Buparvaquone Nanostructured Lipid Carrier: Development of an Affordable Delivery System for the Treatment of Leishmaniases

Buparvaquone (BPQ), a veterinary drug, was formulated as nanostructured lipid carriers (NLC) for leishmaniases treatment. The formulation design addressed poor water solubility of BPQ and lack of human drug delivery system. The DSC/TG and microscopy methods were used for solid lipids screening. Softisan® 154 showed highest BPQ solubility in both methods. The BPQ solubility in liquid lipids using HPLC revealed Miglyol® 812 as the best option. Response surface methodology (RSM) was used to identify the optimal Softisan154 : Miglyol 812 ratios (7 : 10 to 2 : 1) and Kolliphor® P188 and Tween® 80 concentration (>3.0% w/w) aiming for z-average in the range of 100–300 nm for macrophage delivery. The NLC obtained by high-pressure homogenization showed low z-averages (<350 nm), polydispersity (<0.3), and encapsulation efficiency close to 100%. DSC/TG and microscopy in combination proved to be a powerful tool to select the solid lipid. The relationship among the variables, demonstrated by a linear mathematical model using RSM, allowed generating a design space. This design space showed the limits in which changes in the variables influenced the z-average. Therefore, these drug delivery systems have the potential to improve the availability of affordable medicines due to the low cost of raw materials, using well established, reliable, and feasible scale-up technology.


Introduction
Over the past decades, improper prescription and lack of patient compliance, due to severe toxicity of conventional medicines, have caused the development of widespread parasite resistance in leishmaniases. Also, no new drug substances have been introduced in the therapy, and there are no effective vaccines available to prevent or heal this neglected tropical disease [1,2]. Buparvaquone (BPQ) is used in the treatment of blood parasites in cattle and showed promising antileishmanial in vitro activity. Even though there is a formulation for veterinary use, no suitable pharmaceutical dosage form for a human is available. Also, the low water solubility of BPQ limits its bioavailability, which is the main challenge for developing a drug delivery system for enhancing its therapeutic efficacy.
The development of lipid nanoparticles is one of the most promising alternatives to overcome this limitation. Such delivery systems have the ability to overcome biological barriers, increase therapeutic efficacy, reduce toxicity, and allow drug release at the specific site [3,4]. Moreover, these colloidal carriers are quickly taken up by cells from the mononuclear phagocyte system [5], which are the host cells of the leishmaniases parasite infection. This mechanism makes nanoparticles an attractive vector for passive targeting of antileishmanial drug substances [6]. Therefore, lipid-based nanoparticles represent an opportunity to improve the leishmaniases therapy.

Selection of Liquid and Solid Lipids for BPQ-NLCs Preparation.
Regarding the liquid lipids, BPQ in excess was added to the liquid lipid and placed in orbital shaker at 25 ∘ C. After 24 hours, aliquots were filtered through a PVDF membrane of 0.22 m pore size and were injected into HPLC to determine the amount of solubilized BPQ, using the method described below.
BPQ solubility in solid lipids was evaluated according to Kasongo et al. [15]. Briefly, buparvaquone was added to the solid lipid in a concentration of 0.001% (w/w); this preparation was heated 10 ∘ C above the melting point of each lipid, under constant shaking for 24 hours. Aliquots were taken and observed by optical microscope; the BPQ concentration was adjusted until crystals observation. The results were ranked by the amount of crystals, the higher position in the rank, and the lower amount of crystals was observed.
Drug solubility and crystallization behavior in solid lipids were carried out by using DSC/TG. The BPQ thermal behavior and solid lipids and 1 : 1 physical mixtures (PM) were characterized in a DSC 4,000 Perkin Elmer cell (Perkin Elmer Corp., Norwalk, CT, USA), under a dynamic N 2 atmosphere (50 mL⋅min −1 ), using sealed aluminum capsules with about 2 mg of samples. DSC curves were obtained at a heating rate of 10 ∘ C min −1 in the temperature range from 25 to 290 ∘ C. An empty sealed pan was used as a reference.
The onset temperature, maximum peak in the melting range ( peak ), and melting enthalpy (Δ m ) were calculated using the software provided by PerkinElmer. The crystallinity indexes (CI) of BPQ were calculated in percentage according to the following equation [16][17][18][19]: where ΔH BPQ PM is the enthalpy of fusion (J⋅mg −1 ) of BPQ in the binary physical mixture of BPQ and solid lipid. ΔH BPQ 100% is the enthalpy of fusion (J⋅mg −1 ) of pure drug. is the dilution of BPQ in the PM, for example, 1 : 1: dilution of 2.

