Modeling the In Vitro Hydrolysis of Nano-Emulsified Rapeseed Oil Digested with Intestinal Lipases of the Rainbow Trout Oncorhynchus mykiss Through Response Surface Methodology: Effect of the Emulsifier

Abstract

Lipolysis is an interfacial reaction. Lecithins are natural emulsifiers containing a mixture of phospholipids (PL). Lecithin composition can be modified via enzymatic hydrolysis of PLs to produce lysophospholipids (LPL). The quantities of PL and LPL and the PL/LPL ratio are related to the emulsifying properties and interfacial activity of digestive lipases. This study aims to: (i) produce oil-in-water nanoemulsions of rapeseed oil (RSO) with soybean lecithin (SBL) and hydrolyzed lecithin (HL) at different concentrations and homogenization pressures and measure the mean droplet diameter (MDD) and polydispersity index (PdI) by dynamic light scattering; (ii) hydrolyze the emulsions in vitro with intestinal extracts of rainbow trout and estimate the degree of hydrolysis of lipids (DH) by the pH-stat method; and (iii) model the results on MDD, PdI, and DH through the response surface methodology (RSM). When HL was used as an emulsifier, DH, MDD, and PdI were fitted to polynomial quadratic, two-factor interaction, and linear models, respectively. MDD, PdI, and DH were fitted to polynomial quadratic SBL models. The optimal conditions were emulsifier concentrations of 0.45% and 0.76% w/w and homogenization pressures of 10,790 and 10,781 psi for HL and SBL, respectively. Under these conditions, DH = 34.9% and 33.08%, MDD = 241.9 and 543.6 nm, and PdI = 0.29 and 0.52 for HL and SBL, respectively.

1. Introduction

The rapid increase in the global production of farmed fish during the last two decades has imposed a high demand for fish meal and fish oil to produce aquafeeds, leading to high prices of these ingredients [1]. Simultaneously, the search for alternative ingredients has been intensified. The inclusion rates of marine protein and oil sources in Norwegian Atlantic salmon changed from approximately 33.5% and 31.1% in 2000 to 12.1% and 10.3% in 2020, respectively [2]. Presently, the inclusion rates of the same ingredients in feeds for rainbow trout cultured in Norway are 13.4% and 10.8% for marine proteins and lipids, respectively [3]. The replacement of marine ingredients with products from terrestrial food systems also affects the feedstuffs of species other than salmonid fish [4]. Rapeseed oil (RSO) is one of the most used terrestrial oil sources for aquafeeds, representing approximately 18% (w/w) in the formulation of commercial diets for Norwegian salmon and trout [2,3]. In addition, as a result of a meta-analysis work, Qian et al. [5] recommended RSO as the most suitable lipid source to replace fish oil in the diets of Atlantic salmon. In general, fish diets containing RSO are well ingested and digested [6]. However, digestibility coefficients do not provide specific information about the efficiency of lipid hydrolysis. Lipid digestion hydrolysis requires lipid emulsification. The process of emulsification implies an input of mechanical energy to disrupt the surface of lipid masses so that microdroplets are produced and the presence of an emulsifier capable of stabilizing those droplets [7]. Physiological emulsification can occur through gut peristalsis, duodenal constriction of the gastric chyme [8], and the presence of dietary and endogenous emulsifiers. The emulsifying agents are amphipathic molecules that cover the lipid droplets surface. Emulsifiers can interact with the core of the droplet through its hydrophobic side and with the aqueous medium of the gut lumen through its hydrophilic side [7,9]. Emulsifiers can be exogenous, such as proteins and phospholipids from the diet (or included as dietary additives), or endogenous, mainly in the form of bile salts [9,10]. Emulsifiers exert a key role in the hydrolysis of triglycerides by fish lipases, which can only work at the interface between the hydrophobic core of lipid droplets and the aqueous medium in the gut [11]. Based on this rationale, several studies have explored the effects of dietary phospholipids (PL) and lysophospholipids (LPL) on growth performance and lipid digestibility in cultured fish. In some of those articles, the inclusion rate of PL was so large that the experimental design can be considered to test the effects of lipid replacement [12,13,14]. Other authors included PL or LPL at low rates typical of dietary additives, in the range 0.03–2% [15,16,17,18,19,20]. Some of these last experiments indicate that PL and LPL can increase the apparent digestibility of lipids in vivo [16,18,19] and intestinal lipase activity [17,18] in fish. Other experiments have not shown such effects on lipid digestibility [20] or lipase activity [15]. However, the emulsifying capacities of PL and LPL are different. For example, Cabezas et al. [21] reported that hydrolyzed sunflower lecithin (enriched in LPL) has a better capacity to emulsify sunflower oil than non-hydrolyzed sunflower lecithin, leading to better emulsion stability and smaller lipid droplets. Although the emulsifying properties of PL and LPL are probably involved in the reported improvement of lipid digestibility, this point cannot be investigated in detail from digestibility coefficients or lipase activities. Alternatively, the in vitro modeling of lipid hydrolysis is a complementary approach to understand the physical-chemical details of the phenomenon. Thus, it is possible to manipulate key variables, such as the size of lipid droplets, emulsifier type, or lipid type [22]. This experimental approach has been extensively applied to simulate the digestion of emulsified lipids in the human gut [23,24], but much more infrequently in fish [25]. In addition, the experimental design of digestion modeling can be optimized using the response surface methodology (RSM). This methodology makes it possible to simultaneously study more than one factor with a minimum of experimental runs [26]. This analytical tool has been used to model macronutrient digestion in humans [27] and fish [28]. Considering all these reasons, the present study aimed to test if hydrolyzed soybean lecithin (HL) is more efficient than non-hydrolyzed soybean lecithin (SBL) at producing small lipid droplets and promoting in vitro lipolysis of rapeseed oil digested with intestinal extracts of rainbow trout. Another goal is to model the results using the response surface methodology.

2. Materials and Methods

2.1. Chemicals

Sodium taurocholate monohydrate (96%) was purchased from Alfa Aesar (Thermo Fisher (Kandel) GmbH, Kandel, Germany). Calcium chloride anhydrous, disodium hydrogenphosphate 12-hydrate (99%), sodium dihydrogenphosphate 2-hydrate (99%), and sodium azide (99%) were supplied by PanReac Química SLU (Barcelona, Spain). Sodium chloride was purchased from Merck KGaA (Darmstadt, Germany). Volumetric solutions of sodium hydroxide 0.1 M and hydrochloric acid 0.1 M were provided by Fisher Scientific (Leics, Loughborough, UK).

