Cellulose Nanocrystals-Stabilized Acidic W₁/O/W₂ Emulsions for Anthocyanins Encapsulation

Abstract

The limited stability of anthocyanins restricts their application in the food industry, necessitating encapsulation to prevent degradation. This study fabricated an anthocyanin-rich acidic water-in-oil-in-water (W₁/O/W₂) emulsion system stabilized by cellulose nanocrystals (CNCs). Anthocyanins extracted from the by-product peels of ‘France’ Prunus domestica L. were incorporated into the inner aqueous phase (W₁). The internal phase (W₁/O) ratio was increased to 40% (w/w) to enhance anthocyanin loading capacity. CNCs were sonicated to reduce their size and improve their interfacial properties, thereby enhancing the emulsifying capacity. Sonicated CNCs combined with whey protein isolate (WPI) significantly improved double emulsion performance compared to the non-sonicated CNCs–WPI system: (1) reduced D₄₃ from 8.50 µm to 4.35 µm; (2) elevated ζ-potential from 7.49 ± 0.99 mV to 10.07 ± 1.50 mV; and (3) improved encapsulation efficiency from 52.96 ± 2.60% to 83.39 ± 0.96%. Furthermore, encapsulated anthocyanins exhibited significantly enhanced thermal stability compared to free anthocyanins, with the half-life at 50 °C increasing from 14.72 ± 0.35 h to 70.37 ± 0.51 h. This study demonstrates that modifying nanoparticle interfacial properties provides valuable insights for designing stable emulsions and enhancing anthocyanin stability.

1. Introduction

Using nanoparticle surfactants to stabilize the liquid–liquid interface has gathered considerable interest in the field of emulsion development. Cellulose is the most abundant, biodegradable, renewable, and sustainable biopolymer in nature [1]. Nanocelluloses, which possess distinctive characteristics resulting from their structural and surface chemical properties, have been widely employed as solid “particle” stabilizers for various types of emulsions [2]. Cellulose nanofibers (CNFs) and cellulose nanocrystals (CNCs) are two main forms of nanocelluloses. CNFs possess a larger aspect ratio and greater surface charges, primarily relying on an entangled network structure to increase viscosity and form a 3D network around oil droplets, thereby stabilizing emulsions. In contrast, CNCs—obtained through acid hydrolysis of CNFs—are highly crystalline and nanoscale in size. CNCs exhibit amphiphilic properties, allowing them to adsorb at the oil–water interface and form a stable interfacial film [3]. Calabrese et al. [4] found that shorter CNCs lengths lead to higher interfacial coverage and better emulsification efficiency. Ultrasonic treatment can further reduce the size of CNCs, making them a promising stabilizer for emulsion-based delivery systems of bioactive compounds in foods.

Anthocyanidins, a group of water-soluble natural pigments widely present in plants, naturally exist as glycosides known as anthocyanins. As phenolic compounds belonging to the flavonoid class [5], anthocyanins demonstrate color characteristics and functional properties including anti-diabetic, antioxidant, and anti-inflammatory activities [6,7]. However, the practical utilization of anthocyanins in the food industry is limited due to the influence of temperature, pH, light, oxygen, metal ions, and enzymes [8]. Prunus domestica L. originates from Europe and is mainly cultivated in Germany, the United States, Romania, Bulgaria, and Serbia, and is also called the European plum [9]. Nowadays, Prunus domestica L. has been introduced and cultivated in multiple areas of China. ‘France’ Prunus domestica L., widely planted in Xinjiang, China, is particularly valued for its superior storage stability, soft texture, sweet flavor, and nutritional profile rich in vitamins, minerals, and bioactive compounds [10]. Agricultural by-products could be utilized as sources of bioactive compounds such as anthocyanins for more sustainable food production. During ‘France’ Prunus domestica L. processing (including dried fruit, juice, and candied fruit production), the peels are typically discarded as waste. Studies have revealed significantly higher anthocyanins content in peels compared to flesh tissues [11].

It is of great value to develop efficient extraction methods for anthocyanins from ‘France’ Prunus domestica L. processing by-products (peels) and to address their critical stability challenges. Acidic organic solvents combined with emerging technologies (ultrasound/microwave-assisted extraction) efficiently extract anthocyanins by disrupting cell walls and enhancing mass transfer. Encapsulation technology offers an effective approach to stabilize anthocyanins. Emulsions are dispersion systems composed of two or more immiscible liquids, including O/W (oil-in-water) emulsions, W/O (water-in-oil) emulsions, and other complex structures such as W₁/O/W₂ (water-in-oil-in-water) and O₁/W/O₂ (oil-in-water-in-oil) double emulsions [12]. By encapsulating anthocyanins within the innermost aqueous phase (W₁), the W₁/O/W₂ emulsions provide unique advantages for stability and bioavailability improvement. Huang and Zhou [13] developed a W₁/O/W₂ emulsion encapsulating black rice anthocyanins, achieving exceptional encapsulation efficiency (99.45 ± 0.24%). Teixé-Roig et al. [14] engineered a carboxymethyl cellulose sodium–lecithin co-stabilized W₁/O/W₂ emulsion, which significantly enhanced anthocyanin bioavailability to 31.78 ± 1.73% compared to free anthocyanins (2.0%).

This study used ultrasonicated CNCs to stabilize a W₁/O/W₂ emulsion system for encapsulating anthocyanins extracted from ‘France’ Prunus domestica L. peels. Current emulsion systems predominantly employ deionized water as the aqueous phase for anthocyanin protection [15,16,17]. However, anthocyanins exhibit greater stability in acidic conditions (pH < 3), where they predominantly exist as the red flavylium cation, compared to neutral or alkaline conditions [18]. In this study, the internal (W₁) and external (W₂) aqueous phases were both phosphate-buffered solution (pH = 2.5). This pH-controlled dual-phase design effectively minimized anthocyanin degradation and trans-phase migration. A novel whey protein isolate–cellulose nanocrystals from ginkgo seed shells (WPI–CNCs) complex was engineered as emulsifiers. Ultrasound-mediated CNCs length optimization was implemented for enhanced emulsifying capacity. Comprehensive characterization (particle size, ζ-potential, anthocyanin encapsulation efficiency, color parameters, and confocal/brightfield microscopy) was conducted to evaluate the emulsions stability. Furthermore, thermal degradation kinetics of both free anthocyanins and double emulsion-encapsulated anthocyanins were investigated to reveal the enhancement of stability. This research offers an effective strategy for constructing bioactive compounds delivery systems through nanoparticle interfacial engineering, providing technical insights into anthocyanin encapsulation and stability enhancement.

2. Materials and Methods

2.1. Reagents and Raw Materials Treatment

Fresh ‘France’ Prunus domestica L. fruits from Xinjiang were peeled, dried at 40 °C for 12 h in an electric blast drying oven (FD 115, Binder, Tuttlingen, Germany) to a final moisture content of 3.11 ± 0.07%, then ground using a grinder (A11, Ika, Staufen, Germany) and passed through a 50-mesh sieve. The fine powder was stored in amber glass vials at −20 °C. Ginkgo seeds were provided from Jiangsu, shelled, dried, pulverized, and sieved through a 100-mesh sieve prior to storage.

