In vitro and in silico studies of the antiviral activity of polyhydrated fullerenes against influenza A (H1N1) virus

As of today, influenza viruses remain a relevant target for the development of antiviral compounds due to their rapid evolution and acquisition of the resistance to existing drugs. Fullerene derivatives have already shown the ability to successfully interact with viruses, and polyhydrated fullerenes (or fullerenols) are particularly attractive due to their compatibility with biological fluids and low toxicity. Therefore, the goal of this work was to study the effect of two batches of a mixture of polyhydrated fullerenes with a mass ratio of 78.1% C60/C70 and 21.9% C76/C78/C84 on the influenza A (H1N1) virus. It was determined that the mixture of fullerenols, along with the low toxicity, showed high antiviral activity with a decrease in the viral infectious titer up to 4 orders of magnitude. In addition, studied fullerenols did not affect the hemagglutination process and did not show any significant prophylactic activity. With the help of molecular docking and molecular dynamics simulation, the likely target of fullerenols' action was determined—the binding site of the RNA primer of the viral RNA-dependent RNA polymerase. Therefore, we assume that the high antiviral effect of polyhydrated fullerenes on influenza A virus is related to their interaction with the viral RNA polymerase.

endonuclease inhibitor of influenza virus, was the last to enter the pharmaceutical market, but it was approved for treatment and prevention only in people over 12 years of age 5 . Moreover, the situation with SARS-CoV-2 and a phenomenon of its co-infection with influenza viruses acutely demonstrate the need for antiviral agents with different mechanisms of action.
The need for safe drugs that can be used for all vulnerable categories remains, and recently scientists have paid more attention to nanomaterials, which have found wide application not only in the medical field. Unique carbon nanostructures-fullerenes-attract the attention of many scientists and specialists, primarily because of their chemical and physical properties, such as size, symmetry, stability, three-dimensionality and electronic configurations. As of today, there are already confirmations of the effect of fullerenes and their derivatives on HIV 6 , hepatitis C virus 7 , influenza A virus 8 , herpesviruses 9 , as well as an in silico studies of the effect on SARS-CoV-2 10 .
The main problem of using fullerenes in biology and medicine is high hydrophobicity, and, as a result, very low solubility in water 11 . However, this problem is solved by modifications of the molecule. To date, hydrated fullerenes or fullerenols with the general formula C n (OH) x are sufficiently well studied 12 . The formation of soluble fullerenes makes it possible to combine them with liquids, including physiological ones (blood, lymph), which expands the range of applications 13 . In addition, hydrated fullerenes are stable and have relatively low toxicity, which undoubtedly makes them attractive candidates for the role of drugs, in particular, antiviral agents. That is why the aim of our work was to study the effect of two forms of the mixture of polyhydrated fullerenes K050 and K400 on the influenza A (H1N1) virus.

Results
Based on the results obtained after the MTT assay, CC50 indicators (50% cytotoxic concentration) were calculated. For HyFn400 and for the reference-drug oseltamivir the indicators were 11.79 ± 0.58 mg/ml and 0.29 ± 0.01 mg/ml, respectively. HyFn050 at its maximum concentration of 2.5 mg/ml did not show any toxic effect on cells. Accordingly, due to the absence of toxicity it was impossible to calculate CC50 for the latter. Antiviral activity and determination of infectious titer de novo. The results of the calculation of the antiviral activity showed a clear dose-dependent inhibitory effect for the studied polyhydrated fullerenes (Fig. 1A,C). We observed the greatest activity at the highest tested concentrations: HyFn050 (1.5 mg/ml) inhibited H1N1 reproduction by 34.07 ± 1.70%, and HyFn400 (5 mg/ml) by 96.48 ± 4.82%. Oseltamivir at a concentration of 0.15 mg/ml showed the result of 22.33 ± 1.11% antiviral activity.
Determination of the infectious titer of the virus after the antiviral activity study of the fullerenols (that is, in the cell supernatant) showed its decrease in general by 2-4 orders of magnitude. The HyFn400 mixture showed a decrease in the infectious titer at all tested concentrations (Fig. 1D). The greatest drop in the infectious titer by 4.02 orders of magnitude was observed at a concentration of 2.5 mg/ml compared to the virus control (TCID50 = 10 10.61 ). HyFn050 showed an effect on the TCID50 value only at the highest tested concentration of 1.5 mg/ml: a decrease of 2.13 orders of magnitude at the TCID50 of the virus control of 10 7.65 (Fig. 1B). The activity of oseltamivir at a concentration of 0.15 mg/ml was manifested in a decrease in the infectious titer of the virus by 3.14 orders of magnitude compared to the control (TCID50 = 10 10.61 ) (Fig. 1D).

