Computational simulation of nanostructured lipid carrier containing lipids from Cupuassu ( Theobroma grandiflorum ) seed fat : Design , interaction and molecular dynamic study

Drug delivery systems are constantly evolving and developing, as well as the search for promising and effective formulations for drug delivery. Computational simulation methods enable the development of complex systems, such as nanostructured lipid carriers (NLC), the understanding of interaction and dynamics between drug molecule and its transporter. In this work, aimed to simulate a NLC containing cupuassu fat triacylglycerols, carnauba wax and caprylic/capric acid triacylglycerol, stabilized with Tween 80 and Pluronic and ketoconazole enantiomer as drug was simulated. Initially, lipid mixtures were studied by Differential Scanning Calorimetry and X-ray diffraction. Subsequently, computational studies were carried out, among which Molecular Docking of ketoconazole to the lipid mixture and Molecular Dynamics of NLC system containing ketoconazole. From the results obtained it was possible to observe the main binding affinities of the drug and provide a better NLC formulation. It was also possible to propose a three-dimensional NLC model that was stable after molecular dynamics and ideal for future experimental studies.


Introduction
The search for new drug delivery systems (DDS) is boosted by the possibility of obtaining pharmaceutical forms with higher drug bioavailability, specific targeting, prolonged effect, controlled and constant release, low toxicity and good biodegradability, as well as being versatile and adaptable to various drugs (Puri, et al., 2009;Ramezani & Shamsara, 2016). However, time and cost involved in the preliminary experimental tests are inconvenient factors for the preparation of formulations. In this way, Molecular Modeling and Computational Simulation have been highlighted as powerful tools for drug design.
Computational models allow targeting promising formulations, anticipate molecular and biological properties, decrease the amount of test compositions, facilitate screening for formulation experimental design and optimize those that have already been developed but presented limitations (Carvalho, et al., 2003;Ramezanpour, et al., 2016;Sant'Anna, 2009).
Interest in lipid-based DDS, such as liposomes (Jämbeck, et al., 2014), micelles (Lee & Pastor, 2011) and lipid nanoparticles (Galindo, et al., 2020;Rosa, et al., 2020;Santos, et al., 2012), is increasing not only because of their ability to encapsulate and transport drugs and biomolecules but also because of the versatility of compositions that can be generated (Ramezanpour, et al., 2016). Nanostructured Lipid Carriers (NLC) are an evolution of lipid nanoparticles and represent an alternative system with solid and liquid lipids composing the lipid matrix, with greater storage capacity and stability, and release of the encapsulated drugs (Haider, et al., 2020;Salvi & Pawar, 2019). Several types of lipids can compose the lipid matrix, but depending on the chemical affinity for the drug, they may directly influence the success of DDS. Lipids of natural origin stand out mainly for modern medicine, as they have additional properties and are sources of bioactive compounds with multiple health benefits (Lacatusu et al., 2014), such as fat extracted from Cupuassu (Theobroma grandiflorum) seeds, with a strong tendency towards topical application due to the possibility of obtaining phytonutrients as a balanced composition of saturated and unsaturated fatty acids, predominantly of oleic acid and stearic acid, amino acids, vitamins (A, C, B1, B2 and B3) and important flavonoids (catechin, epicatechin, quercetin, kaempferol and theograndins (I and II) (Costa, et al., 2020;Quast, et al., 2011).
In this work, a nanostructured lipid carrier was proposed using Cupuassu fat triacylglycerols, carnauba wax and caprylic/capric acid triacylglycerol (TAC) integrating the lipid matrix, Tween 80 and Pluronic as surfactants and a ketoconazole enantiomer as a drug model representative. Computational simulations were performed for NLC in order to elucidate the three-dimensional structures, study molecular and electrostatic properties, as well as intermolecular interactions between drug and carrier. Results of this study can be used as support for guiding future development of NLC formulations, anticipating experimental tests and helping to obtain thriving formulations. Development, v. 9, n. 11, e92191110433, 2020 (CC BY 4.0) | ISSN 2525-3409 | DOI: http://dx.Doi.org/10.33448/rsd-v9i11.10433 5

Methodology
This is a semi-empirical study of quantitative and descriptive character (Pereira, et al., 2018), which use in silico and experimentals approaches. Therefore, this study was subdivided in experimental section and computational section.