Gallic acid (≥99%), 6-hydroxy-2,5,7,8-tetramethylchroman-2-carboxylic acid (Trolox, ≥98%), and 2,4,6-tris(2-pyridyl)-1,3,5-triazine (TPTZ, ≥98.5%) were obtained from J&K Scientific Ltd. (Beijing, China). 1,1-diphenyl-2-trinitrophenylhydrazine (DPPH, ≥98%), 2,2′-azino-bis(3-ethylbenzothiazoline-6-sulfonic acid) (ABTS, ≥98%), polyglycerol polyricinoleate (PGPR), and whey protein isolate (≥80%) were obtained from Yuanye Bio-Technology Co., Ltd. (Shanghai, China). Folin-phenol (1 mol/L, BR), rutin (BR), phosphoric acid (GR), and dialysis membrane (8–14 kDa molecular weight cut-off) were provided by Sinopharm Chemical Reagent Co., Ltd. (Shanghai, China). Soybean oil was obtained from Cofco Group Co., Ltd. (Beijing, China). Nile red and Nile blue were obtained from Sigma-Aldrich (St. Louis, MO, USA). The rest of the chemicals were of analytical grade.

2.2. Extraction of Anthocyanins

2.2.1. Solvents Extraction Method

The procedure was adapted from Johnson et al. [19] with some modifications. Anthocyanins were extracted from ‘France’ Prunus domestica L. peel powder using 80% (v/v) methanol (pH 2.0), in triplicate. The powder was mixed with the solvent at a solid–liquid ratio of 1:15 (w/v) and magnetically stirred (500 rpm, 1 h) using a homogenizer (Whmix 6, Wiggens, Straubenhardt, Germany) at room temperature (25 °C). The mixture was centrifuged (25 °C, 3500× g, 15 min) in a refrigerated centrifuge (Velocity 14R, Dynamica, Mablethorpe, UK), and the supernatant was collected. The residue was re-extracted, and the combined supernatants were filtered and concentrated under vacuum (45 °C, −0.1 MPa, 10 min) using a rotary evaporator (RV8, IKA, Staufen, Germany). The extract was diluted with 0.05% HCl (v/v). The 80% ethanol extraction method and 70% acetone extraction method were carried out following the same method.

The optimal solvent was applied to subsequent ultrasonic-assisted and microwave-assisted extraction methods. The conventional solvent extraction procedure described in this section was designated as the control.

2.2.2. Ultrasound-Assisted Extraction

The procedure was adapted from Andrade et al. [18] and Bu et al. [20] with some modifications. The peel powder was mixed with the extraction solvent (1:15, w/v), then subjected to ultrasonic extraction (25 °C, 50 kHz, 360 W, 20 min) in an ultrasonic bath (TL-615HTD, Tianling Ltd., Shenzhen, China), in triplicate, followed by centrifugation and re-extraction of the residue. Operations (centrifugation, filtration, evaporation, and dilution) were identical to those described in Section 2.2.1.

2.2.3. Microwave-Assisted Extraction

The procedure was adapted from Liazid et al. [21] with some modifications. The peel powder was mixed with solvent (1:15, w/v), then subjected to microwave extraction (280 W, 5 min; P70F23P-G5 microwave oven, Galanz, Foshan, China), in triplicate. The residue was re-extracted after centrifugation. Subsequent steps were identical to those described in Section 2.2.1. All final extracts were stored at −20 °C in airtight and light-free containers.

2.3. Characterization of Prunus domestica L. Peel Extracts

2.3.1. Determination of Total Anthocyanins Content (TAC)

TAC was determined using the pH differential method [22]. Buffer solutions were prepared as follows:

pH 1: 0.025 M KCl adjusted with 32% HCl

pH 4.5: 0.4 M sodium acetate adjusted with 32% HCl

A sample extract (200 µL) was mixed with 3.8 mL of each buffer and equilibrated in the dark at 25 °C for 15 min. Absorbance was measured at 510 nm and 700 nm using a UV–visible spectrophotometer (L8 UV–Vis, INESA Scientific Instrument Ltd., Shanghai, China) with ultrapure water as the blank. TAC was expressed as “mg cyanidin-3-glucoside equivalents (CGE) 100 g⁻¹ dry weight (DW) of ‘France’ Prunus domestica L. peel powder”, using the following formulas.

$$\mathrm{T}\mathrm{A}\mathrm{C}={\displaystyle \frac{\mathrm{m}\mathrm{g} \mathrm{C}\mathrm{G}\mathrm{E} {\mathrm{L}}^{-1}\times \mathrm{V}\times 100}{\mathrm{m}}}$$

(1)

$$\mathrm{m}\mathrm{g} \mathrm{C}\mathrm{G}\mathrm{E} {\mathrm{L}}^{-1}={\displaystyle \frac{A\times \mathrm{M}\mathrm{W}\times \mathrm{D}\mathrm{F}\times 1000}{\mathsf{\epsilon }\times 1}},$$

(2)

$$A=\left({{\mathrm{p}\mathrm{H}}_1:ABS}_{510 \mathrm{n}\mathrm{m}}-{ABS}_{700 \mathrm{n}\mathrm{m}}\right)-\left({{\mathrm{p}\mathrm{H}}_{4.5}:ABS}_{510 \mathrm{n}\mathrm{m}}-{ABS}_{700 \mathrm{n}\mathrm{m}}\right)$$

(3)

where DF is the dilution factor (20×), MW is the molecular weight of CGE (449.38 g mol⁻¹), ε is the molar extinction coefficient of CGE (26,900 M⁻¹ cm⁻¹), 1 is the path length (1 cm), 1000 is the conversion of ‘g’ to ‘mg’, V is extract volume, and m is powder weight.

2.3.2. Determination of Total Phenolic Content (TPC)

TPC was determined using the Folin–Ciocalteau method as described by Wang et al. [23] and Zhao et al. [24] with some modifications. 1 mL of the diluted extract was mixed with 0.5 mL Folin–Ciocalteau reagent (1 mol/L) for 6 min. Then, 1.5 mL of 7.5% (w/v) sodium carbonate solution was added to the mixture. The mixture was incubated in a water bath at 70 °C for 30 min, cooled and diluted to a final volume of 10 mL with deionized water. Absorbance was measured at 760 nm. A standard curve was constructed using gallic acid solutions, subjected to the same reaction conditions. TPC was expressed as “mg gallic acid equivalents (GAE) g⁻¹ DW”.

2.3.3. Determination of Total Flavonoid Content (TFC)

TFC was determined according to the method described by Xie et al. [25] with some modifications. 1 mL of diluted extract was mixed with 0.3 mL 5% (w/v) sodium nitrite. After 5 min of reaction, 0.3 mL 10% (w/v) aluminum chloride was added and allowed to react for 6 min. Subsequently, 4 mL 1 mol/L NaOH was added, and the mixture was diluted to 10 mL with deionized water. The reaction system was equilibrated at room temperature (25 °C) for 12 min before measuring absorbance at 510 nm. A standard curve was constructed using rutin methanol solution. TFC was expressed as “mg rutin equivalents (RE) g⁻¹ DW”.