2.2. Oil

Rapeseed oil (RSO) (100% pure, supplied by HiPP, Gmunden, Austria) was used in this study. It has a triglyceride content of 98%, with triolein (OOO) and dioleyl-linoleyl-glycerol (OOL) as the two primary molecular species. The fatty acid profile is as follows: palmitic (4.5%), stearic (2.0%), oleic (62.5%), linoleic (20.0%), alpha-linolenic (8.0%), and arachidonic (0.6%) acids. The oil density ranges from 0.910 to 0.920 g/cm³. The weighted average molecular weight of the triglyceride was 879.62 g/mol based on the fatty acid composition.

2.3. Emulsifiers

Two emulsifiers were tested: (i) soybean lecithin (SBL) and (ii) hydrolyzed lecithin (HL). Soybean lecithin (L-alpha-lecithin, granular, from soybean oil, Lot A0428004, supplied by ACROS ORGANIC, Geel, Belgium) contained >97% phospholipids (as acetone insoluble) and <1.1% water. SBL is a complex natural surfactant comprising a mixture of phospholipids, glycolipids, triglycerides, sterols, and fatty acids. The quantity and ratio of phospholipids—specifically phosphatidylcholine (PC), phosphatidylethanolamine (PE), phosphatidylinositol (PI), and phosphatidic acid (PA)—determine the affinity of the dispersed phase and its emulsifying properties [29]. The phospholipid profile of the SBL used in this study was as follows: PC: 36.92%, PE: 40.28%, and PI: 22.80%, as determined by 31P-NMR analysis [30].

Hydrolyzed lecithin was a modified soybean lecithin (supplied by Nutri-Ad, ADISSEO, Alcalá de Gurrea, Huesca, Spain) with a PL to LPL ratio of 27%:37%. The specific composition was as follows: PC: 5.53%, PE: 11.60%, PI: 9.83%, LPC: 22.70%, LPE: 9.40%, and LPI: 4.46%, as determined by 31P-NMR analysis [30].

2.4. Preparation of the Emulsions

Different oil-in-water (O/W) emulsions were prepared. For unifactorial experiments, the emulsions contained 9% (w/w) RSO and 1% (w/w) of each emulsifier (SBL and HL) in phosphate buffer (5 mM, pH 8.0, NaCl 150 mM, sodium azide 0.02%). The mixture was pre-homogenized by sonication (Sonifier Branson SFX 150, Branson Ultrasonics, Fisher Scientific S.L., Alcobendas, Madrid, Spain): continuous sonication for 4 min (amplitude 60%) under magnetic stirring at room temperature. The following step consisted of homogenizing the pre-emulsion with a rotor homogenizer (IKA T18 Ultra Turrax, Staufen, Germany) at 15,000 rpm for 2 min. A second sonication cycle was conducted for 10 min under the conditions of the first sonication treatment. Finally, the ultimate homogenization step involved the application of a high-pressure homogenizer (Microfluidizer LM-20, Microfluidics, Middleborough, MA, USA, with a 75 µm Y-Type Interaction Chamber F20Y) to produce nanoemulsions, adjusting the shear force and the number of homogenization cycles according to the experiment.

In the first unifactorial experiment, the shear force was set at 12,000 psi (pounds of force per square inch), and the number of cycles varied from 0 to 10. In the second unifactorial experiment, the shear force was varied from 3000 to 15,000 psi, and the number of cycles was set at 5. For the two response surface methodology (RSM) experiments (one for each emulsifier), emulsions were prepared in a similar way, but the emulsifier concentration ranged between 0.08% and 1.16%, the shear force varied between 2550 and 12,450 psi, and the number of homogenization cycles was set at 3.

2.5. Preparation of Intestinal Crude Extracts

Intestinal crude extracts (ICE) were prepared from fresh rainbow trout (average weight 400 g) provided by a local commercial farm (Piscifactoria Las Fuentes, Huéscar, Granada, Spain). The fish were handled at the facility by the farm’s personnel in accordance with EU animal welfare regulation for commercial farms and in compliance with relevant ethical authorization. According to Serrano et al. [31], adequate enzyme activities are obtained when the anterior intestine of rainbow trout is sampled before the 4th hour of digestion, fish were subjected to a 3 h fasting period. Afterward, animals were sacrificed and stored at 0 °C in batches of five immediately before market dispatch to ensure maximum freshness. The pyloric caeca and anterior intestine were dissected and weighed. The extraction medium consisted of a solution of NaCl (150 mM) and glycerol (20% w/w), according to the following protocol:

(i)

The biological material, obtained from three batches of five individuals each (n = 15), was cut into small pieces with scissors.

(ii)

The chopped tissue was mixed with the extraction solution at a ratio 1:3 (w/v) at 0 °C.

(iii)

The mixture was homogenized using an Ultra-Turrax homogenizer (IKA T18, IKA-Werke GmbH & Co., Staufen, Germany) at 7500 rpm for 3–5 min. To prevent foam formation, the process was conducted in a beaker submerged in an ice-water bath.

(iv)

The homogenate pH was adjusted to 8.0 and maintained in the ice/water bath for 30 min.

(v)

The mixture was centrifuged at 3220× g for 15 m at 4 °C.

(vi)

The upper fat layer in the tubes was discarded after centrifugation, and the aqueous phase was collected and maintained in an ice/water bath. Afterward, the aqueous phase was aliquoted in 2 mL vials and centrifuged at 20,817× g rpm for 10 min at 4 °C. Subsequently, the supernatants were mixed in a beaker under stirring and maintained in an ice/water bath.

(vii)

At last, the ICE was aliquoted in 2 mL vials and stored at −80 °C.

2.6. Protocol to Measure the Lipase Activity of Intestinal Crude Extract

The ICE lipase activity was determined using the pH-Stat method on two emulsified substrates (tributyrin and glyceryl trioleate) [22]. Two oil-in-water stock emulsions with 3% (w/w) substrate and 9% (w/w) Arabic gum as emulsifier. The mixtures were homogenized (Ultra-Turrax at 15,000 rpm for 2 min) and sonicated (4 min) to prepare the substrate emulsions.

