Fullerenemalonates inhibit amyloid beta aggregation, in vitro and in silico evaluation

The onset of Alzheimer's disease (AD) is associated with the presence of neurofibrillary pathology such as amyloid β (Aβ) plaques. Different therapeutic strategies have focused on the inhibition of Aβ aggregate formation; these pathological structures lead to neuronal disorder and cognitive impairment. Fullerene C60 has demonstrated the ability to interact and prevent Aβ fibril development; however, its low solubility and toxicity to cells remain significant problems. In this study, we synthesized, characterized and compared diethyl fullerenemalonates and the corresponding sodium salts, adducts of C60 bearing 1 to 3 diethyl malonyl and disodium malonyl substituents to evaluate the potential inhibitory effect on the aggregation of Aβ42 and their biocompatibility. The dose-dependent inhibitory effect of fullerenes on Aβ42 aggregation was studied using a thioflavin T fluorescent assay, and the IC50 value demonstrated a low range of fullerene concentration for inhibition, as confirmed by electron microscopy. The exposure of neuroblastoma to fullerenemalonates showed low toxicity, primarily in the presence of the sodium salt-adducts. An isomeric mixture of bisadducts, trisadducts and a C3-symetrical trisadduct demonstrated the highest efficacy among the tests. In silico calculations were performed to complement the experimental data, obtaining a deeper understanding of the Aβ inhibitory mechanism; indicating that C3-symetrical trisadduct interacts mainly with 1D to 16K residues of Aβ42 peptide. These data suggest that fullerenemalonates require specific substituents designed as sodium salt molecules to inhibit Aβ fibrillization and perform with low toxicity. These are promising molecules for developing future therapies involving Aβ aggregates in diseases such as AD and other types of dementia.


