The potential inhibitory mechanism of EGCG against the Chikungunya virus targeting non-structural protein 2 through molecular dynamics simulation

Abstract

This study explores the potential of Indonesian herbal compounds against the chikungunya virus (CHIKV), which causes widespread illness without a specific cure known as chikungunya fever (CHIKF). By focusing on the nsP2 protein, crucial for the virus’s replication, the research utilizes computational methods identifying inhibitor compounds with high binding affinity. These promising candidates are further analyzed through 1 µs of molecular dynamic (MD) simulation studies, aiming to find effective inhibitors to control the chikungunya spread, leveraging Indonesia’s rich biodiversity for novel anti-CHIKV therapies. The results of our study highlight the molecular mechanism of the potential of epigallocatechin 3-O-gallate (EGCG) from Camelia sinensis in inhibiting nsP2 protease by binding to essential catalytic residues and exploring more energetically favorable orientations within the catalytic pocket. This dynamic binding process suggests that EGCG may disrupt the protease’s catalytic functions, potentially altering domain interactions without compromising the protein’s overall structure. Given nsP2’s minimal homology with human proteins, the risk of cross-reactivity is reduced, making it a suitable target for CHIKV therapy. This study suggests EGCG as a prime candidate for further development as a broad-spectrum inhibitor against CHIKF.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-024-81287-0.

Introduction

Chikungunya, a mosquito-borne illness caused by the chikungunya virus (CHIKV), is primarily transmitted by Aedes aegypti and Aedes albopictus mosquitoes¹. Initially detected in Tanzania in 1952, CHIKV has since spread to numerous countries, predominantly in Africa and Asia². The virus was first reported in the Americas in December 2013 and has expanded to over 31 states³. Over the past 10 years, there have been approximately 3.7 million reported chikungunya cases in 50 countries or territories in the Americas⁴. Chikungunya is now considered endemic in the Southeast Asian region⁵.

Chikungunya infection, while typically not fatal, can lead to chronic symptoms such as persistent joint pain that may last for months to years, as observed in 43–72% of La Réunion’s population infected with CHIKV, with symptoms persisting up to two years^6–8. These prolonged symptoms, including joint pain, brain disorders, and neural sensory weakening, detrimentally impact patients’ quality of life and productivity⁹.

Currently, a vaccine, Ixchiq (VLA1553), has been approved for the prevention of Chikungunya virus (CHIKV) infection. However, there are significant concerns regarding its safety. Ixchiq may induce severe, chikungunya-like adverse reactions in some recipients, and the vaccine’s impact on pregnant individuals and the potential risk of transmitting the vaccine virus to newborns, may lead to uncertain adverse outcomes^10,11. While anti-malarial drugs like chloroquine have been suggested as possible treatments, their efficacy in halting CHIKV replication lacks clinical validation. Data from two studies, one involving a human clinical trial and the other an NHP (non-human primates) trial, strongly suggest that chloroquine is ineffective as a prophylactic or therapeutic measure against CHIKV infection¹².

The design of anti-CHIKV agents requires a thorough analysis and identification of essential viral proteins that are integral to CHIKV’s replication process and can serve as precise drug targets. These targets include proteins involved in host cell entry, viral component synthesis, and virus release^13,14. This research started with the identification of essential CHIKV proteins, specifically the E2 capsid and nsP2 proteins, which have shown the most interactions with human proteins and are potential drug targets¹⁵.

The nsP2 was selected based on its similarity to virus-specific proteases, like those of HIV-1 and hepatitis C, previously identified as effective drug targets^16,17. The nsP2 protease (nsP2pro) of CHIKV, a papain-like cysteine protease, is instrumental in cleaving viral non-structural polyproteins into discrete replication proteins—nsP1, nsP2, nsP3, and nsP4—thereby playing a pivotal role in the virus’s life cycle^18,19. The absence of an existing commercial drug for chikungunya²⁰ makes CHIKV-specific enzymes, such as the capsid protease and nsP2pro, prime candidates for developing antiviral medications. Our approach involves employing virtual molecular docking methods to discover ligands capable of hindering the activity of the nsP2 protease, potentially inhibiting the virus’s propagation.

Herbal plants, abundant in bioactive compounds, offer promising prospects for developing unique anti-CHIKV herbal medicines, particularly in Indonesia due to their rich biodiversity. Our research utilizes computational techniques, employing Autodock Vina and Autodock4, to investigate the interaction between Indonesian herbal compounds and CHIKV nsP2, ranking compounds based on their binding affinities to nsP2’s active sites. In this study, one of the highest docking scores as an inhibitor of CHIKV nsP2 protease isepigallocatechin gallate (EGCG).

