Preparation of iron-loaded water-in-oil-in-water (W1/O/W2) double emulsions: optimization using response surface methodology

Abstract Iron deficiency anemia is a major public health problem caused by low iron intake or bioavailability. To overcome the problem of anemia, iron fortification of a variety of food products is being tested worldwide. Ferric sodium EDTA is a desirable iron source for food fortification due to its high bioavailability and ability to prevent binding of iron to phytates during absorption. This study investigated the application of double emulsions for encapsulation and controlled delivery of ferric sodium EDTA stabilized by the lipophilic Span®80 and hydrophilic complex of sodium alginate and sodium caseinate emulsifiers. The influence of factors such as the ratio of internal water to the oil phase, concentrations of alginate or Span®80 and the ratio of water-in-oil emulsion to the external water on internal (din ) and external emulsion droplet sizes (dout ) was studied by response surface methodology. The feasible range for selection of factors for achieving target values of din and dout was proposed by the response optimizer confirmed by the results of experiments. The results indicate that double emulsion systems may be useful for iron fortification of food products. Graphical Abstract


Introduction
Iron plays an indispensable role in myoglobin and hemoglobin formation and transport of oxygen in the human body. Low intake and bioavailability of this mineral are the main causes of iron deficiency anemia in industrialized countries. [1] In addition, in developing countries of Africa, South, and Southeast Asia, such as Maldives, India, and Myanmar, [2] the consumption of polyphenol-rich vegetables such as beans, [3] peppermint, and turmeric [4] with iron-chelating properties, and beverages such as tea, [5] can play a major role in the inhibition of iron absorption. To increase the intake of this micronutrient and overcome the problem of iron deficiency, fortification or enrichment of foods through addition of various iron sources is being considered worldwide. [6,7] The iron sources used for food fortification are classified into (i) water-soluble iron compounds such as ferrous sulfate, ferrous gluconate, ferric ammonium citrate, and ferrous ammonium sulfate. (ii) Iron compounds with poor water solubility that are soluble in dilute acids, such as ferrous fumarate, and ferrous succinate. (iii) Iron compounds that are insoluble in water and poorly soluble in dilute acids such as ferric pyrophosphate and ferric orthophosphate. (iv) Other iron sources such as ferric sodium EDTA. Iron compounds are selected for fortification based on their bioavailability, cost, and organoleptic properties. While water-soluble iron sources show high bioavailability, they contribute to an undesired change in color or taste of the food product. Conversely, the water-insoluble iron types that are organoleptically inert, do not influence the color and taste of the food product, but are less bioavailable. [8] Ferric sodium EDTA is an iron source that is 2-3 times more bioavailable than other iron sources such as ferrous sulfate since it prevents binding of iron to phytates [9] and can be efficiently incorporated into hemoglobin. Through consumption of foods fortified with this iron compound an additional iron uptake of 2.2 mg/day for children and 4.8 mg/day for male adults can be achieved. [10] Ferric sodium EDTA had fewer side-effects such as gastrointestinal problems and produced no metallic taste.
The ultimate aim of fortification requires encapsulation of this micronutrient to protect it from other ingredients present in the food matrix and achieve its controlled delivery. Among techniques used for encapsulation of micronutrients, emulsion systems can be potentially used as an encapsulation technique for confinement of a micronutrient in an aqueous or organic phase stabilized by biopolymeric or synthetic emulsifiers. Ferric sodium EDTA can be encapsulated in a water-in-oil (W/O) emulsion stabilized by a lipophilic emulsifier with a low hydrophilic lipophilic balance (HLB) value, conventionally in the range of 1-6 and the prepared W/O emulsion can be incorporated into a second external aqueous phase for the preparation of waterin-oil-in-water (W/O/W) emulsions. The oil phase can be stabilized by a hydrophilic emulsifier with a higher HLB value, commonly in the range of 10-12. [11] Various hydrophilic emulsifiers and biopolymers have been used for encapsulation of iron in emulsion systems. [11] Several reports contributed to our enhanced understanding of the application of mono or polysaccharides such as glucose, or alginate for the preparation of double emulsions loaded with ferrous sulfate. [11][12][13] Proteins such as whey protein isolate or sodium caseinate have been also utilized for fabrication of double emulsions for encapsulation of ferrous sulfate and ferric chloride. [14,15] The application of other plant-based emulsifiers such as Quillaja saponin was also reported for delivery of ferrous sulfate [16,17] and enhanced delivery with higher encapsulation efficiency could be achieved when multilayer of Quillaja saponin was used together with chitosan. [17] A complex of polysaccharides and proteins can provide higher stability to the emulsions and complexation of various proteins such as whey protein isolate or sodium caseinate with polysaccharides such as carboxymethyl cellulose (CMC), chitosan, gum arabic, or j-Carrageenan is reported in the literature. [18] Among the reported proteinpolysaccharide complexes, the complex of sodium alginate with sodium caseinate showed potential for achieving higher emulsion stability. Hydrogen bonds formed between carboxyl or hydroxyl groups of sodium alginate with amide groups of sodium caseinate and the formed viscous solution provided higher stability to the oil interface when applied as the hydrophilic emulsifier. Sodium caseinate with the isoelectric point between 4.6 and 4.8 and molecular weight of 19-25 kDa can self-assemble to form stable micellar structures which makes it a suitable candidate for delivery of micronutrients. [19] Considering the effectiveness of this protein-polysaccharide complex for production of a more stable emulsion with higher stability of the interface, this study aims to investigate the application of sodium alginate-sodium caseinate complex as the hydrophilic emulsifier and SpanV R 80 as the lipophilic emulsifier for the fabrication of double emulsions for achieving enhanced delivery of ferric sodium EDTA with the aim of food fortification. The influence of various parameters such as the ratio of the water to the oil phase (W 1 :O), the concentration of emulsifiers and the ratio of internal emulsion (W 1 /O) to the external aqueous phase (W 2 ) on the droplet size of the fabricated emulsions were investigated and response surface methodology (RSM) was used for the optimization of experimental conditions. Based on the results achieved from the response optimizer, strategies were developed for the fabrication of double emulsions to achieve the desired target values for emulsion droplet size and achieve enhanced bioaccessibility of the micronutrient during gastrointestinal absorption.