Prophylactic activity and determination of infectious titer de novo. During 24 h of incubation
with cells before influenza infection, hydrated fullerenes did not show high prophylactic activity. HyFn050 and oseltamivir did not affect the inhibition of the infection process at all ( Fig. 2A). Fullerenols HyFn400 demonstrated a slight prophylactic activity at the lowest tested concentration of 0.5 mg/ml with the result of 4.00 ± 0.20% (Fig. 2C).
The calculation of the infectious titer after the prophylactic activity study of the compounds showed the following effect of fullerenols. HyFn0400 showed a slight decrease in the TCID50 value by 0.89 orders of magnitude compared to the virus control (TCID50 = 10 7.59 ) at a concentration of 2.5 mg/ml (Fig. 2D). Similar activity was observed for HyFn050: in the concentration range of 0.05-1.5 mg/ml, the infectious titer decreased by an average of 0.9 orders of magnitude, compared to the virus control (TCID50 = 10 7.52 ) (Fig. 2B). And oseltamivir at a concentration of 0.15 mg/ml showed a decrease in titer by 0.75 orders of magnitude (Fig. 2B).

Hemagglutination inhibition reaction.
Previously, before the HIR (hemagglutination inhibition reaction) with hydrated fullerenes, the titer of the virus was determined using the hemagglutination reaction, which was 64 HAU/25 μl. The ability of the studied structures to directly affect erythrocytes in the concentration range of 0.01-10 mg/ml for HyFn400, 0.01-1.25 mg/ml for HyFn050 and 0.015-0.15 mg/ml for oseltamivir was also tested. Neither compound was found to cause agglutination of the erythrocytes. The HIR was carried out in the same range of concentrations. The results showed that hydrated fullerenes and oseltamivir do not affect the ability of the influenza A (H1N1) virus to agglutinate guinea pig erythrocytes.
The molecular dynamics of polyhydrated C 60 in the complex with H1N1 RdRp was relatively homogeneous and was characterized by a stable preservation of the ligand position throughout the simulation period. The www.nature.com/scientificreports/ root-mean-square deviation (RMSD) of C 60 (OH) 60 during the first 20 ns of the simulation stabilized at a value of ~ 3 Å and remained until the end of the simulation with the exception of minor fluctuations (Fig. 4). The range of RMSD fluctuations was about 1 Å. That, together with the constant preservation of the general level of RMSD, may indicate the sufficient stability of the studied ligand in the complex with RdRp. The root-mean-square fluctuation (RMSF) distribution of amino acids of the RdRp PA subunit showed mostly its minimum value for amino acids involved in the interaction with polyhydrated C 60 . Only Y393, S395, D396, and P398 had a high RMSF. However, it should be noted that these amino acids are a part of the disordered and labile loop that directly interacts with the solvent during simulation.
Among the main orienting interactions, hydrogen bonds of C 60 (OH) 60 with amino group K281 and carbonyl G569 (which are a part of PA subunit) should be highlighted. These connections are stable enough to be observed throughout the simulation. Additional hydrogen bonds with: R279, S395, D396, R566, G569, T570 and S571-PA, H32 and R238-PB1 were also present in some periods of the simulation. In general, the number of hydrogen bonds between the ligand and the receptor ranged from 2 to 3 during the simulation (Fig. 5) and after some rare jumps up to 6-7.
The value of the interaction enthalpy between C 60 (OH) 60 and RdRp, calculated on the basis of the molecular modeling trajectory, was sufficiently significant and equaled to − 129.55 ± 14.52 kcal/mol. That indicates a potentially high gain of the system in case of successful formation of the RdRp/C 60 (OH) 60 complex. Moreover, the contribution of both van der Waals and electrostatic energy to the ligand-receptor interaction were close and provided the basis for the stability of the formed associate: − 61.88 ± 7.62 kcal/mol and − 58.59 ± 18.71 kcal/mol, respectively. A much smaller contribution was provided by solvation free energy -9.08 kcal/mol. In Table 1, it is represented by polar (− 0.58 ± 19.65 kcal/mol) and non-polar (-8.5 ± 1.14 kcal/mol) components. According

