A stock solution of each extract (5 mg/mL) was prepared by dissolving the indicated mass of extract in methanol. Next, serial dilutions (0.1, 0.25, 0.5, 1, 2.5 and 5 mg/mL) were prepared from the stock solution and used in determination of phenolic content and antioxidant activity.

The total phenolic content of the extracts was determined as reported previously [14] using the Folin-Ciocalteu phenol reagent, with gallic acid as standard. The total phenolic content was expressed as milligram gallic acid equivalence (GAE)/mg of extract.

Flavonoids were determined according to the aluminum chloride method as described previously [14]. Catechin (0.01%) was used as standard. The total flavonoids content of extracts was expressed as milligram Catechin Equivalence (CE) mg extract.

The DPPH scavenging activity was measured using Brand-Williams [15] method as described by [14]. Ascorbic acid (0.01 - 0.1 mg/mL) was used as standard. The DPPH scavenging activity (%) was plotted against extract concentration and the IC₅₀ value (mg/mL) which denotes the concentration of extract required to scavenge 50% of free radicals was calculated.

Experiments were performed according to Re et al. [16] as described by Nobossé et al. [2]. The ABTS radical scavenging activity was calculated as percent inhibition and plotted against sample concentrations. Hence, the IC₅₀ values were calculated. Ascorbic acid was used as reference antioxidant.

The total antioxidant capacity of Moringa leaves was evaluated by the method of [17] as described by [2]. Ascorbic acid was used as standard antioxidant and the total antioxidant capacity was expressed as milligram ascorbic acid equivalence per milligram of extract (mg AAE/mg extract).

The antioxidant potential of M. oleifera L. leaf extracts was equally determined by their ability to reduce iron using the ferrous ion reducing capacity method as described by Yen and Chen [18]. Ascorbic acid was used as reference and the FRAP was expressed as milligram ascorbic acid equivalence per milligram of extract (mg AAE/mg extract).

Sodium alginate solution (0.1%, w/v) was prepared by dispersing the required amount of sodium alginate in 0.1% acetic acid solution. Chitosan solution (0.1%, w/v) was prepared by agitating chitosan in an acetic acid solution (0.1%, v/v) at 23˚C - 27˚C for 12 hours. Polyvinyl alcohol (0.2%, w/v) was used as aqueous emulsifier while soy lecithin (1%) was used as oil emulsifier. Stock solutions of calcium chloride (1%, w/v) and sodium tripolyphosphate (1%, w/v) were used as cross-linking agents. All solutions were prepared using double distilled water and filtered through 0.45 µm pore size nylon filter (Axiva) prior to use.

Moringa-loaded nanocapsules were prepared by emulsion-coacervation which consist in preparing an oil-in-water emulsion, followed by crosslinking using calcium chloride or sodium tripolyphosphate (TPP). Three types of nanocapsules, namely alginate nanocapsules, chitosan nanocapsules and alginate/chitosan nanocapsules were prepared. The oil phase was made of soy lecithin (0.5%, w/v) and the antioxidant-rich fraction of the roasted Moringa leaf extract (5%, w/v) in corn oil [19].

Alginate nanocapsules were prepared according to the method described by [20] with modifications. The aqueous phase was made of a mixture of sodium alginate and polyvinyl alcohol (1:1, v/v). The oil phase was gradually dropped into the aqueous phase under continuous mechanical stirring followed by high shear homogenization (Ultraturrax T-10 B, IKA, Germany) at 10,000 rpm for 5 minutes. Thereafter, CaCl₂ (0% - 0.025%) was added dropwise into the resulting emulsion under stirring and homogenized for an additional 5 minutes. Finally, the preparation was subjected to a probing sonication process in pulsed mode at 40% amplitude for 3 minutes with a sequence of 10 s of sonication and 5 s of rest using a sonicator (Sonics Vibra cell, S & M 220 probe, Newtown, Australia) to form nano-sized droplets.

Chitosan nanocapsules were prepared according to the method described by [11] with modifications. The aqueous phase was made of a mixture of chitosan and polyvinyl alcohol (1:1, v/v). The chitosan oil/water emulsion was made by dispersing the oil phase into the aqueous phase under continuous mechanical stirring followed by high shear homogenization (Ultraturrax T-10 B, IKA, Germany) at 10,000 rpm for 5 minutes. Furthermore, TPP (0% - 0.025%, w/v, final concentration) was added dropwise into the resulting emulsion while stirring and homogenized for an additional 5 minutes. Finally, the preparation was subjected to a probing sonication process as described above for alginate nanocapsules.