Experimental section
Characterization and Selection of Lipid Mixtures: Three lipid mixtures (coded as M01, M02 and M03) were initially developed with ratios following the fat/wax 2:1 and the amount of Capric/Caprylic acid triacylglycerol was proportional to 100% of the mixture. M01 with proportion of 40% Cupuassu fat and 20% carnauba wax; For M02, 30% Cupuassu fat and 15% carnauba wax; and for M03 20% Cupuassu fat and 10% carnauba wax. Selection of lipid mixture was performed from the experimental results of Differential Scanning Calorimetry (DSC) and X-ray Diffraction (XRD) from isolated samples and mixtures.

Molecular Electrostatic Potential (MEP)
The Molecular Electrostatic Potential (MEP) was applied to study molecular reactivity patterns. This is a highly informative tool for electronic charges distribution of a molecule of interest. Here, MEP surfaces were derived from DTF method with hybrid functional B3LYP and basis set 6-31G (d, p) (Plumley & Dannenberg, 2011) calculated and generated in Spartan Student 8 software (wavefunction, Inc.) (Engel & Reid, 2006). These surfaces correspond to an isodensity value of 0.05 a.u.

Lipid Matrix Simulation
To simulate a three-dimensional matrix, number of molecules used, corresponded to the percentage of the selected lipid mixture component, as described in the Table 1. Lipid matrix was simulated in Packmol Software Package (Martínez et al., 2009) with simple mixture model and 70Å orthorhombic box. Three-dimensional image was generated in Visual Molecular Dynamycs (VMD) program (Xu, et al., 1996).

Molecular Docking
Molecular Docking study was performed with AutoDock 4.2 software (Morris, et al., 2009) for the selected lipid mixture with regular precision with a maximum of 100 conformations per candidate. Kollman charges required for calculation were added and nonpolar hydrogens suppressed. Rotatable bonds of the ligand (ketoconazole) were automatically defined (Silva, et al., 2017;Silva, et al., 2020). Conformations were classified using scoring function and Lamarckian genetic algorithm. After locating possible binding sites, conformations of the doped complex were optimized using the steepest decent algorithm until convergence, with a maximum of 20 interactions. The grid box dimensions were (X = 22 / Y = 14 / Z = 14).
Files of automated Topology Builder (ATB) topologies were generated for system constituents which were later minimized and thermalized to 298K. Coordinates of each atom of the system remained fixed during thermalization, with a constant force of 1.0 x 103 KJ.mol -1 nm -2 . Aqueous systems were solvated with water molecules through the simple point charge model: Simple Point Charge (SPC/TIP3). Verlet-Verlet algorithm and canonical ensemble (Namba, et al., 2008) were used. The long-range electrostatic contributions were treated through Reaction Field. Yet electrostatic and van der Waals interactions used to treat the short-range interactions occur within a cut-off radius of 1.4 nm, avoiding the occurrence of interactions between atoms and their own virtual images three-dimensionally replicated by periodic boundary conditions. Simulation was performed in 20ns, using FF99S parameter set within the Gromacs program (Van Der Spoel, et al., 2005).