Materials
Ethylenediaminetetraacetic acid ferric sodium salt (EDTA Fe (III) sodium salt) and the lipophilic emulsifier, poly-sorbateV R 80 (SpanV R 80), hydrophilic emulsifiers sodium alginate, and sodium caseinate from bovine milk were obtained from Sigma-Aldrich (Canada, Oakville, Ontario). Coldpressed virgin coconut oil was purchased from a local supermarket. Deionized water was used for the preparation of emulsions. For fluorescence imaging and staining of emulsions Nile red dye was used and purchased from Sigma-Aldrich (Canada, Oakville, Ontario).

Preparation of water-in-oil (W 1 /O) emulsions
Deionized water and coconut oil were used for the preparation of water-in-oil (W 1 /O) emulsions. Ferric sodium EDTA (0.5 wt%) was added to the aqueous phase (W 1 ) and W 1 was added to the oil phase (O) at ratios of 0.67 and 1 (w/w). For the stabilization of the internal aqueous phase, SpanV R 80 was added to the aqueous phase at 5, 15, and 20 wt% concentrations. The mixture of the aqueous and organic phases was homogenized using a Silverson L2R Ultra-Turrax homogenizer at 6700 rpm for 20 minutes. During the homogenization emulsions were cooled then stored at 4 C.