2.3.4. Antioxidant Capacity

DPPH radical scavenging activity

DPPH radical scavenging activity was determined according to the method described by Sun et al. [26] with some modifications. A 50 µg/mL DPPH ethanol solution was freshly prepared. Absorbance was measured at 517 nm, using anhydrous ethanol as blank. Three reaction systems were established:

A₁: 1 mL diluted sample + 3 mL DPPH solution

A₂: 1 mL anhydrous ethanol + 3 mL DPPH solution

A₀: 1 mL diluted sample + 3 mL anhydrous ethanol

All mixtures were vortexed and incubated in the dark at 25 °C for 30 min. The scavenging rate (P, %) was calculated using the following formula. A standard curve was constructed using Trolox solutions. Results were expressed as “mg Trolox equivalents (TE) g⁻¹ DW”.

$$P=\left(1-{\displaystyle \frac{A_1-A_0}{A_2}}\right)\times 100\mathrm{\%}$$

(4)

ABTS radical scavenging activity

ABTS radical scavenging activity was determined according to the method described by Vidana Gamage and Choo [27] with some modifications. The ABTS stock solution was prepared by mixing 7 mM ABTS solution with 2.5 mM potassium persulfate at a 1:1 (v/v) ratio, followed by 12–16 h of dark incubation at 25 °C. The working solution was freshly prepared by diluting the stock with anhydrous ethanol to an absorbance of 0.70 ± 0.02 at 734 nm, using distilled water as blank. Reaction systems were established as:

A₁: 0.4 mL diluted sample + 3.6 mL ABTS working solution

A₀: 0.4 mL anhydrous ethanol + 3.6 mL ABTS working solution

All mixtures were vortexed and incubated in the dark at 25 °C for 10 min. P was calculated using the following formula. A standard curve was constructed using Trolox solutions. Results were expressed as “mg TE g⁻¹ DW”.

$$P=\left({\displaystyle \frac{A_0-A_1}{A_0}}\right)\times 100\mathrm{\%}$$

(5)

Ferric reducing antioxidant power (FRAP)

FRAP was determined according to the method described by Benzie and Strain [28] and Szydłowska-Czerniak et al. [29] with some modifications. The FRAP working solution was prepared by mixing: 300 mmol/L acetate buffer (pH 3.6): 3.1 g sodium acetate trihydrate + 16 mL glacial acetic acid per liter, 10 mmol/L TPTZ dissolving in 40 mmol/L hydrochloric acid, and 20 mmol/L hexahydrate ferric chloride. The components were combined at a 10:1:1 (v/v/v) ratio. For analysis, 3 mL of freshly prepared FRAP working solution was preheated to 37 °C. Test samples were prepared by adding 100 µL diluted extract and 300 µL deionized water to 3 mL FRAP solution. Absorbance was measured at 593 nm after 10 min of dark incubation at 25 °C, using 400 µL H₂O + 3 mL FRAP solution as blank. A standard curve was constructed using Trolox solutions. Results were expressed as “mg TE g⁻¹ DW”.

2.3.5. Determination of Color Parameters

Zero calibration and white calibration were performed on the precision colorimeter (UltraScan Pro 1166, HunterLab, Reston, VA, USA). Subsequently, the L*, a*, and b* values were measured. The a* (red–green axis) and b* (yellow–blue axis) values range from −120 to 120. The L* value ranges from 0 to 100, where values closer to 100 indicate higher sample brightness. The hue angle (H*), chroma (C*), and total color difference (ΔE*) were calculated using the following formulas.

$$H^{*}={\mathrm{t}\mathrm{a}\mathrm{n}}^{-1}(b^{*}/a^{*}),$$

(6)

$$C^{*}=\sqrt{a^{*2}+b^{*2}},$$

(7)

$$∆E^{*}=\sqrt{∆L^{*2}+∆a^{*2}+∆b^{*2}}$$

(8)

2.4. Purification and Ultrasonication of Cellulose Nanocrystals (CNCs) from Ginkgo Seed Shells

The purification process followed Ni et al. [30]. Ginkgo seed shells were sequentially treated as follows: Shell powder was mixed with 4% (w/v) NaOH solution and stirred at 90 °C for 2 h. This alkaline washing was repeated four times. Then, the dried residue was treated with a bleaching solution (1.7% (w/v) NaClO, 2.7% (w/v) NaOH, and 7.5% (v/v) CH₃COOH) and stirred at 80 °C for 2 h. This bleaching step was repeated four times, followed by thorough rinsing with deionized water until neutral pH was obtained. The final residue was dried.

The purified cellulose was dispersed in 62% (v/v) sulfuric acid at a 1:15 (w/v) ratio and stirred at 45 °C for 30 min. The hydrolysis reaction was stopped with 10-fold volume of cold deionized water. After centrifugation (10,000× g, 10 min), the sediments were re-dispersed in deionized water and then dialyzed (MWCO: 8–14 kDa) until a constant pH was obtained. The obtained CNCs suspension (0.3%, w/w, dispersed in pH 2.5 phosphate buffer) was ultrasonicated using an ultrasonic probe (TL-1200Y, Tianling Ltd., Shenzhen, China) at 240 W with 1 s on/off pulses for 45 min under ice-water bath cooling, yielding ultrasonicated CNCs (denoted “Ultra CNCs”).

2.5. Characterization of CNCs

2.5.1. Morphological Characterization

The CNCs suspension was diluted to 0.005% (w/w) with deionized water and homogenized by sonication (150 W, 20 min). Subsequently, 10 μL sample was deposited onto silicon wafers and dried at room temperature. Morphological analysis was performed using atomic force microscopy (Dimension FastScan, Bruker, Ettlingen, Germany). Image processing was conducted with NanoScope Analysis software (v3.0, Bruker, Billerica, MA, USA).

2.5.2. Contact Angle

Wettability differences between untreated and Ultra CNCs were evaluated via water contact angle measurements. Uniform thin films were prepared by adding 1 mL CNCs suspension onto glass slides, followed by overnight drying at 50 °C. Contact angles were determined using the sessile drop method: 2 μL deionized water droplets were deposited onto films via precision syringe (needle diameter: 0.51 mm), with imaging and angle calculation performed using a contact angle measuring device (OCA15EC, Dataphysics, Filderstadt, Germany).

2.5.3. Interfacial Tension

The interfacial tension between Ultra CNCs suspension and soybean oil was quantified via pendant drop method using the contact angle measuring device. Suspension was loaded into a steel syringe (needle diameter: 1.83 mm) and slowly extruded into an oil-filled quartz cuvette. Time-dependent tension values were recorded for 30 min at 25 °C.

2.5.4. X-Ray Diffraction (XRD)

The X-ray diffraction (D2 PHASER, AXS, Berlin, Germany) was used to compare the crystallinity between CNCs and Ultra CNCs. The 2θ range from 5° to 50° with a scan rate of 1° min⁻¹. The conditions of operation were 40 kV, 40 mA, and Cu Kα radiation (wavelength 0.154 nm). The crystallinity index (CrI) of cellulose was calculated by the following formula.

$$\mathrm{C}\mathrm{r}\mathrm{I}\left(\mathrm{\%}\right)={\displaystyle \frac{I_{002}-I_{am}}{I_{002}}}\times 100\mathrm{\%}$$

(9)

2.5.5. X-Ray Photoelectron Spectroscopy (XPS)

Chemical surface analysis was carried out using an X-ray photoelectron spectrometer (ESCALAB 250Xi, Thermo Fisher Scientific, Waltham, MA, USA) with Al Kα X-ray source (1486.6 eV) under ultra-high vacuum (<10⁻⁹ mbar). Charge compensation was applied during the measurement. The energy scale was calibrated by referencing the C1s peak at 284.8 eV. The survey and high-resolution XPS spectra were recorded at pass energies of 200 eV and 50 eV, with resolutions of 1.0 eV and 0.1 eV, respectively. The spot size for analysis was 500 μm in diameter.