Lipid matrix selection
Best DSC and XRD results were established as a criterion to select the lipid matrix.
DSC curves are demonstrated in Figure 1 and in addition, Table 3 shows data concerning melting temperature, on-set temperature and enthalpy. Source: Authors. Research, Society and Development, v. 9, n. 11, e92191110433, 2020 (CC BY 4.0) | ISSN 2525-3409 | DOI: http://dx.Doi.org/10.33448/rsd-v9i11.10433 9 In Figure 1, shows the DSC curves the mixtures lipids (M01, M02 and M03), Cupuassu fat and carnauba wax. Each peak observed in the curves, representing a thermal event that occurred by the temperature variation. In Table 3, shows temperature values of the beginning of the thermal event (on-set), the moment of fusion (melting) and the heat enthalpy generated for each component. values found by other authors (Quast, et al., 2011;Silva et al., 2009). In the Figure 2b the Carnauba wax presented two high intensity peaks, one at 21.51° and another at 23.82°, characteristic of this wax (Villalobos-Hernández & Müller-Goymann, 2006). One can clearly see changes in peak intensity in the diffractogram of the mixtures (according the Figure 2c).
Mixture M01 presented two peaks of medium intensity at 19.51º and 23.86º and one of high intensity at 21.56º. In mixture M02 two peaks of medium intensity were also observed at 19.45º and 23.97º and one of low intensity at 21.56º. In mixture M03 there is a peak broadening at 22º followed by a peak of medium intensity at 21.54º and a decrease at 23.97º.
Among the three mixtures, M03 presented lower intensity peaks, suggesting structures with lower crystallinity. Observed in Figure 2, that the intensity of each peak generated samples studies. This is due to the fact the samples have varied structures crystallinity. Some tend to be more Research, Society and Development, v. 9, n. 11, e92191110433, 2020 (CC BY 4.0) | ISSN 2525-3409 | DOI: http://dx.Doi.org/10.33448/rsd-v9i11.10433 crystalline and others more amorphize, this directly refletct the spacing and intensity peak in the diffractograms the Figure 2.
Changes observed in diffractograms when comparing fat, wax and mixtures may indicate that there has been a polymorphic change of the lipid mixtures in order to decrease crystalline arrangement and there was a greater tendency to amorphize, especially in mixture M03. Relating DSC results to those of XRD one can say that M03 is the mixture of greater interest to produce NLC, once the crystalline form of the lipid matrix directly influences drug release profile, as in general, very ordered lipid-solid matrices have a lower diffusion rate of active substance, and amorphous regions facilitate drug incorporation.

Obtaining the three-dimensional structures models
Initially, structures representing NLC components were established, as observed in the  (Novotná, et al., 2014). Considering the frequent use of ketoconazole, its chiral structure and numerous drug interactions, it is important to study the enantiospecific interactions, mainly because individual drug enantiomers may exhibit different kinetic and dynamic behavior. Studies of antifungal activities of ketoconazole cis enantiomers against seven strains of Candida spp. (Novotná, et al., 2014) demonstrated that for five strains, the (2S, 4R) -(-) KET was about seven times more potent than (2R, 4S) -(+) KET, so it seems to be a more potent inhibitor for the tested strains. In this work the enantiomer chosen to be representative of ketoconazole in NLC was (2S, 4R) -(-) KET ( Figure 4d). All seven structures were optimized with DFT method using B3LYP 6-31G (d, p) basis set in the Gaussian program.  Research, Society and Development, v. 9, n. 11, e92191110433, 2020 (CC BY 4.0) | ISSN 2525-3409 | DOI: http://dx.Doi.org/10.33448/rsd-v9i11.10433

Molecular Electrostatic Potential (mep)
MEP surfaces were calculated to understand groups with the highest affinity for intermolecular interactions established between NLC constituents and ketoconazole. Figure 5 and Figure 6 show MEP surfaces generated for each structure demonstrate nucleophilic regions (negative electrostatic potential) and electrophilic regions (positive electrostatic potential). In all structures, positive electrostatic potentials were observed around hydrogen atoms, which are more susceptible to nucleophilic attack ( Figure 5