EGCG, a polyphenol abundantly found in green tea (comprising 59% of all polyphenols), exhibits various health benefits, including antitumor, antimicrobial, antioxidative, and antiviral properties²¹. Its antiviral efficacy has been well-demonstrated against several viruses, including hepatitis C virus (HCV), human immunodeficiency virus (HIV), influenza virus (FLU), dengue virus (DENV), and chikungunya virus (CHIKV)^22–26. However, while EGCG emerges as a potential ligand for combating CHIKV infection based on computational ranking and previous in vitro studies, the precise mechanism underlying its inhibition of CHIKV remains incompletely understood.

There was no in vitro study specifically examining the effect of EGCG on nsP2. The existing in vitro studies demonstrate that EGCG inhibits chikungunya virus (CHIKV) infection. In our study, we provide new insights by revealing the potential molecular interactions of EGCG with the CHIKV nsP2 protease, which may contribute to its antiviral activity. This represents a novel finding regarding the molecular mechanism by which EGCG could exert its inhibitory effects on CHIKV. By employing MD simulations, this study aims to elucidate the precise interactions between EGCG and nsP2, emphasizing how these interactions contribute to the inhibition of CHIKV. The insights gained could pave the way for novel therapeutic strategies against chikungunya, leveraging the antiviral properties of naturally occurring compounds like EGCG.

Results and discussion

The homology of CHIKV proteomes with the human and gut bacteria

The CHIKV proteome consists of four non-structural proteins (nsP1, nsP2, nsP3, and nsP4) and five structural proteins (E1, E2, E3, C, and 6k). Of these, nsP2 has the longest sequence with 798 residues. E3 and 6k, having less than 100 residues, were excluded from the analyses of homology and protein-protein interactions with human proteins, which often results in poorly defined binding pockets crucial for effective ligand binding in molecular docking studies. Seven non-paralogous proteins of CHIKV (nsP1, nsP2, nsP3, nsP4, E1, E2, and C) were analyzed for homology using the BLASTp server against the Homo sapiens proteome, which contains 1,523,243 proteins. The analysis revealed no significant similarity for any of the seven CHIKV proteins, indicating they are non-homologous to the human proteomes.

The analysis of the intestinal bacterial proteome included 82 bacterial species²⁷ encompassing 199,541,572 proteins. Among the seven proteins, only nsP3 showed similarity to 13 bacterial proteins, with seven having more than 30% identity (Table 1). Consequently, nsP3 was considered homologous to the intestinal bacterial proteome and was excluded from further analysis for potential drug targets.

Table 1.

Results of BLASTp analysis of nsP3 protein.

Protein	Spesies	Bit score	Query coverage	E-value	Percent Identity
O-acetyl-ADP-ribose deacetylase	Peptacetobacter hiranonis	47.4	20%	4  × 106	34.78%
Macro domain-containing protein	Bifidobacterium	55.8	21%	5 × 109	31.45%
Macro domain-containing protein	Bifidobacterium breve	55.1	21%	9 × 109	31.45%
Macro domain protein	Eubacterium siraeum	48.5	22%	6 × 106	31.45%
Macro domain-containing protein	Clostridium spiroforme	54.3	21%	1 × 108	31.40%
ADP-ribose-binding protein	Clostridium sporogenes	48.9	21%	2 × 106	31.15%
Macro domain protein	Tyzzerella nexilis	52.4	29%	4  × 107	30.25%

The potential of EGCG to inhibit nsP2 protease activity

The analysis of homology and interactions between CHIKV proteins and the human proteome highlighted nsP2 as an essential protein, emphasizing its significance as a potential drug target. Protein–protein interaction (PPI) analysis further revealed that nsP2, along with E2, exhibited the most substantial interactions, marking them as particularly promising drug targets compared to other non-structural proteins¹⁵. The nsP2 protein comprises a helicase domain (nsP2h) at the N-terminus and a protease domain (nsP2p) at the C-terminus, connected by a flexible linker of 14 amino acid residues^28,29. Given nsP2’s critical role in viral replication and its dual function, it stands out as a unique and highly attractive target for antiviral strategies. The decision to focus solely on nsP2, after initially considering E2, was driven by its essential nature and its broader therapeutic potential due to its involvement in multiple stages of the virus’s life cycle.

In this study, drug design strategies utilized a structural approach, involving the virtual screening of 5550 compounds through molecular docking techniques. These simulations investigated the binding potential of small, drug-like molecules derived from Indonesian herbs to the nsP2 protease’s catalytic site and surrounding cavity, aiming to interfere with the catalytic function. The catalytic site was identified at the interface between the helicase and protease subdomains, characterized by the catalytic dyad comprising Cys478 and His548³⁰. Emphasizing the importance of a model receptor that encompasses both subdomains for MD simulations, we resorted to using an AlphaFold2 (AF2)-predicted model as the receptor in our molecular docking and MD simulations, due to the absence of an experimental nsP2 structure containing both the helicase and protease subdomains. In addition, the AF2 prediction results indicate a high confidence level for nsP2, as nearly all parts of the protein have pLDDT scores above 70, as shown in Fig. 1. In AF2 structure prediction, a low pLDDT score suggests low accuracy and potential flexibility in that region, which can lead to poor docking performance. For the nsP2 AF2 structure, despite the docking region being located in the cleft between the helicase and protease domains, the pLDDT score in this region remains high(70–90).