Preparation of water-in
emulsions The external aqueous phase was prepared by the addition of hydrophilic emulsifiers, sodium alginate, and sodium caseinate to the deionized water. A 4 wt% sodium caseinate was dissolved in 100 mL deionized water. Various concentrations of sodium alginate, that is, 0.5, 1, 2, and 3 wt% were added to 100 mL of deionized water likewise. The prepared solutions of sodium caseinate and sodium alginate were combined in the volumetric ratio of 1:1 (v/v). The initially prepared W 1 /O emulsions were added to the external aqueous phase (W 2 ) in ratios of 10:90, 20:80, and 30:70. The emulsions were homogenized using a Silverson L2R Ultra-Turrax homogenizer at 4500 rpm for 10 min and cooled during homogenization. The prepared emulsions were stored at 4 C for further analysis.

Brightfield and fluorescence microscopy imaging
The brightfield and fluorescence microscopy images of the produced double emulsions were captured by a fluorescence microscope equipped with an Olympus BX51 digital camera (Japan) with the magnification of 100Â. For staining of the oil phase, Nile red was used as a coloring dye with the excitation spectrum of 543 nm and emission spectrum of 605 nm. 0.02 mg of Nile red was dissolved in 20 mL of ethanol and 100 mL of each prepared emulsion was stained with 10 mL of Nile red solution for fluorescence imaging. Fluorescence images of double emulsions were taken by fluorescence microscopy and further processed using the CellSens software. The emulsion droplet size for the internal aqueous and external organic phase was measured from the brightfield and fluorescence microscopy images utilizing the ImageJ software.

Creaming index (CI) (%) measurement
The creaming behavior of emulsions was monitored for one month and at 25 and 45 C. These two temperatures were selected to study how W 1 /O emulsions behave at the room temperature and if they will be resistant to slightly elevated temperatures upon heating and the duration that they will remain stable. It was required to achieve the fabrication of stable W 1 /O emulsions that can be utilized for fabrication of double emulsions. The height of the creamed layer was measured using the ImageJ software and from Equation (1).
where H s is the height from the bottom of the iron containing serum layer and H t is the height of the total emulsion in each vial. The initial design was a full factorial design with the value of a equal to 2, the number of points and total runs was equal to 31, number of cube points was 16, and number of center points in cubes was 7 with 8 axial points. The number of candidate design points of 31 in the initial design was reduced to 20 points in the optimal design with the selected terms, A, B, C, D, AB, AC, AD, BC, BD, and CD. The effect of the four factors on the two responses was evaluated using the regression equation presented in Equation (2).
where R n is the response (d in or d out ), b 0 , b 1 , b 2 , b 3 , and b 4 are the coefficients of linear effects and b

Experimental design for responses
The experimental values for the responses for each generated condition by the experimental design, were obtained. The images of double emulsions produced at each condition were taken by fluorescence microscopy and the internal and external emulsion droplet sizes were measured using the ImageJ software. The mean value of the measured d in and d out for each tested condition were obtained and used as responses for experimental design.

Influence of spanV R 80 concentration on internal emulsion droplet size (d in )
An increase in the concentration of SpanV R 80 resulted in the presence of more emulsifier for the stabilization of the aqueous interface. With the presence of more emulsifier, smaller internal emulsion droplet size was obtained. Among the tested concentrations of SpanV R 80, larger emulsion droplet sizes with the mean value of 5.82 mm were obtained and when the amount of surfactant was increased to 15 and 20 wt%, smaller emulsion droplet sizes with the mean value of 2.49 mm could be produced. For fabrication of water-in-oil emulsions with the size of water droplets below 1 mm that enable achieving higher bioaccessibility during gastrointestinal absorption, high amount of SpanV R 80 (i.e. 15 and 20 wt%) was required and low stability and larger emulsion droplet sizes were achieved at lower concentrations of this emulsifier. The Span@80 emulsifier shows a tendency to migrate between the phases and to increase the stability of the internal aqueous phase other emulsifiers with higher molecular weight such as polyglycerol polyricinoleate (PGPR) could have been tested for the fabrication of double emulsions. However, due to toxicity the use of PGPR is limited in food formulations. [20] Therefore, for this optimization study Span@80 was selected. The emulsions prepared at SpanV R 80 concentrations of 5 and 15 wt% and their droplet size distribution are presented in Figure 1.