The lipase activity assay comprised a simulated intestinal medium (SIM) and substrate emulsion. According to Bucking and Wood [32], the fluid phase of the anterior intestine in juvenile rainbow trout during digestion contained between 4 and 15 mM Ca²⁺ and a minimal ionic strength between 120 and 160 mM. In addition, Romarheim et al. [33] reported a content of bile acids of 25–100 mg per g of solid phase in the proximal intestine of rainbow trout. Considering that the molecular weight of taurocholate is 537 g/mol, and the moisture in the proximal intestine in that species is close to 90% [32], the maximal concentration of bile acids would range between 5 and 21 mM. Based on these figures, the simulated intestinal fluid contained 150 mM NaCl, 10 mM CaCl₂, 5 mM sodium taurocholate, and the triglyceride emulsion diluted to a final concentration of 1.25% of the substrate lipid. The reaction medium was stirred for 5 min. at 30 °C (to reduce the reaction time). The pH of the reaction was adjusted to 8.0 for tributyrin and 9.0 for glyceryl trioleate. Finally, 1 mL of the intestinal crude extract was added to initiate the reaction. The hydrolysis of fatty acids was monitored by continuous titration with NaOH (pH-Stat method) (902 Titrando, Metrohm, Herisau, Switzerland), and the lipase activity per mL of intestinal crude extract was calculated after 10 min of reaction, according to the following equation:

$$\mathrm{Lipase} \mathrm{activity} (\mathrm{U} {\mathrm{mL}}^{-1})={\displaystyle \frac{{\mathrm{Vol}}_{\mathrm{NaOH}}\times {\mathrm{M}}_{\mathrm{NaOH}}\times 1000}{{\mathrm{Vol}}_{\mathrm{ICE}}\times \mathrm{t}}}$$

where Vol_NaOH is the volume (mL) of NaOH used in the titration procedure, M_NaOH is the molarity (M) of NaOH volumetric solution (titrating), Vol_ICE is the volume (mL) of the intestinal crude extract added, and t is the time of reaction (min). The unit of lipase activity (U) is defined as 1 μmol of released fatty acids per minute. As a result, the lipase activity of ICE was equal to 61.65 ± 1.62 U mL⁻¹ with tributyrin as substrate, and 70.77 ± 2.51 U mL⁻¹ with glyceryl trioleate as substrate.

2.7. Measurement Protocols for Response Variables

2.7.1. Mean Droplet Diameter (MDD) and Polydispersity Index (PdI)

The MDD and PdI of emulsions were measured through dynamic light scattering (DLS) (Malvern Zetasizer Nano S, Malvern Instruments Limited, Worcestershire, UK) [34] after diluting emulsion samples to 1:100 in phosphate buffer (5 mM, pH 8.0) to avoid multiple scattering effects. Measurements were performed at 25 °C using a scattering angle of 173° (backscatter detection). The mean droplet size was expressed as the Z-average diameter (hydrodynamic diameter). This value was calculated via the Stokes-Einstein equation using the DLS software (v 7.13, 2018), with the viscosity of the continuous phase (water or buffer, 0.8872 mPa s at 25 °C) and the refractive index (RI) of the rapeseed oil (approx. 1.47) as input parameters. The polydispersity index (PdI) estimates the dispersion in the size of emulsion droplets. The PdI values were derived from the cumulant analysis according to the ISO 22412 standard. This index can range from 0 to 1. When PdI is below 0.3, the emulsion can be defined as monodisperse. Each sample was measured three times.

2.7.2. Determination of the Degree of Hydrolysis (DH) of in Vitro Digested Emulsions Through the pH-Stat Method

The reaction medium of in vitro digestions contained 0.67 mL of the o/w emulsion (9% of RSO), 7.33 mL of distilled water, 1 mL of sodium taurocholate (75 mM), 3 mL of CaCl₂ (50 mM), and 2 mL of NaCl (1.12 M). The mixture was stirred for 5 min at 30 °C in the pH-Stat beaker, and pH was adjusted to 8.0. The reaction started by adding 1 mL of rainbow trout ICE. Final concentrations in the reaction medium were: 15 mL of volume, 5 mM taurocholate, 150 mM NaCl, 10 mM CaCl₂, and 0.4% RSO emulsified. The hydrolysis of fatty acids was monitored by continuous titration at pH 8.0 (pH-Stat method) with NaOH (0.1 M) for 60 min. The degree of hydrolysis (DH) after 60 min was calculated as follows [22]:

$$\mathrm{D}\mathrm{H}\left(\%\right)=100\times {\displaystyle \frac{{\mathrm{Vol}}_{\mathrm{NaOH}}\times {\mathrm{M}}_{\mathrm{NaOH}}\times {\mathrm{M}\mathrm{W}}_{\mathrm{lipid}}}{{\mathrm{W}(\mathrm{g})}_{\mathrm{lipid}}\times 2}}$$

where Vol_NaOH is the volume (L) of NaOH used in the titration process, M_NaOH is the concentration of NaOH in the titrating solution (0.1 M), MW_lipid is the molecular weight of the emulsified lipid (879.62 g mol⁻¹ in our case), and W(g)_lipid is the weight of lipid in the reaction medium (0.06 g). It was assumed that two free fatty acid molecules are produced per molecule of triacylglycerol.

En DH measure includes three blank assays: (i) a blank assay in which the emulsified substrate was substituted by distilled water. This blank corrects the presence of proton-consuming reactions solely due to molecules contained in the ICE, (ii) a blank assay similar to the first one but includes the emulsifiers. This blank corrects the portion of the titration volume due to the hydrolysis of the emulsifier, and (iii) a blank assay in which ICE was substituted by a solution of NaCl 150 mM: zero titration volumes were obtained in this case.

