Introduction

Mayaro virus (MAYV), which in Brazil was previously considered to be restricted to Amazon riverside regions, has been increasingly studied after being isolated in recent years in urban centres of the states of Mato Grosso [1], Goiás, Tocantins [2] and São Paulo [3]. In forest areas, mosquitoes of the genus Haemagogus are the principal vectors for MAYV. However, the confirmation of their potential to be transmitted by Aedes aegypti reveals a latent public health threat [4]. Mayaro fever is a debilitating and widely neglected disease that is beginning to threaten urban areas [5]. In the endemic region of Amazônia, many cases of Mayaro fever go undiagnosed. It is estimated that approximately 1% of all cases diagnosed as dengue in the northern regions of South America are caused by MAYV [6]. This fact, together with insufficient epidemiological surveillance and the lack of precise diagnostic methods, is among the reasons for the low level of recognition of this disease.

MAYV is closely related to chikungunya virus (CHIKV) and produces a highly debilitating clinical condition, characterized by severe arthralgia that can last for weeks or months [7]. According Halsey et al., 54% of the patients infected by MAYV develop persistent arthralgia affecting the major joints [8]. Little is known regarding the process of induction of arthralgia in infected individuals. There is evidence that arthropathy caused by arthritogenic viruses is due to inflammatory responses induced by the virus or the viral products inside the joint tissues [9].

Currently, there is no vaccine or therapy available for MAYV infection. Symptoms are treated using simple analgesics and/or non-steroidal anti-inflammatory drugs, which provide relief to the patients [10]. Santiago et al. studied the long-term immune response in patients infected by MAYV and showed that all of the patients produced antibody against MAYV [11]. Additionally, the intensity of the response did not differ between patients who developed severe arthralgia and those who recovered completely. This finding revealed that a neutralizing antibody response was not sufficient to protect the individuals from developing arthralgia, especially because the illness correlated with the capacity of the antibodies to reach and neutralize the virus possibly located in various joints. These authors therefore suggest that the development of vaccines seems to be an ineffective option for protection against chronic arthralgia and that antiviral substances still represent a better option for the treatment of infections by arthritogenic alphaviruses. In this context, the identification of compounds with antiviral potential as well as anti-MAYV targets is of great importance.

There is still a lack of data available in the literature on the three-dimensional structures of MAYV proteins. Most of the information available is related to other alphaviruses. The genome of MAYV consists of a single, positive-sense RNA strand containing two open reading frames (ORF) that are separated by an intergenic region. The first ORF of the virus genome can be read directly and codes for four non-structural proteins, nsP1, nsP2, nsP3 and nsP4, which are involved in the transcription and replication of the viral RNA [12]. The second ORF is translated as a result of programmed ribosomal frameshifting events from a subgenomic 26S RNA. After undergoing cleavage by enzymes present in the host and by autoproteolytic activity of the C-terminal domain, the structural proteins E1, E2, E3, 6k and C or TF, E2, E3, 6k and C [12,13,14,15] are produced in the host cell. The C protein is a component of the nucleocapsid, responsible for the packaging of the viral RNA [16] and is involved in viral particle formation through its interaction with the cytoplasmic domain of glycoprotein E2 in the cell membrane [17]. The structure of the C protein has been determined for several alphaviruses and has been found to contain a conserved hydrophobic pocket that is involved in protein-protein interactions and represents a promising target for the development of antiviral agents [18,19,20].

Natural products have a large variety in terms of their structure and physicochemical and biological properties, which makes them attractive in drug development studies. Compounds obtained from members of the genus Salacia are utilized in different countries in folk medicine, primarily against diabetes and as anti-inflammatory drugs. Compounds of the pentacyclic triterpene (TTPC) and flavonoid classes are primarily found in Salacia crassifolia leaf extracts [21].

In view of the important role of protein C in the replication cycle of alphaviruses, in this work, we attempted to model the three-dimensional structure of the MAYV protein C and use it as target for virtual screening studies for different compounds. Modelling results were confirmed by bioassays in order to search for novel drugs against MAYV.