Stability and creaming index of W 1 /O emulsions
The stability of the emulsions was monitored during 24 h at 25 and 45 C and the creaming index was measured according to Equation (1). During the first 6 h, emulsions were very stable with negligible values of creaming index. After 21 h, the emulsions stored at 45 C experienced creaming and phase separation. Emulsions stored at 25 C were highly stable over one-month storage with negligible creaming index value. The comparison of the measured creaming index values is presented in Figure S1 Figure 3. Reduction in the oil droplet size resulted in incorporation of less internal water droplets including the encapsulated iron. Therefore, although increase in alginate content could reduce the oil droplet size, it did not allow achieving high encapsulation efficiency. Simultaneous influence of increase in SpanV R 80 and alginate concentration is shown in Figure 4. When the concentration of SpanV R 80 was increased from 15 to 20 wt% and simultaneously the alginate concentration was increased from 2 to 3 wt%, smaller oil droplets were formed with smaller incorporated internal water droplets. The results are presented in Figure 4 for the W 1 : O ratio of 0.67 and W 1 /O: W 2 ratio of 20:80.

Influence of water to oil ratio (W 1 : O)
An increase in the ratio of water to oil results in the presence of more water in the emulsion system that has to be stabilized with the surfactant, SpanV R 80. When the ratio of W 1 : O was increased from 0.67 to 1, larger internal water droplets were obtained as the amount of emulsifier was not enough to stabilize the aqueous interface. The fabricated emulsion at W 1 : O ratio of 1 is compared to 0.67 presented in Figure 5. When comparing the emulsion in Figure 5 with the emulsion in Figure 4b, we can observe that smaller internal aqueous droplets were produced at lower ratios of water to oil, when less water had to be stabilized by SpanV R 80.

Influence of water-in-Oil (W 1 /O) ratio to W 2
When the ratio of W 1 /O to W 2 increases, more oil is present in the system that has to be stabilized by the hydrophilic emulsifiers, that is, sodium caseinate and sodium alginate. With an increase in this ratio, higher amount of alginate is required to cover the oil interface and stabilize the oil droplets. It can be observed in Figure 5 that the simultaneous increase in the W 1 : O ratio and W 1 /O: W 2 ratio results in the formation of unstable oil droplets with the internal water droplets that have the tendency to escape the oil phase.

Response surface methodology optimization
The results of the measurements calculated for the responses for each experimental design condition are presented in Table S1 for the two responses of d in and d out .

Analysis of response surface design for internal diameter (d in )
Analysis of variance for the internal diameter (d in ) showed the p-value of 0.478 (see Table S2) and as this value is greater than 0.05, it shows that the regression equation predicts the values for the internal diameter with accuracy. The analysis of variance for the internal emulsion droplet size (d in ) is presented in Table S2. Higher F-values and lower pvalues demonstrate the more significant influence of the independent variable on the response. As can be observed from the presented F-values, alginate concentration, and the ratio of W 1 /O: W 2 have the most significant influence on the internal emulsion droplet size.
The regressed coefficients for the internal diameter (d in ) and the Equation (2) are presented in Table 1.
The value of R 2 of 97.35% shows the high accuracy in the prediction of the model. The value of 85.78% for the R 2 (pred) shows that the model can accurately perform new predictions. Larger number of points might be required to increase the predictive ability of the model. Moreover, the low value of S is an indicative that the model can also predict the values for d in .
The obtained response surface plots for the internal diameter (d in ) are presented in Figure S2 Table S3) and as this value is greater than 0.05, it shows that the regression equation predicts the values for the external diameter with accuracy. The analysis of variance for the emulsion outlet diameter d out is presented in Table S3.
The high F-values and low p-values for alginate concentration and W 1 /O: W 2 highlight the major influence of these independent variables on the external emulsion droplet size. The combined effect of these two factors in two-way interaction analysis also shows the high influence of these independent variables on the response as is demonstrated by the high F-value.
The regressed coefficients for the external diameter (d out ) and Equation (2) are presented in Table 2.
The value of R 2 of 99.62% shows the high accuracy in the prediction of the model for the external diameter. The value of 98.04% for the R 2 (pred) shows that the model can perform new predictions with high accuracy. Moreover, the low value of S is an indicative that the model can predict the values for d out with accuracy.
The obtained response surface plots for the external diameter (d out ) are presented in Figure S3.