For the alginate-chitosan nanocapsules, the aqueous phase was made of an equal volume of polyvinyl alcohol solution and alginate/chitosan mixture in a ratio varying from 4:0 to 1:3, w/w). The primary emulsion made of oil phase in aqueous phase was prepared by high shear homogenization (Ultraturrax T-10 B, IKA) at 10,000 rpm for 5 minutes. Furthermore, CaCl₂ (0.0125%, final concentration) was added dropwise into the resulting emulsion under stirring and the preparation was then subjected to a probing sonication process under similar conditions as with alginate nanocapsules. Blank nanocapsules were prepared similarly but without the addition of the fraction.

Nanocapsules (100 mg) were dispersed into 3 mL of methanol. The mixture was vortexed for 5 min and the resulting suspension was stirred for 1 h on a magnetic stirrer. The suspension was then centrifuged at 3000 rpm for 5 minutes and the supernatant collected. The procedure was repeated and both supernatants were pooled. Antioxidants in supernatant were quantified by spectrophotometry at 260 nm (UV-VIS spectrophotometer, Eppendorf, Germany). The amount of antioxidant-rich fraction entrapped into the nanocapsules (NC) was calculated according to a calibration curve of UV absorption at 260 nm versus concentration of AO-rich fraction. Encapsulation efficiency (EE%) was determined as the proportion of entrapped amount on initial amount of antioxidants (Equation (1)) and percent loading capacity (LC%) was calculated as the amount of entrapped antioxidants to amount of nanocapsules (Equation (2)).

EE ( % ) = Amount of entrapped antioxidants Initial amount of antioxidants × 100 (1)

LC ( % ) = Amount of entrapped antioxidants Amount of NanoCapsules × 100 (2)

The mean particle diameters (Z average) and polydispersity index of nanocapsules were measured using a dynamic light scattering instrument (Zetasizer Nano ZS, Malvern Instruments, Malvern, UK). The samples were diluted 10 times using double distilled water and then loaded in a measurement cell without introducing air bubbles. Samples were equilibrated for 1 min at 25˚C inside the instrument before dynamic light backscattering (detection angle: 173˚) data were collected. Each individual measurement was an average of 13 runs and measurements were performed in triplicate. The average particle size was calculated by the instrument using the Einstein-Stokes equation, assuming that the nanoemulsion droplets were spherical.

The microstructure of the nanoemulsions was observed by transmission electron microscopy (TEM). A drop of nanoemulsion was diluted by a factor of 20 and deposited onto a carbon support film on a gold grid. Excess sample was removed using adsorbent paper. After drying the sample at 25˚C for 15 min, micrographs were taken by a transmission electron microscope (Philips Tecnai 12 Bio Twin, UK) operating at 80 kV. Images were registered on Kodak film SO163 and the negative scanned (Canon, Canoscan 9950F) at a resolution of 600 dpi.

The in Vitro extract release study was carried out using treated dialysis bag diffusion technique. The dialysis membrane was soaked in double distilled water for 12 h prior to use. In Vitro release studies of alginate/chitosan nanoparticles of antioxidant-rich fraction of Moringa were performed in simulated gastric fluid (SGF) made of KCl-HCl buffer (0.1 M, pH 2) and simulated intestinal fluid (SIF) made of phosphate buffer (0.1 M, pH 8). Nanocapsules (100 mg) were dispersed in 4 mL of double distilled water. The dialysis tubing containing the dispersed nanocapsules was placed in a conical flask containing 25 mL of corresponding buffer. The flask was shaken at 200 rpm on a horizontal shaking incubator (New Brunswick scientific, Edison, NJ, USA) maintained at 37˚C for 8 hours. Every 30 minutes, 500 µL of dissolution medium was withdrawn and was replenished with same volume of fresh media to maintain sink conditions. The concentration of AO-rich fraction of Moringa leaf extract released from nanocapsules was assayed at 260 nm (UV/Visible spectrophotometer, Eppendorf, Germany). The standard curve was plotted to represent the intensity of UV absorbance as a function of the concentration of the AO-rich fraction of Moringa.