Materials and methods

Source of the primary sequence and building the model

The primary sequence of MAYV protein C was obtained from the National Center for Biotechnology Information (NCBI) database [22] in the FASTA format. The MAYV protein C model was built using the Swiss-Model portal, based on a target-template alignment, using the structure of the Aura virus C protein complexed to a dioxane molecule (PDB accession no. 4AGJ_A) [23] as a model. The search of suitable templates was performed using the Basic Local Alignment Search Tool (BLASTp).

Evaluation of the quality of the model

The global quality of the model was evaluated based on the root-mean-square deviation (RMSD) value, obtained by overlapping the peptide chain of the conserved regions of the generated model using the Swiss-PdbViewer programme [24]. Validation of the stereochemical quality of the model was performed using a Ramachandran plot, which was generated using RAMPAGE [25], and by the QMEAN Z-score [26], which relates the scores obtained for parameters such as local geometry, distance-dependent interaction potential, and solvation potential of the constructed model, with scores being obtained for reference structures determined experimentally by X-ray crystallography.

Inverse virtual screening

Inverse virtual screening was performed using the Octopus 1.0 framework [27]. The ligands were obtained from the ZINC database [28] and the related literature. The predicted charges of the compounds at physiological pH were calculated, and possible tautomers were evaluated using the Marvin Sketch programme [29]. Since the protein C model had already been built and validated, re-docking of the crystallographic ligand present in the template (dioxane) was performed in order to validate the methodology. To explore the binding site of the dioxane and the flexibility of the ligand, a box was built for the modelled protein C of MAYV using the AutoDock Tools programme [30]. This box was defined as a cube with dioxane-centric geometry with dimensions of 14 x 14 x 14 Å, spacing of 1 Å and coordinates X = -23.833, Y = 11.281 and Z = -9.142. Next, the other ligands were docked against the MAYV protein C model utilizing the AutoDockVina 1.0.2 programme [31].

Phytochemical study

The leaves of Salacia crassifolia (Mart. ex Schult.) G. Don, Celastraceae, were collected in Montes Claros, Minas Gerais, Brazil, and the species was identified by Dr. Maria Olívia Mercadante-Simões of Universidade Estadual de Montes Claros, Brazil. A voucher specimen (No. HBCB 144624) was preserved in the herbarium of Instituto de Ciências Biológicas, UFMG. The plant materials were collected as authorized by the Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq), Brazil (Process: 010119/2014-0).

The purification processes were performed by column chromatography (CC) using silica gel 60 (0.063-0.200 mm) as the stationary phase. Pure organic solvents or organic solvents in mixtures of increasing polarity were used as mobile phases. Silica gel 60 (Merck) was used to prepare plates (0.25 mm) for analytical thin-layer chromatography (TLC).

The ethyl acetate extract (10 g) from S. crassifolia was subjected to silica gel CC eluted with hexane (Hex), chloroform (CHCl3), ethyl acetate (EtOAc), and pure methanol (MeOH) or in mixtures of increasing polarity. The fraction (439.1 mg) obtained with EtOAc/MeOH (4:6) after the removal of the solvent was subjected to silica gel CC, eluted with Hex, CHCl3, EtOAc and MeOH, pure or in mixtures of increasing polarity. An AcOEt/MeOH (25:75) fraction was isolated as a solid. The weight of the sample was 139.0 mg, and the melting point was 154.2-158.8°C.

1H (400 MHz) and 13C (100 MHz) NMR experiments were carried out on a Bruker Avance DRX-400 instrument operating at 300 K. The chemical shift assignments (δ) were expressed in parts per million (ppm), and coupling constants (J) were expressed in Hertz (Hz). Tetramethylsilane was used as an internal standard (δ H = δ C= 0). Melting points were determined on an MQAPF-302 apparatus (Microquímica Equipamentos Ltda) [32].

Cells and viruses

Vero cells (derived from African green monkey kidney, ATCC CCL-81) were cultured in Dulbecco’s modified Eagle’s medium (DMEM) (Cultilab, BRA) supplemented with 5% fetal bovine serum (FBS) (Cultilab, BRA) and 100 U of penicillin, 100 μg of streptomycin, and 2.5 μg of amphotericin B (Sigma-Aldrich, USA) per mL. Cells were maintained at 37°C in 5% CO2.

