In vitro Antioxidant Potential Evaluation of Non-Functional- ized Fullerenes and Endofullerene

The antioxidant properties of unmodified aqueous fullerene dispersions (AFD) of C60, C70, Gd@C82 (in μM concentration range) were studied in the model of generation of organic radicals by 2,2’-azobis(2-amidinopropane) dihydrochloride (ABAP) and luminol. Purification protocols for AFDs are proposed. Based on chemiluminescence (CL) measurements, the concentration-dependent of the CL signal for any AFDs is revealed. Suppression of CL signals is up to 5 times better than Mexidol® antioxidant. The concentration of half-suppression increased in row C60<C70<Gd@C82. We further demonstrated mathematical modeling of the long-term kinetics data for C60, C70, Gd@C82 allowed fitting of CL curves. Kinetic schemes are proposed, and the constants of each reaction are estimated (~n×μM–1min–1). C60, C70, and Gd@C82 exhibit a quenching mechanism that is not an antioxidant effect. For C60 and C70, similar behavior is shown that is not quantitatively different in kinetic modeling. On the contrary, Gd@C82 has a double action: (1) quenching and (2) actual antioxidant action. Estimating quenching constants for all of the AFD types showed the same magnitude. Moreover, the antioxidant activity in Gd@C82 is 300 times greater than in the case of C60 and C70.


Introduction
Water-soluble fullerenes are promising candidates for different medical applications, and they have been proposed as vital components for humans and environmental systems [1]. Fullerenes, in particular, their water-soluble derivatives are considered as radical scavenging agents [2], possess antioxidant activity [3], acquire remarkable antimicrobial properties [4], cytotoxicity [5], DNA cleavage, and lipid peroxidation mediated by reactive oxygen species (ROS) [6]. As for non-functionalized aqueous dispersions (AFD) (without addends) [7], their antioxidant and superoxide anion-radical (SAR) scavenging properties have not been thoroughly studied. First attempts at study superoxide dismutase (SOD) mimic activity have yet been made [8] in vitro and cell models. It is known that unsaturated lipids are a target of free radicals. The mechanism of their oxidation (lipid peroxidation) has been described and proved [6]. The result is the accumulation of lipid hydroperoxides, intermediate stable products. The ability of fullerenes to initiate lipid oxidation has not been widely assessed before. There is a slight report on their ability to trap lipid peroxyl radicals and act as chain-breaking antioxidants [9]. The impact of fullerenes in in 5 of 5 vivo and in vitro experiments for Cyprinus carpio brains confirmed the absence of lipid peroxidation [10]. Recent reviews discussed [11] that fullerenes possess an inert scaffold with antioxidant functionalities. Biologically active water-soluble fullerene adducts are also discussed in [12]. C60 is an extremely weak chain-breaking antioxidant with an inherent rate constant for trapping peroxyl radicals per se (kinh=0.3×10 3 M -1 s -1 ). However, some antioxidants covalently bound to fullerenes increase antioxidant activity insignificantly. Grafting their surface with small-molecule antioxidants such as synthetic phenols (2,6-di-tert-butyl-4methylphenol) broadens their antioxidant behavior conveying peroxyl radical-trapping activity up to 30 times [13]. C60 conjugated with phenols indicates a significant improvement of oxidative stability [14]. A C60 derivative with covalently bonded (an analog of αtocopherol with hydroxychromanyl moiety) is an effective antioxidant acting in two model lipid matrices: saturated stearic acid and unsaturated linolenic acid during the nonisothermal oxidation tested by differential scanning calorimetry [15]. There is an ambiguity in information about the ability of fullerenes to generate ROS. Several studies deals with С60 solutions stimulating ROS generation [16]. Another study evidenced the antioxidative properties of fullerenes [17].
The conventional approach to describing the antioxidant properties of low molecular weight free radical scavengers is based on quantitative determination of their ability to terminate free radical chain reactions, relatively a standard antioxidant compound. Evaluation of the antioxidant status of compounds, the total radical-trapping potential (TRAP) method, and total antioxidant reactivity (TAR) from luminol-enhanced chemiluminescence (CL) measurements is previously developed [18]. This approach is based on the ability to trap radicals formed during the decomposition of thermolabile azo compounds in radical-induced reactions. However, this technique does not consider the individual physicochemical characteristics of the antioxidant. A more rigorous approach to the description of antioxidant properties considers the determination of the antioxidant concentration and the mathematical modeling estimation [19] of the rate constant of the interaction with the radicals [20]. Different antioxidants possess different chemiluminescence curves, making it impossible to use any universal parameter to characterize the amount or activity of antioxidants of different chemical nature [21].
Fullerene soot C60 and C70 not only retards oxidation in the mode of an alkyl radical quencher but also operates as a peroxy radical scavenger [22] in the model reaction of initiated (2,2′-azobisisobutyronitrile, AIBN). For reactivity of C60 during oxidation of a series of hydrocarbons shows that the fullerene does not react with the RO2• radicals indicate an extremely weak rate constant estimated from CL [23].
Of fundamental importance is had the functional fullerenes derivatives, but it is beyond the scope of this article to study these derivatives. There are almost no works related to the study of antioxidant properties in the generation of organic radicals for unmodified fullerenes. As far as we are concerned, no one has ever used kinetics modeling for fullerenes to assess the interception ability of radicals. This paper reports data on the antioxidant properties of aqueous fullerene dispersions. Using computer simulation, prediction of the rate constants and the reaction scheme has been proposed.