Fig. 1.

The docking model of EGCG within the nsP2 complex, based on Autodock4 simulations: (A) the binding site of EGCG on nsP2 in a cartoon representation, with the nsP2 structure derived from AlphaFold 2 modeling, coloured with pLDDT scoring, and EGCG depicted in stick form; (B) the EGCG binding site on nsP2 in an electrostatic surface representation; (C) the interactions between EGCG and the residues at the nsP2 binding site, represented in 2D plots.

After conducting a virtual screening with AutoDock Vina, the 100 compounds with the highest binding affinities were further analyzed using the AutoDock4 algorithm. This assessment identified the top 20 compounds (Table 2), chosen for their potential to bind to the nsP2 protease catalytic site, with molecular weights below 500 Da. Notably, catechins constitute eight of these top twenty compounds. To refine our findings, we re-ranked the docking results using the MM/GBSA tool (Uni-GBSA), which provided a more accurate evaluation of binding affinities. The MM/GBSA analysis revealed eriodictyol 5,4′-dimethyl ether 7-O-glucoside (from Cleome viscosa) as the top candidate, followed by catechin-4-ol 3′-methyl ether 3-O-alpha-L-rhamnopyranoside (from Cassia javanica), eriodictin (from Phyllanthus niruri), and isoneorautenol (from Calopogonium mucunoides).

Table 2.

The top 20 compounds were identified through virtual screening with AutoDock Vina and further analysis using Autodock GPU.

Compounds	Plants	ΔG Docking (kcal/mol)		MMGBSA (kcal/mol)
		AutoDock Vina	AutoDock 4	AutoDock Vina	AutoDock 4
Epigallocatechin 3-O-(3-O-methyl)-gallate (EGCMG)	Camellia sinensis	− 8.0	− 9.44	− 39.37	− 31.76
Epigallocatechin 3-O-gallate (EGCG)	Camellia sinensis/ Averrhoa carambola	− 7.9	− 9.88	− 41.50	− 33.29
Epicatechin 3-O-gallate (ECG)	Camellia sinensis	− 7.8	− 9.19	− 36.12	− 22.98
(2R,3R)-3,5,7,3’,5’-Pentahydroxyflavanone 3-rhamnoside (PHFR)	Excoecaria agallocha	− 7.5	− 7.47	− 41.30	− 34.97
Epigallocatechin 3-O-caffeate	Camellia sinensis	− 7.5	− 9.85	− 34.75	− 19.25
Gallocatechin 3’-O-gallate	Camellia sinensis	− 7.4	− 9.29	− 37.05	− 32.48
Eriodictin	Phyllanthus niruri	− 7.4	− 8.34	− 40.35	− 39.49
Isoneorautenol	Calopogonium mucunoides	− 7.4	− 8.34	− 43.24	− 38.26
Maritimein	Zinnia linearis	− 7.4	− 7.67	− 42.00	− 36.65
Catechin-4-ol 3’-methyl ether 3-O-alpha-L-rhamnopyranoside	Cassia javanica	− 7.4	− 8.18	− 42.54	− 43.32
Eriodictyol 5,4’-dimethyl ether 7-O-glucoside	Cleome viscosa	− 7.4	− 7.98	− 47.82	− 45.16
Epicatechin 3-O-(3-O-methylgallate)	Camellia sinensis	− 7.3	− 9.22	− 37.37	− 25.54
Isoengeletin	Smilax china	− 7.3	− 7.48	− 37.76	− 29.82
Murrafoline D	Murraya koenigii	− 7.3	− 9.91	− 39.59	− 39.99
Remerin	Cyperus longus	− 7.3	− 7.53	− 37.31	− 31.23
Paratocarpin	Artocarpus venenosa	− 7.2	− 8.41	− 26.90	− 27.78
Dihydrokaempferide 3-glucuronide	Cleome viscosa	− 7.2	− 7.18	− 36.66	− 32.34
Epigallocatechin	Camellia sinensis/ Averrhoa carambola	− 7.1	− 7.88	− 28.92	− 21.39
Gossypetin 3-galactoside	Crinum latifolium	− 7.1	− 7.42	− 37.88	− 35.01
Reynoutrin	Ricinus communis	− 7.1	− 7.11	− 34.47	− 24.37