2.5.6. Fourier Transform Infrared Spectroscopy (FTIR)

The molecular structure of CNCs was analyzed by ATR-FTIR (Nicolet iS 50, Thermo Fisher Scientific, Waltham, MA, USA). Spectra were recorded from 4000 to 400 cm⁻¹, with 64 scans for both background and sample signal.

2.6. Preparation of W₁/O/W₂ Double Emulsions

2.6.1. W₁/O Emulsions Preparation

The procedure was adapted from Huang and Zhou [13] with some modifications. Polyglycerol polyricinoleate (PGPR, 3%, 4%, or 5%, w/w) was blended with soybean oil (62%, 61%, or 60% w/w, respectively) under magnetic stirring (50 °C, 500 rpm, 30 min). The anthocyanin-rich inner aqueous phase (W₁) was prepared by dissolving ‘France’ Prunus domestica L. peel extracts in phosphate buffer (pH 2.5). The W₁ phase (35% w/w) was combined with the oil phase (65% w/w) and mixed at 10,000 rpm for 3 min using a rotor–stator homogenizer (Ultra Turrax T18, IKA, Staufen, Germany), followed by three cycles (40 MPa) through a high-pressure homogenizer (AH-2010, Antuos Nanotechnology, Suzhou, China). The resulting W₁/O emulsions were stored at 4 °C.

2.6.2. W₁/O/W₂ Double Emulsions Preparation

WPI–Ultra CNCs complex was prepared by dispersing 2.4% (w/w) whey protein isolate (WPI) into Ultra CNCs suspension under magnetic stirring (600 rpm, 2 h). The control complex (WPI–CNCs) used untreated CNCs. Primary W₁/O emulsions (20%, 25%, 30%, 35%, and 40% w/w) were combined with W₂ phases [WPI–CNCs or WPI–Ultra CNCs complex; 80%, 75%, 70%, 65%, and 60% w/w], homogenized under 10,000 rpm for 3 min and processed through a high-pressure homogenizer at 10 MPa for 3 cycles to form W₁/O/W₂ emulsions.

2.7. Characterization of Emulsions

2.7.1. Particle Size and ζ-Potential

The procedure was adapted from Shaddel et al. [15] with some modifications. The ζ-potential was measured using a multi-angle nanoparticle size and zeta potential analyzer (Nano Brook Omni, Brookhaven Instruments, Nashua, NH, USA). Emulsions were diluted 100-fold with deionized water, and the measurements were replicated three times. Particle size distribution was determined by a laser diffraction analyzer (S3500, Microtrac, Montgomeryville, PA, USA).

2.7.2. Confocal Laser Scanning Microscopy (CLSM) and Optical Microscopy

The procedure was adapted from Jiang et al. [31] with some modifications. Nile blue (0.10% w/w) and Nile red (0.10% w/w) were used to stain WPI and the oil phase. The images were observed by fluorescence CLSM (LSM 880, Carl Zeiss, Oberkochen, Germany) with a 40× objective lens. Double excitation wavelengths (543 nm, 633 nm) were used to obtain images with Nile red and Nile blue. For brightfield observations, emulsions were 10-fold diluted with deionized water, and the samples were imaged using an inverted fluorescence microscope (Axio Vert A1, Carl Zeiss, Germany).

2.7.3. Encapsulation Efficiency (EE) of Anthocyanins

The procedure was adapted from Li et al. [32] with some modifications. The double emulsions were diluted 4-fold with deionized water and centrifuged (5000× g, 10 min). The W₂ phase was carefully collected via syringe aspiration from the lower aqueous layer. Anthocyanins concentration in W₂ was quantified using the pH differential method as described in Section 2.3.1. EE was calculated by the following formula.

$$\mathrm{E}\mathrm{E}\left(\mathrm{\%}\right)=100\times {\displaystyle \frac{{\mathrm{C}}_{{\mathrm{W}}_1}\times {\mathrm{X}}_{{\mathrm{W}}_1}-C_i\times (\mathrm{D}+{\mathrm{X}}_{{\mathrm{W}}_2})}{{\mathrm{C}}_{{\mathrm{W}}_1}\times {\mathrm{X}}_{{\mathrm{W}}_1}}}\times 100\mathrm{\%}$$

(10)

where ${\mathrm{C}}_{{\mathrm{W}}_1}$ is the initial anthocyanin concentration in W₁ phase, $C_i$ is the anthocyanin concentration in W₂ phase after the formation of W₁/O/W₂ emulsions, ${\mathrm{X}}_{{\mathrm{W}}_1}$ is the mass fraction of W₁ phase, ${\mathrm{X}}_{{\mathrm{W}}_2}$ is the mass fraction of W₂ phase, and $\mathrm{D}$ is the dilution factor (buffer volume/emulsions volume).

2.7.4. Storage Stability of W₁/O/W₂ Double Emulsions

The visual appearance, droplet size, ζ–potential, and anthocyanin retention of the double emulsions (40% W₁/O, w/w) were periodically detected during seven days of dark storage at 25 °C, 37 °C, and 50 °C, respectively. The upper emulsions layer was collected for analysis after phase separation occurred.

2.8. The Thermal Degradation Kinetics of Anthocyanins

The procedure was adapted from Slavu et al. [33] with some modifications. Double emulsions (40% W₁/O, w/w) and free anthocyanin solution adjusted to identical pH (3.3) were heated at 50, 60, 70, 80, and 90 °C in a water bath protected from light for 6 h. Samples were collected at 0, 1, 2, 3, 4, 5, and 6 h, rapidly cooled in an ice-water bath, and immediately analyzed for anthocyanin content. The thermal degradation kinetics were fitted to a first-order model, and the half-life (t_1/2) was calculated by the following formulas.

$$\mathrm{l}\mathrm{n}{(C}_t/{\mathrm{C}}_0)=-kt$$

(11)

$$t_{1/2}={\displaystyle \frac{\mathrm{l}\mathrm{n}2}{k}}$$

(12)

where $C_t$ is anthocyanin concentration at time “t” (mg/L), ${\mathrm{C}}_0$ is initial anthocyanin concentration (mg/L), k is first-order rate constant (h⁻¹), t is time (h), and $t_{1/2}$ is half-life (h).

2.9. Statistical Analysis

One-way ANOVA with Tukey’s post hoc test was conducted to determine significant differences at p < 0.05 (95% confidence level), using SPSS Statistics 26.0 (IBM, Chicago, IL, USA). Data were expressed as mean ± standard deviation (SD) of three replicate determinations.

