Fluorographene-derived colloidal dispersions for donepezil delivery: From chemical synthesis to engineering of drug carriers

Supporting Information Fluorographene-derived colloidal dispersions for donepezil delivery: From chemical synthesis to engineering of drug carriers Sofia-Anna Ntousia, Martin Pykalb, Maria-Anna Gatouc, Vítězslav Hrubýb Paraskevi Papakyriakopouloua, Sergii Kalytchukb,d, Themistoklis Aivaliotisa, Nefeli Lagopatie,f, Pavlos Pantelisg, Evangelia A. Pavlatouc, Vassilis G. Gorgoulisf,g,h,i, Natassa Pippaa, Demetrios D. Chronopoulosk,* aSection of Pharmaceutical Technology, Department of Pharmacy, School of Health Sciences, National and Kapodistrian University of Athens, 15771 Athens, Greece bRegional Centre of Advanced Technologies and Materials, Czech Advanced Technology and Research Institute (CATRIN), Palacký University Olomouc, Olomouc 779 00, Czech Republic cLaboratory of General Chemistry, School of Chemical Engineering, National Technical University of Athens, Zografou Campus, 15772 Athens, Greece dNanotechnology Centre, Centre for Energy and Environmental Technologies, VSB–Technical University of Ostrava, Ostrava 708 00, Czech Republic eLaboratory of Biology, Department of Basic Medical Sciences, Medical School, National and Kapodistrian University of Athens, 11527 Athens, Greece fBiomedical Research Foundation, Academy of Athens, 11527 Athens, Greece gMolecular Carcinogenesis Group, Department of Histology and Embryology, Medical School, National Kapodistrian University of Athens (NKUA), 11527 Athens, Greece hNinewells Hospital and Medical School, University of Dundee, Dundee, UK. iFaculty Institute for Cancer Sciences, Manchester Academic Health Sciences Centre, University of Manchester, Manchester, UK. kLaboratory of Organic Chemistry, Department of Chemistry, National and Kapodistrian University of Athens, Panepistimiopolis, Athens 15771, Greece 1. Experimental section 1.1 Reagents and Materials Fluorinated graphite (GrF) (C:F 1:1.1, >61 wt. %F), sodium cyanide (NaCN) and dimethylformamide (DMF) were purchased from Sigma-Aldrich. Isopropanol (IPA), acetone, ethanol and nitric acid (HNO3) were purchased from Penta. All reagents were used as received without further purification. Donepezil hydrochloride (DH; molecular weight: 379.50 g/mol) was obtained from Cipla Ltd. (India) and utilized for both incorporation into the graphene-based delivery system and as a reference standard in high-performance liquid chromatography (HPLC) analysis. Regenerated cellulose syringe filters (Whatman Spartan, 0.45 μm) were employed to filter samples prior to HPLC injection. HPLC-grade solvents, including water, methanol, and acetonitrile, as well as other reagents, were sourced from Fischer Scientific (Pittsburgh, PA, USA). All preparations were made using triple-deionized water also purchased from Fischer Scientific. 1.2 Synthesis of graphene derivatives Exfoliation of Fluorographene Commercially available GrF (1 g) was placed in a round-bottom glass flask containing 2 L of IPA, and the mixture was sonicated for 6 h. Afterwards, the suspension was centrifuged at 2000 rpm for 30 min, and then the supernatant was subjected to filtration on a Whatman Nylon membrane filter (0.2 μm). Preparation of Graphene Acid (GA) Exfoliated FG by the aforementioned method (120 mg, ~4.00 mmol) was added to a roundbottom glass flask containing DMF (15 mL). After 72 h stirring and 4 h sonication of the mixture under N2 atmosphere, NaCN (800 mg, ~16.00 mmol) was added, and the mixture was heated at 130 °C for 24 h. The prepared GCN was obtained by centrifugations and consecutive washes with DMF, acetone, ethanol, and water. (Caution: The overall manipulation with open NaCN and the reaction itself have to be carried out in a fume hood, since in contact with air moisture, toxic hydrogen cyanide gas is generated. We recommend that the individuals involved in the synthesis and treatment of the reaction be equipped with respiratory protection. Furthermore, all waste from the reaction supernatant and supernatants from the DMF washing steps must be discarded as toxic organic waste and handled accordingly. All the equipment used for handling NaCN has to be treated with aqueous sodium hypochlorite solution after handling NaCN is finished.) For the synthesis of GA, a proper amount of HNO3 (65%) was slowly added at room temperature under stirring to a suspension of GCN in ultrapure H2O (ratio of suspension: 1 mL of H2O for 10 mg of GCN) in a round-bottom glass flask, until the final concentration of HNO3 in the reaction mixture reached 20%, and the mixture was refluxed at 100 °C for 24 h. Finally, purified GA was received after dialysis and freeze-drying. Preparation of Fluorinated Graphene Acid (FGA) Following the above protocol for the synthesis of GA, a lower amount of NaCN (400 mg, ~8.00 mmol), with respect to that used for the synthesis of GCN, was added into a FG suspension, resulting in the preparation of FGCN, which was converted to the corresponding FGA by following the above-mentioned procedure for synthesizing GA. The desired FGA was obtained in powder after dialysis and freeze drying. 