Krill (KO, Superba^TM 2) and astaxanthin (AO, Qrill^TM) oils were kindly donated by Aker Biomarine Manufacturing LLC in Houston (Texas, US) and Aker biomarine Antarctic (Lysaker, Norway), respectively. According to the supplier's specifications, the total phospholipid content of KO was about 48.0 wt% comprising phosphatidylcholine (about 44.0 wt%) and choline (about 6.0 wt%). It has a total ω-3 PUFA content of 24.2 wt% including EPA (14.8 wt%) and DHA (6.7 wt%) (Table 1). In addition, it contains a residual amount of omega-6 (ω-6) fatty acids (about 1.0 wt%) and about 287 (μg/g oil) astaxanthin. AO contains a greater concentration of astaxanthin (≥750 μg/g oil), and consists of about 16.8 wt% ω-3 PUFAs, and about 0.5 wt% free fatty acids (FFAs) (Table 1). Among ω-3 PUFAs, AO contains about 2.8 wt% DHA and about 4.3 wt% EPA (Table 1). According to the supplier's specifications, ω-3 PUFAs exist in the form of triacylglycerols in AO. The food emulsifier sodium caseinate from bovine milk, and sodium azide (SA) that was used as an antimicrobial preservative were purchased from Sigma-Aldrich (St. Louis, USA). Grinsted® citrem LR10, a commercial anionic food emulsifier made from sunflower oil, was received as a gift from Danisco A/S (Copenhagen, Denmark). It consists of citric acid esters of mono- and di-glycerides, where oleic acid content is 79.1% of its total fatty acid content (Amara et al., 2014). For producing KO and AO emulsions, 67 mM phosphate buffer (PBS) at pH 7.40 was prepared by using Dulbecco's saline tablets purchased from Sigma Aldrich (St. Louis, Missouri, USA). In the in vitro digestion study, pH was adjusted to 6.5 using either 0.2 M NaOH or 0.2 M HCl (p.a. grade, Sigma-Aldrich, St. Louis, MO, USA). Pancreatin from porcine pancreas (8 × USP grade pancreatin activity) was purchased from Sigma-Aldrich (St. Louis, MO, USA). It is a commercial mixture of digestive enzymes produced by the exocrine cells of the pancreas and contains components such as lipase, amylase, trypsin, ribonuclease, and protease (Salentinig et al., 2017). For preparing the buffer, Milli-Q water was collected from Millipore Direct-Q3 UV system (Billerica, MA, USA). All materials were used without further purification.

Three oil-in-water (O/W) emulsions stabilized using sodium caseinate (samples S1 and S2) or citrem (sample S3) were prepared by means of ultrasonication (Ultrasonic Processor Qsonica 500, Qsonica LLC, Newtown, CT, USA) for 5 min in pulse mode (5 s pulses interrupted by 2 s breaks) at 30% of its maximum power (500 W). For the preparation of these samples, the solutions of binary mixtures of KO/sodium caseinate, AO/sodium caseinate, and AO/citrem were prepared at room temperature at a fixed KO (or AO)/sodium caseinate (or citrem) weight ratio of 4:1 and dispersed in excess PBS with pH of 7.4. They were composed of 5.0 wt% binary oil/emulsifier mixture of KO/sodium caseinate (sample S1), AO/sodium caseinate (sample S2), and AO/citrem (sample S3) and a small amount of sodium azide (0.01 wt%), that was used as an antimicrobial preservative. In addition, they were flushed with nitrogen gas for 2 min, capped, covered by aluminum foil, and incubated at room temperature prior to the planned investigations.

The zeta potentials and the mean hydrodynamic diameters were measured using Zetasizer Nano ZS (Malvern Instruments, Worcestershire, UK) equipped with a 633 nm laser and 173° detection optics. The measurements were performed at room temperature on samples diluted 100x in 67 mM PBS, and data acquisition and analysis were conducted using Malvern DTS v.6.34 software (Malvern instruments, Worcestershire, UK). For the viscosity and refractive index, the values of pure water were used.

The in vitro digestion was studied using online synchrotron SAXS similar to our previous studies (Salentinig et al., 2011, 2015a, 2017). The digestion reaction was performed at 37°C in a thermostated 20 mL reaction vessel under constant magnetic stirring. This vessel was connected to a flow-through 1.5 mm quartz capillary using HPLC tubing and pH was maintained at 6.5 in all experiments with an autotitrator (Titrando 906, Metrohm AG, Herisau, Switzerland). In the small-angle X-ray scattering (SAXS) set-up, the capillary was placed in the capillary holder and the digestion medium was circulated using a peristaltic pump, while recording SAXS in real time for monitoring the evolution of the self-assembled nanostructure during the sample digestion.