3. Results and Discussion

3.1. Characterization of ‘France’ Prunus domestica L. Peel Extracts

3.1.1. Effects of Solvents on Bioactive Components and Antioxidant Capacity of ‘France’ Prunus domestica L. Peel Extracts

Solvent extraction remains the conventional method for anthocyanins extraction. Anthocyanins exhibit different chemical forms and colors in different pH solutions, existing as red flavylium cation when pH is less than 3, while converting to colorless carbinol pseudobase and chalcone at pH > 4 [18]. Acidic solvents (pH < 3) are commonly employed to enhance membrane permeability and maintain anthocyanins stability during extraction [34]. Aqueous-organic solvent systems generally have higher extraction efficiency than pure organic solvents due to the high polarity of water [22]. As shown in Figure 1A, methanol extraction yielded the highest total anthocyanins content (TAC) in ‘France’ Prunus domestica L. peel powder (288.98 ± 3.00 mg CGE 100 g⁻¹ DW) compared to ethanol and acetone, which is similar to the results of Boulekbache-Makhlouf et al. [35]. Comparative studies on Prunus species revealed substantial TAC variations: Myrobalan Plum (Prunus cerasifera Ehrh.) peels showed 193–1986 mg CGE 100 g⁻¹ FW [36], while Sanhua plum (Prunus salicina L.) exhibited 689.5 ± 10.3 mg CGE 100 g⁻¹ DW [37]. These discrepancies likely arise from genetic differences and harvest maturity, as anthocyanin accumulation peaks at full ripeness [38]. Structurally, anthocyanins consist of aromatic rings with polar substituents (hydroxyl and methoxy groups) and glucosyl moieties [39], rendering them more soluble in highly polar solvents. Methanol’s superior performance aligns with its higher polarity (dielectric constant ε = 32.7) compared to ethanol (ε = 24.3) and acetone (ε = 20.7), facilitating the dissolution of anthocyanin glycosides through enhanced hydrogen bonding.

No significant differences (p > 0.05) were observed in total phenolic content (TPC) across solvents extractions (Figure 1B), likely due to the broad solubility of phenolics in polar media. However, acetone extraction achieved the highest total flavonoid content (TFC, 16.97 ± 0.03 mg RE g⁻¹ DW) (Figure 1C), which is similar to the results of Carmona-Hernandez et al. [40]. The conjugated double bonds in flavonoid molecules promote structural planarity, which hinders solvent penetration into the molecular interior, resulting in lower solubility in methanol and ethanol [41]. In contrast, acetone optimally disrupts the cell wall matrix and dissolve phospholipids of the cell membrane, facilitating efficient flavonoid extraction.

Acetone extracts demonstrated the strongest FRAP (Figure 1D), correlating with their highest TPC (14.01 ± 0.39 mg GAE g⁻¹ DW) and TFC, which is consistent with the results of Wijekoon et al. [42]. This aligns with reports that acetone’s low viscosity (0.32 cP) enhances mass transfer efficiency, increasing bioactive compound yield [42]. The synergistic effect between flavonoids and phenolic may further amplify antioxidant potential. Ngolo et al. [43] reported a synergistic antioxidant activity (CI = 0.57) resulting from the combination of T. riparia (with the highest phenolic content) and O. gratissimum (with the highest flavonoid content).

3.1.2. Effects of Extraction Solvents on Color Parameters

The color parameters of solvent extracts from ‘France’ Prunus domestica L. peels are summarized in

Table 1. Effect of different extraction solvents on the color parameters of ‘France’ Prunus domestica L. peel extracts.

Solvents	L*	a*	b*	C*	H* (°)
80% Methanol	41.26 ± 0.25 a	13.62 ± 0.28 a	3.98 ± 0.22 a	14.19 ± 0.33 a	16.29 ± 0.55 a
80% Ethanol	40.97 ± 0.07 a	12.89 ± 0.16 b	3.53 ± 0.10 b	13.37 ± 0.18 b	15.32 ± 0.28 b
70% Acetone	41.10 ± 0.02 a	12.67 ± 0.03 b	3.54 ± 0.01 b	13.15 ± 0.03 b	15.59 ± 0.07 ab

The results are presented as mean ± standard deviations (n = 3). Different superscript letters in the same column indicate significant differences (p < 0.05).

. A chroma (C*) value slightly above 0 represents low saturation, while C* values between 70 and 90 indicate high saturation, meaning more vivid colors. Hue angle H*: 0° or 360° = red, 90° = yellow, 180° = green, and 270° = blue [44]. The C* values for all extracts were between 13 and 15, suggesting low color saturation. Their H* values were close to 0°, indicating a dominant red color. The a* and b* values of three extracts were all positive, indicating dominant red–yellow hues. The a* value of methanol extracts (13.62 ± 0.28) was significantly higher than ethanol (12.89 ± 0.16) and acetone (12.67 ± 0.03) (p < 0.05), its superior redness corresponded with the highest TAC.

Given the study’s focus on anthocyanin extraction, 80% (v/v) methanol was selected as the optimal solvent, delivering the highest TAC (288.98 ± 3.00 mg CGE 100 g⁻¹ DW) and ideal color parameters (highest a* value). Although acetone extracts showed slightly higher FRAP activity compared to methanol extracts (24.85 ± 0.95 vs. 23.46 ± 0.37 mg TE/g DW), its 17.98% lower TAC and inferior redness disqualified it for applications requiring anthocyanin-rich formulations. The selected solvent was subsequently applied in ultrasound- and microwave-assisted extraction.

3.1.3. Effects of Assisted Methods on Bioactive Components and Antioxidant Capacity of ‘France’ Prunus domestica L. Peel Extracts

Conventional solvent extraction suffers from prolonged duration, low yield, and potential degradation of bioactive compounds [45]. In recent years, rapid and efficient techniques are commonly used for anthocyanin extraction [46]. Ultrasound is a form of mechanical vibration with frequencies exceeding 20 kHz that enhances extraction by accelerating solvent penetration and mass transfer. Microwave energy functions by inducing the movement of molecules with permanent dipole moments, enabling rapid energy absorption [47]. As shown in Figure 2A, ultrasound-assisted extraction (UAE) achieved the highest TAC (765.41 ± 3.72 mg CGE 100 g⁻¹ DW), surpassing microwave-assisted extraction (MAE, 732.09 ± 8.37 mg CGE 100 g⁻¹ DW) and conventional method (694.67 ± 11.84 mg CGE 100 g⁻¹ DW). The 10.2% TAC improvement by UAE originates from cavitation-induced cell wall disruption.

No significant differences (p > 0.05) were observed in TPC (Figure 2B) and TFC (Figure 2C) across extraction methods. This suggests that 80% methanol effectively dissolves phenolics under conventional conditions, while UAE/MAE primarily enhance anthocyanin release without altering phenolic solubility equilibria. MAE’s dielectric heating may even degrade some heat-sensitive compounds, offsetting its mass transfer advantages.

Ultrasonic shear forces and microwave dielectric heating can enhance the dissolution of bioactive compounds but may simultaneously cause molecular fragmentation, thereby partially compromising their antioxidant capacity. The antioxidant capacity results (Figure 2D) showed negligible variation among methods (p > 0.05), aligning with TPC/TFC data. This implies that anthocyanins contribute less to total antioxidant activity compared to phenolics. The study by Flores et al. [48] demonstrated that the correlation coefficient between FRAP and TAC was 0.717, while the correlation between FRAP and TPC was 0.947.

3.1.4. Effects of Assisted Methods on Color Parameters

The color difference (ΔE*) of the UAE and MAE samples was calculated against the conventional extracts (control). ΔE* can be classified into three categories. ∆E* < 1.5: minor color difference; 1.5 < ∆E* < 3: noticeable color difference; and ΔE* > 3: significant color difference [49]. As shown in

Table 2. Effect of assistant methods on the color parameters of ‘France’ Prunus domestica L. peel extracts.