Introduction
Amyloid plaques are the primary hallmark of Alzheimer's disease (AD), the most prevalent type of dementia worldwide. The aggregation of the amyloid beta (Ab) peptide leads to the onset of extracellular plaque throughout the cortical mantle. The band g-secretases sequentially proteolyze the amyloid precursor protein (APP), releasing a 40 or 42 amino acid peptide. In AD, Ab aggregates extracellularly form soluble oligomers, insoluble b-sheet protobrils, brils and plaques. Ab plays an important role in the onset of AD, described in the amyloid cascade hypothesis, resulting from a chronic imbalance between Ab production and Ab clearance 1 that turns into: neuronal loss, neurobrillary tangle formation, vascular damage, and dementia that correlates directly to Ab deposition. Despite Ab plaques showing a low correlation with dementia, Ab oligomers display high toxicity to neurons 2,3 suggesting that Ab brillogenesis plays an important role in AD-induced toxicity. The Ab aggregation involves the C-terminus of the peptide that determines the rate of bril formation while the N-terminus promotes Ab-Ab interaction for polymerization, leading to a random coil or a-helix to b-sheet transition via a nucleation mediated process. 4 The extended b-sheets promote homophilic interactions and eventually lead to Ab oligomer formation. Kinetic studies demonstrated the monomeric Ab requirement for oligomer formation as seeds/nuclei, rich in b-sheets, for accelerated bril growth. 5 Several strategies for targeting Ab production and clearance have failed, since Ab immunotherapies induced encephalomyelitis and possible microhemorrhages, 6,7 and the inhibition of secretases disrupts important metabolic processes. 8 Therefore, another strategy is based on inhibition of the Ab peptide self-assembly. Dyes and small molecules, [9][10][11] peptides 12-14 and nanoparticles [15][16][17][18] were identied as effective inhibitors of Ab aggregation, ameliorating cell survival and cognitive decit. Fullerene C 60 , a carbon nanomaterial with a symmetric nanostructure, has been extensively used in areas of science, particularly in biomedics. Even though it has great potential in biological applications, the solubility of fullerenes has shown low compatibility in biological systems because of its hydrophobicity; so different strategies have been developed to achieve soluble fullerenes in aqueous dispersions for biological applications, including as an inhibitor of human immunodeciency virus-1 (HIV-1) protease, 19 an antioxidant, 20 and a cancer therapeutic. 21 C 60 functions as both a reactive oxygen species (ROS) producer under UV or visible light, and a ROS scavenger in the dark. This dual property of fullerenes to either quench or generate cell-damaging ROS has been applied as a cytoprotective or cytotoxic anticancer/antimicrobial/anti-Ab agent. [22][23][24][25][26] Previous reports indicated that fullerenes, and certain fullerene derivatives, inhibit Ab aggregation much more efficiently under photo-irradiation with visible light. 24,25 The dual property of C 60 to either scavenge or produce ROS has been used for a synergistic therapy for Alzheimer's disease (AD). 26 Fullerenes have two important advantages in AD research: their structure allows them to cross the blood-brain barrier 27 and they show a high capacity to inhibit Ab bril formation. 16,17,24,25,[28][29][30] However, there is controversy regarding the biocompatibility of fullerenes: some groups report nontoxic effects in different tested models, such as the LLC-PK1 proximal tubule cell model 31 or L929 mouse subcutaneous connective cells; 32 however, a high concentration of fullerenes induces toxicity in the same studies. Therefore, the correlation between the amount of fullerene necessary to perform its biological activity and biocompatibility remains unclear. Recently, impaired spatial memory with a signicant decrease in BDNF protein levels and gene expression has been demonstrated in rats injected with C 60 , but in contrast, it showed high antioxidant capacity. 33 Fullerene might be an important molecule for the treatment of neurodegenerative disorders, but the molecular design requires further research. An interesting approach to improve fullerenes for AD treatment is based on functionalization with different substituents that promote its stability, biocompatibility, capacity to cross the blood-brain barrier and ability to inhibit Ab bril formation. One successful method to obtain more polar fullerenes introducing functional groups consists in exohedral functionalization by the Bingel reaction. 34,35 In particular, functionalization with a malonate group led to water-soluble fullerenes and in addition serves as a precursor for further functionalization, so it can be used to link a variety of functional groups to obtain different derivatives.
In this study we synthesized diethyl fullerenemalonates and the corresponding sodium salts, through the Bingel reaction to obtain adducts of C 60 holding 1 to 3 diethyl malonyl and disodium malonyl substituents (C 60+n (COOR) 2n , where n ¼ 1-3 and R ¼ -CH 2 CH 3 , -Na) (Fig. 1). This is the rst report to demonstrate the effect of the type and number of organic addends of fullerenemalonates on anti-Ab activity and their biocompatibility with cells. The potential inhibitory effect of the fullerenes on Ab 42 bril formation was shown by thioavin uorescence assay and electron microscopy. Low cytotoxicity was shown in neuroblastoma SH-SY5Y cells exposed to the fullerenemalonates during 24 h, and the cytotoxic effect decreased even more in the presence of the corresponding sodium salt molecules. The in silico data obtained by atomistic molecular dynamics showed that the puried C 3 trisadduct binds to the Ab 42 monomer, mostly to 1D, 5R, 16K residues by means of hydrogen bonds and to 2A, 4F, 6H, 8S, 12V, 15Q residues.