Five milliliters of sample (emulsion) was mixed with 4 mL of buffer at pH 6.5 and circulated during the real-time SAXS measurements. In this study, the pancreatin solution was prepared by mixing 5 g pancreatin extract into 20 mL of buffer at pH 6.5 and stirring using a magnetic stirrer for 30 min at 4°C. The mixture was then centrifuged at 5,000 xg for 20 min at 4°C, and the supernatant filled into an 1 mL syringe. The digestion reaction was started by using a syringe pump (neMesys, centoni GmbH, Korbußen, Germany) for remotely adding 1 mL of the prepared pancreatin solution with a speed of approximately 0.2 mL/s. Then, the pH-stat titrated the digestion mixture with 0.2 M NaOH solution in order to maintain pH at 6.5.

For gaining insight into the effect of in vitro digestion on the structural features of the three prepared oil-in-water emulsions, the lipolysis model was combined with synchrotron SAXS. The measurements were performed on the cSAXS beamline at the Swiss Light Source, Paul Scherrer Institute (PSI) (Villigen, Switzerland). An X-ray beam having a wavelength of 1.107 Å at X-ray energy of 11.2 keV was used. The sample-detector distance was 2152.4 mm and the scattering vector, q, covered a range of 0.1–5.0 nm⁻¹ (q = 4π/λ sinθ, where λ is the wavelength and 2θ is the scattering angle). All experiments were done at 37°C (±0.1°C). The 2D SAXS patterns were acquired over 3,500 s with an exposure time of 1 s with 9 s delay between frames, using an Pilatus 2M detector (in-house prototype) with an active area of 254 × 289 mm² and a pixel size of 172 x 172 μm². The 2D scattering frames were radially integrated into one-dimensional (1-D) curves and plotted as a function of relative intensity I(q) vs. q. After subtracting the background scattering (PBS buffer), all Bragg peaks were fitted by Lorentzian distributions. The lattice parameter (a) and d-spacing of the inverse hexagonal (H₂) and lamellar (L_a) liquid crystalline nanostructures, respectively, were gleaned from SAXS reflections.

The mean droplet diameters for sodium caseinate-stabilized KO and AO emulsions (samples S1 and S2) were 294.6 ± 22.7 and 296.8 ± 25.8 nm, respectively, and consistent with those previously reported for KO dispersions prepared in the presence of lysolecithin, glycerin, and hydroxypropyl methylcellulose (Seto et al., 2018). The polydispersity indices (PDIs) for both samples were 0.26 ± 0.08 and 0.20 ± 0.11, respectively (Table 2). The presence of marine phospholipids in KO, and natural emulsifiers including diacylgylcerols, monoacylglycerols (monoglycerides), phospholipids, and FFAs in AO improves the emulsifying properties of these oils, and enhances, therefore, the colloidal stabilization of the prepared oil-in-water (O/W) emulsions. For citrem-stabilized AO emulsion (sample S3), the use of the anionic emulsifier citrem instead of sodium caseinate was associated with a decrease in the mean droplet size and PDI to 226.4 ± 7.9 nm and 0.152 ± 0.07, respectively. The droplets in the three samples were negatively charged based on zeta potential measurements (Table 2). Clearly, the replacement of sodium caseinate by citrem led to a decrease in zeta potential from −19.85 ± 1.49 to −31.8 ± 1.86 mV owing to the adsorption of the negatively charged citrem molecules into the outer surfaces of AO droplets. In literature, it was reported that a good colloidal stability for ionic aqueous dispersions including emulsions can be achieved in the presence of single charged emulsifiers such as citrem on dispersing droplets with zeta potential values around ±30 mV in excess water (Honary and Zahir, 2013). In addition to its use as stabilizer with an anti-oxidative property for various food emulsions (Gudipati et al., 2010; Berton et al., 2011), citrem is attractive for use as a safe emulsifier for stabilizing lamellar and non-lamellar liquid crystalline nanoparticles owing to its hemocompatiblity and poor activation of the complement system (Nilsson et al., 2012; Hedegaard et al., 2013; Wibroe et al., 2015; Prajapati et al., 2018). In our study, the three investigated samples were visually inspected and found to be colloidally stable at room temperature for at least 3 months after preparation, see images in Figure 2B.