Preparation of BPQ-NLC.
After lipids selection, a mixture of solid (SL) and liquid (LL) lipids was tested in different ratios of SL : LL, from 1.5 (3 : 2 SL : LL) to 4 (4 : 1). The melting point of the mixture was determined by DSC as described  [20] by dispersing the lipid phase in the aqueous phase. The system was mixed by a mechanical homogenizer at 8000 rpm (Ultraturrax, IKA) for 5 min. The preemulsion was passed through a temperature-controlled high-pressure homogenizer (Nano DeBEE, BEE International), 60 bar and five cycles.

NLC Formulation and Optimization by Response Surface
Design. A three-factor (2 3 ) central composite design was employed for optimization of the nanostructured lipid carriers. The study was performed to evaluate the influence of the following independent variables: solid and liquid lipids ratio (SL : LL) and surfactants concentrations (independent variables) in the -average (response or dependent variable). The response surface method was applied to define the optimal conditions of the process and identification of relevant variables and their working range. The SL : LL range was 1.5 (3 : 2) (w/w) to 4.0 (4 : 1) (w/w); the surfactants concentrations were from 1.0 to 3.0% (w/w). Minitab5 17 software was used to generate the design matrix and data analyses.
The desirability method was used for the optimization process. The desirability function is based on the optimization techniques developed by Derringer and Suich [21]. In this approach, each response is first converted into a of each desirability function ranging between 0 and 1 (0 ≤ ≤ 1), where ( ) = 0 represents an entirely undesirable value , while ( ) = 1 represents a completely desirable or ideal response value. The individual values are then combined using a geometric mean to give a general desirability function [22].

Determination of -Average, Polydispersity Index, and
Zeta Potential. The -average of the nanostructured lipid carriers was determined immediately after high-pressure homogenization and was assessed by photon correlation spectroscopy (PCS) using Zetasizer ZS90 (Malvern Instruments, Malvern, UK) at 25 ∘ C and 90 ∘ ( = 10). The PCS technique calculates the -average as light-weighted intensity and polydispersity index (PDI) as a measure for the width of the -average distribution. The equipment was also used to measure the zeta potential (ZP). ZP measurements were carried out in purified water with a conductivity adjusted to 50 S⋅cm −1 by the addition of NaCl 0.1% (w/v), to avoid ZP fluctuations caused by the difference in conductivity ( = 3).
The pH was adjusted to 6.5±0.2 by the addition of 0.01 N HCl or 0.01 N NaOH solution.