Interestingly, seven of the top-binding compounds were from C. sinensis, including epigallocatechin 3-O-(3-O-methyl)-gallate (EGCMG), epigallocatechin 3-O-gallate (EGCG), epicatechin 3-O-gallate (ECG), epigallocatechin 3-O-caffeate, gallocatechin 3’-O-gallate, epicatechin 3-O-(3-O-methylgallate), and epigallocatechin, with EGCG showing the highest binding affinity (-41.50 and − 33.29 in AutoDock Vina and AutoDock4 MM/GBSA, respectively). Despite slight differences in docking poses between AutoDock Vina and AutoDock4, the binding energies remained comparable, as shown in Table 2. The docking position and orientation of EGCG (Fig. 1) and the data of the other compounds (Fig. 2-SI) suggest the potential of these compounds to inhibit nsP2 protease activity.

Fig. 2.

The various aspects of molecular dynamics simulations of CHIKV nsP2 in complex with the EGCG: (A) the distance between Cys478 and EGCGH17; (B) the distance between His548 and EGCGH16; (C) the root mean square fluctuation (RMSF) of N-terminal residues during the molecular dynamics simulation (1000 ns).

All the C. sinensis compounds share several key structural features. They all possess a flavan-3-ol core structure, which consists of two aromatic rings (A and B rings) connected by a three-carbon chain forming a dihydropyran ring (C ring). This core structure is characteristic of catechins. Additionally, each compound has hydroxyl groups attached to the B ring, typically in the 3′, 4′, and sometimes 5’ positions, contributing to their antioxidant properties and molecular interactions. Another common feature is the substitution at the 3-position on the C ring with different groups, such as gallate (a gallic acid ester), methyl gallate, or caffeate. These substitutions are crucial for their binding affinity and specificity in molecular interactions.

The interaction between EGCG and the nsP2 of CHIKV

Given EGCG’s strong binding affinity, its high abundance in C. sinensis³¹, and its well-documented broad-spectrum antiviral activities, it was selected for further MD simulations. Additionally, prior research on EGCG’s effectiveness in inhibiting CHIKV infection²⁶ provided a solid foundation for this choice. Although EGCG was not the top candidate in the MM/GBSA analysis, its significant potential as an nsP2 protease inhibitor made it a compelling focus for MD simulations.

An all-atom MD simulation study was conducted to validate the stability of the predicted EGCG-nsP2 complex and elucidate the potential interactions between EGCG and the nsP2 protease. This approach not only confirms the complex’s stability but also sheds light on the dynamic behaviour of both the EGCG molecule and the nsP2 protease, as well as the crucial binding interactions of the ligand with key residues at the catalytic site. For this purpose, the predicted complexes of nsP2 and EGCG were subjected to AAMD simulation, using the ligand-protein complex structure obtained from Autodock4 docking as the initial configuration. In addition, we performed an independent simulation in which the EGCG ligand was randomly added to the simulation box to independently determine the binding site and observe the binding events.

To quantify the interaction of the EGCG-nsP2 complex, we analyzed the distance measurements of the nsP2 catalytic residues (Cys478 and His548) with EGCG for 1 µs of simulation. Figure 2A,B represents the renewable interactions between these amino acids and EGCG throughout the simulation, representing the binding and unbinding events of EGCG with the protein. For both amino acids, a distance of less than 3 Å is observed, which is achieved after 200 ns for Cys478 and about 340 for His548. This may be explained by the closer location of the Cys within the pocket, which could act as a tethering hook to form initial interactions followed by more stringent interactions. This can also be indicated by the dynamics of binding, where in the studied time interval the binding event was repeated 3 times. Besides, Fig. 2B exhibits a more consistent distance with reduced fluctuations, suggesting a potentially stronger or more stable interaction between His548 and EGCG. The protein-ligand distances generally range between 2.7 Å to 4 Å, indicating a possibly more fixed interaction with different conformation, potentially acting as a secondary pose where EGCG’s mobility is limited or its association is tighter than interactions with Cys.

To assess the structural stability of the protein during binding events, we examined the Root Mean Square Deviation (RMSD) graph (Fig. S3). RMSD value indicates that the protein’s tertiary structure remains consistent throughout the simulation, maintaining its conformation despite potential interactions with EGCG. This further supports the notion of conformational stability within the nsP2 protein complex, implying that EGCG does not trigger significant structural changes and protein stability during the simulation. Additionally, the low RMSD values might hint at the energetic favorability of the protein-ligand complex³², an essential factor for the effectiveness of potential drug interactions.

Combined with Fig. 2A,B, the RMSD analysis presented in Fig. S3 (SI) indicates that EGCG dynamically interacts with the nsP2 without inducing significant structural changes. This suggests that EGCG’s binding impacts local regions of the protein while preserving its overall conformation. Such a dynamic yet stable pattern of interaction is advantageous for drug design, highlighting the potential for effective targeting without destabilizing the protein.

