Tunable Release of Curcumin with an In Silico-Supported Approach from Mixtures of Highly Porous PLGA Microparticles.

In recent years, drug delivery systems have become some of the main topics within the biomedical field. In this scenario, polymeric microparticles (MPs) are often used as carriers to improve drug stability and drug pharmacokinetics in agreement with this kind of treatment. To avoid a mere and time-consuming empirical approach for the optimization of the pharmacokinetics of an MP-based formulation, here, we propose a simple predictive in silico-supported approach. As an example, in this study, we report the ability to predict and tune the release of curcumin (CUR), used as a model drug, from a designed combination of different poly(d,l-lactide-co-glycolide) (PLGA) MPs kinds. In detail, all CUR–PLGA MPs were synthesized by double emulsion technique and their chemical–physical properties were characterized by Mastersizer and scanning electron microscopy (SEM). Moreover, for all the MPs, CUR encapsulation efficiency and kinetic release were investigated through the UV–vis spectroscopy. This approach, based on the combination of in silico and experimental methods, could be a promising platform in several biomedical applications such as vaccinations, cancer-treatment, diabetes therapy and so on.


Introduction
In recent decades, polymeric microparticles (MPs) have been widely used as drug delivery systems for the controlled release of small molecules, proteins or peptides [1][2][3][4][5]. The reason of this great diffusion is due to several attractive features such as the use of non-laborious techniques [6], low production costs [7], simplicity in industrial scale-up [8] and possibilities of different ways of administration (oral, ocular, parental, inhalation) [9,10]. Particularly, biodegradable MPs composed of PLGA, a random copolymer of poly(glycolic acid) (PGA) and poly(lactic acid) (PLA), are well-established drug delivery systems for small macromolecules involved in the treatment of several important diseases including cancer [11]. PLGA is also a Food and Drug Administration (FDA) and European Medicine Agency (EMA) approved polymer for ophthalmologic, and other medical applications. PLGA shows relatively high miscibility with other polymers and solvents [11], and, moreover, it is able to encapsulate both hydrophilic and lipophilic drugs [12]. The use of PLGA MPs has many advantages, Materials 2020, 13, 1807 3 of 12

CUR-MPs Production
CUR in Oil Phase (CUR-Oil) and Water Phase (CUR-Water) Formulation CUR-MPs were prepared by the water/oil/water double emulsion/solvent evaporation technique as already reported [36]. In particular, a gas foaming porous agent, ammonium bicarbonate (ABC), at a concentration of 7.5 mg/mL was added in the first emulsion to obtain highly porous particles. Twelve milligrams of curcumin were loaded in DCM for the oil phase preparation, while the same amount of drug was dissolved in 100 µL of ethanol and 900 µL of water plus 1 mL of DMSO in the water phase formulation.

CUR-o/w 20% Oil Nano-Emulsion (CUR-NE) as Water Phase
The CUR-NE was prepared as previously reported [33,37]. Briefly, 1.2 mg of egg-lecithin (surfactant) was dissolved in 5 mL of soybean oil (oil phase). After, 20.83 mg of CUR was added to the mixture [38]. The final emulsion was obtained by adding 19.3 mL of Milli-Q water to oil phase [34]. After the process, 100 µL of CUR-NE was used to produce CUR-NE-MPs as described in the previous paragraph.

Confocal Microscopy
All three CUR-MPs formulations were characterized by confocal microscopy (Leica SP5 microscope (Wetzlar, Germania)) in order to evaluate the signal of the molecule inside them. In detail, fluorescence images were acquired using an HCX IRAPO L 40×/0.95 water objective and a 488 nm laser as an excitation source as already described [39].

Microparticle Size and Polydispersity Index (PDI)
The mean size and the PDI of all CUR-MPs were determined by static light scattering (Mastersizer 3000, Malvern Instruments, Malvern, UK) using a concentration of 3 mg/mL in water.

Scanning Electron Microscopy (SEM)
CUR-MPs morphology was evaluated by SEM microscopy as already described [40]. Concisely, 20 µL were deposited on a standard SEM pin stub and analyzed by FESEM ULTRA-PLUS (Zeiss) (Milan, Italy) at 5 kV with the SE2 detector. Moreover, the internal porous structure of the MPs was investigated using a PDMS 2 mm in thickness cured at 80 • C for 30 min. After cooling, MPs were deposited on it and another PDMS layer 2mm in thickness was used to cover them up. Finally, the solid PDMS block was frozen in liquid nitrogen (-196 • C) and sectioned using the Leica CryoUltra Microtome EM-FC7-UC7 (Milan, Italy).