Methods	L*	a*	b*	ΔE*
Control (60 min)	39.22 ± 0.01 a	6.06 ± 0.27 a	0.60 ± 0.12 a	-
Ultrasound-assisted (20 min)	38.82 ± 0.12 b	5.03 ± 0.04 b	0.27 ± 0.01 b	1.16 ± 0.01 a
Microwave-assisted (5 min)	38.93 ± 0.14 ab	5.35 ± 0.05 b	0.35 ± 0.04 b	0.81 ± 0.01 b

The results are presented as mean ± standard deviations (n = 3). Different superscript letters in the same column indicate significant differences (p < 0.05).

, UAE and MAE exhibited significant differences (p < 0.05) in a* and b* values compared to the control, yet their ΔE* achieved 1.16 ± 0.01 and 0.81 ± 0.01, respectively, indicating minor differences.

While TPC, TFC, and antioxidant capacities showed no significant differences (p > 0.05), UAE delivered the highest TAC (765.41 vs. 694.67 mg CGE 100 g⁻¹ DW) with 66.7% shorter extraction time (20 vs. 60 min) and preserved color fidelity (ΔE* < 1.5) compared to conventional methods. The improvement in efficiency confirmed UAE’s superiority for scaled production, meeting the demands for efficiency, yield, and appearance quality.

3.2. Characterization of Untreated and Ultrasonicated CNCs

3.2.1. Morphological and Interfacial Properties of CNCs

CNCs have been widely employed as nanoscale stabilizers for emulsions. Ginkgo seed shells, which are typically incinerated or buried and cause environmental damage and resource waste, can serve as an excellent source for CNCs extraction [30]. CNFs are obtained through sequential alkali treatment and bleaching to remove lignin, hemicellulose, and pigments. Subsequent acid hydrolysis, dialysis, and ultrasonication yield highly crystalline regions of cellulose, known as CNCs [50,51]. Meirelles et al. [50] demonstrated that ultrasonication (20 kHz, 675 W) reduced CNCs particle size from 400.13 ± 17.38 nm to 326.88 ± 26.26 nm while narrowing polydispersity index (PDI) from 0.73 to 0.29, producing nanoscale cellulose particles with unimodal distribution. Smaller nanocrystals are more conducive to adsorbing at interfaces and stabilizing emulsions.

The morphology and properties of cellulose particles critically affect their performance as emulsion stabilizers. This study subjected CNCs, obtained through acid hydrolysis and dialysis, to ultrasonication. Atomic force microscopy (AFM) images (Figure 3A,B) confirmed the fragmentation of original CNCs aggregates (≈11 μm) into discrete rod-shaped nanoparticles (≈340 nm) (Figure 3D). This morphological transformation is attributed to acoustic cavitation effects, where violent bubble collapse generates microjets and shockwaves that fracture suspended cellulose particles. The disruptive forces overcome interfacial cohesion between cellulose fibers, which is primarily governed by van der Waals forces and hydrogen bonding [52]. Ultrasound energy effectively disintegrated micron-scale aggregates into nanocrystals [53]. Ultrasonicated CNCs (Ultra CNCs) reduced the oil–water interfacial tension from 21.01 mN/m to 19.4 mN/m (Figure 3C). The ζ-potential varied from −40.97 ± 1.30 mV to −44.74 ± 1.90 mV (Figure 3E), indicating increased exposure of negatively charged sulfonate groups (-OSO₃⁻) due to cavitation-induced surface erosion. Water contact angle measurements revealed increased hydrophobicity, with values rising from 31.02 ± 1.72° to 59.29 ± 0.69° (Figure 3F). This result correlates with the exposure of hydrophobic (200)β crystallographic planes and cavitation-induced partial crystal disordering [54]. Ultra CNCs exhibited shortened dimensions and enhanced hydrophobicity, improving their adsorption kinetics at oil–water interfaces, thereby reducing interfacial tension more efficiently than original CNCs.

3.2.2. Molecular Characterization of CNCs

The crystal structures of CNCs before and after ultrasonication were evaluated using X-ray diffraction (XRD). As shown in Figure 3G, all diffraction patterns consisted of a broad amorphous halo and a distinct crystalline peak. Cellulose possesses a crystalline structure due to intermolecular hydrogen bonds and van der Waals forces, whereas hemicellulose and lignin are amorphous. The XRD pattern of cellulose from ginkgo seed shells exhibited two prominent peaks at 2θ = 15.6° and 22.0°, along with a minor peak at 34.5°, which were assigned to be cellulose I [55]. The peaks at 15.6°, 22.0°, and 34.5°, corresponded to (101), (002), and (023)/(004) planes, respectively [55]. Ultrasonication may have removed part of the amorphous regions, resulting in an increase in the crystallinity of CNCs from 69.56% to 72.98%.

The FTIR spectra of CNCs in the range of 4000–400 cm⁻¹ before and after ultrasonication are shown in Figure 3H. The strong and broad absorption peak in the 3500–3000 cm⁻¹ range was attributed to the O-H stretching vibration. The characteristic peak near 2900 cm⁻¹ was assigned to the C-H stretching vibration of saturated hydrocarbons. The peak near 1715 cm⁻¹ likely corresponded to the C=O stretching vibration from carbonyl, ester, and acetyl groups in the xylan component of residual hemicellulose and lignin [30]. The absorption peak near 1640 cm⁻¹ was assigned to the O-H bending vibration, though it may also be attributed to the C=O stretching vibration of carboxyl groups introduced during the purification process. The absorption peak at 1161 cm⁻¹ was likely the result of the stretching vibration of the pyranose ring (C-O-C), while the peak near 1030 cm⁻¹ was attributed to skeletal vibrations of bonds in the C-C ring [56]. The small peak near 900 cm⁻¹ was typical of the asymmetric valence vibration of the β-glycosidic bond [56]. In the spectrum of Ultra CNCs, the peaks near 3340 cm⁻¹ and 1030 cm⁻¹ were more intense compared to CNCs, reflecting an increase in hydrogen bonds and the bonds in C-C ring. In general, no significant differences were observed between the two spectra.

As shown in Figure 3I, the wide-scan XPS spectra of CNCs and Ultra CNCs (Figure 3I) displayed the characteristic C 1s and O 1s peaks. In the high-resolution C 1s spectra (Figure 3J,K), the carbon signals were resolved into several component peaks, which were corresponded with different local chemical environments of carbon atoms, including C-C, C-O, and C=O bonds [57]. There were no significant changes in the surface elemental composition or molecular structure following ultrasonication. Overall, ultrasonication treatment primarily induced physical change rather than chemical modification.