Influence of various factors on responses
The Pareto charts of the standardized effects are shown and compared in Figure S4 for both the internal and external diameter. Figure S4a shows that W 1 /O: W 2 ratio, sodium alginate concentration, and W 1 : O ratio have the most significant influence on the internal droplet diameter. The concentration of SpanV R 80 has a major influence on the internal emulsion droplet size. However, compared to the other factors it had only a minor influence on d in in this limited concentration range tested. In contrast, the concentration of sodium alginate and the ratio of W 1 /O: W 2 and their combined effect (AB) significantly influenced the external emulsion droplet size (d out ). With an increase in the concentration of the hydrophilic emulsifier, more emulsifier was available to cover the oil interface and contribute to its stabilization. Therefore, smaller oil droplets can be obtained at higher alginate concentration. With an increase in the W 1 /O: W 2 ratio, more oil will be present in the system that is required to be stabilized by the hydrophilic emulsifiers and as this ratio increases, larger external oil droplets are expected as predicted and confirmed by the response surface plot of the   external diameter in Figure S3. The response surface plots in Figure S2 present the influence of the SpanV R 80 concentration and the ratio of W 1 : O on the size of the internal emulsion droplets (d in ). With an increase in the concentration of SpanV R 80 and the presence of more emulsifier, smaller internal water droplets can be obtained and as the ratio of W 1 to oil increases, more surfactant is required to stabilize the aqueous phase. Therefore, the size of the internal diameter increases with an increase in this ratio. An increase in the concentration of the hydrophilic emulsifier, sodium alginate, results in the formation of smaller oil droplets that can carry smaller internal droplets and can simultaneously influence both the size of the oil and internal water droplets.