Aqueous Fullerenes Material Safety Data Sheet
In this study, ultrasound-assisted preparation of AFD for pristine C60, C70, Gd@C82 was used. The AFD showed excellent stability for more than 20 months; all of the information about the characterization of AFD was previously published elsewhere [7]. The main drawback of this process is an accumulation of titanium dioxide nanoparticles (TiO2 NPs) due to cavitation sonotrode erosion. It is confirmed by ICP-OES analysis for metal impurities in AFDs (  Table 1). A syringe polyvinylidene fluoride (PVDF) filter has been used to remove particles from AFDs. The filter removes large fullerene nanoparticles (more than 1 μm) and finally cleans titanium nanoparticles (less ca. 1 ppb) from the ultrasonic probe. The ICP-OES showed that AFDs contain silicon, which cannot be removed by syringe filtering.

Scavenger Antioxidant Properties in The Model of ABAP Decay
We studied the antioxidant properties of AFDs by activated kinetic chemiluminescence in a molecular model of organic radical generation is thermal-induced decomposition of 2,2'-azobis(2-amidinopropane)dihydrochloride (ABAP) in the presence of luminol. Based on the data obtained, we evaluated the antioxidant capacity and mathematical modeling of the reaction kinetics (see below). A common antioxidant, Trolox® [24], was used as a standard comparison substance. Also, to compare the behavior of fullerenes in the ABAP system in the presence of luminol, we compared the behavior with the medium strength antioxidant, Mexidol® [25].
The shape of the analytical CL signal for Trolox (100 and 200 nM) is shown in Figure  1. The signal behavior of the chemiluminogram in Figure 1 is typical for strong antioxidants; different curves were obtained for AFDs Figure 3  The shape of the analytical signal suppression, in this case, coincides with the shape of the curves for medium-strength antioxidants. The general form of chemiluminescence diagrams for all AFDs in the narrow (up to 20 min) and long time intervals (up to 720 min) is presented for C60 at  Table 2. It is shown that AFDs can be ranked by their ability to suppress chemiluminescence as C60 > C70 > Gd@C82. Ratios of half-suppression signal concentrations were 1:1.7:3.5.