Mayaro virus strain BeAr 20290, kindly provided by Dr. Maurício Lacerda Nogueira (Faculdade de Medicina de São José do Rio Preto / Famerp), was propagated in Vero cells and titrated using a plaque assay [33]. The viral stock was kept at -80°C until the time of use.

Cytotoxicity assay

Cytotoxicity was evaluated using the colorimetric MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) method [34]. Vero cells grown in 96-well microplates (5 × 104 cells/well) were treated with different concentrations of compounds (0 to 1.723 μmol/mL) diluted in DMEM (containing 2.5% FBS) and incubated at 37°C. After 48 hours, the medium of the wells was replaced by 25 μL of MTT solution (2 mg/mL in 1x PBS), and the plates were incubated for 2 hours at 37°C in 5% CO2 to allow the tetrazolium salt to be metabolized by the viable cells, after which 100 μL of DMSO was added. The optical density of the solution in the wells of the microplate was determined by spectrophotometry at 492 nm. Using regression analysis, absorbance values to determine of the concentration of the compound that reduced cell viability by 50% (CC50) compared to the viability of the untreated cells.

Global antiviral activity assay

For the global antiviral activity, 90% confluent Vero cells (5 × 104 cells/well) in 96-well microplates and the viral inoculum were pre-treated for 30 minutes at 37°C with different dilutions of the compounds, ranging from 0.013 to 0.574 μmol/mL. This pre-treatment was performed to include all of the stages of the viral cycle at which the compound could act, i.e., in adsorption, penetration, viral replication or virucidal effect. Next, the pre-treated cells were infected with the pretreated viral inoculum at a multiplicity of infection (moi) of 0.1 or 5. To evaluate other conditions, the treatment was done at different times in the viral life cycle. To analyze the virucidal effect, the drug was mixed with the viral inoculum 1 hour before infection. After 48 hours of incubation at 37°C in 5% CO2, cell viability was also determined using the MTT colorimetric method [34]. The optical density of the solution was determined by spectrophotometry at 492 nm. Using regression analysis, absorbance values were used to calculate the 50% effective concentration (EC50), the concentration that resulted in protection of 50% of the cells from viral infection. Ribavirin (Sigma-Aldrich, USA) and DMSO (Sigma-Aldrich, USA) were used as a positive and a negative control, respectively. The selectivity index (SI) was expressed as the CC50/EC50 ratio, and the relative potency (RP) was expressed as the ratio of the EC50 of ribavirin (reference antiviral) to that of the tested compound. When appropriate, the virus concentration (in plaque-forming units [PFU] per mL), was determined using the same culture supernatant.

Replication kinetics

To examine the replication kinetics of Mayaro virus in the absence or presence of epicatechin, 90% confluent Vero cells (5 × 104 cells/well) in 96-well microplates and the viral inoculum (moi, 5) were pre-treated for 30 minutes at 37°C with epicatechin at a concentration of 0.344 µmol/mL. The pre-treated cells were then infected with the pre-treated viral inoculum for 1 hour at 37°C. After viral adsorption, the inoculum was removed and the cells were washed with 1x PBS, after which DMEM medium containing epicatechin at a concentration of 0.344 µmol/mL was added. The supernatants from the wells were collected at 1, 3, 6, 12, 24 and 36 hours postinfection (hpi), and the virus yield was determined by plaque assay [33].

Statistical analysis

Regression analysis was performed and the mean ± standard deviation of triplicate samples was determined using the statistical software GraphPad Prism® v. 7 (GraphPad Software, Inc. La Jolla, California, USA). A one-way ANOVA test was used to compare treatments. Dunnett’s test was used to determine differences between experimental groups and a control group (Statistica software 7.0, StatSoft, Inc., 2004). All tests were considered statistically significant when the P-value was less than 0.05.