Synchrotron SAXS coupled with flow-through lipolysis model was used to gain insight into the fate of KO and AO emulsion droplets under simulated conditions of the small intestine, which is considered the main site for lipid digestion that leads to structural alterations or generation of nano-self-assemblies (Salentinig et al., 2015a, 2017; Salentinig, 2019; Vithani et al., 2019). At pH of 6.5, where pancreatic lipase has its maximum activity, the digestion-triggering dynamic structural alterations in the three KO and AO emulsions (samples S1–S3) were investigated. Figure 3 presents the time-resolved SAXS intensity profiles for the digested KO and OA emulsions. Before the addition of pancreatin extract containing the lipase, a weak reflection at q of about 0.83 nm⁻¹ can be observed in the SAXS pattern of the KO emulsion (Figure 3A), that is absent in the AO systems (Figures 3B,C). This reflection may arise from the formation of traces of multi-lamellar structures during emulsification of KO that contains up to 48.0 wt% phospholipids. Representative SAXS diffraction patterns at different elapsed times of lipolysis are also presented in Figure 4. Before the addition of pancreatin extract, the detected upturn of the SAXS patterns of the three investigated emulsions, as shown in Figure 3, at relatively low q values is most likely attributed to scattering from emulsion droplets with dimensions larger than the resolution of the SAXS setup. For these emulsions, it was evident from the SAXS patterns presented in Figures 3, 4 that KO and AO digestion was associated with the appearance of either a biphasic feature of a lamellar (L_α) phase coexisting with an inverse hexagonal (H₂) phase (Figures 3A, 4A) or a neat L_α phase (Figures 3B,C, 4B,C). Depending on the lipid structure (e.g., unsaturation degree, acyl chain length, and structure of polar headgroup) and composition, the trends in appearance and evolving behavior of the lyotropic lamellar and non-lamellar liquid crystalline phases during lipolysis in these emulsion droplets were consistent with previous investigations on the in situ lipolysis of triolein emulsions, milk, and triglyceride-containing emulsified food products such as mayonnaise (Salentinig et al., 2011, 2015a, 2017) (Table 2). The rapid occurrence of self-assembled interiors in these KO and AO emulsion droplets (Figure 3) is attributed to the lipolysis-mediated generation of amphiphilic lipids and the involved fast molecular re-distribution and alterations in the lipid composition. Referring back to the lipolysis products, the adsorption of pancreatic lipase into the emulsion droplets' surfaces enhances the hydrolysis of triglycerides resulting in monoglycerides, diglycerides, and FFAs. Depending on molecular structure, ionic strength, temperature, and pH, these amphiphilic compounds and their combinations have the propensity to adopt lamellar and non-lamellar liquid crystalline phases on exposure to excess water (Nakano et al., 2002; de Campo et al., 2004; Yaghmur et al., 2005, 2017; Yaghmur and Glatter, 2009; Azmi et al., 2016; Gontsarik et al., 2016, 2018, 2019; Angelova et al., 2017). Thus, the lipolysis-triggered transition from emulsified oil droplets to lamellar or non-lamellar liquid crystalline interiors (Figure 3) was affected by the type of oil and attributed to the self-assembly of the digestion generated amphiphilic lipids in the lipolysis medium. The evolved scattering curves resembled those typically reported for liposomes, cubosomes, hexosomes, amongst others (Nakano et al., 2002; de Campo et al., 2004; Yaghmur et al., 2005; Yaghmur and Glatter, 2009; Azmi et al., 2016; Angelova et al., 2017; Gontsarik et al., 2018). Taking into account the suggested important role of these produced self-assembled nanostructures in facilitating the delivery and distribution of nutritional compounds throughout the body (Salentinig et al., 2011, 2015a, 2017; Salentinig, 2019; Vithani et al., 2019), there is a growing interest in exploring the implications for future development of advanced functional food and drug nanocarriers (Azmi et al., 2015; Angelova et al., 2017; Shao et al., 2018; Bor et al., 2019; Yaghmur, 2019).