Fullerene synthesis
The Bingel-type adducts were synthesized by cyclopropanation of C 60 with different equivalents of diethyl malonate, CBr 4 and 1,8-diazabicyclo- [5,4,0]undec-7-ene (DBU) as an auxiliary base, following the procedure published by X. Camps et al. and E. Straface et al. 36,37 Diethylmalonate mono-and bisadducts were isolated and puried by a chromatography column on silica gel. Chromatographic separation of the reaction mixture with nhexane/toluene (65 : 35) generated a rst fraction consisting of a residual amount of C 60 . Subsequently, pure toluene was used to yield a second fraction that consisted of the monoadduct contaminated with traces of C 60 , and nally there was a third fraction containing an isomeric mixture of the corresponding bisadducts. Seven regioisomers of the bisadducts (of the eight possible) have been isolated and characterized by Hirsch et al. 38 The second and third fractions were also puried by chromatography, using n-hexane/toluene (65 : 35) and pure toluene to remove the traces of C 60 or monoadduct to obtain pure monoadduct, C 61 (COOEt) 2 , and an isomeric mixture of bisadducts, C 62 (COOEt) 4 . The trisadducts were puried using elution of the reaction mixture with n-hexane/toluene (65 : 35) to pure toluene, allowing the separation of three fractions containing enriched samples of monoadduct, bisadducts and trisadducts as the major components, respectively. The third fraction containing semipure trisadducts was rechromatographed on silica gel using toluene as the mobile phase to obtain the puried isomeric mixture, C 63 (COOEt) 6 . Seven regioisomers of the trisadducts (of the 10 possible considering the restriction that only e or trans additions to bisadducts having e-and trans-positional relationships are considered) have been isolated and characterized by Djojo et al. 39 A puried sample of the C 3 trisadduct (C 3 -C 63 (COOEt) 6 ) was obtained by elution of the fraction containing the isomeric mixture of trisadducts with toluene/acetonitrile (99.5 : 0.5), producing a last fraction (red band) enriched in the regioisomeric trisadduct with C 3 symmetry. The fraction containing semipure isomer C 3 was further rechromatographed twice using toluene/ acetonitrile (99.5 : 0.5) thus obtaining a pure C 3 adduct (C 3 -C 63 (COOEt) 6 ). The identity of all compounds was conrmed by electrospray ionization time-of-ight mass spectrometry (ESI-TOF-MS), giving molecular ions identical to those calculated. The obtained data was as follow: C 61 (COOEt) 2 (m/z) 878.059 (M À ), calculated for C 67  The disodium fullerenemalonates were synthesized by hydrolysis of the respective diethyl fullerenemalonates, with a 1.5-fold (relative to ester groups) molar amount of NaOH (1 M) in tetrahydrofuran : methanol : water for the case of monoadduct C 61 (COONa) 2 and bisadducts C 62 (COONa) 4 and toluene : methanol : water in the case of the trisadducts C 63 (COONa) 6 and C 3 -C 63 (COONa) 6 , according to a method described already. 43 The reaction was stopped aer all of the starting diethyl malonate was consumed, as monitored by thinlayer chromatography. The sodium salts were triturated from toluene and water or methanol and isolated by centrifugation. Finally, the product was evaporated, and dried under vacuum. The yields were nearly quantitative. The samples were characterized by UV-vis and IR spectroscopies. Disodium fullerenemalonates were identied by the shi in the characteristic and strong iC]O vibration in the IR spectrum (KBr pellet) of the ester from v ¼ 1743 cm À1 to v ¼ 1638 cm À1 , 1621 cm À1 , 1626 cm À1 and 1638 cm À1 for the monoadduct, isomeric mixture of bisadducts and trisadducts, and C 3 -trisadduct, respectively; due to the mesomeric weakening of the iC]O double bond in the dicarboxylate.

Preparation and brillization of Ab 42 peptide
The Ab 42 monomer (1 mg) was resuspended in 1 ml of cold 1,1,1,3,3,3-hexauoro-2-propanol (HFIP), and it was kept in the dark for 30 min at 4 C to solubilize the peptide in the monomeric stage. 44,45 The solution was aliquoted (25 ml) and dried under vacuum in a rotary evaporator for 1 h at room temperature; the resultant transparent lm was stored at À80 C for further experiments. The amyloid brils were formed as previously described; 46 briey, the solubilized Ab 42 was dissolved in polymerization buffer containing PBS 1Â (nal concentration of 10 mM PO 4 3À , 137 mM NaCl, and 2.7 mM KCl) and 1% DMSO and incubated at 37 C at different time points. The required volume of polymerization buffer was adjusted according to each experiment.