Results

Homology modelling

A template search using an X-ray crystal structure of Aura virus protein C complexed to a dioxane molecule (PDB 4AGJ_A com) at 1.98 Å resolution, a similarity of 63% with the protein C of the MAYV was determined with a satisfactory score of 201 and an e-value of 3e−64 with limit values of 30% and 50e−7, respectively [35]. The values obtained corresponded to a root-mean-square deviation (RMSD) of 0.44 Å for the carbon alpha atoms (αC) (Fig. 1a), which was lower than that of 2 Å and therefore, considered of good quality [36]. The stereochemical quality was confirmed by a Ramachandran plot in which 93.2% of the amino acid residues were in favourable regions, 4.7% were in allowed regions, and only 2.0% were in disallowed regions (Fig. 2). The quality of the model was confirmed by a QMEAN Z-score that was close to zero, as would be expected for crystallographic structures with high resolution [26] (Fig. 3). The model was therefore regarded as appropriate for use in the subsequent docking study.

Fig. 1
figure 1

Construction of a three-dimensional model of MAYV protein C by homology modelling. (a) Superimposition of the crystal structure of Aura virus protein C (red) used as a template and the model of the MAYV protein C (yellow) (b) Three-dimensional structure of the model of MAYV protein C (ribbon representation) complexed with dioxane (represented by tubes)

Full size image
Fig. 2
figure 2

Ramachandran plot of the modelled protein C of MAYV. Evaluation of the stereochemical quality by the angles φ and ψ of the amino acid residues of the model generated using the RAMPAGE programme

Full size image
Fig. 3
figure 3

Quality estimate of the stereochemistry of the model generated for MAYV protein C. (a) Z-score and (b) QMEAN calculated using the Swiss-Model Workspace

Full size image

Inverse virtual screening

Re-docking was performed to validate the virtual screening method. A comparison of the crystallographic and the docked dioxane conformations yielded an RMSD value of 1.92 Å (Fig. 4a). Values around 2 Å are considered satisfactory for re-docking, [37]. The binding of the dioxane to MAYV protein C exhibited a binding energy value of -2.8 kcal/mol. Then, 590 molecules were docked to the protein C site, and their respective binding energy was compared to the value (-4.9 kcal/mol) obtained for the docking of the 1,3-bis{(R)-[1,4]dioxan-2-il}propane (bisdioxane) molecule to MAYV protein C. According to Kim et al. [38], this substance exhibited potential for interaction with a similar binding site on the Sindbis virus protein C and induced antiviral effects in bioassays performed by the same authors. Of the 590 molecules evaluated, various molecules were predicted to have higher affinity for the binding site of the MAYV protein C than that of bisdioxane (Online Resource 1). Among the substances with potential antiviral activity, flavonoids stood out as a promising compound class. Six flavonoids compounds (1 to 6) were therefore docked to the binding site of the MAYV protein C with the same conformation (Fig. 4b). All of the flavonoids evaluated bound to the binding site with favourable free energy values (Online Resource 2). Of these, epicatechin (-6.8 kcal/mol) was considered the most promising. Thus, to confirm the antiviral activity in this compound, we performed bioassays using epicatechin isolated from S. crassifolia (Celestraceae).

Fig. 4
figure 4

Configuration of the three-dimensional structure of the molecules docked to MAYV protein C. (a) Re-docking of dioxane in the model of MAYV protein C. Red, crystallographic structure; yellow,: structure docked. (b) Cluster of flavonoids in the active site of the modelled MAYV protein C

Full size image

Cytotoxicity, antiviral activity and replication kinetics

Epicatechin was used in biological assays, as it was one of the six flavonoids with favourable binding energy to MAYV protein C (-6.8 kcal/mol). Its purity was confirmed by hydrogen and carbon (1H and 13C NMR) nuclear magnetic resonance spectroscopy (Online Resource 3).

In the cytotoxicity assay, treatment of cells with epicatechin, even at the highest concentration used, did not reduce the cell viability to 50% (CC50). However, potential cytotoxicity of epicatechin was detected at 1.704 ± 0.101 μmol/mL, with only 10% reduction in cell viability.