3.3. Characterization of W₁/O Emulsions and W₁/O/W₂ Emuslions

3.3.1. Optimization of the Emulsifier Concentration of the W₁/O Emulsions

Surfactants with low hydrophilic–lipophilic balance (HLB < 6) preferentially stabilize water-in-oil (W/O) emulsions due to their oil-soluble nature, while surfactants with high HLB (>8) favor the formation of O/W emulsions. Examples of low HLB surfactants are polyglycerol polyricinoleate (PGPR), fatty acid esters, and sorbitan fatty acid esters [58]. As a non-ionic surfactant, PGPR adsorbs at the oil–water interface and resists droplet coalescence through steric repulsion [58]. The stability of W₁/O emulsions correlates inversely with droplet size. Larger droplets accelerate flocculation and coalescence, while smaller droplets with narrow polydispersity (PDI) can resist Ostwald ripening [13]. The equilibrium interfacial tension at the oil–water interface decreases with increasing PGPR concentration [59]. In the W₁/O/W₂ double emulsions, stable W₁/O emulsions prevent anthocyanins diffusion from the inner aqueous phase to the outer aqueous phase; thus, the amount of emulsifier PGPR was optimized. As shown in Figure 4, fresh W₁/O emulsions (containing 3%, 4%, and 5% of PGPR, w/w, respectively) initially displayed homogeneous appearances without phase separation or creaming. After seven days of storage at 4 °C, phase separation occurred in the W₁/O emulsions containing 3% PGPR, while the homogeneity of the 4% PGPR-stabilized W₁/O emulsions decreased. Optical microscopy and CLSM images revealed that the droplet size in W₁/O emulsions increased as PGPR concentration decreased. Larger droplets are more prone to collision in the dispersed phase due to increased gravitational effects and reduced collision distances within the confined container space, ultimately leading to droplets flocculation and coalescence. The mean droplet size of W₁/O emulsions stabilized by 3% PGPR increased from 2216 nm to 2296 nm after seven days of storage at 4 °C. For 4% PGPR, a significant increase (1062 nm to 1622 nm) was observed. However, 5% PGPR-stabilized emulsions remained stable and exhibited uniform particle size distribution (662 nm varied to 668 nm) during storage. Therefore, 5% was subsequently used for emulsion preparation.

3.3.2. Characterization of W₁/O/W₂ Emulsions with Different Ratio of Internal Phase (W₁/O)

Proteins with distinctive surface activity can be spontaneously adsorbed at the oil–water interfaces, while polysaccharides primarily act as thickeners or stabilizers that delay phase separation and gravity-induced creaming by controlling the rheological properties and network structure of the continuous phase [60]. Protein–polysaccharide complexes resist droplet coalescence through steric hindrance and electrostatic repulsion, effectively reducing interfacial tension and improving emulsion stability. Whey protein isolate (WPI) is widely used as an emulsifying stabilizer in the food industry due to its superior emulsifying capacity [61]. Cellulose nanocrystals (CNCs) can irreversibly assemble and adsorb at oil–water interfaces to stabilize emulsions due to their amphiphilic properties [30].

This study utilized a complex of CNCs/Ultra CNCs and WPI to stabilize W₁/O/W₂ emulsions. The emulsion stability was investigated to compare the emulsifying capacity of CNC and Ultra CNCs. The internal phase (W₁/O) ratio was increased to enhance anthocyanin loading capacity. However, the freshly prepared emulsion exhibited phase separation immediately when the ratio reached 45% (w/w). Therefore, the study focused on double emulsions with internal phase ratios ranging from 20% to 40%. According to Figure 5, higher W₁/O emulsions ratios resulted in darker emulsion colors that manifested as increased a* values, which correlated with elevated anthocyanins content.

The final W₁/O/W₂ emulsions had a pH of 3.3, avoiding WPI’s isoelectric point (pH 4.5–5.5) and stabilizing anthocyanins. At this pH, WPI carries a positive charge, while sulfuric acid–hydrolyzed cellulose bears sulfonic acid groups, conferring a negative charge. Double emulsions stabilized with WPI–CNCs complex were positively charged due to the electrostatic complexation. A higher absolute ζ-potential indicates stronger electrostatic stabilization in the emulsions, preventing droplet aggregation. The 100-fold dilution with deionized water reduced ionic strength and lowered measured ζ-potential magnitudes. Notably, the nanocrystal properties directly influenced emulsion electrostatic performance. As the W₁/O emulsions ratio increased from 20% to 40% (w/w)—meaning a decrease in the W₂ phase ratio and a reduction in WPI–CNCs content—the ζ-potential of the W₁/O/W₂ emulsions decreased from 16.68 ± 1.26 mV to 7.49 ± 0.99 mV (Figure 6B), indicating reduced electrostatic repulsion and lower emulsion stability. In contrast, W₁/O/W₂ emulsions stabilized with WPI–Ultra CNCs exhibited significantly higher ζ-potentials due to the smaller nanoparticle size and higher surface charge density of Ultra CNCs. The ζ-potentials of double emulsions containing 20–40% (w/w) W₁/O emulsions increased from 16.68 ± 1.26 mV, 12.52 ± 1.44 mV, 10.66 ± 0.20 mV, 8.97 ± 1.96 mV, and 7.49 ± 0.99 mV to 17.96 ± 1.39 mV, 17.19 ± 1.92 mV, 13.23 ± 1.27 mV, 12.76 ± 1.49 mV, and 10.07 ± 1.50 mV, respectively (Figure 6B,E). This demonstrated that the nanoscale dimensions and interfacial properties of CNCs are critical in determining the electrostatic and stabilizing performance of the emulsion system.

WPI–CNCs-stabilized double emulsions exhibited a bimodal size distribution (Figure 6C): a dominant peak at 1–10 µm and a minor peak at 10–100 µm. As the W₁/O ratio increased (20–35% w/w), the dominant peak shifted toward larger sizes, with median particle sizes (d₅₀) of 2.84, 3.46, 4.12, and 4.31 µm (Figure 7). This trend was attributed to the reduced interfacial coverage by CNC-based emulsifiers. Higher oil droplet concentrations in the dispersed phase significantly increased collision frequency, accelerating coalescence. In the double emulsions containing 40% (w/w) W₁/O, simultaneous presence of oil droplets lacking aqueous cores and oil droplets containing internal aqueous phase led to a small d₅₀ of 2.95 µm but a high SPAN of 7.08 (Figure 7), indicating poor dispersion homogeneity. Owing to their optimized nanoscale dimensions and higher surface activity, WPI–Ultra CNCs-stabilized double emulsions had significantly smaller particle sizes, with d₅₀ values of 2.07, 2.93, 3.04, 3.27, and 3.39 µm for 20–40% W₁/O ratios (Figure 8).

The leakage of anthocyanins from the W₁ to the W₂ phase may be attributed to the following pathways: (a) rapid coalescence of large oil droplets; (b) coalescence of small internal water droplets; (c) lamellar transport from the internal to the external aqueous phase; (d) osmotic pressure gradient-induced transport of the internal aqueous phase to the external one, leading to the shrinkage of oil droplets, or vice versa; and (e) reverse micellar transport of water and water-soluble molecules [62]. EE, defined as the ability of the double emulsions to retain anthocyanins within the inner aqueous phase (W₁), serves as a critical parameter for evaluating emulsions integrity [63]. Untreated CNCs with micrometric lengths (11,434 ± 276 nm) exhibited limited interfacial adsorption capacity. When the oil phase proportion increased, CNCs struggled to stabilize the oil–water interface, which likely led to the leakage of anthocyanins into the W₂ phase via pathway (a) due to insufficient emulsifier coverage. The WPI–CNCs-stabilized double emulsions containing 40% W₁/O displayed a heterogeneous droplet distribution (SPAN = 7.08). This heterogeneity accelerated Ostwald ripening, driving progressive droplet enlargement, and potentially exacerbating leakage through pathway (d). Concurrently, reduced absolute ζ-potential values weakened inter-droplet electrostatic repulsion. According to DLVO theory, van der Waals attractions dominated at shorter inter-droplet distances, promoting leakage through coalescence-related pathways. As shown in Figure 6A, the EE decreased from 90.24 ± 1.69% to 52.96 ± 2.60% as the W₁/O ratio increased from 20% to 40%. In contrast, using Ultra CNCs with nanoscale size and enhanced amphiphilicity as emulsifiers significantly improved emulsion performance, achieving 83.39 ± 0.96% EE for the 40% (w/w) W₁/O ratio (Figure 6D). The densely adsorbed rod-like CNCs at the interface formed a denser and more rigid film, effectively slowing anthocyanin leakage.