2.8. Experimental Design and Statistical Analysis

2.8.1. Unifactorial Experiments

Preliminary one-factor-at-a-time (OFAT) experiments were conducted—varying one parameter while keeping others constant—to establish the experimental ranges for the factors subsequently optimized via Response Surface Methodology (RSM), following the method described by Senanayake and Shahidi [35]. Two series of preliminary experiments were conducted to study the effects of the homogenization pressure and the number of homogenizing cycles during RSO emulsion production. The first experimental series was designed to study the effect of homogenization pressure on MDD and PdI of RSO (9%) emulsified with SBL or HL (1%) in phosphate buffer (5 mM, pH 8.0) at a homogenization pressure ranging from 3000 to 15,000 psi, and 5 homogenizing cycles.

The second experimental series was conducted to study the effect of the number of homogenizing cycles on the MDD and PdI of RSO (9%) emulsified with SBL or HL (1%) in phosphate buffer (5 mM, pH 8.0) at a homogenization pressure of 12,000 psi, and homogenizing cycles ranging from 0 to 10.

2.8.2. Response Surface Methodology

The response surface methodology (RSM) [26] was applied to investigate and model the effects of the emulsifier concentration and homogenization pressure on three response variables of RSO emulsions: (i) MDD, (ii) PdI of emulsion droplets, and (iii) DH (%) of the emulsion when it is subjected to in vitro digestion with ICEs from rainbow trout. Two independent RSM series were conducted: one with SBL as the emulsifier ingredient and a second with HL as the emulsifier ingredient.

A 2 × 2 factorial composite central design (CCD) with factorial (6 replicas), central (2 replicas), and axial (2 replicas) points was applied to each RSM series. As a result, 22 combinations of levels (runs) and 2 blocks were conducted for each emulsifier (Supplementary Tables S1 and S2, Supplementary Figure S1). Different combinations of levels (runs) were conducted in a random order. The non-codified and codified levels for each experimental factor (−alpha, −1, 0, 1, alpha, with alpha = √2), are shown in

Table 1. Independent variables (factors) and their corresponding coded and uncoded levels were used in the RSM-CCD design for the preparation of 9% (w/w) rapeseed oil-in-water (O/W) emulsions.

		Codified Levels
	Design Points	Axial (2)	Factorial (2)	Center (6)	Factorial (2)	Axial (2)
Independent Variables (Numeric Factors)	Symbol	−alfa	−1	0	+1	+alfa
Homogenization pressure (psi)	A	2550	4000	7500	11,000	12,450
Emulsifier concentration (% w/w)	B	0.08	0.24	0.62	1.00	1.16

. The design was orthogonal and rotable. The homogenization pressure levels were set according to the previous unifactorial experiment (Supplementary Figures S2 and S3). According to these unifactorial experiments, in the homogenization pressure range of 2500–12,500 psi, the particle size was approximately constant with HL, but it peaked at 6000–9000 psi in the case of SBL. Unifactorial experiments also showed that PdI decreased with pressure (in the range 2500–12,000) when using HL, but it increased with pressure in the case of SBL. Thus, since the homogenization pressure interval of 2500–12,500 psi captured the different behaviors of the two emulsifiers (Table 1), it was set as the range pressure to be studied in the CCD design. The emulsifier concentration range was set between 0.08% and 1.16% because this is within the typical interval for the inclusion rate of emulsifiers when used as feed additives in the recent literature [15,16,17,18,19,20]. The number of homogenization cycles was set at 3 since, according to the preliminary unifactorial experiments, more cycles seem to promote coalescence phenomena (an increase in droplet size) with both emulsifiers.

The response variables (MDD, PdI, and DH) were fitted to the general equation of a quadratic polynomial expression:

$$Y={\beta }_0+{\sum }_i{\beta }_i{X}_i+{\sum }_i{\beta }_{ii}{X}_i^2+{\sum }_i{\sum }_j{\beta }_{ij}{X}_i{X}_j+ϵ$$

where β₀, β_i, β_ii, and β_ij are the intercept and regression coefficients, respectively, for linear, quadratic, and interactive terms. X_i and X_j are the codes of the independent variables (factors). The model is subjected to an analysis of variance (ANOVA) to simplify it when some regression coefficients are not significant.

2.8.3. Multi-Response Optimization of RSM Models

The response surface methodology makes it possible to simultaneously optimize several response variables. For our experiment, we used the optimization method based on a desirability function [36] as explained by Pulido and Salazar [37]. The procedure starts by imposing quantitative limits and a given objective or restriction (maximize, minimize, and be within the limits) for each selected response variable: MDD, PdI, and DH. Therefore, after checking the RSM plots, we chose the optimization limits and objectives and restrictions shown in

Table 2. Optimization parameters selected for the three response variables produced with each emulsifier, soybean lecithin, and hydrolyzed lecithin.

Response Variables	Soybean Lecithin (SBL)		Hydrolyzed Lecithin (HL)
	Objective	Limits	Objective	Limits
GH (%)	Maximize	32–34	Maximize	31–34
MDD (nm)	Within the limits	450–550	Within the limits	240–280
PdI (dimensionless)	Within the limits	0.45–0.55	Within the limits	0.30–0.35

. Next, for each response variable, an individual desirability function is defined as a function of the experimental factors (homogenization pressure and emulsifier dose). Individual desirability functions return values in the interval [0, 1] by considering the restrictions imposed on the response variable. The closer the individual desirability function is to 1.0, the closer the response variable is to the optimal value. Next, a global desirability function is defined as a weighed geometric mean of individual desirability functions. The global desirability function is also in the interval [0, 1] (where 0 indicates an inacceptable combination of experimental factors, and 1 indicates an optimal combination of them). During the optimization process, the software calculates the value of the global desirability for random values of the experimental factors (in our case, in the ranges 4000–11,000 psi and 0.24–1.00% for the homogenization pressure and emulsifier dose, respectively). Then, it creates a contour plot for global desirability and finally estimates the combination of experimental factors that maximize the global desirability function.

2.8.4. Validation of the RSM Model Optimization

After defining the combination of experimental factors that maximize the global desirability function (both for SBL and HL) that combination was experimentally validated. The validation protocol followed the experimental sequence already explained for measuring the response variables for any combination of factor values: (i) preparing a 9% RSO emulsion (with the dose of emulsifier and at the homogenization pressure that maximizes the global desirability function), (ii) measuring MDD and PdI of lipid droplets through dynamic light scattering techniques (7 replicates for SBL and 8 replicates for HL), and (iii) measuring the DH of the prepared emulsion by simulating the digestion of the emulsion with an intestinal extract of rainbow trout in vitro and estimating the percentage of released fatty acids following the pH-stat method (Supplementary Tables S3–S5).