1.3 Instrumentation Raman spectroscopy was performed on a DXR Raman microscope using the 633 nm excitation line of a diode laser. FT-IR spectra were recorded on an iS5 FTIR spectrometer (Thermo Nicolet) using the Smart Orbit ZnSe ATR accessory. Briefly, a droplet of an aqueous dispersion of the relevant material was placed on a ZnSe crystal and left to dry and form a film. Spectra were acquired by summing 32 scans recorded under a nitrogen gas flow through the ATR accessory. ATR and baseline corrections were applied to the collected spectra. X-ray photoelectron spectroscopy (XPS) was carried out with a PHI VersaProbe II (Physical Electronics) spectrometer using an Al Kα source (15 kV, 50 W). The obtained data were evaluated using the MultiPak (Ulvac - PHI, Inc.) software package. The charge correction of all the XPS spectra was done by positioning the maximum of the C 1s line to 284.8 eV. Thermogravimetric analysis (TGA) with evolved gas analysis (EGA) was performed using a Netzsch STA 449C Jupiter thermo-microbalance coupled with a QMS 403C Aëolos quadrupole mass spectrometer. Measurements were carried out in an α-Al2O3 open crucible under N2 flow. A temperature program from 40 to 1000 °C with a heating rate of 10 °C min-1 was used. Before each experiment, the crucible was heated to 1340 °C and then cooled to room temperature. Scanning electron microscopy (SEM) was carried out with a Hitachi SU6600 instrument with an accelerating voltage of 5 kV. The hydrodynamic diameter, as well as the distribution of the samples’ particles in dispersion, was assessed through dynamic light scattering (DLS) (Malvern Zetasizer Nano ZS, Malvern Panalytical Ltd., Malvern, UK). The scattering intensity’s recording was achieved using a 633 nm laser and a 173° scattering angle. Photoluminescence (PL) spectroscopy was performed using an FLS980 fluorescence spectrometer (Edinburgh Instruments) equipped with double monochromators on both the excitation and emission sides. A thermoelectrically cooled R928P photomultiplier (Hamamatsu Photonics) was employed as the detector, and a 450 W xenon arc lamp served as the excitation source for steady-state spectra. Spectral correction curves were provided by Edinburgh Instruments. Liquid samples were placed in a quartz microcuvette with a 102 mm pathlength (Hellma). Table S1. Stability assessment of FGA. T (days) Z-Average (nm) PDI Zeta Potential (mV) 0 1925 0.817 -14.2 10 1932 0.847 -15.1 20 2019 1.000 -14.7 30 2126 1.000 -13.2 Figure S1. Photoluminescence quenching of donepezil (Don) conjugated with FGA. (a) PL excitation spectra (λem = 474 nm) and (b) PL emission spectra (λex = 355 nm) of Don, FGA, and FGA_Don. 2. In vitro toxicity HEK293 cells were cultivated using DMEM High Glucose culture medium (provided by BioSera) that contained 10% FBS, 2 mmol/L glutamine, 100 IU/mL penicillin, and 100 μg/mL streptomycin at 37 °C. Every 48 h (approximately), medium was replaced and cells were passaged on a weekly basis using the trypsin/EDTA method. When cells reached a sufficient confluency, they were transferred to a 96-well plate, and 5000 cells/well were seeded. Cells were incubated for 24 h with a range of concentrations of 0-25 mg/mL of the samples. The MTT toxicity assay was performed as previously described [1]. All samples were held in duplicate. HPLC-PDA-based quantification of donepezil in FG systems The quantification of donepezil loading (%) in the FG systems was carried out using a highperformance liquid chromatography system coupled with a photodiode array detector (HPLCPDA), specifically the Shimadzu Prominence system. This system included an LC-20AD quaternary gradient pump with a built-in degasser, an SIL-HT auto-sampler, and an SPD-M20A photodiode array detector. Data collection and processing were conducted using LabSolutions® software (version 1.25 SP4, Shimadzu, Kyoto, Japan). Chromatographic separation was performed using a reverse-phase analytical column (Nucleosil 100-5 C18, 125 × 4.6 mm, 5 μm particle size) from MZ Analysentechnik, along with a matching C18 guard column (12.5 × 4.6 mm, 5 μm particle size). The mobile phase comprised a phosphate buffer, methanol, and acetonitrile in a 50:40:10 ratio, adjusted to pH 2.8 using 85% orthophosphoric acid. Isocratic elution was applied at a flow rate of 0.8 mL/min. The column temperature was maintained at 28 °C, and each sample injection was 30 μL. Spectral data were recorded with a resolution of 4 nm over the 200–400 nm wavelength range, with the detection wavelength set at 268 nm. The method described by Papakyriakopoulou et al. [2] was adapted and optimized for this study. The calibration curve was established using donepezil concentrations ranging from 0.5 to 7 μg/mL. To ensure thorough homogenization, the samples were sonicated and then filtered through 0.45 μm regenerated cellulose syringe filters to remove any debris or sediment while maintaining their original volume. Each solution underwent three HPLC measurements after appropriate dilutions. 