Entrapment Efficiency (%ï) of CUR inside MPs
The %ïof curcumin inside the three formulations of MPs was measured dissolving 10 mg of MPs in 1 mL of DMSO for 30 min, at room temperature. The solution was then analyzed by UV-vis (UV-Visible-V-730 UV-Visible Spectrophotometer, Jasco, (Cremella, (LC), Italy) following the signal at 426 nm. The quantity of curcumin-loaded was obtained through the Beer-Lambert law using 58,547 dm 3 ·mol −1 ·cm −1 as the molar extinction coefficient of curcumin in DMSO [41]. All experiments were performed in triplicate.

In Vitro Release Study
In Silico Approach Curcumin in oil phase, water phase and nano-emulsion experimental release data were fitted using MATLAB®(v.R2019a) (Turin, Italy) employing an exponential growth model. In particular, the curcumin release C r was described by: where a and b are the model parameters, and y = 0 at t = 0 the initial conditions. A simple release kinetics prediction under the non-linear first-order assumption of Equation (1) was developed by Equation (2): where a n and b n are the model parameters, C n is the percentage of weighted curcumin MPs, n is the number of different MP formulations considered.

In Vitro Cumulative Release of CUR from MPs
In vitro curcumin release profile was obtained by the UV-vis technique (V-730 UV-Visible Spectrophotometer, Jasco, Cremella, (LC), Italy)).Aliquots of 5 mg of the three different microparticle formulations were suspended in 1.5 mL of phosphate buffer saline PBS at pH 7.2, vortexed under magnetic stirring at 550 rpm and incubated at 37 • C. At defined time intervals, 1 mL of PBS was removed without removing particles. The supernatants were then diluted 1:1 in ethanol and analyzed by UV-vis using as molar extinction coefficient 28,648 dm 3 ·mol −1 ·cm −1 [42].The experiments were achieved in triplicate.

CUR-Microparticle Production and Morphological Characterization
CUR-MPs were produced through the double emulsion technique using the method of solvent evaporation [36]. Three different configurations were obtained (CUR-NE, CUR-oil and CUR-water) in order to produce microspheres with curcumin molecules embedded inside or outside the porous structures. This strategy was designed to produce MPs with different drug release kinetics in physiological conditions. To compare the differences between the microparticles, their morphology was evaluated using Confocal and SEM microscopies. In particular, as shown in Figure 1A,B, CUR-oil and CUR-water microparticles were visibly homogeneous with a high fluorescent signal corresponding to the embedded curcumin molecules. In detail, in the oil configuration, all curcumin was outside the porous structure as expected, while in the water conformation most of it was located inside the pores. Some aggregation phenomena were instead visible in the CUR-NE microparticles, probably due to the instability of the nano-emulsion during the production process ( Figure 1C). Similar considerations can be made analyzing microparticles by SEM microscopy. In general, CUR-oil and CUR-water microparticles showed a homogeneous polymeric surface (Figure 2A,B) and by investigating their internal porous structure it was possible to confirm an open porosity for both the strategies of production ( Figure 2D,E). As forCUR-NE microparticles, they displayed good open porosity as the configurations just described ( Figure 2F) but their polymeric surface showed a slightly porous structure ( Figure 2C), maybe some nano-emulsion droplets aggregated on the surface generating a closed superficial porosity. The homogeneity of MPs was also confirmed by analyzing their size with a Malvern Mastersizer. The obtained results showed that CUR-oil and CUR-water microparticles have a uniform distribution with a mean diameter of 13.36µm and 9.32 µm ( Figure 3A,B).Contrarily, as forthe CUR-NE, despite having an average diameter of 7.9 µm, they expose a less sharp curve with a large peak at ≅1 µm typical of the nano-emulsion used in this study, corroborating our hypothesis about its instability during the production phase of the microparticles ( Figure 3C). Similar considerations can be made analyzing microparticles by SEM microscopy. In general, CUR-oil and CUR-water microparticles showed a homogeneous polymeric surface (Figure 2A Similar considerations can be made analyzing microparticles by SEM microscopy. In general, CUR-oil and CUR-water microparticles showed a homogeneous polymeric surface (Figure 2A,B) and by investigating their internal porous structure it was possible to confirm an open porosity for both the strategies of production ( Figure 2D,E). As forCUR-NE microparticles, they displayed good open porosity as the configurations just described ( Figure 2F) but their polymeric surface showed a slightly porous structure ( Figure 2C), maybe some nano-emulsion droplets aggregated on the surface generating a closed superficial porosity. The homogeneity of MPs was also confirmed by analyzing their size with a Malvern Mastersizer. The obtained results showed that CUR-oil and CUR-water microparticles have a uniform distribution with a mean diameter of 13.36µm and 9.32 µm ( Figure 3A,B).Contrarily, as forthe CUR-NE, despite having an average diameter of 7.9 µm, they expose a less sharp curve with a large peak at ≅1 µm typical of the nano-emulsion used in this study, corroborating our hypothesis about its instability during the production phase of the microparticles ( Figure 3C). The homogeneity of MPs was also confirmed by analyzing their size with a Malvern Mastersizer. The obtained results showed that CUR-oil and CUR-water microparticles have a uniform distribution with a mean diameter of 13.36 µm and 9.32 µm ( Figure 3A,B).Contrarily, as for the CUR-NE, despite having an average diameter of 7.9 µm, they expose a less sharp curve with a large peak at 1 µm typical of the nano-emulsion used in this study, corroborating our hypothesis about its instability during the production phase of the microparticles ( Figure 3C).

