 Å, p = 0.05) and to residue 321, where a smaller but statistically significant reduction was noted (−0.14 Å, p = 0.04). However, residue 342 exhibited significantly increased fluctuations in the mutant structure, with a mean difference of +0.72 Å (p = 0.02), indicating localized destabilization (Table S1). These findings indicate that the mutation induces a propagated stabilization effect across multiple structural elements. This suggests that these regions may be functionally interconnected and that the mutation could influence the protein’s overall dynamic properties and stability.

The residue-by-residue solvent accessibility (SASA) analysis reveals localized structural changes in the mutant protein, with statistically significant reductions in solvent exposure at residues 302, 318, 338, and 339 (p < 0.05; Figure C,D and Table S2). Residue 318 (V318A mutation site) shows a minor but significant SASA reduction (−0.01 Ų, p = 0.03). This suggests that the mutation introduces subtle steric or conformational changes that reduce solvent exposure while paradoxically increasing the local flexibility. Residues 338–339, adjacent to the destabilized residue 342 (RMSF +0.72 Å), exhibit reduced SASA (−0.15 Ų and −0.05 Ų, respectively), implying that decreased solvent access in this region may destabilize interactions, contributing to the heightened flexibility observed at 342. Residue 302 (SASA −0.10 Ų, p = 0.03) lies near the stabilized RMSF cluster (310–312), potentially indicating coupled stabilization of this subdomain through reduced flexibility and solvent shielding.

Together, the RMSF and SASA analyses reveal that the mutation not only induces localized stabilization and reduced flexibility in key structural regions (residues 302, 310–312, 315, 318, 321, and 338) but also subtly alters solvent accessibility at functionally important sites. Since terminal regions of the protein are more flexible, less constrained, and more exposed to the solvent, stability cannot be ascertained from the RMSF and SASA values. These findings collectively indicate that the mutation has propagated effects on the protein’s stability and solvent exposure, likely impacting its functional architecture and potentially biological activity.

Clustering and Conformational Dynamics of Wild-Type and Mutant Ensembles

Clustering analysis of the ensembles of each replicate for both wild-type and mutant conformation samples during 300 ns simulations was performed using TTClust. The clustering segregated the ensembles of both wild-type and mutant trajectories into 2–5 optimal clusters by elbow statistics, based on Cα atom RMSD (Figure A). Clustering analysis segregated wild-type trajectories into 2–5 clusters (R1:2, R2:5, and R3:3), while mutant trajectories formed 3–5 clusters (R1:3, R2:5, and R3:4). The average Simpson’s diversity index was 0.56 ± 0.13 for wild-type and 0.58 ± 0.10 for mutant, indicating slightly higher conformational diversity in the mutant (Table S3). Notably, wild-type R2 showed exceptional diversity (0.72) from balanced cluster populations, whereas mutant replicates avoided such extremes, suggesting that the mutation constrains heterogeneity. Further, the centroids of the most populated cluster of wild-type and mutant trajectories were superimposed to observe the 3D structural deviation (Figure B,C). The pairwise RMSD values between centroid structures for wild-type replicates were 1.85, 2.04, and 1.73 Å, while those for mutant replicates were 1.99, 2.49, and 1.78 Å. Overall, both wild-type and mutant structures show some conformational variability, with the wild-type exhibiting a slightly broader range between pairs (notably, R1 vs R3). In contrast, the mutant’s variability is mostly lower except for a higher R1 vs R3 RMSD. This suggests that the mutant may be more structurally constrained but retains some conformational diversity.

4.

Molecular dynamics clustering analysis and superimposition of dominant cluster representative. (A) Linear projections of the clusters in each replicate. (B) Structural superimposition of the most populated cluster representatives from wild-type simulations, demonstrating conformational convergence and structural stability and (C) structural superimposition of the most populated cluster representatives from mutant simulations, highlighting mutation-induced conformational changes and altered protein dynamics compared to wild-type.