The optimized response surface solution
To obtain double emulsions with high encapsulation efficiency, with enhanced bioaccessibility of the encapsulated micronutrient, smaller inner droplet diameters (in nm size range) are generally preferred and oil emulsion droplets in the size range below 10 mm are desired. To optimize various factors for the production of double emulsions with an internal diameter between 0.9 and 1.2 mm and the external diameter size between 6.5-8 mm, overlaid contour plots were generated, which show the feasible region for the fabrication of double emulsions considering the binary effect of various factors. The obtained overlaid contour plots from the response surface design are shown in Figure 6.
To optimize the response of the model prediction, target values of 1 and 7 mm were considered for d in and d out respectively in the response optimizer with lower and upper limits of 0.95 and 1.05 mm for the internal diameter and 6.90 and 7.10 for the external diameter. The solution of the response optimizer proposed the conditions for various factors with the high composite desirability of 1 as is presented in Table 3.
The standard error of the fit for the internal diameter was 0.051 with 95% confidence interval of (0.885, 1.115) and for the external diameter the standard error of 0.137 with 95% confidence interval of (6.080, 7.920). The proposed condition by the response optimizer is in the feasible region (white region) presented in Figure 6 for the selection of factors for experimental design.
A double emulsion is prepared experimentally with the proposed condition by the response optimizer and the emulsion droplet size is measured for the internal aqueous phase and external oil phase. The images of the emulsion fabricated at the proposed condition obtained by fluorescence microscopy and brightfield imaging are presented in Figure 7.
The average values of measured internal emulsion droplet size (d in ) was 0.85 mm and the external emulsion droplet size was 7.53 mm. The obtained emulsion droplet sizes from experiments are close to the desired target values of 1 and 7 mm for d in and d out , respectively, which implies that the response optimizer can propose the desired condition with accuracy.
The results of this RSM study as is neatly demonstrated in Figures S4 and 6, show the combined influence of the concentration of the hydrophilic emulsifier, sodium alginate, and the ratio of the phases, mainly the ratio of W 1 : O to the second external aqueous phase (W 2 ) on emulsion droplet size. The effect of these two factors is also demonstrated in Table S3 by the high F-value of the 2-way interaction of these two factors. Reports in the literature refer to the application of RSM for selection of emulsifier concentrations for optimizing the double emulsion droplet size. [21,22] However, there are few studies that report on the combined influence of phase ratios and emulsifier concentration on the internal (d in ) and external (d out ) emulsion droplet sizes, which can greatly influence the bioavailability of the encapsulated molecule. In a s imilar study on the delivery of polyphenols using W 1 /O/ W 2 double emulsion systems [23] it was also shown that the content of W 1 /O emulsion in the second aqueous phase (W 2 ) and emulsifier concentration greatly influenced the emulsion droplet size. However, the authors did not study the effect of surfactant content in W 1 /O and W 1 /O/W 2 emulsions on the droplet size of the internal aqueous phase (d in ) and external oil phase (d out ) separately.

Conclusions
Water-in-oil-in-water (W 1 /O/W 2 ) double emulsion systems show potential for encapsulation of ferric sodium EDTA for fortification of food products. Fabrication of W 1 /O emulsions with coconut oil with high melting point allowed the production of emulsions that were stable during one-month storage at room temperature and the emulsions were stable at elevated temperatures up to 45 C when stored for 21 h. The stable W 1 /O emulsions at room temperature enabled the fabrication of W 1 /O/W 2 double emulsions. Although the synthetic lipophilic emulsifier, polysorbate80 (SpanV R 80) could stabilize the aqueous interface and protect the encapsulated ferric sodium EDTA in the internal aqueous phase, high concentrations of this surfactant (15-20 wt%) were required to achieve high stability of the internal aqueous droplet containing iron and simultaneously achieving small internal emulsion droplet size to enable enhancing the bioaccessibility of iron during gastrointestinal absorption. The results obtained from experimental design and RSM show that the concentration of alginate as the hydrophilic emulsifier and the selection of the ratio of W 1 /O: W 2 , which specifies the amount of oil present in the final double emulsion that has to be stabilized by the hydrophilic emulsifier, have major impact on the external emulsion droplet size and can reduce the oil droplet sizes. However, increasing the amount of alginate to 3 wt% resulted in significant decrease in the size of the oil droplets which then could not hold the internal aqueous phase containing iron. The concentration of the lipophilic emulsifier SpanV R 80 has major influence on the size of the internal aqueous phase. Nevertheless, in the tested range of 15-20 wt% this emulsifier did not have a major impact on emulsion droplet size and was not selected as the most important factor influencing d in in the response surface design. The results of the response surface design show the influence of phase ratios and emulsifier concentrations and their synergistic combined effect on internal and external emulsion droplet sizes that is scarcely investigated and reported in the optimization studies in the literature for double emulsion fabrication. To enhance the bioaccessibility of iron during gastrointestinal absorption, smaller internal diameter size (nm range) is desired. The predicted feasible range for the selection of various factors proposed by the response optimizer can be used for the preparation of double emulsion to achieve the desired droplet sizes. The complex of sodium caseinate with sodium alginate could provide high stability to the emulsion and can be used as a promising polysaccharide-protein complex for encapsulation of iron.