Referring again to Figures 3A, 4A, the fast molecular re-distribution and self-assembly of the generated amphiphilic lipolysis products in the emulsified KO droplets led first to the appearance of the H₂ phase. After about 80 s, the first order reflection (100) of this phase started to be visible as a very weak peak at q-value of about 1.01 nm⁻¹ (Figure 4A). Its intensity increased with time of digestion and the two additional characteristic (110) and (200) reflections started to be evolved at q-values of about 1.74 and 2.01 nm⁻¹, respectively, after about 280 s of lipolysis. At this lipolysis time, the first order reflection of the coexisting L_α phase started also to appear at q value of about 1.40 nm^−l. Its identification was based on the appearance of very weak third characteristic reflection after 2,980 s of lipolysis at q-value of about 4.20 nm⁻¹. Based on the position of the first order reflection that was clear to identify, the second order peak position at q-value (~2.80 nm⁻¹) was calculated and marked in Figure 4. As shown in Figure 5A, the lattice parameter and d-spacing of the H₂ and L_α phases were about 7.21 ± 0.04 and 4.50 ± 0.10 nm, respectively, and were almost not affected with increasing the lipolysis time up to 2,980 s. In this H₂/L_α phase coexistence regime, the H₂ phase became more ordered and developed with increasing lipolysis time as indicated from the time evolution of the intensity of its first order diffraction peak (Figure 5B). However, less time was required for the coexisting L_α phase to become fully developed as the intensity of its first order diffraction peak was increased up to ~910 s and then reached plateau with increasing further the lipolysis time (Figure 5B).

To compare the effect of ω-3 PUFA-containing oil type on the lipolysis-triggered evolving behavior of lyotropic liquid crystalline phases in the emulsified emulsion droplets, two additional sets of time-resolved SAXS experiments were carried out on AO emulsions stabilized by using sodium caseinate (sample S₂) and citrem (sample S₃). As presented in Table 1, AO exists similar to FO in the form of triacylglycerols; whereas most of ω-3 PUFAs as mentioned above are bound to phospholipids in KO. This significant difference between these two oils in lipid composition, phospholipid content, molecular structure of ω-3 PUFAs (triacylglycerols as compared to phospholipids) was reflected in the SAXS profiles presented in Figure 3 for the three investigated emulsions under identical experimental conditions. For AO emulsions (samples S₂ and S₃, Table 2), it was evident from the SAXS profiles presented in Figures 3B,C that the associated changes in the lipid composition of the emulsified AO during lipolysis led to the appearance of a neat L_α phase. Replacing sodium caseinate with citrem did not have a significant effect on the kinetics of structure formation. However, it is important to bear in mind that the replacement of sodium caseinate and citrem by non-digestible polymeric stabilizers could lead to a significant decrease in the digestibility of these emulsions due to restrictions on lipase accessibility to the oil localized in the droplets' interiors. In a previous study, Gudipati et al. (2010) reported on the important role of the stabilizer adsorption pattern in modulating the digestibility of FO emulsions, and showed that the replacement of citrem by non-digestible polymeric stabilizers is associated with a decrease in the digestion rate.

The experimental findings in Figures 5C,D provide more insight into the evolved neat L_α phase in both samples. Similar to sodium caseinate-stabilized KO emulsion (Figure 5A), the d-spacing of the L_α phase in samples S₂ and S₃ was 4.48 ± 0.02 and 4.52 ± 0.10 nm, respectively, and was almost not affected within the lipolysis periods of 3,010 and 3,410 s, respectively (Figure 5C). Thus, the replacement of sodium caseinate by citrem did not induce any significant change in the d-spacing of the L_α phase. These results indicate that there is no significant incorporation of sodium caseinate or citrem into AO during the emulsion preparation and digestion, and are in contrast with our previous findings on the significant effect of this emulsifier (citrem) on the structural features of lamellar and non-lamellar liquid crystalline nanoparticles based on soybean phospholipid, ω-3 PUFA monoglycerides, or monoolein (Hedegaard et al., 2013; Azmi et al., 2015, 2016; Wibroe et al., 2015; Yaghmur et al., 2017; Prajapati et al., 2018; Shao et al., 2018). The time evolution of the first order reflection of the L_α phase and the single exponential function used to fit the SAXS data (full blue line) are presented in Figure 5D. The formation of the developed L_α phase in samples S₂ and S₃ was accomplished after ~1,500 and 1,250 s, respectively. It is interesting that the oil type did not have any significant influence on the d-spacing of the L_α phase in the three investigated samples (Figures 5A,C). The reason for this similarity in the d-spacing of the L_α phase in emulsified KO and AO droplets stabilized either by sodium caseinate or citrem is unclear but could be attributed to the presence of relatively similar lipid composition (mainly FFAs) in the lipolysis-triggered L_α phase in these samples. It is clear that the effects of lipid composition of KO and AO, and the concentration and type of ω-3 PUFA-containing lipolysis products on modulating their digestibility and the associated dynamic structural features warrant further investigation.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.