3. Time-dependent release studies Materials The dialysis bags with a MW cutoff of 6kDa were purchased from Sigma-Aldrich Chemical Co., St. Louis, MO, USA, while phosphate-buffered saline (PBS) was obtained from Fischer Scientific (Pittsburgh, PA, USA). Methods A permeation experiment was conducted using dialysis bags set up to compare the release profiles of Don and FGA-Don. Briefly, 1 mL of each formulation (1 mg/mL in PBS, pH 7.4) was loaded into pre-wetted dialysis bags (Sigma-Aldrich; molecular weight cut-off 6 kDa). Each sack was placed in a beaker containing 15 mL of PBS (pH 7.4, mimicking the hydrodynamic conditions of human plasma) at room temperature (25 °C). At each time point, 0.5 mL of the external medium was withdrawn and immediately replaced with an equal volume of fresh PBS to maintain sink conditions. The samples were collected from the external medium at 15, 30, 45, 60, 90, 120, 180, and 240 min. Samples were stored in -70ºC until the day of the analysis. The samples were analyzed by the HPLC-PDA method described in the Supplementary material. 4. Computational studies Molecular dynamics (MD) simulations Four graphene-based flakes - graphene acid (GA) or fluorinated graphene acid (FGA) – were constructed with identical elemental composition to mimic experimental systems. Nitrogens were modeled as graphitic. The flakes were individually placed randomly in a cubic simulation box of 10 × 10 × 10 nm, together with 25 molecules of donepezil. The flakes were simulated in their dissociated forms, consistent with the measured pKa of GA of ~5.2 [3]. Donepezil molecules were modeled in their protonated form. Atomic charges for donepezil and the graphene-based materials were derived using the RESP procedure.[4] For graphene flakes, charges of functional groups were derived from small polycyclic aromatic hydrocarbon fragments (pyrene) at the HF/6-31G* level of theory. Bonded parameters for perfluorinated groups were taken from Watkins et al. [5]. The system was solvated with SPC/E water molecules [6]. The resulting excess of negative charge of the dissociated graphene flakes was neutralized by adding sodium counterions to maintain overall charge neutrality. All simulations were performed in GROMACS 5.1.2. [7] using the Verlet cutoff scheme. Electrostatic interactions were computed with the particle-mesh Ewald (PME) method [8] with a real-space cutoff of 1.0 nm. Van der Waals interactions were truncated at 1.0 nm. Hydrogen bonds were constrained with the LINCS algorithm [9]. Energy minimization was performed using steepest descent with a maximum of 20 000 steps. The system was thermalized by heating from 10 K to 300 K over 8 ns under an NPT ensemble, controlled with a velocity-rescale thermostat [10] (𝜏T = 0.1 ps) and the Berendsen barostat [11] (𝜏p = 20.0 ps, p = 1.0 bar, isotropic coupling). Production simulations were run for 50 ns in the NVT ensemble at 300 K, using a 2 fs timestep. Molecular graphics and visualization were generated using PyMOL [12]. Additional insights from the MD A more detailed analysis of the MD trajectories provided additional insight into the adsorption behavior of donepezil on GA and FGA flakes. Over the last 25 ns of simulation, the average number of bound molecules per flake was 4.1 for FGA and 3.1 for GA. The mean surface-tomolecule distances were 0.38 nm on FGA and 0.35 nm on GA, suggesting slightly looser yet more frequent binding on the fluorinated surface. Visual inspection of the trajectories showed that adsorption occurred primarily at the flake edges rather than on the basal plane. In these regions, donepezil often acted as a bridging agent between neighboring flakes (Figure 4b). Higher-order aggregates of donepezil molecules were also observed, typically stabilized through stacking of their indanone moieties. Direct basal-plane adsorption was rare, and when present, the molecules oriented away from carboxyl or fluorine groups to maximize π–π interactions with graphene-like domains. Regarding flake-flake interactions, most flakes remained dispersed, despite a relatively high concentration; however, in the GA system, we observed the formation of a two-flake stack. This aggregate was stabilized by Na+ ions positioned between the layers, which reduced electrostatic repulsion among negatively