Determination of BPQ Solubility in Liquid Lipids
and Encapsulation Efficiency. The method described by Venkatesh et al. [23] was used as a basis for quantification method for solubility in liquid lipids and encapsulation efficiency: in brief, C4 250 mm × 4.6 mm ID, 5 m particle size column; 1% glacial acetic acid in water, acetonitrile, and methanol 30 : 60 : 10 (v/v/v) mobile phase; = 251 nm; 50 L injection volume; 35 ∘ C column temperature. The final conditions were as follows: column Waters Xterra, C8 100 × 4.6 mm, 3.5 M particle size; a mobile phase of 1% glacial acetic acid, acetonitrile, and methanol (25 : 65 : 10); a column temperature of 35 ∘ C; injection volume 50 L and UV detection = 251 nm. The sample to assess the total drug amount was prepared in the following steps: acetonitrile was added to the NLCs and sonicated for 30 minutes. After cooling, the volume was completed with the same solvent, an aliquot was centrifuged at 376 RCF for 15 min, and the supernatant was diluted with mobile phase to appropriate concentration. Aliquots of the NLC were filtered through Amicon Ultra-0.5 mL Ultracel-10 membrane 50 kDa centrifugal filters (Millipore Merck). The filter was checked for BPQ adsorption previously. The filtrate was directly injected to quantify the supernatant content. Both samples were analyzed by HPLC as described above ( = 3). Buparvaquone encapsulation efficiency was calculated by the difference between total BPQ amount and the content in the supernatant.
In the matter to test if the NLC structure could increase the drug loading, two formulations with augmented amount of BPQ were prepared, and their encapsulation efficiency was determined. A stability study was conducted at 8 ∘ C for three months; the samples were stored in 8 mL borosilicate sealed vials. Table 1 shows the results of BPQ solubility in liquid lipids  determined by HPLC. Miglyol5 182 presented the highest solubility (11.55 g⋅kg −1 ). This lipid is widely used for parenteral, oral, and topical formulations; it is resistant to oxidation and it is a synthesized medium-chain triglyceride, which ensures uniformity of product batch by batch. Table 2 shows the results for solid lipids using the microscopy method. The lipids were ranked by the likely presence of crystals. Figures 1(a) and 1(b) show the criteria for crystals observation in the mixture. Figure 1(a) shows the lipid mixture, where BPQ was not soluble, and Figure 1(b) shows the lipid mixture, where BPQ was soluble.

Evaluation of BPQ Solubility in Solid and Liquid Lipids.
Microscopic assessment for the determination of drug solubility in a solid lipid was used by several authors [24][25][26][27][28][29]. However, it is time-consuming; the results are semiquantitative and the interpretation can vary depending on the qualification of the analyst.
To find a potential substitute or an auxiliary method for microscopic evaluation, differential scanning calorimetry and thermogravimetry (DSC/TG) were used as a faster and reproducible approach for evaluating the ability of the solid lipids to solubilize BPQ. Kovács et al. [13], Severino et al. [30], and Liu et al. [31], in their studies, provided the rational for the correlation of DSC data with solubility. The greater the reduction in CI of a drug peak, in a physical mixture with the SL, the better the ability of the lipid to solubilize the drug. Table 3 shows CI of BPQ and SL mixtures. Figure 2 shows Softisan 154 DSC/TG curves. For the other lipids, the same approach was performed. In the first heating cycle, the drug was more soluble in Gelucire 50/13 and Witepsol E85. The crystallinity index of BPQ was reduced to 40.6 and 48.8% for these lipids, respectively. In the second heating cycle, the Witepsol E85 and Softisan 154 showed almost the same CI, 49.0 and 49.1%, respectively. A second heating cycle was studied to simulate the manufacturing conditions of NLC by high-pressure homogenization [15,26]. By quench-cooling the sample, the contact time between drug and molten lipid is increased, and it is possible to evaluate better the capacity of the lipid to solubilize the drug. Additionally, it is possible to evaluate the drug and lipid recrystallization behavior. This method can also provide information about drug degradation and or drug-lipid incompatibility.
Microscopy revealed that Softisan 154, among the other tested lipids, presented highest BPQ solubility. However, the DSC results confirmed its high solubility as the second rank, Witepsol being the first ( Table 3). The divergence in the results revealed the relevance of the two methods.
By these findings, a rational for using the combined methods involves two steps: the screening of a high number of lipids with DSC/TG and microscopy evaluation for the most suitable ones to confirm and determine the drug amount that  is soluble in the lipid. This combination of methods can save time, materials, and work labor and can generate important information about the drug and lipids.