Finally, the validation values of MDD, PdI, and DH obtained for the homogenization pressure and emulsifier dose that maximize the global desirability function were compared with the values of MDD, PdI, and DH predicted by their respective RSM equations under the same pressure and emulsifier conditions.

2.8.5. Statistics

The experimental data produced by the two CCD designs (SBL and HL) were fitted to a second-order polynomial by multiple regression analysis to estimate the coefficients β₀, β_i, β_ii, and β_ij (software Stat-Ease Design Expert, v10.0.3). An analysis of variance was conducted for each response variable to obtain the significance of regression coefficients and assess the fitting goodness of the obtained model. The fitting goodness was evaluated using parameters such as the determination coefficient (R²), adjusted R², p-value, and lack-of-fit. In general, the model shows a suitable fitting to experimental data when R² is close to 1.0 and the lack-of-fit test is not significant. The effects of the two studied factors (emulsifier concentration and homogenizing pressure) on the three response variables (MDD, PdI, and DH) were visualized using contour and overlay plots.

3. Results

3.1. Unifactorial Experiments

3.1.1. Effect of Number of Cycles

The first unifactorial experiment evaluates the effect of the number of homogenization cycles on MDD, and PdI of RSO emulsified with SBL, or HL (Supplementary Figure S2).

SBL-based emulsions showed droplets with an MDD between 530 and 850 nm. MDD was below 550 nm when the number of homogenizing cycles was 3 or below 3. From the 4th cycle onwards, it began to increase to 850 nm on the 10th cycle. The second response variables, PdI, ranged between 0.3 and 0.5 with SBL, and it seemed to show a minimum valuer, 0.3, after 3 cycles of homogenization.

Hydrolyzed lecithin-based emulsions contained droplets with an MDD between 240 and 260 nm from the 1st cycle onwards. MDD was maintained below 245 nm for cycles 1–3 and showed a slow increase from the 4th cycle. In addition, the PdI index varied between 0.22 and 0.26, but it was in a narrow interval (0.225–0.235) between the 2nd and 9th cycles of homogenization.

3.1.2. Effect of the Pressure

The second unifactorial evaluation evaluated the effect of homogenization pressure on MDD and PdI of RSO emulsified with SBL or hydrolyzed lecithin (Supplementary Figure S3).

SBL-based emulsions showed droplets with a MDD between 690 and 1150 nm. MDD appeared to peak at 1100–1200 nm when the homogenization pressure was between 6000 and 9000 psi. With SBL, the PdI index ranged between 0.25 and 0.5, and it increased from 0.25 to nearly 0.5 when the pressure varied from 3000 to 9000 psi, and it leveled off from 9000 psi onwards.

The MDD of droplets in emulsions based on hydrolyzed lecithin was quite constant at 245–275 nm for the entire range of pressures (3000–15,000 psi). The PdI index varied between 0.23 and 0.28. When the pressure increased from 3000 to 9000 psi, PdI slowly decreased and then leveled off and maintained between 0.23 and 0.25.

3.2. CCD Experiments and Response Surface Models

3.2.1. Response Surface Models for MDD, PdI, and DH When SBL Was Used as an Emulsifier

ANOVAs for the RSM models obtained for MDD, PdI and DH in SBL-based emulsions are shown in Supplementary Tables S6–S9.

The data for the MDD variable were normalized using the Box–Cox transformation: MDD′ = (MDD + k)^λ, with λ = −0.77 and k = 0 [38]. The quadratic RSM model was highly significant for MDD (p < 0.001) (Supplementary Tables S6 and S9), with an adjusted R² of 0.592. The lack of fit was significant (p < 0.001), indicating that the model could be improved using cubic terms. However, CCD designs cannot provide single solutions for the coefficients of cubic terms [26]; therefore, we selected the quadratic model with the following equation:

$$\begin{array}{cc}{MDD}^{\prime }(\mathit{diameter}\_SBL)\hfill & \\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}=5.46\times {10}^{-3}+1.18\times {10}^{-6}\mathit{Pressure}-3.28\times {10}^{-3}\mathit{Emulsifier}\hfill & \\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}+2.66\times {10}^{-7}\left(\mathit{Pressure}\times \mathit{Emulsifier}\right)-8.05\times {10}^{-11}{\mathit{Pressure}}^2\hfill & \end{array}$$

MDD with SBL varied in the 400–900 nm interval. It showed a zone of minimal MDD (400–450 nm) for pressures between 5500 and 9500 psi and SBL concentrations below 0.6% (Figure 1). An interesting feature was an increment of MDD with SBL dose when the homogenization pressure was low (4000–5000 psi). On the other hand, MDD was nearly independent of the SBL dose with the highest homogenization pressures of 10,000–11,000.

The RSM model for the PdI index with SBL was highly significant (p < 0.001) (Supplementary Tables S3 and S5). Both factors exerted a significant effect on PdI. The coefficient of determination, R², was 0.68, the adjusted R² was 0.64, and the lack of fit was significant, as in the case of the MDD model. A quadratic model was selected with the following mathematical expression:

$$\begin{array}{cc}PdI\left(SBL\right)=0.780-6.55\times {10}^{-5}\mathit{Pressure}-0.285\mathit{Emulsifier}+3.75\times {10}^{-9}{\mathit{Pressure}}^2\hfill & \\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em} +0.326{\mathit{Emulsifier}}^2\hfill & \end{array}$$

In general, the value of the PdI index varied between 0.43 and 0.62 in the pressure and emulsifier dose ranges tested (Figure 1). The response surface was concave upwards and showed a zone of minimal PdI values, 0.44–0.45, when pressure and SBL ranged between 7000 and 10,000 psi and 0.2% and 0.6%, respectively.