Computer Simulation of The Fullerene Aqueous Dispersions Antioxidant Activity (Complete chemiluminescence kinetics before antioxidant depletion)
The use of rate constants of the inhibition reaction can be used as an indicator of antioxidant activity. We recorded the kinetics of chemiluminescence for mathematical modeling of antioxidant action to the moment of depletion (disappearance) of the antioxidant, ~100 min. Dependences for standard substances are shown in Figure 1, 2, for AFDs, in Figure 3b, 4b, and 5b. Experimental and calculated curves have a sufficient degree of coincidence. The effect of the strong antioxidant, Trolox, on CL kinetics is expressed in the complete suppression of CL (a "latent period" [26]); depletion of the antioxidant is characterized by an increase in the intensity of CL and return to the previous stationary (control) level. Thus, for the Trolox concentration of 200 nM, the latent period was ~3.5 sec, the time of radical depletion in the system is ~90 min (Figure 1). In contrast, the effect of Mexidol shows that for concentrations of ~10 μM, the effect of complete signal suppression occurs at more than 400 min, with an increase in concentration to 20 μM this period is >700 min.
For the AFDs in the long-term experiment, the experimental curves recorded the moment of antioxidant depletion, which for all types of AFDs was from 85 to 100 minutes ( Figure 3b, 4b, and 5b). In this regard, it is possible to compare the mathematical model and experimental data for all AFD types studied in this work. To carry out the kinetics simulation, we selected the optimum concentration ranges of the investigated AFDs (Table 2). We also chose the most suitable exact concentration values for which we observed a suppression of CL intensity ~2 times. From the experimental data presented, all AFDs satisfy the requirements necessary for mathematical modeling: (1) the moment of antioxidant depletion is registered; (2) the concentration dependence is traced. (c) Figure 6. The experimental and simulated chemiluminescence plots for aqueous fullerene dispersions C60 (a), C70 (b), Gd@C82 (c) for one-stage mechanism. In all cases, solid lines have a specific color: black is the control experiment, blue is the modeling data of the control experiment, red is experimental data for aqueous fullerene dispersions, green is modeling data for aqueous fullerene dispersions.
A simple mathematical model was used to simulate the steady-state level of chemiluminescence (without adding antioxidants), whose kinetic scheme consisted of two reactions: (1) the free radical generation reaction from ABAP; (2) chemiluminescence reactions. Combined reaction schemes: where R• is a free radical or reaction product in the electronically excited state, which reacts with the antioxidants, P is the stable product of the free-radical reaction.
To model the action of antioxidants, reactions with fullerenes (3) were added to reactions (1), (2, 2a): (3), which is general for all AFDs, was included in all kinetic schemes. The principle is the interaction of AFDs with the excited product of luminol. For more coincidence of experimental and model data only in the AFD Gd@C82 reaction (3a), intercepting obtained radicals from ABAP was added. An acceptable match between the experimental CL data and the mathematical modeling data for AFDs C60, C70, Gd@C82 was obtained under the following conditions ( Table 3). AO2 + R → ... n/a n/a n/a 30 n/a Additional reaction n/a -not applied Thus, it is possible to arrange AFDs in a sequence by strength С60 > C70 ~ Gd@C82, the ratio of reaction rate constants between antioxidant (AFD) and free radical for С60, C70, and Gd@C82 are 1.5:1:1.

Discussion
Since the priority of this work is to model the process kinetics and a deeper understanding of the mechanism of the process, we did the steps below.
(1) Chemiluminescent systems can be influenced by the purity of the AFDs objects under study. Moreover, the action of the impurity component can be taken as a false positive mechanism, so we carried out a deep purification of the AFDs (as in the section above). Thus, eliminating the influence of this factor.

Peculiarities of Preparation Procedure and Aqueous Fullerene Dispersions Purification
Our experience with the implementation of the sonication system for the preparation of carbonaceous nanomaterials (hydrosols of nanodiamond) reveals titanium impurities [31] even for a short time (e.g., several minutes) of ultrasound treatment. Ti accumulation rate was proportional to the prolonged ultrasound exposure time [32].
Our previous experiments proved that titanium and silicon are in the form of oxide [8]. However, processes of purification of AFD would not be resolved by now. It is associated with transition metal (TM)-carbon interactions [33]. It is essential to choose the right filter system. We suggest using commercially available syringe PVDF membrane filters 0.22 or 0.45 μm. PVDF membrane can be attributed to the high affinity of TiO2 for water [34]; the presence of hydroxyl moieties from TiO2 NPs contributed to improving the hydrophilicity, thereby improving the water flux [35]. As shown in Table 1, filtration with a 0.22 or 0.45 μm PVDF filter completely removes TiO2 NPs fullerene dispersions.
Moreover, [32] showed that 0.2 μm PTFE filter has low filtration efficiency from titanium. This is due to redox reactions between TiO2 NPs and Pu(III)/Pu(IV) at low pH solutions. However, for practically neutral AFDs Table 1, the efficiency of using PVDF filters is high. The presence of impurities can reduce or increase the effectiveness of antioxidants. The ability of TiO2 NPs to enhance the production of SAR and alter the antioxidant system in human osteoblast cells is shown in [36]. As well impurities may adversely affect the reproducibility of results [37]. To evaluate the efficiency of fullerene antioxidant capacity, preparation of pure dispersion is an essential item. As for SiO2 NPs, it remains in solution (approximately 65%, see Table 1) after filtration using PVDF filters. In [38] showed that SiO2 NPs pass through the membrane during filtering with polytetrafluoroethylene (PTFE) membranes in a range 0.1-1.0 μm of pore size. In our opinion, we find a residual silica content due to: (1) Typically silica SiO2 NPs can be formed with less 220 nm in diameter [39]; Intermolecular and intramolecular interactions deal with (2) Probable silica surface with an adsorbed fullerene monolayer [40]; (3) Sorption SiO2 NPs on the surface of fullerene [41]. The addition of SiO2 NPs increased the capacity of the antioxidant enzymes superoxide dismutase (SOD), catalase (CAT), glutathione reductase (GR), and phenylalanine ammonia-lyase (PAL) in plants [42]. In our work, the total content of SiO2 NPs is less than the total fullerene content ratio in-stock solution from 10 to 170 times, which remains diluted. Therefore, it should not affect the enhancement of antioxidant activity. The study of such a mechanism was beyond the scope of this work.