Electronic properties (homo and lumo)
The highest occupied molecular orbital (HOMO) and the lowest unoccupied molecular orbital (LUMO) are border orbitals and are related to the energy that the molecular orbital occupies (Namba, et al., 2008). They provide information about the electron/donor or electron/acceptor character of an atom and a complex formed by charge transfer. HOMO energy is related to the ability to donate electrons, so the higher the HOMO energy, the greater the ability to donate electrons. LUMO energy is related to the ability to receive electrons from a molecule, the lower the LUMO energy, the lower the resistance to accepting electrons (Arroio, et al., 2010).
In the Table 4 the values are described the optmization energy (Hatree) of HOMO energy, LUMO energy and Gap values. Research, Society and Development, v. 9, n. 11, e92191110433, 2020 (CC BY 4.0) | ISSN 2525-3409 | DOI: http://dx.Doi.org/10.33448/rsd-v9i11.10433 Table 4. Orbitals HOMO, LUMO and interval Gap (H-L). Molecules with lower Gap values (H-L) are more reactive and have lower stability (Zhang & Musgrave, 2007). According to Table 4 it can be observed ketoconazole, SOO and SOS had lower module values, so they prove to be the most reactive compounds.

Model of lipid matrix
The three-dimensional model of mixture M03 (Figure 7) was generated using Packmol program and snapshot generated in VMD. Packmol creates an optimized model for further molecular docking studies and molecular dynamics simulations. The generated model aims to ensure that short-range repulsive interactions do not disturb or interfere with simulations. Analysis of Figure 7 shows that the model for the proposed mixture was quite satisfactory, since there was no shock between components of the mixtures, molecules were well distributed in the box space, and it was possible to create appropriate initial configurations. This molecular arrangement is fundamental to generate suitable systems for subsequent computational simulations, since it minimizes the possibility of destabilization of the system because of great repulsive interactions, which is one of the reasons for failure of more complex simulations such as molecular dynamics.

Molecular docking
Result obtained from Packmol software was used as a starting point to determine the best conformation and preferred site of ketoconazole in the selected mixture (M03). The model generated twenty iteractions, among which the one that presented the best conformation (model 10) had the lowest affinity energy, -5.3 Kcal/mol, best spatial conformation and greatest stability. The Figure 8 shows the main interactions established between ketoconazole and the other lipid matrix compounds. Research, Society and Development, v. 9, n. 11, e92191110433, 2020 (CC BY 4.0) | ISSN 2525-3409 | DOI: http://dx.Doi.org/10.33448/rsd-v9i11.10433

Molecular dynamics simulation
Initially a NLC three-dimensional model was generated with Gromacs software, using 120Å orthorhombic box and the following components: SOS, SOO, wax, TAC, Tween 80, pluronic, ketoconazole and water. After 20ns of molecular dynamics (MD) it was observed, in the Figure 9, that the formed system stabilized in such a way that structures agglomerated, tending to a spherical shape, similar to droplets of oily phase when dispersed in aqueous phase, stabilized by surfactants, in emulsions of lipid nanoparticles (Bruxel et al., 2012). Total energy of the system was -1895.47 KJ/mol, density 0.9765 kg/m 3 and 100nm of diameter medium. Total energy remained stable during simulation. After MD simulations, hydrogen bond type interactions with ketoconazole were observed. Many biological processes are directly related to solubility of substances, which, by turn, are related to bond formation, therefore hydrogen bonds are some of the most relevant (Martins, et al., 2013). Hydrogen bonds with distances between 2.2-2.5Å are considered strong, those of 2.5-3.2Å are moderate and 3.2-4.0Å are considered weak. Hydrogen bonds are among the interactions responsible for the crystalline ordering of a system, the attraction forces that bind atoms in a crystalline lattice are directly proportional to the crystalline order, which means that the stronger the bonds, the more cohesive are the atoms of the lattice and the higher the crystal order (Klein & Dutrow, 2009), which could influence encapsulation, permeability, release and diffusion of the drug in lipid nanoparticles (Desai, et al., 2012; In the Table 5 shows the main interactions that occurred between ketoconazole and other NLC components with their respective binding distances, obtained after dynamics simulation. and H-alkyl of TAC (TAC 76). Research, Society and Development, v. 9, n. 11, e92191110433, 2020 (CC BY 4.0) | ISSN 2525-3409 | DOI: http://dx.Doi.org/10.33448/rsd-v9i11.10433 20