To understand the interaction between EGCG and the nsP2 protease, we assessed the flexibility of residues Cys478 and His548 during the MD simulation, as depicted in the Root Mean Square Fluctuation (RMSF) graph (Fig. 2C). The RMSF graph is crucial for identifying regions within the protein that exhibit significant flexibility or dynamism. Notably, the graph reveals increased mobility within the linker region, which aligns with expectations, given that linker regions are typically flexible.

However, while Cys478 and His548 are pivotal due to their interactions with EGCG, the RMSF data alone do not conclusively suggest that EGCG stabilizes these residues by reducing their fluctuations. The stability of Cys478, in particular, appears unaffected by the ligand, indicating that its high stability in both complexes is likely due to factors other than EGCG binding. For a more comprehensive analysis, we included a comparison of the RMSF values for the unbound protein, which provides further context for evaluating the impact of EGCG on residue stability. Interestingly, the ligand’s stability of His548 seems to be enhanced, as indicated by the lower RMSF values in the ligand-bound protein compared to the apo-protein. Furthermore, the RMSF analysis suggests specificity in the interaction between EGCG and the nsP2 protein complex, with EGCG predominantly associating with a specific binding region on the protein. The altered dynamics of certain residues, especially those proximal to the EGCG binding site, could reflect the ligand’s effect on the protein’s conformation or flexibility.

To predict whether EGCG interacts with the catalytic residues Cys478 and His548, we examined the ΔG_bind values associated with these residues. The molecular mechanics/Poisson-Boltzmann surface area (MMPBSA) analysis, which evaluates the contribution of individual residues to the binding free energy (ΔG_bind) between EGCG and the nsP2 protease (Fig. 3A), reveals that while Cys478 and His548 interact with the ligand, other residues (Leu670, Ala511, Asn547) exhibit more negative ΔGbind values. While Cys478 and His548 are crucial for enzymatic activity, Leu670, Ala511, and Asn547 likely play a key role in stabilizing EGCG within the binding pocket. Their contributions enhance the overall binding affinity, indirectly supporting the inhibitory mechanism. These residues seem to be the most energetically significant for EGCG binding, suggesting they play a more prominent role in stabilizing the interaction than the catalytic residues Cys478 and His548. This implies that the inhibitory effect of EGCG may depend more on interactions with these stabilizing residues rather than direct binding to the catalytic residues.

Fig. 3.

Molecular mechanics/Poisson-Boltzmann surface area (MMPBSA) analysis: (A) Evaluation of individual residue contributions to the binding free energy (ΔG_bind) for the nsP2 protein in complex with EGCG; (B) Overall ΔG_bind for the EGCG-bound nsP2 complex, highlighting the stability and energetics of the ligand–protein interaction across the simulation period.

The MMPBSA analysis (Fig. 3B) indicates that the binding interaction between EGCG and the nsP2 complex is stable, with a slight trend toward stronger binding over time. The moderate and consistent variability in binding energy suggests a stable complex, with no significant disruptions in the interaction throughout the simulation. This stability is further supported by the progression observed in Fig. 3A,B, which illustrate the positioning of EGCG relative to the nsP2 complex at different simulation points.

Initially, at 0 ns (Fig. 4A), EGCG is positioned within the catalytic pocket of the nsP2 protease, close to the surface. As the MD simulation progresses to 1000 ns (Fig. 4B, top), EGCG remains anchored within this catalytic pocket, undergoing subtle yet significant adjustments in its orientation and interactions with the surrounding residues as shown in Fig. S4. These adjustments reflect EGCG’s exploration of the conformational space within the pocket, seeking to optimize its binding interactions.

Fig. 4.

Positional changes of the EGCG-bound nsP2 complex during the molecular dynamics simulation at various time points (top) and the free energy surface (FES) projections along two principal components (PC1 and PC2) of the EGCG-bound nsP2 complex (bottom). EGCG is represented as spheres, nsP2 as a cartoon, and Leu670, Asn547, Cys478, and His548 as magenta, blue, yellow, and orange sticks, respectively. All minima in the FES are highlighted with the corresponding conformations of the inhibitor within the catalytic pocket of the nsP2 protease. In kcal/mol, free energy values are indicated by a color palette ranging from lower (deep purple) to higher (yellow) values.

The principal component analysis (PCA) combined with the free energy surface (FES) mapping further elucidates this process (Fig. 4, bottom). The FES plot, which projects the movement of EGCG along the first two principal components (PC1 and PC2), reveals that the ligand navigates through several low-energy states during the simulation. These energy minima correspond to different stable conformations of EGCG within the same binding pocket. The color gradient in the FES plot and its conformation within this site is dynamically optimized.