Antioxidant Activity in the System of Generation of Organic Radicals and Simulations of chemiluminescence kinetic for aqueous fullerene dispersions
Analysis showed that the data on the activity of unmodified fullerenes and their aqueous dispersions are practically absent. The closest fullerene derivatives in the structure are fullerenols, which have up to 42 hydroxyl groups in their structure, depending on the type of fullerene in the series from C60 to Gd@C82. Hydroxylation is one of the cheapest and most straightforward approaches to dissolving fullerenes in water and does not require deep purification of the resulting product. However, it should not be excluded that the appearance of even minor surface derivatization can lead to an increase in the antioxidant activity of fullerenes [43]. It was shown that [44] less toxicity and greater antioxidant capacity is proven for fullerenols С60Oy(OH)x, С60,70Oy(OH)x, x+y=24÷28. There are two limitations of using any derivatization of fullerene surface they are related to the fact that: (1) additional groups can be involved in metabolic processes, groups on the surface can reduce the availability of the π-electron system, leading to reversible free radical capture [45], differently affect the spin environment [46]; (2) fullerene derivatives could act as potent oxidizing agents under excitation with light in the presence of oxygen [47]. We only studied AFDs C60, C70, and Gd@C82 without addends in our work. Moreover, first of all, we proceeded to the comparison of known antioxidant systems. Measurements of the antioxidant action of Mexidol are used to calibrate the chemiluminescence dependences for determining the level of free radical generation. For bioluminescence and chemiluminescence, half-maximum signal suppression concentration is widely used to assess the overall fullerene toxicity and antioxidant properties [44]. Moreover, the half-maximum signal suppression concentration for fullerenes is 1.5 to 5.0 times lower than Mexidol, which shows their better antioxidant activity. Comparison of fullerenes with each other shows an increase in the level of half-maximum signal suppression concentration for AFD in the series ( ⁄ С60<C70<Gd@C82) this is because the proportion of "active" fullerene molecules decreases in the series С60>C70>Gd@C82. Previously, we showed [8] that the fraction of active molecules on the surface (% of total concentration) is 1.6 : 1.5 : 0.9, which agrees with the obtained chemiluminescence suppression series in the organic radical generation model.
From the presented experimental data for fullerenes, we can make a preliminary conclusion (without mathematical modeling) that the antioxidant activity (relative radical interception rate) of fullerenes differs insignificantly (within one order). A significant difference shown in the antioxidant capacity is the number of intercepted radicals per 1 particle of a substance. In order to simulate the kinetics of reactions, the whole CL curve was recorded to establish the moment of antioxidant depletion. This effect is seen in the convergence of chemiluminescence curves: (a) control experiment (without antioxidant) and (b) with antioxidant. This point is very crucial for testing the mathematical model. In the experiment and the mathematical model, the antioxidant should run out after the same 5 of 5 time after addition. If the moment of antioxidant depletion is not registered in the experimental data, it is impossible to assert the adequacy of the mathematical model parameters.
Antioxidant rate constants allow us to discriminate fullerenes by antioxidant strength. AFDs C60 and C70 are weak antioxidants and approximately equal in strength; Gd@C82 exhibits two antioxidant components. Numerically AFDs C60 appeared slightly more active antioxidant than C70, as for Gd@C82 activity is comparable with C70. In [48] classified the strength of antioxidants according to constants obtained from process kinetics simulations, strong antioxidants from 2000 nM -1 min -1 , medium-strength from 100 nM -1 min -1 , and weak ones below 10 nM -1 min -1 . The C60, C70, and Gd@C82 constants obtained in Table 3 are comparable and refer to medium strength antioxidants. As shown in Table  3, the added concentration of Gd@C82 was 4 times higher than C70 and C60 (17.2 and ~4 μM, respectively). For these initial conditions, the effect of the 2 nd antioxidant contained in Gd@C82 was evident. Moreover, the content of the 2 nd antioxidant was 100 times less, while the activity was about 300 times higher. The 2 nd antioxidant in Gd@C82 is characterized by a different mechanism of action dealing with the interception of ABAP radicals rather than quenching of excited luminol molecules. AFDs C60 and C70 likely contain such an antioxidant, but their action is little noticeable due to their low concentration.