Preparation of BPQ-NCL. The Softisan 154 (SL) and
Miglyol 182 (LL) were mixed in different ratios to determine the maximum proportion of Miglyol 182. The SL : LL mixtures were visually homogeneous (no phase separation), and their melting points (MP) were above 40 ∘ C. For example, SL : LL 1.5 (3 : 2), the MP was 48.8 ∘ C. The SL : LL ratios where the mixtures showed melting points higher than 40 ∘ C were selected to ensure the NLC solid matrix structure [32]. Figure 3 shows visual aspect of the formulations, before homogenization (Figure 3(a)) and after homogenization (Figure 3(b)). The high-pressure homogenization proved to be reliable for BPQ-NLC preparation; all formulations showed low -averages <350 nm and low polydispersity <0.3 ( Table 4). The yellow color of the drug was not observed after homogenization for F1 (Figure 3), indicating the drug internalization into the lipid matrix, which was confirmed by encapsulation efficiency. The stability of the formulations was accessed; all of them presented -average and polydispersity variation less than 10% during three months (data now shown).

BPQ-NLC Optimization by Response Surface Design.
The response surface methodology (RSM) is an efficient and powerful statistical method to optimize processes. Using the RSM, it is possible to reduce the number of experimental tests, with consequent savings in time and materials. The statistical design identifies component interactions and critical process parameters. The methodology involves three steps: (1) the development of rotary central composed design; (2) the response surface modeling by regression analysis (ANOVA); and (3) the optimization of the process using the model [33].
For this work, 20 experiments (Table 4) were used to evaluate the influence of the SL : LL and concentrations of Kolliphor P188 and Tween 80 in the -average. The best fitting model from ANOVA was selected based on the comparison of statistical parameters including the -square ( 2 ), adjusted -square (adj-2 ), predicted -square (pred-2 ), lack-of-fit test, and values. The linear model and its significance were revealed by the value < 0.05 ( = 0.05) for each term (Table 5).
If the experimental environment is very noisy or significant variables are not considered in the experiment, it is possible that fraction of the observed data cannot be representative of the product/process. No significant lack-offit alone ensures model adequacy and therefore, measures of overall performance, referred to as -squares ( 2 ), were evaluated. For this work, 2 of 94.41%, adjusted 2 (adj-2 ) of 92.92% and predicted 2 (pred-2 ) of 91.04% show a suitable model from the data collected (Table 6).
When 2 , adj-2 , and pred-2 differ dramatically, there is an indication that nonsignificant terms were included in the model [34]. For the developed -average model, the three   (Table 6), which means the model is representative of the data and variables were representative of the product, regarding -average.
Additionally, the pred-2 indicates the high probability of the practical values being close to the theoretical calculations from the model. In the present study, the observed values were within the predicted range (95% CI) ( Table 6). Therefore, the model can predict the -average as a function of SL : LL ratio and surfactants concentrations. The -average is a critical quality attribute for intravenous administration and macrophage uptake. It was found that nanoparticles between 100 and 300 nm are more appropriate for macrophage internalization by phagocytosis [35,36]. Smaller particles (<100 nm) tend to be taken up by endocytosis, a process which occurs in virtually all cells where is the -average; SL : LL is the solid and liquid ratio; T80 is the % of Tween 80 (w/v); and K188 is the % of Kolliphor P188 (w/v).    Residuals investigation is an essential part of all statistical modeling. The residual analysis for the ANOVA model was verified by the normal probability plot and the histogram (Figure 4). In normal probability plot, the residue is represented according to its expected value, calculated from the assumption that the residues follow a normal distribution [37]. Considering the results, the location of points on the probability curve, and the histogram, the normality assumption is valid. Also, the graph of the individual observations showed the random behavior of the residuals. There was no phenomenon of heteroscedasticity. The analysis of the residuals confirmed that the choice of this model was appropriate [37]. Figure 5 showed that Kolliphor P188 reduced -average at a lower concentration than Tween 80. The result showed the area of -average <220 nm for Kolliphor P188 is larger than Tween 80. Likewise, the coded coefficient for Kolliphor P188 and Tween 80 is −33.78 and −16.85 (Table 6), which means that, in the same concentration, the formulation prepared using Kolliphor P188 will present lower -average than the one prepared using Tween 80. This finding can be explained by the longer central hydrophobic chain of polyoxypropylene from Kolliphor. Cohesive energy of lipophilic chain is stronger when compared to short lipophilic chain from Tween 80 [38,39]. Moreover, two hydrophilic chains of polyoxyethylene oxide (PEO) exhibit gelation phenomena when heated [40], which contributes to integrity and stability of the surfactant film.
According to Figure 5, when adjusting the SL : LL ratio to 2.75 (11 : 4), surfactants above 3% (w/w) are required to reach low -averages (see Table 7). The interaction between surfactants can be observed by the plot Kolliphor * Tween and by the Kolliphor * Tween coefficient interaction term, which presents a positive value of +28.88 ( Table 6). As a consequence, for the optimization, low amounts of Tween 80 were applied to reduce the impact of this interaction. Furthermore, it can be seen that -average <220 nm are only achieved at low ratios of SL : LL (<2.5).
For the mathematical model confirmation, two formulations were prepared: F1 and F2, according to Figures 6 and  7, respectively. The measured values of -average for F1 were 215.4 nm ± 1.7% (183.8 to 217.4 nm; CI 95%). For F2, they   were 231.2 nm ± 1.0% (200.8 to 232.4 nm; CI 95%), which are within the predicted range (Table 6), confirming the validity of the mathematical model. Table 8 shows the BPQ EEs for three formulations containing 320 (F1), 400 (F3), and 500 g⋅mL −1 (F4) of BPQ. These formulations were prepared using the same SL : LL ratio and surfactant concentration. They were prepared aiming to increase the drug loading since the structure of the matrix is disorganized by the presence of the liquid lipid and more drugs could be accommodated, as reported by Chinsriwongkul et al. [41] and Yuan et al. [42]. For these three formulations, the EEs were close to 100%, even in the F4, which was prepared with 156% of calculated drug loading. At the end of 3 months of stability, for all formulations, no drug was precipitated and no significant changes in the EE were found (data not shown). These findings revealed the role of NLCs for BPQ encapsulation, comparing to SLN and NE. The preparation of SLN would compromise the drug loading and encapsulation efficiency during stability since BPQ has low solubility in solid lipids. The development of NE could increase the drug loading for using only liquid lipid. However, the stability of the nanoparticles and the increasing amount of surfactants, or even a required replacement for more sophisticated and expensive surfactants, such as lecithin, would lead to complex formulations and, consequently, expensive medicines, which is a big concern for neglected diseases.