Discussion
Studies of the activity of polyhydrated fullerenes against influenza A (H1N1) virus indicate their ability to influence the process of the virus reproduction. Considering the lack of interaction with the hemagglutinin according to the results of the hemagglutination inhibition reaction, we assume that HyFn050 and HyFn400 do not interact directly with surface structures of the virus particle. Also, under the prophylactic scheme of the study, mixtures of fullerenols did not show a significant activity with a decrease in the infectious titer by less than 1 order of magnitude. Considering this, we believe that the studied pohydrated fullerenes do not have an immunomodulatory effect at the cellular level similar to interferon-inducing antiviral compounds 14 .   www.nature.com/scientificreports/ Hydrated fullerenes HyFn050 and HyFn400 demonstrated the greatest effect on influenza virus reproduction after the establishment of infection in MDCK cells. We assume that a high dose-dependent antiviral effect, together with a decrease in the infectious titer up to 4 orders of magnitude, may be due to the interaction of fullerenes with influenza structures during the reproduction of the latter in the cell.
Some studies indirectly testify to the significant role of charge distribution on the surface of a fullerene in this process. In particular, Kataoka et al. 7 demonstrated the ability of fullerene derivatives to inhibit RNA polymerase and protease of hepatitis C virus. At the same time, the C 60 fullerene derivative which contained an additional carboxyl group showed three times greater activity against polymerase compared to protease. And Shoji et al. 8 proved the ability of fullerene derivatives to inhibit the activity of the PA subunit of RdRp of influenza A H1N1 and H3N2 in their study. In this case, derivatives of C 60 with a small negatively charged substituent on the surface showed a stronger inhibitory effect compared to derivatives with more massive and/or positively charged substituents. Therefore, we can assume that the presence of a small negatively charged substituent on the surface of C 60 fullerenes may be an important factor for effective inhibition of viral RNA polymerases. The mixture studied in our work (C 60 /C 70 /C 76 /C 78 /C 84 )(OH) 40 contains fullerenes of different sizes, but with a similarly hydrated surface. Each of the carbon atoms was bonded to one hydroxyl group and three other carbon atoms. Thus, the polyhydrated fullerenes that appear in our work had a relatively homogeneous surface lined with highly electronegative oxygen atoms. Taking into account the above-mentioned studies and the results obtained by us, we assume that the drug forms of fullerenols K050 and K400, represented by two batches HyFn050 and HyFn400, can have an inhibitory effect on the activity of the RdRp of the influenza A virus. In addition, the high hydrophilicity of the surface makes it difficult for fullerenols to pass through the cell membrane, which explains the activity at high concentrations and the dose-dependent effect.
In order to understand the direct mechanisms underlying the interaction between the RdRp of influenza A virus and the investigated hydrated fullerenes, molecular docking was performed. According to its results it was determined that the most likely site of fullerenol binding is the pocket of RdRp interaction with the RNA primer, more precisely, with the hook structure of its 5′-end 15 . In our opinion, this is logical because the charge distribution on the nucleic acid surface can be considered similar to the charge distribution on the surface of hydrated fullerenes. A further simulation study confirmed the high affinity of C 60 (OH) 60 to the previously identified pocket. The RMSD of the ligand and the drop in the free energy of the system are the first evidence. The RMSD of hydrated C 60 quickly stabilizes within ~ 3 Å (range of oscillations ~ 1 Å). Which, according to modern data, is included in the RMSD optimum for a ligand in a complex with its potential receptor 16 . At the same time, the significantly negative value of the binding free energy (− 73.7 ± 14.52 kcal/mol) indicates the high stability of the RdRp/C 60 (OH) 60 complex. This is also confirmed by the low RMSF value characteristic of RdRp PA amino acids interacting with the ligand. The only exceptions are amino acids Y393, S395, D396 and P398. However, they are included in the highly labile disordered part of the PA-subunit, which, moreover, directly interacts with the solvent. Therefore, they can be considered as an exception in this case.
The main forces stabilizing the studied complex are hydrogen bonds, as well as a steric compatibility. This is reflected in the calculated high values of both van der Waals interactions and electrostatic energy as the main components of the enthalpy of the system. The interaction of C 60 (OH) 60 with PA K281 and G569 was the most stable: all analyzed frames of the trajectory contained them. The formation of additional interactions with many of the above-described amino acid residues was also observed locally. Thus, the average number of hydrogen bonds fluctuated around 2-3, although significant excesses of this number were also observed.
In general, the placement of C 60 (OH) 60 in the binding site of the 5′-end of the RNA primer, as well as its affinity to this site, in our opinion, may be related to its previously determined biological properties.