AFDs turned out to be very weak antioxidants; they do not give a latent period, as in the strong antioxidant Trolox [48]. Also, the fullerene in AFDs differ from Mexidol in the mechanism of action: (1) Mexidol intercepts degradable ABAP radicals; (2) On the contrary, fullerenes C60, C70, and Gd@C82 intercept the excited product of luminol. There is the quenching of chemiluminescence, rather than competition with luminol molecules for free radicals, seen by the magnitude of the constant values (Table 3). This part of the quenching mechanism and the role of fullerenes in reducing the fluorescence signal are still unclear [49]. In some cases, it has been shown that fullerenes and fullerenols can non-covalently bind molecules, exhibiting dynamic luminescence quenching, for example, with Ribonuclease A [50]. However, the binding sites in each case are individual; for a more detailed study of the system and the nature of the binding, molecular dynamics simulations should be performed [51]. In additional experiments (see Supporting Information), we evaluated the luminescence quenching constants for the ABAP and luminol mixture studied, as well as for luminol alone. In both cases, we observed comparable orders of magnitude values of Stern-Volmer quenching constants for C60, C70, Gd@C82: C70 > C60 > Gd@C82.
This data can be explained by the polarizability (α) of the fullerene molecules for C60 and C70. The more significant efficiency of the quenching of the electronically excited states by the C70 can be attributed to the higher average polarizability of the C70 molecule [52]. The polarizability is for the C60 (82.7 Å 3 ) and C70 (102.7 Å 3 ) [53], Gd@C82 114.67 Å 3 [54] at that αEMF<αatom+αFullerene. In addition, (1) the reactivity upon AE-type reactions for the fullerenes decreased in the order C60 > Gd@C82 [55]; (2) C82 is a better free radical scavenger than C60, the presence of endoatoms improves the fullerenes antiradical capacity [56].
However, a 1.3-fold decrease in the quenching constants for Gd@C82 could explain a different behavior in the chemiluminescence reaction, a change in the quenching mechanism to a bimolecular one [57], or the involvement of the molecule in another competing process at multiple positions of the fullerene cage [55]. Thus, according to the kinetic simulation results, we obtained: (3) The Gd@C82 AFDs have a tiny amount of "true antioxidant" admixture. In general, for Gd@C82, we observe the same background process of action as quenchers. However, on the other hand, we obtain an antioxidant effect based on calculations based on chemiluminescence curves (order of constants in mathematical modeling) and the experimental course of the curves. From the simulation, we can conclude that the effect is comparable to Mexidol, which intercepts exactly ABAP radicals, but because of the primary and predominant action of CL quenching, it is pretty difficult to separate and measure the signals. On the other hand, the different behavior for Gd@C82 AFD can be explained by (1) the presence of a carbon frame that contains the inner paramagnetic metal ion Gd 3+ with spin S=7/2, (2) the presence of anion Gd@C82 3-acting as a radical located on the outer shell [58]. Gd@C82 3-can involve in free-radical addition reaction, which can change the electronic structure of the inner cluster and subsequently affect its configuration [59]. The electron affinity values of Gd@C82 are more significant than those for pristine C60 and C70 (1.25 and 1.19 times greater for Gd@C82). It can be indicated that Gd@C82 can act as a strong electron donor and an electron acceptor [60].