Essential Dynamics Analysis Reveals the Significant Changes in the Motion of Atoms due to Mutations

Mutations can perturb residue and atomic contacts, altering the protein structure and function. To investigate these effects, essential dynamics (ED) analysis was performed to characterize the collective motions of wild-type and mutant trajectories. The scatter plots of PCs revealed distinct conformational distributions between the wild-type and mutant, highlighting significant differences in their dynamic behaviors (Figure ). The wild-type exhibited a trace value of 16.09 ± 0.29 nm², indicating greater conformational flexibility. In contrast, the mutant showed a lower trace value (14.47 ± 1.02 nm²), suggesting reduced structural diversity and increased rigidity due to the mutation. The first three PCs captured 44.33% of the total variance in the wild-type compared to 42.29% in the mutant. The first 10 PCs collectively explained 91.16–92.59% of the variance in the wild-type and 90.80–92.01% in the mutant (Figure G,H). The higher trace value and broader variance distribution across PCs reflect greater conformational plasticity, consistent with the wild-type system. Meanwhile, in the case of the mutant system, the reduced trace value and narrower variance contributions suggest that the mutation restricts collective motions, stabilizing a subset of dominant conformations. These findings align with the clustering analysis, reinforcing that the mutation imposes structural rigidity and potentially modulates functional dynamics.

5.

Comparative essential dynamic analysis (principle component analysis) of the wild-type and mutant variant structure of the YFV EDIII domain. (A–C): Projections of wild-type trajectories (R1, R2, and R3) onto the first two principal components (PC1 vs PC2), illustrating conformational dynamics. (D–F): Projections of mutant trajectories (R1, R2, and R3) onto PC1 and PC2, highlighting reduced motion compared to wild-type. (G): Combined scatter plot of wild-type replicates, emphasizing explained variance (%) of the first 10 principal components. (H): Combined scatter plot of mutant replicates, emphasizing explained variance (%) of the first 10 principal components.

To study the quantitative motion in Cα atoms throughout the simulation captured by the first two PCs, Porcupine plots were generated using the ProDy plugin of the VMD program. The arrow’s directions and length represent the motion’s directions and amplitude. The motion of regions 311–315, 318, 319, and 393 in the mutant structure showed significantly (p-value <0.05) lower motions on both components (PC1 and PC2) as compared to the wild-type structure (Figure A,B). In contrast, residues 339 and 342 show very significant motions on both components (PC1 and PC2) in the mutant structure. From Porcupine plots, the changes in directions and amplitude of the motions reveal the conformational changes in the mutant compared to those in the wild structure over the 300 ns simulations.

6.

Comparative essential dynamics analysis (Porcupine plot) of wild-type and mutant YFV EDIII domain structures using Porcupine plots. (A) Directional displacements (cones) along the first principal component (PC1) for the wild-type structure, illustrating collective atomic motions. (B) Porcupine plot of the mutant structure along PC1, showing restricted motion compared to the wild-type. The shorter cones and reduced directional variability suggest that the mutation stabilizes specific structural regions. The height and orientation of cones reflect the magnitude and direction of dominant fluctuations, highlighting regions of high flexibility.

Changes in the Secondary Structure Elements

The analysis of changes in the secondary structure throughout the simulation time provides insights into the structural stability of the protein. The DSSP algorithm was used to examine changes in various secondary structure elements, including coils, β-sheets, β-bridges, bends, turns, α-helices, 3₁₀-helices, and β-helices, across all trajectories for both the wild-type and mutant proteins. Overall, there was an average decrease in the coil and turn structures. In contrast, the β-sheet and 3₁₀-helix structures increased in the mutant variant compared to the wild-type (Figure S3). Notably, after 140 ns, there were specific changes in the residues: the coil structure decreased by 2.34 residues, the β-sheet structure increased by 1.96 residues, the β-bridge structure increased by 0.85 residues, and the 3₁₀-helix structure increased by 1.18 residues. Moreover, as per the t-test analysis, in the mutant variant, β-sheet, β-bridge, and 3₁₀-helix structure types were significantly (p-value < 0.05) increased, while coil, bend, and turn structure types were significantly (p-value < 0.05) decreased (Table S4). In summary, the overall secondary structure elements in the mutant variant increased by 2.34 residues after 140 ns; this makes the mutant structure more compact and stable than that of the wild-type.

Free-Energy Landscape

The FEL analysis revealed distinct conformational preferences between wild-type and mutant YFV EDIII domain structures, with notable differences in PCA space and structural parameter distributions (RMSD and RoG). FEL calculations are highly relevant for distinguishing the kinetic and thermodynamic characteristics of wild-type and mutant protein forms. FEL is based on the probability of the occurrence of combinations of data points, which are then transformed into free-energy values using a simple relationship. Energy minima of the landscape were visualized and mapped to representative configurations occurring during the trajectory. Qualitative inspection of these configurations indicates the energy barriers between different conformational basins (Figure and Table S5).

7.