The fluctuations observed in the distance plots (Fig. 2A) likely represent these fine-tuning movements within the same binding site, rather than transitions to entirely different binding regions. As EGCG moves within the catalytic pocket, it refines its interactions with key residues, enhancing the stability of its binding. Notably, the stable binding configurations observed at two-time points (128.840 and 630.380 ns) align with the low-energy conformations identified in the FES, supporting the notion that EGCG is engaged in a dynamic but consistent interaction with the nsP2 protease.

Overall, these findings suggest that EGCG’s interaction with the nsP2 complex is both stable and dynamic. The ligand does not migrate between different binding sites instead, it optimizes its binding conformation within the catalytic pocket over time. This dynamic optimization is critical for the effective inhibition of the protease, as it allows EGCG to maintain a strong and stable interaction with the protein, ultimately enhancing the stability and effectiveness of thecomplex.

In vitro studies have further demonstrated EGCG’s effectiveness in inhibiting CHIKV infection, with a notable 60% inhibition rate observed at a 10 µg/ml concentration in HEK293T cells²⁶. Furthermore, EGCG is combined with suramin, it exhibits a synergistic effect against CHIKV in human osteosarcoma cells, providing enhanced protection against the CHIKV strain S27 and two clinical isolates compared to the use of either compound alone³³. However, the precise mechanism behind this inhibition remains unclear. These findings underline the potential biological mechanisms of EGCG action, particularly its ability to modulate protein activity through targeted interactions with the nsP2 protein complex, which may alter the dynamics of specific protein residues and affect their functionality or stability.

Moreover, EGCG has also shown antiviral properties against various virus families, including Retroviridae, Orthomyxoviridae, and Flaviviridae, targeting significant pathogens such as HIV, influenza A virus, and hepatitis C virus. It also disrupts the replication cycles of DNA viruses, including hepatitis B virus, herpes simplex virus, and adenovirus³⁴. However, a limitation of EGCG is its low bioavailability, only 0.14%³⁵. Future research, such as using selenium nanoparticles³⁶, could enhance EGCG’s bioavailability and therapeutic efficacy.

Our findings also differ from prior in silico and in vitro studies. Ivanova et al. (2021) identified compounds 10 and 10c as potential nsP2 inhibitors through ligand-based drug design, demonstrating antiviral activity with IC₅₀ values of 13.1 and 8.3 µM, respectively. These compounds were found to block the active site and catalytic dyad of nsP2, a mechanism that is similar to the way EGCG modulates nsP2 activity³⁷. Kumar et al. (2020) used selective docking and MD simulations to identify pyranooxazole-based inhibitors of CHIKV nsP2, focusing on binding energy and drug-likeness screening, which led to the identification of CMPD167 as a potential nsP2 inhibitor³⁸. Similarly, Nguyen et al. (2015) employed structure-based approaches, including molecular docking and MD simulations, to identify inhibitors targeting multiple binding sites on CHIKV nsP2, with their top hit compounds emerging from database searches using the NCI Diversity Set II³⁰. These insights cover the way for future research on the therapeutic potential of EGCG and similar compounds in the context of CHIKV inhibition.

However, our study has several limitations. First, it relies solely on computational predictions without in vitro experimental validation of the ligand-protein interactions, limiting the confidence and applicability of our findings. Additionally, EGCG’s bioavailability may vary, potentially affecting its efficacy as an inhibitor. Furthermore, our study did not encompass all stages of the viral lifecycle or host interactions, which could influence the overall antiviral potential of EGCG. Despite these limitations, our findings provide valuable insights that pave the way for future research on the therapeutic potential of EGCG and similar compounds in the context of CHIKV inhibition.

Conclusion

This study highlights the significance of nsP2 as a potential drug target for chikungunya virus (CHIKV) inhibition, particularly through its interaction with epigallocatechin gallate (EGCG). The homology analysis revealed that while most CHIKV proteins are non-homologous to human and gut bacterial proteomes, nsP3 showed some similarity to intestinal bacteria, and nsP2 emerged as a key non-structural protein for drug targeting. Molecular docking and dynamics simulations demonstrated that EGCG, a compound found in C. sinensis, has a strong and stable binding affinity to the nsP2 protease, especially around the catalytic site. The research underscores EGCG’s potential as a CHIKV inhibitor, providing a promising foundation for further exploration of its therapeutic applications in antiviral drug development. The mechanism by which EGCG inhibits nsP2 protease activity involves its stable binding within the catalytic pocket of the nsP2 protease, specifically interacting with key residues crucial for enzymatic function. The MD simulations demonstrate that EGCG consistently remains within this pocket throughout the simulation, refining its interactions rather than migrating to a different site. Snapshots from the simulation, taken at energetically stable time points, reveal that while EGCG’s position within the catalytic pocket remains stable, the ligand undergoes subtle adjustments in orientation and interaction networks to optimize its binding.