Conclusions
The studied batches of the mixture of polyhydrated fullerenes K050 and K400 have an antiviral effect against influenza A (H1N1) virus, which is manifested in the suppression of viral reproduction by a maximum of > 94% and a decrease in the infectious titer de novo by 4 orders of magnitude. Since in the simulation experiment C 60 (OH) 60 demonstrated a significant level of affinity to RNA-dependent RNA polymerase, in particular to the binding site of the RNA primer, we assume that the described effect is due to the interaction of the studied structures with the viral RNA polymerase. Therefore, we believe that this drug is promising for further research as an antiviral agent. The drug of Ukrainian production in two forms K050 and K400 ((C 60 /C 70 /C 76 /C 78 /C 84 )(OH) 40 ) is an aqueous colloidal solution of a polyhydrated mixture of fullerenes with a mass ratio of 78.1% C 60 /C 70 and 21.9% C 76 /C 78 / C 84 . The forms of the drug differ in the initial concentration. Two batches with different concentrations were tested: HyFn050 2.5 mg/ml (FN1120CSMP4K050) and HyFn400 20 mg/ml (FN1120CSMP5K400). Oseltamivir phosphate (Macleods Pharmaceuticals LTD) was used as a reference drug. www.nature.com/scientificreports/ solution (Biowest, France) were used, then cells were resuspended in growth medium. Cells were incubated in a thermostat at 37 °C in an atmosphere of 5% CO2.

Materials. MDCK cell culture and influenza A (H1N1) virus strain
Determination of cytotoxicity. The cytotoxicity was determined using the MTT assay. The supernatant was removed from a 96-well plate (SPL Life Sciences, Republic of Korea) with a monolayer of MDCK cells and the test compounds diluted in the growth medium were added to the wells in the concentration range from 5 to 20 mg/ml for HyFn400, from 0.25 to 2.5 mg/ml for HyFn050, and from 0.01 to 1 mg/ml for oseltamivir. The compounds were incubated together with the cells for 48 h in a thermostat at 37 °C in an atmosphere of 5% CO 2 .