To evaluate global antiviral activity by the MTT technique, 0.430 μmol of epicatechin per mL (Fig. 5d) was able to protect approximately 100% of the MAYV-infected cells when compared to infected and untreated cells (p < 0.001) (Fig. 5b). The nucleoside analogue ribavirin (Fig. 5e), which was used as a classical positive control, was found to have an EC50 above 0.486 ± 0.008 μmol/mL and was therefore less potent than epicatechin. Because reducing the concentration also decreased the protection of cells from viral infection, the antiviral effect of epicatechin appears to be dose dependent (Fig. 6a). The EC50 value was 0.247 ± 0.026 μmol/mL (Table 1). A dose-dependent antiviral effect of epicatechin also was observed at an moi of 5, with an EC50 value of 0.413 ± 0.004 μmol/mL (Fig. 6b).

Fig. 5
figure 5

Antiviral activity of epicatechin in Vero cells infected with MAYV. Vero cells were treated with epicatechin or ribavirin and infected with MAYV (moi, 0.1) and photographed after 48 hours of infection. (a) Uninfected and untreated cells, (b) infected and untreated cells, (c) infected cells treated with DMSO, (d) infected cells treated with 0.430 μmol of epicatechin per mL, (e) infected cells treated with 1.024 μmol of Ribavirin per mL (magnification, 100x)

Full size image
Fig. 6
figure 6

Anti-MAYV activity of epicatechin. The anti-MAYV activity of epicatechin was evaluated by treating MAYV-infected cells with 0–0.574 μmol of this compound per mL for 48 hours. The antiviral effect of epicatechin was dose dependent in (a) cells infected with MAYV at an moi of 0.1 and (b) cells infected with MAYV at a moi 5. MAYV-infected cells at an moi of 0.1 were also treated with ribavirin (reference substance) and DMSO. The data presented are the mean ± SD

Full size image
Table 1 Cytotoxicity and anti-MAYV activity of epicatechin and ribavirin (reference substance)
Full size table

The selectivity index (SI) and relative potency (RP) are used as selection criteria for promising drugs. The SI value for epicatechin was 7, making it twice as potent as ribavirin.

Analysis of replication kinetics showed a strong inhibitory effect of epicatechin on MAYV growth, with reduction of least four logs when compared with the MAYV-infected cells without drug treatment (Fig. 7). In the earliest phase (0 to 6 hours), MAYV was not detected. After 12 hours, virus was detected without significant variation in the yield (less than one log).

Fig. 7
figure 7

Growth kinetics of MAYV in Vero cells. Cells and virus were treated with 0.344 µmol of epicatechin per mL before and during infection (moi, 5). Infected and untreated cells were used as virus controls. Viral progeny in the supernatant was quantified at various time points postinfection by plaque assay

Full size image

Discussion

The three-dimensional structure of MAYV protein C was built from the coordinates of the Aura virus protein C complexed to a dioxane molecule (accession no. PDB 4AGJ_A), to which it had 63% sequence identity. According to Takeda-Shitaka et al., models built with more than 50% sequence identity are sufficiently accurate and suitable for application in the development of new drugs [39]. The evaluated three-dimensional structure confirmed the quality of the model, confirming its possible application in molecular docking studies proposed for antiviral development.

Aggarwal et al. [23] determined the X-ray crystal structure of Aura virus protein C, which was utilized as a template in this work, and found that a dioxane solvent molecule, which was used in the crystallization process, occupied the protein C hydrophobic pocket in the a region that is believed interact the viral glycoprotein E2. According to these authors, the dioxane was inserted in the same position that would be occupied by the pyrrolidine ring of the proline 405 residue (Pro 405), a constituent of the loop region of the helix-turn-helix motif of glycoprotein E2, which is conserved in alphaviruses [23]. This suggests that the dioxane or a similar molecule may compete with binding of Pro 405 in this hydrophobic pocket and disrupt the protein C-E2 interaction, preventing the spread of the infection. The dioxane molecule is smaller than the binding site (Fig. 1b), and therefore, the few contacts that it makes culminate in a binding energy value of only -2.8 kcal/mol. In elucidating the three-dimensional structure of Sindbis virus protein C, Lee et al. also found that dioxane bound to a hydrophobic core [40]. However, in that study, the hydrophobic core was occupied by two dioxane molecules. Based on those results, Kim et al. [38] synthesized molecules containing two covalently bound dioxane molecules that were predicted to interact more strongly than the dioxane molecule alone. In vitro biological tests performed by these authors with this molecule, bisdioxane, demonstrated an antiviral effect of the compound against Sindbis virus [38]. The docking of bisdioxane to the modelled protein of MAYV resulted in a binding energy of -4.9 kcal/mol, and this value was used as a parameter for the selection of promising antiviral molecules against MAYV, since that compound showed activity against a virus that also belongs to the genus Alphavirus. Of the 590 molecules docked to protein C, six flavonoids bound to the binding site with favourable energy, and all showed higher affinity than bisdioxane.