As the simulation progresses, EGCG evolves its interactions, forming more energetically favorable contacts with residues such as Leu670 and Ala511, while maintaining crucial interactions with catalytic residues Cys478 and His548. These interactions, coupled with EGCG’s adaptability within the pocket, contribute to a stable and effective binding pose, ensuring the protease’s inhibition. The ligand’s movement is best described as a refinement within the same binding pocket, rather than a transition to a completely different site, as supported by FES analysis, which shows EGCG residing in energetically favorable states throughout the simulation.

Distance measurements demonstrate consistent interaction between EGCG and the catalytic residues, with reduced fluctuations indicating tighter binding over time. The RMSD and RMSF analyses further support the structural stability of the nsP2-EGCG complex and highlight changes in protein flexibility around the binding site, reflecting the stabilizing effect of EGCG binding. MMPBSA analysis quantitatively confirms stronger binding as EGCG settles into a stable groove, with a gradual increase in binding affinity over time.

In brief, the MD simulation illustrates that EGCG refines its interactions within the catalytic pocket of nsP2 protease, emphasizing the dynamic interplay of residue interactions in achieving stable and effective inhibition. This finding aligns with the observed binding energies and interaction patterns, highlighting EGCG’s potential as a promising inhibitor of CHIKV nsP2 protease.

Methods

Retrieval of non‑paralogous essential proteins

The essential proteins of CHIKV were retrieved using the GenBank accession number KX097982³⁹. To ensure specificity, paralogous proteins were excluded through a CD-HIT clustering analysis⁴⁰, applying a sequence similarity cutoff of 80% or higher.

Human host non‑homologous

The homology between the CHIKV proteome and the human host proteome Homo sapiens (taxid: 9606) was assessed through comparative sequence analysis using the BLASTp tool⁴¹ with cut-off expectation value (E-value) 10^− 4. Threshold values of no more than 35% for both query coverage and sequence identity were applied during this analysis, following the previous parameters⁴². Proteins that showed significant similarity to the human proteome were eliminated, and the remaining non-homologous proteins were shortlisted for additional investigation. Furthermore, a search for non-homology against sequences of proteins from the human gut microbiota was conducted, employing a threshold cutoff defined by an E-value of 0.001 or greater^43,44.

Ligand preparation

To find a potential candidate for treating Chikungunya infection, molecular docking analysis was performed over Indonesian herbal compounds (herbal DB database⁴⁵). The retrieval of ligand data was carried out using a Python program and the Selenium library through scraping techniques. The information collected included the names of the compounds, Knapsack IDs, and names of the plant species. The Knapsack ID was utilized to extract data from the KNApSAcK Family database using the same scraping technique⁴⁶. The extracted information comprised molecular formulae, molecular weights, InChIKeys, SMILES, and references. The InChIKey provided the IUPAC (International Union of Pure and Applied Chemistry) name of the related compound. The 3D structures of the compounds were generated from SMILES using the Gypsum-DL program and saved in SDF and PDB formats⁴⁷. The ligand in .pdb format was converted to .pdbqt using the Raccoon v1.0 program with default settings⁴⁸. A total of 6,092 compounds were obtained from the HerbalDB and KNApSAcK Family databases. After removing duplicates and generating three-dimensional structures, 5,550 compounds were used for virtual screening.

Structure modeling and visualization

The Chikungunya virus nsP2 protease model was generated using AF2⁴⁹, utilizing ColabFold as the platform⁵⁰. A total of 25 predictions were made (five seeds per model). The highest confidence protein model was taken for visualization and analysis.

Molecular docking analysis

Molecular docking studies were executed using Autodock Vina and Autodock4 on two platforms: Ubuntu and Google Colab Pro. The ligand dataset was docked into the catalytic pocket of the non-structural protein 2 (nsP2) protease. The preparation of the receptor model was conducted using the AutoDock Tools program⁵¹.

Virtual screening processes were carried out on both Ubuntu and Google Colab Pro operating systems using AutoDock Vina⁵², included the receptor name, ligand name, grid box size (20 × 20 × 20 Å), and grid box location centered at x -0.279, y − 3.238, and z 24.188. All other parameters were left at default settings. Subsequently, the top 100 ligands, selected based on their optimal binding affinities, were subject to re-docking on the receptor using AutoDock-GPU. This version of AutoDock4, enhanced by OpenCL and Cuda for accelerated performance⁴⁸, was employed on Google Colab Pro. A Grid Parameter File (GPF) was generated using the AutoDock Tools program. The GPF included the receptor name, ligand name, grid box size (60 × 60 × 60 Å), and grid box location at x − 0.279, y − 3.238, and z 24.188. The remaining parameters were set to default values. The GPF was then input into AutoGrid4, which produced several output files, including the Grid Map Field File (.maps.fld). Docking with AutoDock-GPU utilized this Grid Map Field File, with additional parameters such as the best output pose (gbest: 1), the number of docking cycles (nrun: 100), population size (psize: 100), and atom derivative type (derivtype: -T). Finally, to rank our docking results, we employed the MM/GBSA tool (Uni-GBSA)⁵³ for predicting binding affinities.