Radial distribution function (RDF)
Radial distribution function represents the probability of finding an atom or molecule in a spherical shell at some distance away (r) from an atom or reference molecule. RDF, g(r), provides an estimate of spatial conformation of atoms and / or molecules relative to a focal point and demonstrates a long-range arrangement in the structure or system (Razmimanesh, et al., 2015). The figure 10 shows the estimates RDF the molecules of NLC simulated.
In the figure 10 shows RDF diagrams generated by the interaction of NLC proposed, Note that Figure 10a demonstrates RDF diagram using ketoconazole as reference molecule and related to SOS (SOS), SOO (SOO), wax (CER), TAC (TAC) and Tween 80 (T80) in a water free molecular system, calculated with a distance of up to 3.5Å from the reference molecule. No interactions with Pluronic were observed. Around 2Å a well defined and discrete peak with wax was observed, registering its first interaction in the shortest distance. This event is followed by well defined and discrete peaks around 2 Å to 2.5Å of interaction with Tween 80, SOO, SOS and TAC. RDF profile of ketoconazole showed that interactions with higher intensity peaks occurred with Tween 80 molecules suggesting its higher lipophilic affinity. Interactions were generally observed from approximately 2.0Å on.  discrete peaks were observed at a distance around 2.0Å in relation to Pluronic and 2.5Å in relation to Tween 80. It was observed that curves of greater intensity were obtained by interaction with Pluronic, which may suggest molecules with a more hydrophilic character at this distance. Since in aqueous environments, where water is used as a reference, the higher peaks suggest a higher hydrophilic character and the lower ones suggest a greater amount of hydrophobic regions (Wang, et al., 2013). No interactions were obtained with wax and ketoconazole. A discrete peak with water was Research, Society and Development, v. 9, n. 11, e92191110433, 2020 (CC BY 4.0) | ISSN 2525-3409 | DOI: http://dx.Doi.org/10.33448/rsd-v9i11.10433 22 observed at around 1.75Å, followed by a well defined peak of medium intensity with TAC around 2.25Å, and around 2,5Å other well-defined high intensity peaks with SOO and medium intensity with Tween 80 and SOS. In general, interactions of greater intensity occurred in relation to SOO, suggesting greater affinity of its molecules to Pluronic. Figure 10d demonstrates RDF diagram in which Tween 80 (T80) is the reference molecule related to ketoconazole (CET), wax, (CER), SOO (SOO), TAC (TAC), pluronic (PLU) and water (SOL). The first well-defined and low-intensity peak was observed at around 2.0 Å due to interaction with TAC; around 2.25Å the first interaction peak of medium intensity with ketoconazole and SOO was observed and in the range of 2.5Å interaction peaks of medium intensity with wax and another very discrete with Pluronic were also observed. In general, Tween 80 had stronger interactions with SOO and ketoconazole.

Conclusion
In conclusion, we report a three-dimensional model of lipid mixture whose molecules remained well distributed and no destabilization was possible. The NLC system proposed in this work is stable and with interactions of moderate relative strength, and possibly would have little influence on the active principle diffusion. A satisfactory model of NLC containing ketoconazole was generated from Molecular Docking and Molecular Dynamics studies.
The future perspectives of this study include use this model of NLC for more computational studies of the next stages of lipophilic drug delivery system, as the evaluation of the drug release and absorption profile. As well as, make this an experimental model for the rational planning for development of lipid formulations.