In drug delivery systems, mathematical modeling plays an important role in elucidating the main drug release mechanisms [21]. The first order model describes whether bioactive release rate is concentration dependent. On the other hand, the Higuchi model assumes that diffusion (Fick law) is the main mechanism underlying bioactive release. Meanwhile, under experimental conditions the release mechanism can deviate from the Fick equation, following an anomalous behavior (non-Fickian). This anomalous behavior can either be swelling or erosion-controlled release and would follow either Korsmeyer-Peppas or Hixson-Crowell model respectively [22]. It is worth noting that the Higuchi model assumes that swelling or dissolution of the matrix is negligible while the Korsmeyer-Peppas model involves simultaneously diffusion and swelling release mechanism. Meanwhile, during the modelling of bioactive release mechanism using Korsmeyer-Peppas model, only the first 60% of release data should be considered to avoid incorporation of erosion effect.

First order kinetics describes the dissolution profile of bioactive from dosage forms in a way that release rate is proportional to the amount of bioactive remaining in its interior. Hence, the amount of bioactive released per unit of time diminishes with release time (Equation (3)) [22].

Q t = Q 0 e − k t (3)

where Q_t is the amount of bioactive released at time t, Q₀ is the initial amount of bioactive in the solution and K is the first order release constant. The linearization of Equation (3) using decimal logarithm results in Equation (4):

log Q t = log Q 0 − K t 2.303 (4)

In this way a graph of the decimal logarithm of the released amount of bioactive versus time will be linear.

Higuchi model describes the release of a bioactive from an insoluble matrix as a function of the square root of time based on Fickian diffusion (Equation (5)).

Q t = K H ∗ t (5)

where, Q_t is the amount of bioactive released at time t and K_H is the Higuchi dissolution constant.

Korsmeyer-Peppas is a semi-empirical model, relating exponentially the bioactive released from swelling-controlled systems to the elapsed time (t) (Equation (6)). It is used to analyze the release of bioactives from polymeric dosage forms when the release mechanism is not well known or when more than one type of release phenomena could be involved [22].

Q t = K t n (6)

where K is a constant incorporating structural and geometric characteristics of the dosage form, Q_t is the fraction of bioactive released at time t and n is the diffusional exponent, indicative of the drug release mechanism. For a Fickian diffusion, n = 0.5 while 0.5 < n < 1 in case of a non-fickian diffusion [21]. Introducing logarithm on both sides, Equation (6) will be written as Equation (7):

log Q t = log K + n log t (7)

Consequently, a graphic of the logarithm of the released fraction of bioactive versus time will result in a straight line.

Hixson-Crowell model is derived from the assumption that the particle regular area is proportional to the cubic root of its volume [22]. This model is given by Equation (8)

W 0 3 − W t 3 = K H C t (8)

where W₀ is the initial amount of bioactive in the dosage form, W_t is the remaining amount of bioactive in dosage form at time t and K_HC is the Hixson constant incorporating the surface-volume relation. A graphic of the cubic root of the unreleased fraction of bioactive versus time (Equation (9)) will be linear if the equilibrium conditions are not reached and if the geometrical shape of the pharmaceutical dosage form diminishes proportionally over time [22]. In this graph, K_HC is given by the slope of the straight line which intersects the ordinate axis at 1.

W t W 0 3 = 1 − K H C t (9)

Then, introducing logarithm results in Equation (10) which is a linear form of Hixson-Crowell equation.

log ( W t W 0 3 ) = − log K N + log t (10)

where K_N is a modified Hixson constant which could be easily obtained as the intercept of the linear graphic of the logarithm of cubic root of the unreleased fraction of bioactive versus time.

The prepared nanocapsules were characterized based on their particle size, polydispersity index (PDI) and mean zeta potential (Table 3).