%ɳ of CUR inside MPs
In order to evaluate the amount of curcumin encapsulated into MPs, 10 mg of each formulation was dissolved in a basic solution as reported in the Materials and Methods section. Table  1summarizes the %ɳ for all three preparations; in particular, CUR-oil and CUR-water reached encapsulation efficiencies of 40.01 ± 0.3 and 42.30 ± 3.5, respectively. As forthe CUR-NE, the %ɳ was 31.02 ± 0.5and it was maybe due to the instability of the nano-emulsion during the microparticle production steps (e.g., high speed, pH, time for solvent evaporation) as argued in the previous paragraph.

%ïof CUR inside MPs
In order to evaluate the amount of curcumin encapsulated into MPs, 10 mg of each formulation was dissolved in a basic solution as reported in the Materials and Methods section. Table 1summarizes the %ïfor all three preparations; in particular, CUR-oil and CUR-water reached encapsulation efficiencies of 40.01 ± 0.3 and 42.30 ± 3.5, respectively. As for the CUR-NE, the %ïwas 31.02 ± 0.5 and it was maybe due to the instability of the nano-emulsion during the microparticle production steps (e.g., high speed, pH, time for solvent evaporation) as argued in the previous paragraph.

In Silico Prediction
Release rate studies are important to control, tune and adjust the drug dose during a time-long therapy such as for diabetes [43], chemotherapy [44] and other chronic diseases as neurological [45] or inflammation diseases [46,47].To this end, mathematical modeling can provide valuable information on the mechanism of the release process [25].
For measuring curcumin release kinetics, the experimental release data of the three MP formulations were fitted using a non-linear first-order equation. Data fittings are shown in Figure 4 together with the extracted model parameters a and b. The correlation coefficient R 2 and adjusted R 2 values of CUR-water, CUR-NE and CUR-oil were 0.99, 0.95, 0.98 and 0.98, 0.94 and 0.98, respectively. Therefore, the experimental data were not far from the calculated ones, indicating the suitability of the non-linear first-order kinetic equation model. Release rate studies are important to control, tune and adjust the drug dose during a time-long therapy such as for diabetes [43], chemotherapy [44] and other chronic diseases as neurological [45] or inflammation diseases [46,47].To this end, mathematical modeling can provide valuable information on the mechanism of the release process [25].
For measuring curcumin release kinetics, the experimental release data of the three MP formulations were fitted using a non-linear first-order equation. Data fittings are shown in Figure 4 together with the extracted model parameters and . The correlation coefficient R 2 and adjusted R 2 values of CUR-water, CUR-NE and CUR-oil were 0.99, 0.95, 0.98 and 0.98, 0.94 and 0.98, respectively. Therefore, the experimental data were not far from the calculated ones, indicating the suitability of the non-linear first-order kinetic equation model. Once acquired the mathematical equations which describe the dependence of release as a function of time, a quantitative combination of non-linear first-order models (Equation (2)) describing the three CUR-MP kinetics was used for simulating further releases of the encapsulated molecule. Some possible combinations are shown in Figure 5. Mathematical modeling of curcumin release kinetics has been used to design a number of controlled MP-based drug delivery systems in order to release a specific concentration of curcumin in the target tissues with the desired timing. This tool is very useful to predict releases, avoiding the necessity of realizing experiments. Once acquired the mathematical equations which describe the dependence of release as a function of time, a quantitative combination of non-linear first-order models (Equation (2)) describing the three CUR-MP kinetics was used for simulating further releases of the encapsulated molecule. Some possible combinations are shown in Figure 5. Mathematical modeling of curcumin release kinetics has been used to design a number of controlled MP-based drug delivery systems in order to release a specific concentration of curcumin in the target tissues with the desired timing. This tool is very useful to predict releases, avoiding the necessity of realizing experiments. function of time, a quantitative combination of non-linear first-order models (Equation (2)) describing the three CUR-MP kinetics was used for simulating further releases of the encapsulated molecule. Some possible combinations are shown in Figure 5. Mathematical modeling of curcumin release kinetics has been used to design a number of controlled MP-based drug delivery systems in order to release a specific concentration of curcumin in the target tissues with the desired timing. This tool is very useful to predict releases, avoiding the necessity of realizing experiments.