The quadratic RSM model for DH (%) with SBL as an emulsifier was highly significant (p < 0.001) (Supplementary Tables S4 and S5). Both factors were significant with linear and/or quadratic effects. The R² and adjusted R² were 0.67 and 0.63, respectively, and the lack of fit was significant, but the p-value was close to 0.05. The equation of the selected model is as follows:

$$\begin{array}{cc}DH\left(SBL\right)=26.73-2.45\times {10}^{-4}\mathit{Pressure}+14.27\mathit{Emulsifier}\hfill & \\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em} +1.03\times {10}^{-3}\left(\mathit{Pressure}\times \mathit{Emulsifier}\right)-15.64{\mathit{Emulsifier}}^2\hfill & \end{array}$$

The response surface showed a clear curvature, being convex upwards (Figure 1). When the emulsifier dose was low (0.2–0.3%), the effect of homogenization pressure was negligible. On the other hand, at high doses of the emulsifier (0.6–1.0%), the higher the homogenization pressure, the higher the DH. The highest values for DH (33.0–34.5%) were obtained when the homogenization pressure and emulsifier dose were between 9000 and 11,000 psi and 0.6% and 1.0%, respectively.

3.2.2. Response Surface Models for MDD, PdI, and DH When HL Was Used as an Emulsifier

ANOVAs for the RSM models obtained for MDD, PdI, and DH in hydrolyzed lecithin-based emulsions are shown in Supplementary Tables S10–S13.

The data for the MDD variable were normalized using the Box–Cox transformation: MDD′ = (MDD + k)^λ, with λ = −2.77 and k = 0. The Two-Factor Interaction (2FI) model, including both factors and an interactive term, was highly significant for MDD with HL (p < 0.001) (Supplementary Tables S10 and S13). The 2FI model represents an intermediate stage between a simple linear model and a full quadratic model, where the effect of one factor (homogenization pressure) varies depending on the level of another factor (emulsifier concentration). The adjusted R² was 0.822, and the lack of fit was not significant. Therefore, we selected the model that included both factors and the interactive term with the following equation:

$$\begin{array}{cc}{MDD}^{\prime }(\mathit{diameter}\_HL)\hfill & \\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}=6.96\times {10}^{-8}+3.65\times {10}^{-12}\mathit{Pressure}+9.30\times {10}^{-8}\mathit{Emulsifier}\hfill & \\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}+1.54\times {10}^{-11}\left(\mathit{Pressure}\times \mathit{Emulsifier}\right)\hfill & \end{array}$$

MDD varied approximately between 200 and 320 nm for the pressure and HL dose ranges under study. The RSM plot showed a monotonous decrease in MDD with both factors, but the interactive term introduced a mild curvature (Figure 2). Under these conditions, no zone of maximum or minimum values exists. The higher MDD values detected (300–320 nm) were associated with lower pressures and HL doses, and vice versa. The lower MDD values in the RSM plot were found at higher pressures and HL doses tested.

The response surface model for the PdI index when HL was used as an emulsifier was highly significant (p < 0.001) (Supplementary Tables S11 and S13). Both the homogenization pressure and emulsifier dose exerted significant effects on PdI. Only the linear coefficients were significant. The coefficient of determination, R², was 0.64, the adjusted R² was nearly the same, 0.63, and the lack of fit was not significant. The following linear model was selected to model PdI with HL:

$$PdI\left(HL\right)=0.476-1.03\times {10}^{-5}\mathit{Pressure}-0.111\mathit{Emulsifier}$$

The linear RSM model for PdI indicates that the effects of both factors were additive in the pressure and dose ranges shown in Figure 2. Both factors exerted an increasing effect on PdI, which varied from 0.25 to 0.41 in the same pressure and dose ranges.

Data on the DH (%) of RSO emulsified with HL were fitted to a highly significant quadratic RSM (p < 0.001) (Supplementary Tables S12 and S13). Both factors had significant linear and quadratic coefficients. The R² and adjusted R² were 0.66 and 0.63, respectively, and the lack of fit was significant. The model equation was as follows:

$$\begin{array}{cc}DH\left(HL\right)=26.85+9.25\times {10}^{-4}\mathit{Pressure}+8.54\mathit{Emulsifier}-4.29\times {10}^{-8}{\mathit{Pressure}}^2\hfill & \\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em} -9.59{\mathit{Emulsifier}}^2\hfill & \end{array}$$

In this case, the response surface presented a conspicuous curvature, being convex upwards. In the range of factor values shown in Figure 2, DH(HL) varies as a negative parabola with emulsifier dose at any given pressure value. On the other hand, DH(HL) increases with homogenization pressure at any given emulsifier dose value. DH(HL) varied between 28.8% and 33.7% when the pressure and dose ranges were 4000–11,000 psi and 0.24–1.00%, respectively (Figure 2). Higher values of DH(HL) were obtained for pressures between 9000 and 11,000 and doses between 0.3% and 0.6%, with values of 33.4% to 33.7%.

3.3. Model Optimization

The contour plot of the global desirability with SBL as an emulsifier (Figure 3) shows a zone of values above 0.8 with homogenization pressures between 10,000 and 11,000 psi and doses of the emulsifier approximately in the range 0.6–1.0%. The maximum global desirability within that zone was estimated to be 0.916, which is close to 1.0, and it was obtained for a homogenization pressure of 10,781 psi and a SBL dose of 0.763%. Under this combination of experimental factors, the corresponding RSM equations predict MDD and PdI of 492 nm and 0.482, respectively, for lipid droplets and a DH of 34.5% for in vitro digestion.

In the case of the contour plot of the global desirability with the emulsifier HL, the values above 0.8 were in a zone characterized by homogenization pressures between 8000 and 11,000 psi and HL doses ranging between 0.25% and 0.60% (Figure 3). The maximum desirability in this zone was 1.00, which is the maximum possible global desirability, obtained for a homogenization pressure of 10,790 psi and an HL dose of 0.445%. Under these conditions, the RSM equations for MDD, PdI (for lipid droplets), and DH (for in vitro digestion) predicted values of 253 nm, 0.315, and 33.8%, respectively.

In summary, after optimizing the global desirability function for both emulsifiers (SBL and HL), the expected maximum DH was similar for SBL and HL. Although similar DH was obtained with clearly different mean droplet diameters (MDD), MDD at maximum DH with SBL nearly doubled MDD at maximum DH with HL (

Table 3. Comparison between the expected values of MDD (nm), PdI (dimensionless), and DH (%) (mean ± SD) estimated by optimizing the global desirability function (prediction) and the observed values of the same response variables measured from validation experiments (confirmation). SBL = soybean lecithin; HL = hydrolyzed lecithin.