PDI is a quality parameter that indicates the uniformity of droplet size in nanoemulsions. Nanoparticles having PDI ≤ 0.1 are considered to be highly monodisperse while values comprised between 0.1 - 0.4 and >0.4 indicate respectively moderately and highly polydisperse particles [27]. Following addition of calcium

PDI: polydispersity index, TPP: Sodium tripolyphosphate; ALG: Alginate; CTS: chitosan.

chloride in alginate nanocapsules, the PDI dropped in a concentration-dependent manner from 0.46 in nanocapsules without calcium chloride to 0.21 with 0.025% CaCl₂ solution; indicating that the addition of calcium chloride in alginate nanocapsules improves uniformity of particles in the emulsion.

On the contrary, addition of TPP in chitosan nanocapsules resulted in an increase in both particle size and PDI. High concentration of TPP (0.025%) caused formation of large and non-uniform particles with size and PDI as high as 2249 nm and 0.6 respectively (Table 3). The substitution of chitosan with alginate in alginate-chitosan nanocapsules resulted in a reduction of both particle size and PDI. For instance, the mean particle size dropped gradually from 1209 nm in alginate-chitosan nanocapsules (1:3, w/w) to 394 nm in those made exclusively of alginate (Table 3). Generally, in most of the formulations, the PDI was between 0.2 - 0.4, indicating that the nanocapsules were moderately disperse and had a narrow particle size distribution [8] [27].

Smaller particle sizes are desirable as it has been demonstrated that these are taken up and transported more efficiently across the intestinal cells than larger particles [7]. Moreover, it has been found that nano-sized particles are ideal for oral drug delivery since they are preferentially taken up by both enterocytes and epithelial cells, demonstrating higher intestinal transport compared to larger sized particles [9] [28].

Zeta potential is defined as the electric charge on particle surface which creates an electrical barrier and acts as a “repulsive factor” responsible for the stabilization of nanoparticles [29]. The high magnitude of zeta-potential (>10 mV in absolute) is a sign of high-energy barrier between particles and predicts good colloidal stability of prepared nanoemulsions [30]. The particle charge was inherently negative in both alginate and alginate-chitosan nanocapsules due to the presence of free carboxyl residues in alginate. Alginate nanocapsules, made without addition of chitosan and/or calcium chloride were highly negatively charged with zeta potential values of −43.3 and −41.6 mV respectively. Conjugation of alginate with chitosan greatly increased the particle charge from −41.6 to −5.4 mV as well as did crosslinking with calcium chloride (from −43.3 to −19.1 mV). The loss of negative charge in ALG nanocapsules with conjugation or crosslinking suggests successful conjugation of chitosan and calcium to alginate molecules. It is indeed known that the charge of droplets in nanoemulsions can be altered by adsorption of other charged substances onto their surfaces [31]. Similarly, particle charge in chitosan nanocapsules decreased with cross-linking to sodium tripolyphosphate (TPP) from 50 mV in TPP-free nanocapsules to 13 mV in those containing 0.025% of TPP (Table 3); indicating an effective association between positively charged chitosan molecules and negatively charged TPP molecules.

When evaluating the suitability of biopolymer particles as encapsulation systems for a specific bioactive, the loading capacity (LC) and encapsulation efficiency (EE) are typically the most important factors of interest. The EE (%) of chitosan nanocapsules was lower (61%) compared to that of alginate nanocapsules (78%) (Table 4). Addition of a crosslinking agent in both cases improved the EE by 7% - 9% in alginate nanocapsules and 5% - 14% in chitosan nanocapsules. Combination of alginate and chitosan markedly enhanced EE (84% - 88%) compared to CTS nanocapsules (61% - 70%) but was similar to EE of calcium cross-linked ALG nanocapsules (83% - 85%). Similar observations were made with LC. Highest EE (87.6%) and LC (13%) were obtained for alginate-chitosan ratio of 3:1 (w/w) (Table 4). In addition to its low PDI (0.2) and high zeta-potential (−28.1 mV) (Table 3), this formulation stands out as the most relevant for nutraceutical application.

Our results show that both encapsulation efficiency and loading capacity varied with type of coating polymer as well as concentration of cross-linking agent, in accordance with previous observation that encapsulation efficiency may be influenced by both the polymers and the surfactants used [30]. The loading capacity on its part is highly dependent on the affinity between the biopolymer and bioactive compounds [32], as well as the physicochemical properties such as lipophile/hydrophile ratio of the bioactive [33]. Loading capacity as high as 52% was reported for encapsulation of coumarin 6 in poly (lactic-co-glycolic) nanoparticles by emulsification solvent evaporation method [33]. Loading capacity range of 37% - 100% was reported in alginate/chitosan nanoparticles depending on ALG/CTS ratio and calcium chloride proportion [34]. On the other hand, Tan et al. [35] reported maximum loading capacity of 1.6% for incorporation of β-carotene in liposomes. From a practical point of view, it is important to have the highest loading capacity of a nutraceutical within a delivery system [36].