Free-energy landscape (FEL) analysis of wild-type and mutant YFV EDIII domain structures. FEL plots depicting the conformational dynamics of wild-type and mutant YFV envelope domain III (EDIII) structures across three independent simulation replicates. Panels (A–C): PC analysis-based FEL plots (PC1 vs PC2) for wild-type YFV EDIII show conformational state distribution in the first two PC spaces. Panels (D–F): PC1 vs PC2 FEL plots for the mutant variant, revealing altered conformational sampling patterns compared to the wild-type. Panels (G–I): RMSD vs RoG FEL plots for wild-type structures, illustrating the relationship between structural deviation and compactness. Panels (J–L): RMSD vs RoG FEL plots for mutant structures, demonstrating structural flexibility and compactness profile changes. Black dots within each plot mark the locations of the three lowest energy minima, representing the most thermodynamically favorable conformational states. Each panel represents an independent simulation replicate (R1, R2, and R3), statistically validating the observed conformational behavior.

In the PC1 vs PC2 analysis, as presented in Figure A–C and Table S5, wild-type structures exhibited diverse conformational sampling patterns across the three replicates, with global minima located at coordinates ranging from (−2.99, 0.85) to (1.81, 1.12), indicating significant conformational flexibility and multiple stable states. The wild-type system demonstrated relatively shallow energy wells, with secondary minima differing by 0.01–0.14 kcal/mol from the global minimum, suggesting simplistic interconversion between conformational states. In contrast, the mutant structures showed more restricted conformational sampling in PC1 vs PC2 space, with global minima clustered in a narrower coordinate range from (1.92, −0.60) to (2.36, −0.11), as represented in Figure D–F and Table S5. The mutant system exhibited smaller energy differences between minima (0.01–0.08 kcal/mol), indicating a more uniform energy landscape with reduced conformational heterogeneity. This suggests that the mutation stabilizes specific conformational states while restricting access to alternative conformations observed in the wild-type structure.

The RMSD versus RoG analysis further supported these findings, revealing that wild-type structures occupied a broader range of structural deviation and compactness values, with RoG coordinates spanning from 0.38 to 0.71 nm. The wild-type system maintained relatively consistent RMSD values (1.40–1.49 Å) while exhibiting significant variation in the RoG, indicating structural breathing motions and domain flexibility, as represented in Figure G–I and Table S5. On the contrary, mutant structures demonstrated more constrained structural parameters, with RoG values tightly clustered around 0.33–0.40 nm and consistent RMSD values (1.40–1.42 Å), suggesting reduced structural flexibility and increased rigidity, as represented in Figure J–L and Table S5. The smaller free energy differences between minima in the mutant RMSD vs RoG landscapes (0.01–0.07 kcal/mol) compared to those in the wild-type (0.01–0.05 kcal/mol) indicate that while both systems have accessible conformational states (Table S5), the mutant structure exhibits more uniform energetic barriers between these states, potentially affecting the kinetics of conformational transitions and functional dynamics.

Overall, our MD analysis suggests that the four mutations significantly changed the conformational rearrangement and folding of the EDIII domain, as revealed through changes in Gibbs free energy, RMSD, RMSF, RoG, SASA, HBonds, clustering, PCA, Porcupine plots, changes in secondary structure analysis, and distribution of FELs. These comprehensive structural and thermodynamic alterations collectively demonstrate the profound impact of the mutations on protein stability, flexibility, and conformational dynamics, with implications for viral protein function and host–pathogen interactions.

Discussion

The 2017–2018 YFV epidemic in the southeastern states of Brazil marked a significant resurgence of YFV, particularly in metropolitan regions that had been free of the virus for over a century. This re-emergence raised critical questions about the underlying molecular mechanisms driving viral adaptation and immune evasion. Our comprehensive computational analysis of four conserved mutations (V318A, K331R, I335 M, and I344 V) exclusive to Brazilian YFV strains provides compelling evidence for sophisticated evolutionary adaptations that may explain the epidemic’s severity and persistence. Our sequence analysis revealed that these four mutations occur in more than 65% of the analyzed YFV sequences, with higher prevalence in Brazilian isolates than global strains, strongly suggesting their role in region-specific viral adaptation. This geographic clustering indicates that these substitutions likely confer selective advantages that enhance viral fitness in the Brazilian transmission context, potentially contributing to the epidemic’s magnitude and the observed vaccine breakthrough cases.