Aqueous fullerene dispersions preparations
Preparation and complete characterization techniques for aqueous fullerene dispersions were recently described elsewhere [7]. In this work, we used long-term stable AFD of the pristine: (1) C60, (2) C70 (NeoTechProduct LLC (Russia), 99+% HPLC-grade); (3) the enriched soot containing the Gd@C2n EMFs (total content of Gd atoms up to 4 wt. % has been synthesized by the evaporation of the composite graphite electrodes compounded by gadolinium in the electric arc reactor as we previously described [61].

Techniques and additional reagents
The enhanced chemiluminescence protocol for quantification of the antioxidant activity of aqueous fullerene dispersions C60, C70, and Gd@C82 has been used. The chemiluminescent system consisted of a source of free radicals cationic 2,2'-azobis Reference antioxidant compound: Trolox ® (±)-6-Hydroxy-2,5,7,8-tetra-methylchromane-2-carboxylic acid (Sigma, USA); Mexidol ® ethylmethylhydroxypyridine succinate 50 mg/ml solution for intravenous and intramuscular administration (LLC "NPK" PHAR-MASOFT", Russia). The chemiluminescence signal was recorded up to achieving stationary level, and then an aliquot of the antioxidant solution of Trolox or AFDs was added. The registration was performed until the novel steady-state level was reached.

Equipment
The measurements were carried out with a Lum-1200 chemiluminometer with 12channel (DISoft, Russia). The chemiluminometer makes available visible light detection within a range from 300 to 700 nm. No bandpass filters were used. Signal processing was performed via PowerGraph 3.3 Professional software (DISoft, Russia). The relative standard deviation of chemiluminescence intensity did not exceed 0.05. Fluorolog®-2 spectrofluorimeter (Horiba Jobin Yvon, Japan) was used. The statistical processing of the data was performed with STATISTICA v.10.0 software (StatSoft Inc., USA).

Computer simulation and data handling
The computer simulation was carried out with the specially designed computer program "Kinetic Analyzer" (by Dr. D. Izmailov). For a set of the predetermined reactions and the initial concentrations of the reactants, the rate constants were selected, providing the maximal curve fitting of experimental and calculated plots. As a criterion for maximal curve fitting, the minimum sum of squared residuals was calculated with the OriginPro 2015 software (OriginLab, USA).
As an analytical signal characterizing the antioxidant capacity of AFD was the difference between the intensities of stationary chemiluminescence level ∆I = I0 -I (Figure 7). This parameter linearly depended on the concentration of added AFD. Chemiluminograms in the molecular model of generation of organic radicals for С60, С70, Gd@C82 were used to determine the concentration of half-suppression of the chemiluminescent signal for all AFDs ( 1 2 ⁄ , μM). The concentration of half-suppression of luminescence ( 1 2 ⁄ , μM) is a concentration that reduces the amplitude of the signal (∆I) or the light sum of the signal (S) of the response by two times and can hypothetically be taken as a quantitative indicator of the inhibitory activity of a given compound.

Conclusions
This paper is the first work to evaluate the antioxidant activity of fullerenes, which will be helpful in further evaluation of antioxidant properties in a living cell. C60, C70, Gd@C82 AFDs show no antioxidant activity in the system of organic radical-induced ABAP decay. We prove that it is not free radical capture but quenching. In the case of Gd@C82, there is a show of significant antioxidant effect. The results provide insights into the possible mechanism of interactions of fullerenes between free-radicals C60, C70, Gd@C82, which are of fundamental importance for understanding the potential biomedical effects of AFDs.
We believe that this work will help create reference materials for further study of the antioxidant properties of functional fullerene derivatives. To create such platforms, AFDs of C60, C70, Gd@C82 can be proposed as model substances not exhibiting antioxidant properties. Also, the absence of a significant free radical interception effect allows the development of sensors to control impurity composition, acting as a free radical interceptor rather than as a quencher for in vitro and in vivo experiments.
Supplementary Materials: The following are available online at www.mdpi.com/xxx/s1, Figure S1, S2: The fluorescent spectra of aqueous fullerene dispersions C60, C70, and Gd@C82 act as a quencher at different systems ABAP+luminol, and only luminol.