Molecular dynamic simulation

We performed MD simulations of the Chikungunya virus nsP2 protease in aqueous salt solutions by atomistic classical (AA) MD simulation within the presence and absence of the selected above EGCG ligand. For the atomic model of the nsP2 protein, containing both the helicase and protease subdomains, the predicted AF2 structure was used as an initial model.

All-atom MD simulations were performed with the MD package Gromacs (version 2020)^54,55 using the CHARMM36 force field⁵⁶. The protein model was placed in a cubic box, solvated by ∼15,999 waters with the TIP3P⁵⁷ water model, and neutralized, and 0.15 M NaCl was added. The total number of atoms in the simulation box was ∼53,313. After solvation, the system was subjected to energy minimization using the steepest descent algorithm until the maximum force of 1000 kJ mol^− 1 nm^− 1 was achieved. The systems were optimized and equilibrated for at least 1 ns in the NVT ensemble. After the system was simulated with restraints for 20 ns twice in the NPT ensemble with the Nose-Hoover thermostat^58,59 and the Parrinello-Rahman barostat^60,61 with a reference temperature and pressure of 303.15 K and 1 bar, respectively. The production run was carried out for 1 µs using the NPT ensemble (constant pressure, constant temperature) without backbone restraints for each system three times. The non-bonded interactions were treated using Verlet cut-off scheme, with the particle mesh Ewald (PME) method⁶² to treat long-range electrostatic interactions, while the short-range electrostatic and van der Waals interactions were calculated with a (real space) cut-off of 12 Å. Periodic boundary conditions were applied to all simulations, and bonds involving hydrogen atoms were constrained using the linear-constraint-solving (LINCS) algorithm.

MD analysis: We analyzed the temporal evolution of the interaction distances between of CHIKV nsP2 and the EGCG ligand based on the center of geometry of the Cys478 and His548 residues to the center of geometry of the EGCGh17 and EGCGh16 residues, respectively. Visual inspection of the trajectories was performed with VMD⁶³ and PyMOL (DeLano Scientific, Palo Alto, CA, USA).

In addition, the binding free energy of the molecular dynamics (MD) simulations was evaluated using the Molecular Mechanics Poisson-Boltzmann Surface Area (MMPBSA) method. Calculations were performed with the gmx_MMPBSA program, version 1.6.3⁶⁴. For these calculations, the “forcefields” parameter was omitted in the input file to support the CHARMM force field. The residue decomposition analysis was configured with a distance threshold of 4 Å. Additionally, the PCA was used to analyze the conformational changes and key motions of the EGCG-nsP2 complex based on a previous study⁶⁵. The most significant motions were identified by the eigenvectors with the highest eigenvalues. The first two principal components (PC1 and PC2) defined the essential subspace of the protein-ligand dynamics. The lowest energy conformations were evaluated using the free energy surface (FES) projected along these components.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

I.S, A.G.S, and M.N investigated and analyzed the data; I.S and M.N wrote and edited the main manuscript text; I.S, A.G.S, and M.N prepared the figures and Tables; K.V supervised, reviewed, and edited the manuscript.

Funding

This research was funded by Dana Abadi Perguruan Tinggi-Lembaga Pengelola Dana Pendidikan (DAPT-LPDP) through the national research collaboration funding program (Riset Kolaborasi Nasional) with the Grant No. 500/IT3.D10/PT.01.03/P/B/2023.

Data availability

Publicly available essential proteins of CHIKV were deposited into the GenBank database under accession number KX097982 and are available at the following URL: https://www.ncbi.nlm.nih.gov/nuccore/KX097982.1/. The retrieval of the herbal library (ligand data) from the herbal DB database (http://herbaldb.farmasi.ui.ac.id) is not currently available due to the website being down and is available from the corresponding author upon reasonable request. Data are located in controlled access data storage at Universitas Indonesia and data constructed based on this publication can be accessed at the following URL: https://www.phcogj.com/sites/default/files/PJ-11-6-39.pdf. The nsP2 structure model was derived from computational analysis with AlphaFold2, utilizing ColabFold as the platform and are available at the following URL https://colab.research.google.com/github/sokrypton/ColabFold/blob/main/AlphaFold2.ipynb.

Declarations

Competing interests

The authors declare no competing interests.