Our comprehensive stability and vibrational entropy predictions using seven computational methods (DynaMut, DynaMut2, mCSM, CUPSAT, SDM, DUET, and ENCoM), V318A consistently demonstrated destabilizing effects across multiple computational predictions (with ΔΔG values ranging from −0.43 to −5.40 kcal/mol) among the four analyzed variants. This mutation is also reported as the genetic signature of the strains circulating in Americans.

Our postdynamic study suggested that the mutant YFV-EDIII structure exhibited enhanced stability compared to the wild-type, as evidenced by lower mean RMSD values (4.85 ± 0.76 Å vs 5.94 ± 1.06 Å), increased structural compactness (RoG analysis), and enhanced intramolecular hydrogen bonding (54.41 ± 2.59 vs 53.42 ± 2.61 bonds per frame). The ED analysis further supported this interpretation, revealing reduced conformational diversity in the mutant (trace value 14.47 ± 1.02 nm² vs 16.09 ± 0.29 nm²). This suggests that these mutations stabilize preferred conformational states while restricting access to potentially dysfunctional conformations. Additionally, our RMSF analysis showed that residues 310–312, 315, and 321 show significantly decreased fluctuation in the mutant structure. However, residue 342 exhibited significantly increased fluctuations in the mutant structure with a mean difference of +0.72 Å (p = 0.02), indicating localized destabilization.

The structural modifications induced by these mutations have profound implications for viral immune evasion. In flaviviruses, including YFV, cellular entry depends critically on E protein interactions with cognate receptors, while the EDIII domain contains crucial linear antigenic epitopes that interact with neutralizing antibodies. Previous studies in other viruses like SARS-CoV-2 have reported that point mutation in structural proteins has been proven to reduce the neutralizing ability of MAbs produced by vaccines. , Moreover, it has also been reported that any alteration in the motion of critical amino acid residues results in changes in the conformation of the respective domain, eventually affecting the respective protein’s functionality. A recent article published in 2022 reported that the current commercially available vaccines (17D-204, 17DD, and 17D-213) produced mAb 411, which recognized specific epitope positions (299, 305, 325, and 380) located on EDIII; alterations to the residues at these positions influenced the binding of mAbs to the epitopes. In another study, it has been experimentally validated that antibodies in human vaccination sera specific for the EDIII domain could not neutralize the 2017–18 Brazilian YFV strains. Recently, a study was carried out to identify the YFV variants that could avoid being neutralized by the 17D-204 vaccine-specific mAb 864. One amino acid mutation in the envelope (E) protein, at position 325, was responsible for the conformational shift on the viral surface that led to the loss of the mAb 864-defined epitope.

Additionally, the alterations in some of the variants constituted a reversion to the wild-type Asibi virus sequence. The findings suggest that the E protein epitope containing the 325 position recognized by mAb 864 encompasses a region of great functional significance and encodes critical molecular determinants of YFV pathogenesis in vivo. Altogether, sites 305 and 325 of the EDIII domain of the E protein are crucial sites of YFV antigenic epitopes.

Interestingly, our RMSF analysis observed restricted motion or decreased flexibility in the Cα atoms at positions 305 and 325 in the mutant structure compared to the wild-type (mean difference of >0.25 Å). Consequently, it may be hypothesized that a slight rotational alteration in residues at positions (305 and 325) may influence the binding efficacy of pre-existing monoclonal antibodies to these epitopic sites. Supportively, the restricted motion of Cα atoms was also observed in the mutant structure in comparison to the wild-type at the 307 (mean difference of 0.77 Å), 315 (mean difference of 0.90 Å), and 316 sites (mean difference of 0.81 Å). These sites in the YFV EDIII domain are identified as part of the homodimer interface responsible for polypeptide binding, as designated by the NCBI Gen Bank for YFV. Taken together, it may be speculated that the vaccine-induced antibodies may not be able to properly bind and neutralize the Brazilian strains of YFV due to the reduced flexibility of the EDIII domain antibody binding site, unlike other worldwide strains. The above findings may also explain the findings of Haslwanter et al., who experimentally validated that the genotype-specific features reduced the susceptibility of South American YFV strains to vaccine-induced antibodies. This may have accounted for many YFV-positive cases in vaccinated individuals in Brazil during the 2017–19 outbreak. Furthermore, a recent study has also indicated antigenic differences between the WHO-licensed 17D and 17DD vaccines of YFV. They have shown higher titers of neutralizing antibodies to the homologous 17DD strain than to the heterologous 17D-204 strain. This difference is seen in PRNT₅₀ and PRNT₈₀ titers, where the 17DD strain consistently outperformed 17D-204.