Fibril formation analysis by western blot
Western blots were performed using 16% SDS-PAGE, loading 1.5 mg of non-boiled Ab 42 bril samples per line, followed by transfer to a nitrocellulose membrane and blocking with 5% skim milk. The blot was performed with a mouse monoclonal anti-b-amyloid antibody (Sigma-Aldrich) at 1 : 3000 dilution, incubated overnight and probed with goat anti-mouse Ig-G (Millipore) at 1 : 10 000, and, nally, detected using enhanced chemiluminescence reagents (Thermo Scientic).

Fluorescence analysis
Fibrillization of Ab 42 peptide was detected by ThT uorescence intensity that correlates with the number of brils formed. 47 ThT was added to the Ab 42 samples to nal concentrations of 10 and 20 mM, respectively, in a total volume of 200 ml of polymerization buffer. The uorescence was measured in a 96-well plate during 24 h at 37 C using a TECAN Innite (M1000PRO) spectrouorometer at 440 nm excitation and 490 nm emission. The uorescence intensity was the normalized value of Ab 42 aggregates at 24 h (positive control). Likewise, to evaluate the inhibitory effect of the fullerenes, 13 mM (nal concentration in the solution) of the respective fullerene was added to the same reaction. The relative uorescence of fullerenemalonates without Ab 42 was measured and subtracted from the respective assay with the peptide. The IC 50 for Ab 42 amyloid brillization inhibition were determined from the curves obtained by tting the average uorescence values in three independent experiments at the following fullerene concentrations: 2.5, 5, 7.5 and 10 mM. The experiments were done in triplicate.

Electron microscopy
The polymerization assay solution samples and that with disodium fullerenemalonates C 3 -C 63 (COONa) 6 and an isomeric mixture of C 62 (COONa) 4 were sedimented and placed side by side onto Formvar-coated copper grids for 1 min, followed by incubation in 50 mM ammonium bicarbonate for carbon coating of the sample for 3 min and then negatively stained with 2% uranyl acetate for 1 min. This procedure was repeated twice for 2 min and 1 min, respectively. Aer drying, the samples were imaged with a JEOL 1400 EX transmission electron microscope (TEM). The experiment was done in duplicate.

Cell viability assay
The neuroblastoma cell lines SH-SY5Y (ATCC) were grown in 25 cm 2 asks at 37 C with 5% CO 2 in advanced DMEM/F-12 medium (Sigma-Aldrich), supplemented with 10% fetal bovine serum (Invitrogen); for cytotoxicity assay, cells were seeded at a density of 1 Â 10 4 in triplicate per sample in a 96 well microplate 24 h before the treatment. The fullerenemalonates were incubated for 24 h at the corresponding concentration for their ability to inhibit Ab 42 aggregation, 13 mM. Likewise, the vehicles of each fullerenemalonate other than water (acetonitrile and ethanol) were assessed to validate its toxicity. The cell viability assay was evaluated using Thiazolyl Blue Tetrazolium Bromide (MTT, Sigma-Aldrich); based on the conversion of MTT to water-insoluble MTT-formazan of dark blue color by the mitochondrial dehydrogenases of living cells. The absorbance of formazan was measured at a wavelength of 570 nm in a plate reader, iMark Biorad. For a cytotoxic concentration (CC 50 ) the fullerenemalonates were incubated at different concentrations followed by a cell viability assay. The CC 50 was determined from the curves obtained by tting the average absorbance values in three independent experiments in the 10-80 mM fullerenemalonate concentration range.