Determination of infectious titer de novo.
To determine the infectious titer we used supernatant with the virus after 48 h of incubation from the antiviral and the prophylactic activity assays. A tenfold titration of the supernatant was performed using RPMI medium (Sigma, USA). Next, 50 μl of the suspension was added to the monolayer of cells in a 96-well plate (pre-washed with 150 μl of PBS, pH = 7.4). The plate was incubated for 90 min in a thermostat at 37 °C in an atmosphere of 5% CO 2 . After that, the supernatant was removed and 200 μl of the supporting medium (49% DMEM (Sigma, USA), 49% RPMI 1640 (Sigma, USA), 2% fetal bovine serum (Sigma, USA) and 1 μl/ml TPCK-treated trypsin (Sigma, USA)) was added. After the incubation for 48 h at 37 °C and 5% CO 2 , CPE was checked using an inverted microscope and cells were stained with crystal violet (neoFroxx, Germany Hemagglutination reaction. The hemagglutination reaction was carried out using a 2% suspension of guinea pig erythrocytes in PBS (pH = 7.4) according to the standard method 18 . Preparation of erythrocyte suspension: the animal's blood (2-5 ml) was taken postmortem into a test tube with anticoagulant-3.2% tri-sodium citrate dihydrate (Sigma, USA) diluted in sterile water (9 parts of blood + 1 part of anticoagulant). Carbon dioxide was used for euthanasia of the animals, which complies with the American Veterinary Medical Association. PBS (pH = 7.4) was added to the blood-anticoagulant mixture to a general volume of 10-12 ml. The test tube was centrifuged for 10 min at 1400 g and the supernatant was discarded. Centrifugation was repeated 2-3 times until the supernatant was clear. The resulting sediment of erythrocytes was diluted to the 2% suspension in PBS (pH = 7.4). Before use it was checked for autoagglutination: 25 μl PBS + 25 μl suspension in a plate with U-shaped wells left at a room temperature for 60 min. Hemagglutination inhibition reaction. The experiment was conducted according to a modified classical HIR 19 . The influenza virus dilution corresponding to 8 HAU/25 μl (hemagglutination units) was used for performing HIR. Reaction was carried out using a 2% suspension of guinea pig erythrocytes in PBS (pH = 7.4) in a serological 96-well plate with U-shaped wells (SPL Life Sciences, Republic of Korea). 25 μl of PBS (pH = 7.4) was added to the wells. 25 μl of the test compound was added to the first well and 25 μl was transferred to the next one (twofold titration). 25 μl was removed from the last well. 25 μl of virus was added to each well. 25 μl of the erythrocyte suspension was added to all wells. 25 μl of PBS (pH = 7.4) + 25 μl of the suspension were added to the control wells of the quality of erythrocytes. To control the effect of the studied compounds on erythrocytes, 25 μl of virus-free suspension was added to the corresponding dilutions of the substances. It was mixed on a shaker for 1 min. The result was taken into account after 60 min of incubation at a room temperature.
Preparation of input data for the in silico studies. The selected file of the 3D structure of H1N1 RNAdependent RNA polymerase (RdRp) (PDB ID: 7NK1), due to the lack of coordinate data for some amino acids, was completed based on the template using the SWISS-MODEL online resource 20 . For the further studies data of all molecules except the structure of the RdRp heterotrimer were deleted. The structure of the ligand -C 60 fullerenol was generated using the Avogadro molecular editor 21 .
Molecular docking. The molecular docking of fullerenol C 60 (OH) 60 with H1N1 RdRp was performed using the Vina Wizard tool of the PyRx software package 22 . For this study, the grid box included the entire heterotrimeric protein simultaneously. The value of exhaustiveness due to the predicted large protein-ligand interaction site was 40. For further simulation experiments, the position of fullerenol with the best evaluation score was selected.
Simulation of molecular dynamics. The actual construction of the system was carried out by the CHARMM-GUI toolkit 23 using the CHARMM36m force field 24 and involved the creation of a rectangular box with periodic boundary conditions, which contained the RdRp/fullerenol complex, water as a solvent (TIP3P model), K + and Cl − ions at the physiological concentration 0.15 M. After construction, the system underwent the steepest descent minimization to a system energy value of less than 1000 kJ/mol/nm (5000 steps). Next, the equilibration was carried out under conditions of stable volume and temperature of the system (classical NVT, Nose-Hoover thermostat, T-coupling constant of 1 ps, Van-der-Waals with Lennard-Jones potential and a 1.2 nm cut-off, electrostatic with a 1.2 nm cut-off, LINCS algorithm for covalent bonds, 303.15 K, the time step -2 fs and the constraints for the protein) for 125 ps. The main simulation differed from the equilibration stage only by the additional condition of Parrinello-Raman barostat at 1 atm and the absence of restrictions on the protein position. The system temperature was 303.15 K. The equilibration and molecular simulations (100 ns) were performed in GROMACS 2020.6 25 using GPU as an accelerator.
Calculation of the binding free energy. The gmx_MMPBSA 1.5.2 package was used to calculate the free energy using the MM/GBSA (Molecular Mechanics/Generalized Born Surface Area) method 26 . The source of input data was the trajectory (the last 80% of frames -800 in total) generated by GROMACS in the process of the molecular dynamics simulation of target associates. The polar contribution of solvation was estimated on the basis of the GB-OBC2 model 27 , and the non-polar contribution was estimated on the basis of SASA (solvent-accessible surface area). The contribution of the entropy component was calculated by the IE (interaction entropy) method 28 .
Analysis of the results obtained in silico. The results were analyzed using standard software provided by GROMACS (trjconv, rms, rmsf, hbond) and gmx_MMPBSA (gmx_MMPBSA_ana). Visualization was performed using XMGRACE 29 and Pymol 1.8 30 .
Statistical processing of data obtained in vitro. Each experiment was carried out in three independent repetitions of 3 wells for 1 concentration of the compound. Statistical data processing was performed in the Origin 2018 program. According to standard approaches standard deviation was calculated and confidence interval was determined (p ≤ 0.05 for all arrays).

Data availability
All data generated or analysed during this study are included in this published article. www.nature.com/scientificreports/ Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.