Furthermore, it is well-known that the outbreak strains of a virus are considered more virulent than their previous counterparts. The rationale is that the outbreak strains may circumvent the host immune response and reproduce more efficiently inside the host cell. They may use several tactics, such as enhancing host cell entry, accelerating replication, or adapting to a new host or vector by attaching to novel receptors to propagate an epidemic. During the YFV entry phase, the E protein experiences conformational alterations due to low pH, which initiates the fusion of viral and host cell membranes. This fusion step is crucial for releasing the viral genome into the host cell. In the case of YFV, the sites 327, 346–349, and 351 are considered part of the low pH domain interface site of the EDIII domain, which is involved in the fusion of the virus and the host membrane as designated by the NCBI Gen Bank for YFV. Interestingly, our FEL analysis revealed more restricted conformational sampling in mutant structures, with narrower energy wells and reduced conformational heterogeneity. This suggests these mutations have optimized the protein’s energy landscape for enhanced performance.

These findings have significant clinical implications for vaccine development and antiviral strategies. The enhanced stability and altered conformational dynamics of the mutant EDIII domain could affect not only antibody recognition but also the binding of potential therapeutic compounds. Understanding these structural changes is crucial for developing next-generation vaccines that can effectively neutralize circulating Brazilian YFV strains and designing antiviral drugs that remain effective against evolved variants.

The secondary structure analysis showing significant increases in β-sheet and 3₁₀-helix content in the mutant variant further supports the enhanced structural organization and stability. This structural reorganization, combined with the observed changes in surface accessibility and hydrogen bonding patterns, collectively contributes to the altered antigenic profile and enhanced viral fitness observed in Brazilian strains.

Conclusion

In conclusion, our computational analysis identified four key mutations (V318A, K331R, I335M, and I344 V) in the EDIII domain of Brazilian YFV strains that exhibit distinct thermodynamic and structural consequences. Our comprehensive analysis identified V318A as the most deleterious mutation, exhibiting unanimous destabilizing predictions. The extensive structural disruptions associated with this mutation suggest its functional significance among Brazilian YFV strain variants. These findings provide computational evidence that mutations in Brazilian YFV strains may alter the structural and dynamic properties of the EDIII domain, which are critical for antibody recognition and viral entry. However, the functional significance of these predicted changes requires experimental validation through structural studies, binding assays, and immunological analyses. Further investigations, including protein–antibody interaction studies, in vitro mutagenesis experiments, and neutralization assays, are necessary to establish the direct relationship between these mutations and potential immune evasion or altered virulence. Additionally, comparative studies with mutations from other geographic strains would provide a broader context for understanding the evolution and adaptation of YFV in different populations. Finally, these findings encourage vaccine design to ensure adequate neutralization and protection against viral infections.

Availability of data and materials. This published article and its supplementary files include all data generated or analyzed during this study. Additional data sets used and/or analyzed during the current study are available from the corresponding author upon reasonable request.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsomega.5c01600.

• Statistical analysis of RMSF and SASA differences between wild-type and mutant YFV-EDIII structures, clustering analysis of molecular dynamics ensembles, secondary structure element quantification, thermodynamic analysis of conformational minima with free-energy landscapes, vibrational entropy change visualizations for point mutations, time-evolution plots of structural parameters across simulation replicates, and secondary structure content analysis for wild-type and mutant variants. (PDF)

∥.

M.G. and P.M. shared equal authorship. Conceptualization: M.G. and P.M.; data curation: M.G.; formal analysis: M.G. and P.M.; investigation: M.G. and P.M.; methodology: M.G., P.M., and B.B.S.; resources: B.B.S.; supervision: P.M. and B.B.S.; visualization: M.G.; writing-original draft: M.G., P.M., and B.B.S.; writing-review and editing: M.G., P.M., B.M., A.R., S.C., S.R., M.M., and B.B.S.

This research was supported by the Department of Biotechnology (DBT), Ministry of Science and Technology, New Delhi, India, through a fellowship to the first author (Grant Id: BT/INF/22/SP45078/2022). Additionally, the computational resources used in this study (Dell DP 7820 Workstation) were funded by the Indian Council of Medical Research (ICMR), Ministry of Health and Family Welfare, New Delhi, India (Grant Id: AMR/DHR/GIA/4/ECD-II-2020).

The authors declare no competing financial interest.