Emulsifier	Conditions	Estimation	MDD	PdI	DH
SBL	10,781 psi, and 0.76% SBL	Prediction	492 ± 8.8	0.482 ± 0.05	34.4 ± 1.79
		Validation	544 ± 30	0.520 ± 0.02	33.1 ± 0.97
HL	10,790 psi, and 0.45% HL	Prediction	253 ± 2.4	0.315 ± 0.04	33.8 ± 1.29
		Validation	243 ± 8.2	0.294 ± 0.03	34.8 ± 0.68

).

3.4. Validation Assays

As shown in Table 3 and Supplementary Table S5, the values of MDD, PdI, and DH resulting from the validation assays matched the predicted values through the optimization procedure for both emulsifiers. In general, the values observed present a relative difference with the expected values, which were less than 10% and nearly always less than 8%. The errors were particularly small for the response variable DH, with an underestimation of 4.0% for SBL and an overestimation of 3.4% for the emulsifier HL. Interestingly, the underestimation of DH(SBL) was associated with an overestimation of MDD, and vice versa, the overestimation of DH(HL) was associated with an underestimation of MDD, which is consistent with the theory of the interfacial activity of lipases (that depends directly on the droplet surface and inversely on the diameter of emulsion droplets).

4. Discussion

4.1. CCD Experiments and Response Surface Models

The response surface models (MDD, PdI, and DH) with hydrolyzed lecithin did not show a significant lack of fit, whereas the response surface models with soybean lecithin did, particularly in the cases of MDD and PdI. The significant lack of fit indicates that the difference between the mean experimental response and the modeled surface is most likely true. Fitting a third-order model may improve the lack of fit; however, a two-factor CCD experimental design cannot support a full third-order model surface [26]. Thus, in the case of lack of fit, estimating the size of the difference between the experimental and modeled responses are important to evaluate the model’s suitability. Validation assays are useful for obtaining such insights. The results of the validation assays (Table 3) show that the percentage difference between the modeled response variable and the mean experimental value obtained from the validation assays was below 7.5%, excluding MDD with soybean lecithin, for which the difference was 10%. It is interesting to note that the percentage difference was particularly low for DH irrespective of the emulsifier, being between 3% and 4%. Therefore, it is possible to conclude that although the models for soybean lecithin suffer from lack of fit, the differences between the experimental and the modeled responses were reasonable low, particularly for the degree of hydrolysis (the variable we tried to optimize).

4.2. Effects of Emulsifier Type, Emulsifier Dose, and Homogenization Pressure on the Characteristics of Emulsion Droplet

In general, the mean diameters of emulsion droplets (MDD) produced with hydrolyzed lecithin (HL) are smaller than MDD obtained with non-hydrolyzed soybean lecithin (SBL). This difference can be explained by the different compositions of HL and SBL, since SBL is rich in phospholipids (PL) [39], and hydrolyzed lecithins become enriched in lysophospholipids (LPL) [30]. The supramolecular structures that PL and LPL can form in water depend on the chemical properties of those molecules, mainly the so-called hydrophilic-lipophilic balance (HLB), and the packing parameter (P). Therefore, emulsifiers with a higher HLB number tend to produce smaller emulsion droplets, and those with a lower P parameter below 1.0 produce a convex curvature at the surface of emulsion droplets [7,29,40]. Since PL and LPL have HLB numbers in the ranges 7–9, and 10–12, respectively, and P parameters between 0.5 and 1.0 and close to 0.33, respectively [29], HL is expected to form smaller emulsion droplets (and with a more convex surface) than SBL, as obtained by Cabezas et al. [30] with sunflower lecithin and hydrolyzed lecithin and in the present work (Figure 1 and Figure 2).

In addition to HLB and P parameters, other factors affect MDD in an emulsion. For instance, the concentration of the emulsifier (and the emulsifier-to-oil ratio) and the energy input during the emulsification process have also been reported to exert a significant effect. Generally, MDD decreases with emulsifier concentration until a given value (above which it remains approximately constant) and with homogenization pressure (when using a pressure homogenizer) [7,41,42]. However, there are particular cases that do not follow not these general principles [43]. Our data on the emulsions prepared with HL showed that MDD decreased as the homogenization pressure changed from 4000 to 12,500 psi and as the emulsifier concentration varied from 0.24% to 1.0% (Figure 2), just as expected from the general principles discussed above. The MDD values roughly ranged from 200 to 300 nm, which is close to the MDD values obtained by Hu et al. [44] when using corn oil emulsified with 1% lysolecithin at 9000 psi. The same trends were observed for PdI (Figure 2). Thus, the smaller the MDD, the more homogeneous the emulsion, suggesting a simple process of lipid fractionation responsive to the intensity of the factors promoting the phenomenon.

Conversely, the relationships among MDD, homogenization pressure, and emulsifier dose are more complex in the case of SBL-prepared emulsions. The contour plot presented in Figure 1 proved that in SBL emulsions, MDD increased with the concentration of SBL when homogenization pressures were between 4000 and 5000 psi, whereas it presented a minimum value for pressures above 6000 psi. Other authors have also reported “unorthodox” results when SBL is used to emulsify plant oils. For example, Kampa et al. [45] observed that the MDD of olive oil emulsified with SBL increased as the homogenization pressure increased from 0 to 800 bar (0–11,600 psi). It is important to note that, in our SBL emulsions, PdI also increased with SBL dose at low pressures (Figure 1). These two facts suggest the occurrence of coalescence and/or flocculation processes, which could have been promoted by the presence of Ca²⁺ (10 mM) in the water phase, as previously reported by Chum et al. [46] and Hu et al. [44]. In any event, the phenomenon requires further investigation.