charged surface groups. Adsorption analysis Adsorption of donepezil molecules on graphene flakes was defined by a plane-based combined criterion, where for each flake, a local principal component analysis plane was fitted using the aromatic carbons. A donepezil heavy atom was considered adsorbed if it simultaneously satisfied the distance to the plane normal of ≤ 0.40 nm and an in-plane distance of ≤ 0.50 nm to the nearest flake atom. Flake aggregation analysis Aggregation between flakes was evaluated by monitoring inter-flake contacts. Two flakes were defined as aggregated if at least 20 interatomic contacts lay within 1.1 nm together with a minimum inter-flake atom-atom distance of ≤ 1.0 nm. A persistence criterion of 50 ps was applied to ensure only stable aggregates were counted. Interaction Energy Calculations To better characterize binding energetics, interaction energies between donepezil and GA or FGA were evaluated using a multistage approach. A nitrogen-doped graphene nanoflake was constructed with lateral dimensions of ~1.56 × 2.22 nm corresponding to a 9 × 7 hexagonal lattice containing 128 sp2 carbon atoms to be large enough to cover the entire donepezil molecule. Four graphitic nitrogen atoms were introduced into the lattice as substitutions. The flake was functionalized with 14 carboxyl groups, distributed evenly across both faces, with carboxyl positions randomized within the plane. Additional fluorine atoms were added, giving a total of 21 fluorine atoms. Donepezil was modeled in its protonated form, and the flake was considered in several partially dissociated states with 3, 4, 5, or 6 deprotonated carboxyl groups. Initial conformational searches of donepezil were performed using Avogadro software [13] with the general AMBER force field (GAFF) and systematic rotor search. The 10 lowest-energy conformers (five in chair and five in boat conformation of the piperidine ring) were retained. Each conformer was docked onto the differently functionalized flakes using AutoDock Vina [14] to generate starting poses, while the actual energetic ranking was obtained from subsequent semiempirical and DFT calculations. Between 5-9 complexes per system were selected for further semiempirical optimization using MOPAC2016 [15] with the PM6-DH2 method [16], which has been parametrized for noncovalent interactions, including π–π stacking and validated for graphene systems [17]. To explore the effect of fluorination patterns in more detail, four representative systems were considered: i) FGA with fluorine atoms restricted to the edges, ii) FGA, in which two basal carboxyl groups were substituted with fluorine atoms, iii) FGA with CF2-type edge terminations, and iv) GA without fluorination. These different substrates were derived from the same initial graphene nanoflake geometry and re-optimized within the framework. The dielectric constant of water (ε = 78.4) was applied via a continuum solvation model. The interaction energy was calculated as: ∆𝐸𝑖𝑛𝑡 = 𝐸𝐴𝐵 −(𝐸𝐴+ 𝐸𝐵), where EAB is the total energy of the complex and EA and EB are the energies of the isolated fragments with the same geometry as in the complex. For selected low-energy complexes, further calculations were performed with ORCA [18]. Although a broader set of complexes was initially selected based on the semiempirical screening, in practice, only the –4 charged flakes could be successfully optimized at the DFT level. Attempts with –5, or –6 charged flakes frequently failed to converge or led to structural instabilities, preventing reliable energy evaluation. Therefore, the DFT results reported here are limited to the subset of –4 charged FGA/GA flakes (Figure 1). Geometry optimizations were performed at the B3LYP/def2-SVP [19,20] level of theory, including the D3(BJ) dispersion correction [21], with the RIJCOSX approximation [22] and the auxiliary def2/J basis set [23] to accelerate Coulomb and exchange integral evaluation. A tight convergence criterion for the self-consistent field (SCF) procedure was applied (tightscf, MaxIter 300), and the AutoTRAH option was used to stabilize SCF convergence in these large π-systems. Numerical accuracy was improved by raising the integration accuracy (IntAcc 5) and using a dense integration grid (defgrid3). Solvent effects of water were accounted for with the conductor-like polarizable continuum model (CPCM) using the SMD solvation parameters [24], ensuring electrostatic stabilization of the charged flake– donepezil complexes. Following geometry optimization, single-point energy refinements were carried out at B3LYP/def2-TZVP and M06-2X/def2-TZVP levels [25]. Basis set superposition error (BSSE) was not explicitly corrected in DFT single points. With def2-TZVP this error is rather modest, but it could slightly shift absolute interaction energies.