Conclusions
DSC/TG method revealed BPQ best solubility using Witepsol E85 and Softisan 154 as solid lipid. Microscopy supported Softisan 154 suitability for NLC preparation. Both techniques, in combination, proved to be a powerful and consistent tool to select the appropriate solid lipids in the preformulation studies. The Miglyol 182 allowed a drug loading up to 0.5 mg/mL due to the high solubility of BPQ (11.55 g⋅kg −1 ) in this liquid lipid.
The response surface methodology identified the critical surfactants concentration and Softisan 154 and Miglyol 812 ratios and the interaction between Kolliphor P188 and Tween 80. The relationship among these variables, demonstrated by a linear mathematical model, allowed generating a design space. This design space revealed the limits in the variables for the achievement of -averages at submicron size range.
The developed nanostructured lipid carriers showed low -averages (<350 nm), polydispersity (<0.3), and encapsulation efficiency close to 100%, even when the BPQ amount was raised 156%, endorsing the ability of the NCL to increase the drug loading of lipid-based nanoparticles.
Thus, these findings showed the successful development of buparvaquone nanostructured lipid carriers, which are promising formulations for the treatment of leishmaniases. This drug delivery system has the potential to improve the availability of affordable medicines due to the low cost of raw materials, using technology with well-established reliability, efficacy, and scale-up feasibility.