In Vitro CUR Release
With the aim to understand the reliability and accuracy of the in silico studies, experimental in vitro release profiles of curcumin were performed. Particularly, four different combinations used for theoretical studies were analyzed: (i) 50% of CUR-oil plus 50% CUR-water, (ii) 50% of CUR-oil plus 50% CUR-NE, (iii) 50% of CUR-water-50% CUR-NE and (iv) 33% of all three formulations. Moreover, the release of curcumin from the single formulations was also evaluated as control. Interestingly, as shown in Figure 6A,B, a perfect correlation between hypothetical and experimental results was obtained, confirming that mathematical models can be a great support to reduce the number of experiments and to analyze different conditions and strategies.
The amounts of curcumin released from each case coming from in silico and in vitro experiments after 72h, were summarized in Table 2. As we can see, the CUR-water and CUR-NE formulations can guarantee a fast release; after 72 h, all curcumin is released, but they are able to release only 285 ± 2.95 and 45.35 ± 4.21 µg of curcumin, respectively. In addition, for these formulations, a percentage of release more than 100% is reported; numbers over 100%, but still close to this value, are potentially due to random and systematic errors coming from the evaluation methods. Intermediary situations can be achieved by mixing them with CUR-oil configuration, indeed, using the 50% of CUR-oil MPs with 50% of CUR-water or CUR-NE, 60% of the drug can be released after 72 h with an amount of >1 mg. This situation was maintained even by using 33% of the three formulations together. The same results were confirmed by the in silico data ( Table 2). The obtained outcomes underline how, thanks to our approach, we are able to finely regulate the quantity of the drug to be released, generating a powerful platform for the drug delivery field.
for theoretical studies were analyzed: (i) 50% of CUR-oil plus 50% CUR-water, (ii) 50% of CUR-oil plus 50% CUR-NE, (iii) 50% of CUR-water-50% CUR-NE and (iv) 33% of all three formulations. Moreover, the release of curcumin from the single formulations was also evaluated as control. Interestingly, as shown in Figure 6A,B, a perfect correlation between hypothetical and experimental results was obtained, confirming that mathematical models can be a great support to reduce the number of experiments and to analyze different conditions and strategies. The amounts of curcumin released from each case coming from in silico and in vitro experiments after 72h, were summarized in Table 2. As we can see, the CUR-water and CUR-NE formulations can guarantee a fast release; after 72 h, all curcumin is released, but they are able to release only 285 ± 2.95 and 45.35 ± 4.21µg of curcumin, respectively. In addition, for these formulations, a percentage of release more than 100% is reported; numbers over 100%, but still close to this value, are potentially due to random and systematic errors coming from the evaluation methods. Intermediary situations can be achieved by mixing them with CUR-oil configuration, indeed, using the 50% of CUR-oil MPs with 50% of CUR-water or CUR-NE, 60% of the drug can be

Conclusions
This project was undertaken to design curcumin-loaded PLGA MP-based formulations with tunable kinetic release by using a combination of different PLGA MPs. We demonstrated that the rate of curcumin released PLGA MPs can be controlled by playing on the encapsulation strategy which has an effect on the MP microstructure and on the drug distribution within the MPs. Moreover, thanks to the use of a non-linear first-order mathematical model we were able to predict and obtain intermediate situations capable of guaranteeing prolonged or fast drug releases by combining the starting PLGA MPs. The perfect agreement obtained between experimental and in silico methods, confirmed that mathematical modeling could be a valuable support to reduce the number of experiments during the development of novel personalized therapies, especially for the long-term ones.
Our approach can be easily extended to other molecules besides curcumin and it will be useful to release drugs to the target sites with a controlled timing and amount, maximizing the therapeutic efficiency and thus decreasing the side effects.