ALG: Alginate; CTS: chitosan; EE: Encapsulation Efficiency; LC: Loading Capacity; TPP: Sodium tripolyphosphate.

To study the morphology, the shape and size distribution of nanoparticles, transmission electron microscopy (TEM) was used [37]. TEM micrographs of alginate, chitosan and alginate-chitosan nanocapsules (Figure 2) showed that the morphology of nanocapsules depends on the nature of polymer used for encapsulation. Alginate nanocapsules had a spherical shape while both chitosan and alginate-chitosan nanocapsules were ovoid. Both alginate and alginate-chitosan nanocapsules showed black oily inner droplet characteristic of oil-in-water emulsion while chitosan nanocapsules formed water-in-oil emulsion characterized by aqueous whitish inner phase. Moreover, TEM images showed particles size ranging from 100 to 1000 nm, in accordance with the size obtained from dynamic light scattering.

Alginate-chitosan (3:1, w/w) nanocapsules which showed high encapsulation efficiency, loading capacity, zeta potential and low PDI were used to study the release profile of bioactives from nanocapsules in simulated gastric fluid (SGF) and simulated intestinal fluid (SIF). Bioactives release from nanocapsules occurred in a controlled manner and was characterized by three phases (Figure 3). An initial low-release rate phase lasting about 30 minutes with about 5% of bioactives released. This was followed by an exponential-release phase lasting up to 400 minutes with 60% or 70% of bioactives released in SGF and SIF respectively. Beyond this point, was a stable phase where no additional release of bioactives occurred.

In the second phase, the release profile was significantly (p < 0.05) influenced by pH with more bioactives being released in alkaline (SIF) than in acidic (SGF) medium. This observation is in line with the fact that sodium alginate used as coating material shrinks in acidic medium but degrades in alkaline medium [38], thus lowering the diffusion of encapsulated active substance from inner to outer surface in SGF and favoring greater release in the intestine (SIF). The absence of a burst effect at the initial phase is evidence of the absence of uncoated bioactive on the surface of nanocapsules and also of earlier degradation of the coating polymer; given that the burst effect is attributed either to desorption of the drug located on the surface of nanocapsules [39], or to the degradation of the thin polymeric membrane [40]. Similar observation of the absence of initial burst phase was reported during the release of rutin from ferritin-rutin-EGCG nanoparticles [41]. The absence of a burst effect disagrees with previous observations that the release profile from nanocapsules obtained by emulsion-coacervation is biphasic and characterized by a fast initial (burst) phase followed by a slower second phase corresponding to the diffusion of bioactive molecules from the inner compartment to the outer phase [40] [42]. This difference can be explained by the nature of encapsulate and coating polymers, the method of nanocapsules preparation and interactions between encapsulate and coating polymer [30]. The incomplete release observed in this study (Figure 3) may be attributed either to the retention capacity of the targeted antioxidants on polymer matrix [40] or to the saturation of diffusion media [43].

The mathematical modeling of bioactive release profile from the nanocapsules is described by Figure 4 using first order (a), Higuchi (b), Korsmeyer-Peppas (c) and Hixson-Crowell (d) models. The validity of these models was evaluated according to goodness of fit based on correlation coefficients. High correlation coefficients (R² > 0.9) indicate strong fit of antioxidant release profile to the tested models. The release profile was most adequately fitted to Korsmeyer-Peppas model (R² = 0.94 - 0.97). According to values of correlation coefficients, all tested models were valid and could be adequately used to represent the release of antioxidants from Moringa-loaded nanocapsules. Based on these models, it can be affirmed that the release of antioxidants from Moringa-loaded nanocapsules is concentration-dependent (First order model) and involves both diffusion (Higuchi model), swelling (korsmeyer-Peppas model) and erosion (Hixson-Crowell model) mechanisms.