In silico experiments
The topology, pdb le, for the Ab 42 molecule was taken from Crescenzi O. et al.; 48 the derivative fullerene was obtained by editing a pdb topology le of trisadduct with a diethyl malonyl substituents, 49 and further optimized using the Automated Topology Builder. 50 The atomistic force eld used was CHARMM36 51 together with the TIP3P model for water 52 modied 53 for use with this specic force eld. Four systems were considered: one simulation of Ab 42 chains in vacuo, two of Ab 42 in water at two different temperatures and the last one was composed of Ab 42 chains with trisadducts of fullerene, in water. The amount of water added was enough to mimic the experimental value of the water density under thermodynamic conditions of pressure and temperature. In the case of the ternary system, the concentration of Ab 42 and fullerene was kept low, close to the experimental values, preventing an excess of water molecules. The details of the simulation boxes and thermodynamic conditions are shown in Table 1. The pure system and the binary systems were used to set the stability of the peptide structure when we use the force eld CHARMM36. The initial boxes were prepared in an ensemble with a constant number of particles, constant pressure and constant temperature (NPT), using a Berendsen thermostat and barostat; the conditions were then changed to use an NVT ensemble, where V, the volume of the simulation box, is kept xed, using a Nose-Hoover thermostat and turning off the barostat, setting the volume of the box to the average value obtained in the equilibrated NPT simulation. Aer equilibration was reached (around 100 ns), the simulations were sampled for 10 ns. The Verlet algorithm was used to integrate the movement equations with a time step of 0.001 ps. Long-range electrostatic interactions were calculated using the Particle Mesh Ewald method with 1.2 nm as the cut-off. The van der Waals interactions were calculated using a cut-off equal to 1.2 nm. The coupling times of temperature and pressure were xed to 2.0 ps.

Ab polymer formation
The study of Ab aggregation inhibition requires an assay with Ab monomers that interact with molecules that prevent further aggregation. To validate that Ab monomers without preaggregated formation were appropriate for evaluating the adducts of C 60 , we used HFIP that breaks the beta sheet structures, preventing Ab aggregation. The monomer and bril formation were observed by western blot. The Ab monomer without pre-aggregates is shown in Fig. 2A with a single band between 2 and 4 kDa, and aggregates between 40 and 160 kDa were formed under the same conditions except for HFIP treatment, demonstrating the requirements of the treatment. Moreover, we evaluated bril formation during 24 h. Following 3 h of polymerization, low and high molecular weight polymers were seen as well as a reduction in the monomer at 6 h, validating the efficacy of the assay (Fig. 2B).

Fullerenemalonates inhibit amyloid b peptide aggregation
The inhibition of Ab aggregation by fullerene derivatives has been previously demonstrated; however, it has shown solubility complications in water and high toxicity in cells. In this study, we synthesized adducts of C 60 with one to three diethylmalonate substituents and their corresponding sodium salts to increase their biocompatibility and capacity to inhibit Ab aggregation, evaluated by a ThT uorescence assay. In Fig. 3A, the normalized values of the uorescence signal of the Ab aggregates showed signicantly lower aggregation of Ab in the presence of eight different fullerenemalonates compared to the control, in three independent experiments. The highest Ab aggregation inhibition was shown by both isomeric mixtures of bisadducts, C 62 (COONa) 4 and C 3 -symmetrical trisadduct (C 3 -C 63 (COONa) 6 ), with 97% and with 80%, respectively (Fig. 3B). To conrm these results, the polymerization assay was analyzed by TEM, showing scattered brils (Fig. 4A). In the presence of either C 62 (COONa) 4 or C 3 -C 63 (COONa) 6 ( Fig. 4B and C, respectively) Ab brils were not found. These data indicate that fullerenemalonates inhibit and/or delay Ab aggregation. The functionalization of C 60 with two or three disodium malonate substituents increased the efficacy compared to monoaddition.

Inhibitory activity of the fullerenemalonates by IC 50
To evaluate the Ab anti-aggregatory capacity of C 62 (COONa) 4 , the most efficient inhibitory fullerene, we determined the IC 50 value at 24 h by ThT assay, in the concentration range from 2.5 to 13 mM at a xed peptide concentration of 20 mM. The relative uorescence spectra showed that C 62 (COONa) 4 (Fig. 5) inhibits the process of Ab aggregation in a dose-dependent manner and the concentration to inhibit 50% of Ab bril formation determined by IC 50 value is equal to 6.7 mM.  This journal is © The Royal Society of Chemistry 2018