The potential of flavonoids as antivirals has been described in the literature for some viruses. For example, studies have shown antiviral activity of these compounds against dengue virus [41,42,43,44,45], chikungunya virus [46], hepatitis C virus [47] and influenza virus [48]. Flavonoids have been reported to negatively impact the formation of the viral RNA polymerase complex [49] and the hemifusion between virus particles and the cell membrane. This reduces the integrity of the viral membrane, resulting in the loss of the possible internalization of virus particles [50] in the final stages of the viral replication cycle [51]. Additionally, flavonoids are potent inhibitors of entry, acting on the viral particle even before it is adsorbed to cellular receptors [52]. However, it remains unclear whether the antiviral activity of flavonoids is based on their interactions with one or more targets.

In the present study, molecules with promising antiviral properties against Mayaro virus were identified by virtual screening. The activity of epicatechin flavonoid (EC50, 0.247 μmol/mL) isolated from Salacia crassifolia, a plant species of the family Celastraceae, was confirmed by in vitro biological assays. This result was consistent with those reported by Takeshita et al., who demonstrated antiviral activity of epicatechin against hepatitis C virus (HCV), with an EC50 of 0.317 μmol/mL. Like MAYV, HCV is an enveloped virus with a single-stranded positive-sense RNA genome [53]. In cytotoxicity assays, epicatechin was found to be non-cytotoxic to Vero cells (CC50 > 1.723 μmol/mL). The SI value of 7 demonstrates that epicatechin is able to inhibit viral replication at a concentration seven times lower than the cytotoxic dose [54]. Epicatechin showed strong inhibition of MAYV growth at 36 hours postinfection (reduction of ~4 logs). After 12 hpi, MAYV was detected in small amounts without significant variations in its kinetics. The reason why the effect of epicatechin is decreasing after 12 hours is not clear. Since a high moi was used, it could be speculated that some virus evaded the virucidal action of epicatechin and initiated a subtle infection. Moreover, few viruses could escape, due to the presence of epicatechin in the extracellular environment, culminating a single cycle of infection and a low level of virus production.

In the search for mechanisms of action against MAYV, preliminary results in our laboratory indicate that a blockade of infection by epicatechin is easily observed when the compound is added in advance to the viral inoculum (Table 2). When cells and virus inoculum (global antiviral effect) or only viral inoculum (virucidal effect) were pre-treated with epicatechin before infection, a powerful effect was observed (p < 0.001). When only the cells were treated (1 hour) and rinsed prior to viral adsorption (adsorption effect), no inhibitory effects were observed (p > 0.05). Under the effective antiviral conditions, a six-log reduction in virus production was observed. When added after infection, epicatechin on average reduced virus production by two logs in the first 12 hours (data not shown). This is thought to be due to the action of epicatechin in the second cycle of infection, when mature virions are released into the extracellular environment, which is common to alphaviruses. Considering the low moi, we also do not rule out an intracellular action of epicatechin in the late stages of the viral replication. This action would be able to block important stages of the cycle and formation of new viral progeny. Thus, our data together indicate that epicatechin acts by binding to components of the viral particle, but not to cellular components. Details of the mechanism of action, such as the stage of the viral cycle during and/or after viral penetration, possible synergistic interactions with other antivirals, and development of an in vivo assay are part of an upcoming study by our group. This study was able to identify promising antiviral molecules by virtual screening, and the antiviral activity of epicatechin was confirmed by bioassays.

Table 2 Anti-MAYV activity of epicatechin (0.344 µmol/mL) at different times in the viral life cycle
Full size table