4.3. Effects of Emulsifier Type, Emulsifier Dose, and Homogenization Pressure on the Degree of Hydrolysis of Rapeseed Oil Digested with Intestinal Extracts of Rainbow Trout

After 60 min of reaction, the degree of hydrolysis (DH) of emulsified RSO varied between 29% and 35% for both emulsifiers, SBL and HL (Figure 1 and Figure 2). The literature on the in vitro hydrolysis of emulsified oils digested with mammalian lipases reports several DH values. For example, Li and McClements [22] obtained a DH value of 46% by digesting emulsified corn oil with porcine pancreatic lipase. Other experiments with soybean oil emulsions [47] also yielded modest DH values between 8% and 42% (depending on the bile salts dose). Even with complex, multi-step hydrolysis protocols to simulate human digestion in vitro (INFOGEST), the lipolysis levels of fish or plant oils are in the range of 30–40% [48,49], although the pH-stat method did not determine the percentage of hydrolysis in those cases. On the contrary, other studies reported DH above 80% when digesting emulsions of corn oil [50] or triglycerides with medium-length fatty acids [22]. In addition, Hu et al. [44] obtained a wide range of DH, from 10 to more than 90%, depending on the Ca²⁺ concentration in the reaction medium. In those works, reaction times ranged from 20 to 180 min, and there was no clear relationship between reaction time and DH when different articles were compared. Similar studies reporting maximal in vitro DH of lipids with fish lipases are scarce. However, Lie and Lambertsen [51] conducted an in vitro lipolysis of 60 mg of triglycerides from a fish oil with intestinal fluids of the Atlantic cod and determined the percentage in weight of hydrolysis products. Those authors reported a 32–33% (in weight of hydrolysis products) of free fatty acids after 1 h of reaction. Considering that the molecular weight of fish triglycerides is close to 900 g/mol and that the mean molecular weight of its fatty acids is close to 270 g/mol, it is possible to estimate a DH of 53–54%. In addition, Tocher and Sargent [52] obtained 56% of released oleic acid when hydrolyzing triolein with an intestinal extract of the rainbow trout in the presence of 50 mM of cholate, which corresponds to a DH of 84%. However, Tocher and Sargent [52] also reported that lipase activity of rainbow trout intestinal extracts in the presence of 5 mM taurocholate (as in the present work) was approximately 50% of the activity with 50 mM cholate. Thus, a DH of 42% can be estimated under the reaction conditions of Tocher and Sargent [52] but with 5 mM of taurocholate. Therefore, the maximum DH values of 33–35% obtained for rapeseed oil emulsified with SBL or HL are consistent with the results from previous in vitro models implemented to study the lipolysis process.

The similarity in the range of DH and maximal DH obtained in this work, irrespective of the emulsifier, is a result worth discussing. MDD with SBL was always larger than MDD with HL. At maximal DH, the mean droplet diameter of SBL-emulsions was approximately 1.5 times that of HL-emulsions. This implies that the total droplet surface with HL was approximately 50% higher than the total droplet surface with SBL. Since lipolysis occurs at the interface between lipid droplets and the aqueous medium [53], prima facie, the lack of differences in DH could mean that nearly all lipase molecules were effectively adsorbed onto emulsion droplets and that lipase molecules released fatty acids at a similar rate. However, the activity of lipases is known to be affected by the interfacial composition (IFC), which is determined by the type of emulsifier and the concentration of bile salts [54,55]. Thus, since we tested two emulsifiers with important differences in their physicochemical properties, published results on the effects of IFC seem to invalidate the initial hypothesis of nearly complete lipase molecule adsorption with a similar activity. There is substantial research supporting that different emulsifiers lead to different accessibility of the lipase to the emulsion interface and that the activity of mammalian pancreatic lipase can increase when bile is supplied in addition to the emulsifier [34]. These results seem to be a particular case of more general experimental designs in which a series of emulsifier concentrations are tested. For example, Vinarov et al. [56] demonstrated the existence of a threshold emulsifier concentration from which lipase activity decreases with emulsifier concentration to zero or nearly zero, depending on the nature of the emulsifier. When the same research team extended the experimental work to consider the effect of bile, Vinarov et al. [54] concluded that lipase activity depended more on the ratio of emulsifier/bile salts than solely on emulsifier concentration. According to Vinarov et al. [54], when the emulsifier-to-bile salt molar ratio (EBMR) is below 0.01, IFC and DH are largely dominated by bile salts. For EBMR values from 0.01 to a certain value between 0.08 (for non-ionic surfactants) and 1.05 (for anionic surfactants), an increasing fraction of the emulsifier is incorporated into the interface, and DH increases because of the efficient assembly of mixed micelles. Finally, for EBMR values above these values (0.08–1.05), IFC begins to be dominated by the emulsifier, and DH suffers a steep drop until it reaches zero-rate lipolysis. In summary, the results of Vinarov et al. [54] support that DH peaks at a given EBMR value, which depends on the chemical nature of the emulsifier. This work is relevant to the interpretation of the results presented here. In our experiments, EBMR was roughly in the 0.6–2.5 range for SBL (assuming that phosphatidylcholine is the dominant component) and in the ranges 0.2–0.7 and 0.4–1.6 for lysophospholipids (LPL) and phospholipids (PL) contained in HL, respectively. EBMR is just in the range of optimal values described by Vinarov et al. [54] for different emulsifiers, which can explain the presence of dome-shaped surface responses for both SBL and HL. From a practical point of view, the emulsifier concentrations to obtain maximal DH were different for SBL and HL (0.8% and 0.4%, respectively). These concentrations should be considered to design future in vitro and in vivo experiments for studying the digestive lipolysis in rainbow trout.

5. Conclusions

The present work shows that it is possible to model the in vitro digestion of plant triglycerides with fish intestinal extracts in a manner similar to the experimental simulation of lipid digestion in mammals. Both soybean lecithin (SBL) and hydrolyzed lecithin (HL) are useful for emulsifying rapeseed oil with emulsion droplets below 500 nm in diameter, but HL produced smaller droplets than SBL. The response surface methodology showed that the degree of hydrolysis (DH) was significantly affected by the emulsifier concentration and the homogenization pressure applied during the emulsion production. Rainbow trout intestinal extracts effectively hydrolyze rapeseed oil emulsions. DH response surfaces showed a zone of maximum values (close to 35% after 60 min of digestion) at 11,000 psi of homogenization pressure, and approximately 0.4% and 0.8% of SBL and HL, respectively. Finally, DH seems to be more related to the nature of the emulsifier than to the lipid droplet size. The same modeling approach can be applied in future experiments to advance the understanding of lipid digestion in rainbow trout. Studying the effects of the emulsifier to bile salt molar ratio and the concentration of calcium ions on the degree of hydrolysis with rainbow trout digestive lipases will be particularly interesting.