Biocompatibility of fullerenemalonates
To evaluate the cytotoxic effect of the fullerenemalonates, we used the SH-SY5Y cell line that is used as a model for neurodegenerative diseases including AD; 54 they can be differentiated from a dominantly cholinergic phenotype suitable for AD studies 55 and they have been tested for Ab toxicity in a large number of studies. [56][57][58] The cells were incubated for 24 h in the presence of fullerenemalonates at 13 mM, corresponding to an Ab bril inhibition concentration. The viability of the cells was evaluated by MTT assay. The values normalized to control showed the highest cell viability for fullerenemalonates bearing two (C 62 (COONa) 4 ) or three (C 63 (COONa) 6 and C 3 -C 63 (COONa) 6 ) disodium malonyl groups (Table 2). Likewise, disodium    fullerenemalonates demonstrated higher solubility compared to diethyl fullerenemalonates, indicating that this molecular modication is important for biocompatibility. To determine the cytotoxic concentration of C 62 (COONa) 4 and C 3 -C 63 (COONa) 6 to reduce cell viability by 50%, CC 50 was performed (Fig. 6). The dependence of cell viability on fullerene Fig. 6 Cytotoxic concentration of C 62 (COONa) 4 (A) and C 3 -C 63 (COONa) 6 (B) by CC 50 . Fig. 7 Snapshot of an equilibrated configuration of the ternary system: fullerene trisadduct in blue, Ab 42 is represented using different colours for different residues, water molecules are not shown for clarity. derivative concentration values: CC 50 : 38.8 mM and 69.6 mM for C 62 (COONa) 4 and C 3 -C 63 (COONa) 6 , respectively, conrmed that the toxic concentrations for these molecules are 7 and 12 times higher than the required concentration for Ab aggregation inhibition.

In silico interaction of trisadduct with Ab 42
The in vitro inhibition of the Ab 42 bril formation demonstrated an interaction between the fullerene molecules and Ab 42 . To propose a mechanism for this effect, the only isolated fullerene, C 3 -symmetrical trisadduct (C 3 -C 63 (COONa) 6 ), was analysed in silico in the presence of Ab 42 and water to determine the possible interactions leading to inhibition of Ab 42 bril formation. Herein atomistic molecular dynamics simulations of the ternary system were performed and the GROMACS 5.1.4 soware package 59 was used to generate the trajectory les. In Fig. 7 a snapshot of the ternary system can be found; this gure was prepared using the VMD soware package 60 using different colors for different residues; water molecules are not shown for clarity. Radial distribution functions (rdf) were calculated between every residue from Ab 42 , analysing the center of the mass corresponding to each aminoacid together with that of C 3 -C 63 (COONa) 6 . The analysis of the data indicated that residue 1D shows a high pick at 0.83 nm; residue 2A at 0.82 nm; residue 4F has the highest pick at 0.86 nm and a second one at 1.05 nm; 5R has a pick at 0.86 nm, in the same position; residue 6H shows its highest pick at 0.84 nm and a second at 1.16 nm; 8S at 0.88 nm; 12V at 0.82 nm; and nally, residue 15Q shows its highest pick at 0.81 nm. A representative plot of the 5R residue is shown in Fig. 8A. These data suggest a remarkable structure at very short distance (less than 1 nm) between these residues and the fullerene trisadduct, probably binding to the carboxylate group. On average, the fullerene molecules kept their positions during the simulation time. For all the other residues, the picks had an rdf signal lower than 20. The interactions between the residue and the carboxylate group were assessed; the rdf between the center of mass of the two oxygen atoms in the carboxylate group and the center of mass of each residue were also calculated. The analysis of these results indicated that the residues where the rdf was higher than 15 were the 1D, 2A, 4F and 5R residues. Only the plot of the 1D residue is shown in Fig. 8B. 6H, 8S, 12V, 15Q, and 16K residues presented rdf values lower than 20, but they were still well-dened structures. The number of hydrogen bonds and the hydrogen bond distribution as a function of donor-acceptor distance between them and the residues, during the simulation time, were also calculated. The residues that showed hydrogen bonds were 1D, 5R, and 16K. In Fig. 9A and B we show results of the 5R and 16K residues, respectively. 2A and 15Q residues contain a weaker structure of hydrogen bonds, so that they are not retained during the whole simulation time. The minimal distances between each residue and the fullerenemalonate adduct were calculated using the center of mass as reference, and the systems with the shortest distances were the 1D, 5R, and 16K residues. The results for 16K are shown in Fig. 9C. The other residues that retained shorter distances lower than 0.35 nm were 2A, 3E, 4F, 6H, 8S, 11E, 12V, 14H, 15H, 32 G and 35V. These results demonstrated which residues contain strong interactions with C 3 -C 63 (COONa) 6 .

Discussion
In this study, we demonstrated the capacity of fullerenemalonates to inhibit Ab bril formation. Our polymerization assay contains mostly monomers at the starting point, which conrms that the interaction of the fullerenes is not directly with pre-aggregates. This is important in terms of the molecular dynamics, since we evaluated the interaction of C 3 -C 63 (COONa) 6 specically with the Ab monomer in silico. Several experimental and computational studies have demonstrated that the type of substituent inserted on the fullerene surface has a remarkable inuence on their anti-amyloid activity. For example, in vitro experiments showed that 1,2-(dimethoxymethano) fullerene strongly inhibits Ab peptide aggregation in the early stages (IC 50 $ 9 mM, Ab 42 concentration 20 mM). 16 Another group has studied the anti-amyloid activity of the sodium salt of the fullerene polycarboxylic derivative C 60 Cl(C 6 H 4 CH 2 COONa), the sodium fullerenolate and complexes of fullerene with polyvinylpyrrolidone and revealed the existence of a strong antiamyloid activity in Ab 42 , concluding that the latter two had the most effective Ab inhibitory effect in both Ab 42 and muscle amyloid X-protein. 29,61,62 Recently, Bednarikova et al. demonstrated through in vitro and in silico experiments that fullerenol, fullerene C 60 modied with 16 OH groups (C 60 (OH) 16 ), inhibits the brillization of Ab 1-40 in a dose-dependent manner (IC 50 $ 32 mg ml À1 , Ab 1-40 concentration 10 mM) and the inhibition of bsheet formation results from the strong electrostatic interactions of the fullerenol OH groups with the polar, negatively charged amino acids. 28 Zhou et al. performed multiple all-atom explicit solvent molecular dynamics simulations to study the effect of fullerene substituents and concluded that functionalization with a dimethoxymethane group on the fullerene surface retarded the rotation of the fullerene, thus enhancing the binding stability of the 1,2-(dimethoxymethano)fullerene. 17 Xie et al. found by molecular dynamics simulations that the contact surface area between the fullerenes and the Ab [16][17][18][19][20][21][22] octamers is an important factor that affects the Ab-fullerene interaction and a large contact surface area usually implies strong interactions. 63 The results obtained in this study show that all fullerenemalonates inhibit Ab 42 aggregation and their activity depends highly on the number and the nature of the substituents attached to the fullerene surface. In addition, an in silico approach demonstrated that the inhibition of Ab 42 brillization by C 3 -C 63 (COONa) 6 results from strong electrostatic interactions and hydrogen bonding of the fullerenemalonate carboxylate groups predominantly with amino acid groups (residues 1D, 2A, 4F and 5R) and residues 1D, 5R, and 16K, respectively. The higher anti-aggregatory effect of C 62 (COONa) 4 (98% at 24 h, IC 50 6.7 mM, Ab 42 concentration 20 mM) may be explained by a balanced relationship between the number of organic addends and the contact surface area available compared to those of the monoadduct and trisadducts. Moreover, disodium fullerenemalonates exhibit higher anti-Ab activity compared to diethyl fullerenemalonates, which suggests that the addends/surface area ratio, besides the nature of the addends, plays an important role in the anti-Ab activity of the fullerene derivatives. This nding is consistent with previous studies, suggesting that the strong Abfullerene derivative interaction, due to introduced addends and the contact area available in the fullerene, signicantly weakens the Ab-Ab interaction and thus inhibits b-sheet formation. 17,28,63 Likewise, the appended organic addends on the carbon cage in fullerenemalonates make them truly amphiphilic and induce a strong tendency to self-assemble in polar solvents to form stable solutions of nanoaggregates. 25,43,[64][65][66][67][68] Specically, the self-assembly of sodium carboxylated fullerenes (C 61 (COONa) 2 and C 62 (COONa) 4 ) in aqueous solution produces solid spherical particles with an average hydrodynamic radius R h z 32 nm for C 61 (COONa) 2 and hollow shells with mainly two different size scales of R h z 23 nm and R h z 104 nm for the isomeric mixture of bisadducts, C 62 (COONa) 4 . 66 Therefore, it is not surprising that aqueous solutions of disodium fullerenemalonates (C 61 (COONa) 2 , C 62 (COONa) 4 and C 63 (COONa) 6 ) and other water-soluble fullerene derivatives evaluated as inhibitors of Ab aggregation comprise nanoclusters rather than individual solvated molecules. 16,24,25,29,30,61,62,69 Nevertheless, the results obtained in this study and other reports show that self-assembly in polar solvents to form stable nanoaggregates does not affect its anti-Ab activity.
Regarding the biocompatibility of fullerenemalonates, our results showed that modication of the fullerene surface with diethyl malonyl groups causes higher cytotoxicity compared to those with disodium malonyl groups. Moreover, it was found that, excluding the monoadduct sodium salt (C 61 (COONa) 2 ), disodium fullerenemalonates with two or three substituents were not toxic for the SH-SY5Y neuroblastoma cell line at the half maximal inhibitory concentration (IC 50 6.7 mM). Likewise, it was found that the cytotoxicity is reduced as the number of disodium malonyl substituents attached to the fullerene surface and their attachment symmetry increase. The higher biocompatibility of C 3 -symmetrical triadduct (C 3 -C 63 (COONa) 6 , viability 99%) compared to those of monoadduct (C 61 (COONa) 2 , viability 64%) and the isomeric mixture of bisadducts (C 62 (COONa) 4 , viability 91%) and triadducts (C 63 (COONa) 6 , viability 97%) may be attributed to the combined effect of the decrease in hydrophobic surface due to hydrophilic addends attached to the fullerene core and the e,e,e-symmetrical addition pattern of C 3 -C 63 (COONa) 6 . This study agrees with the reported literature, which suggests that a higher abundance of hydrophilic addends on the fullerene surface and a high-symmetry addition pattern, result in a decrease in cytotoxicity. 29,70,71

Conclusions
In this study we have demonstrated that fullerenemalonates bearing 1 to 3 diethyl malonyl substituents and their corresponding sodium salts interact and effectively reduce Ab bril formation in vitro. The bisadduct salts (C 62 (COONa) 4 ) and trisadduct (C 63 (COONa) 6 ) inhibit 98 and 83% of Ab aggregation, respectively. The 6.7 mM IC 50 value of a C 62 (COONa) 4 mixture conrmed one of the highest anti-amyloid capacities that has been reported. The anti-aggregatory effect of the bisadduct salts, C 62 (COONa) 4 , is mostly attributed to the balance between the hydrophobic surface and the number of substituents bound to the fullerene, promoting stability in the interaction with the Ab peptide. The sodium salts C 62 (COONa) 4 , C 63 (COONa) 6 and C 3 -C 63 (COONa) 6 showed low toxicity in neuroblastoma SH-SY5Y cell viability, suggesting that these molecules are highly biocompatible at concentrations which are able to effectively inhibit Ab aggregation. The lowest toxicity presented in the trisadduct salt, C 3 -C 63 (COONa) 6 , is associated with the combined effect of the reduction in the fullerene hydrophobic surface through the addition of hydrophilic substituents and the fullerene symmetry. The effective anti-amyloid activity and low toxicity of the bisadduct isomeric mixture (C 62 (COONa) 4 ) and trisadduct (C 63 (COONa) 6 ) could be promising candidates for further animal studies, and potential therapeutic molecules for the treatment of Alzheimer's disease.

Conflicts of interest
There are no conicts to declare.