Optical microscopy and CLSM images confirmed the hierarchical structure of the double emulsions (Figure 7 and Figure 8). Internal water droplets were dispersed in the oil phase, which was surrounded by CNCs (not shown) and WPI, and further dispersed in the continuous external aqueous phase. CLSM using mixed channels (Nile red-stained oil phase: red; Nile blue-stained protein: blue) revealed that red oil droplet sizes gradually increased in WPI–CNCs-stabilized emulsions with higher W₁/O ratios, consistent with droplet size results. According to Stokes’ law, emulsion stability correlates with particle size. In addition, viscosity is an important factor in retarding creaming phenomenon in emulsions [64]. The study by Kyroglou et al. [65] demonstrated that both the flow consistency coefficient and apparent viscosity increased as the concentration rose. In this study, the increase in the W₁/O ratio led to a higher system concentration and an increase in overall viscosity. This increased viscosity slowed down the creaming kinetics, which explains why oil droplet aggregation was observed under microscopy without immediate macroscopic phase separation. For emulsions stabilized with CNCs, the 40% (w/w) W₁/O ratio group contained both small oil droplets and large oil droplets encapsulating internal aqueous phase, indicating partial inner droplets leakage into W₂ phase, which was consistent with the reduced EE. This suggested that the interfacial coverage was likely insufficient at high W₁/O ratio, ultimately leading to the breakdown of the interfacial film. In contrast, the emulsions stabilized by Ultra CNCs benefit not only from a smaller initial droplet size distribution and potentially stronger interfacial film because of the nanoscale size and optimized interfacial properties of Ultra CNCs, but also from higher viscosity and more favorable flow behavior. This allowed them to maintain effective dispersion uniformity across the 20–40% internal phase ratio range, sustaining a relatively low SPAN (2.00) even at the 40% ratio.

After seven days of dark storage at 25 °C, WPI–CNCs-stabilized double emulsions containing 40% (w/w) W₁/O showed visible phase separation at the bottom (Figure 9A). Macroemulsions are thermodynamically unstable, with common destabilization pathways including coalescence, creaming/sedimentation, and Ostwald ripening. Since molecular motion at 25 °C is faster than at 4 °C, emulsions are more prone to destabilization at 25 °C [32]. All double emulsions stabilized by WPI–Ultra CNCs remained stable without significant delamination during the seven day storage period (Figure 9B), demonstrating that the nano-engineered properties of Ultra CNCs significantly enhanced emulsification performance. The enhancement can be attributed to the modification of CNCs: ultrasonication reduced CNCs lengths, thereby enhancing electrostatic and steric hindrance effects. This improved CNC adsorption at oil–water interfaces, reduced interfacial tension, increased interfacial coverage, and enhanced emulsion stability while reducing anthocyanin leakage.

Subsequently, to achieve a higher anthocyanin load, W₁/O/W₂ emulsions with a 40% (w/w) W₁/O ratio were prepared using WPI–Ultra CNCs as the emulsifier, and their storage stability was investigated. The results showed that after seven days of storage in the dark at 25 °C, 37 °C, and 50 °C, the ζ-potential of the emulsions changed from 22.86 ± 1.43 mV to 22.26 ± 1.73 mV, 20.28 ± 0.34 mV, and 13.03 ± 0.61 mV, respectively (Figure S1A). The significant decrease in ζ-potential at 50 °C indicated a weakening of electrostatic repulsion between droplets, leading to reduced emulsion stability. Concurrently, the mean particle size increased from 3.39 ± 0.37 µm to 4.13 ± 0.19 µm, 5.12 ± 0.14 µm, and 6.37 ± 0.12 µm, respectively (Figure S1B). The increase in particle size was more pronounced at higher temperatures, indicating that droplet coalescence occurred over time and was accelerated by heat. The anthocyanin retention rate decreased from 100% to 97.10 ± 2.41%, 81.28 ± 1.15%, and 21.44 ± 2.18%, respectively (Figure S1C), suggesting potential leakage of anthocyanins from the W₁ to the W₂ phase and their chemical degradation. As shown in Figure S1D, the emulsions stored at 25 °C and 37 °C remained stable throughout the storage period. In contrast, phase separation began in the 50 °C group after three days, with the height of the lower aqueous phase gradually increasing over time. Crucially, even after seven days of storage at 37 °C, the emulsions showed no significant changes in macroscopic appearance, droplet size, or ζ-potential, while maintaining a high anthocyanin retention rate of 81.28 ± 1.15%. This demonstrated that the system possessed robust physical stability and effectively protected the encapsulated bioactive compounds from degradation.

In contrast to conventional emulsions stabilized by synthetic surfactants, the protein-cellulose nanocrystal complex forms a solid-like layer at the interface, granting the emulsion superior resistance against Ostwald ripening and coalescence. The W₁/O/W₂ emulsions stabilized with ultrasonicated CNCs demonstrated superior performance in this study. Even at 40% (w/w) internal phase ratio, the W₁/O/W₂ emulsions maintained a small droplet size (D₄₃) of 4.35 µm and remained stable after seven days of storage at 37 °C. However, the Tween 80-stabilized system reported by Fu et al. [66] exhibited a much larger D₄₃ of 24.9 µm (with a 30% internal phase ratio), and its creaming index (CI) decreased from 100% to 83.2% after just 10 days at 25 °C. Furthermore, our system achieved a high anthocyanin encapsulation efficiency of 83.39 ± 0.96%, surpassing the encapsulation efficiency of tartrazine (approximately 40%) reported in Tween 80-stabilized W₁/O/W₂ emulsions by Tamnak et al. [62]. From a safety and sustainability perspective, synthetic surfactants like Tween 80 raise potential toxicity concerns, whereas the cellulose nanocrystals derived from ginkgo seed shells offer excellent biocompatibility and biodegradability, aligning with the modern food industry’s demand of clean-label ingredients.

3.4. Characterization of Anthocyanin Thermal Stability

Anthocyanin degradation during thermal processing involves multiple mechanisms—including glycosidic bond cleavage, nucleophilic attack by water—leading to significant bioactive loss. For double emulsion-encapsulated anthocyanins (E-ACNs), the nanoscale dimensions and high aspect ratio of Ultra CNCs increased the viscosity of the continuous phase, decelerating water molecule diffusion, while the oil phase, complemented by the interfacial Ultra CNCs layer, provided an enhanced thermal barrier (Figure 10).

As shown in Figure 11A,B, ‘France’ Prunus domestica L. anthocyanin degradation followed first-order kinetics across 50–90 °C, consistent with Slavu et al. [33]. The degradation rate constant (k) increased with temperature, peaking at 90 °C with the shortest half-life (t_1/2) (Figure 11C). Crucially, E-ACNs exhibited markedly lower degradation rates than free-forms (F-ACNs) at all tested temperatures. At 50 °C, k decreased from 0.05045 h⁻¹ (F-ACNs) to 0.01018 h⁻¹ (E-ACNs), extending t_1/2 from 14.72 ± 0.35 h to 70.37 ± 0.51 h (4.8-fold enhancement). Even at 80 °C, E-ACNs maintained a 1.6-fold longer half-life (5.82 ± 0.15 h vs. 3.67 ± 0.45 h for F-ACNs). In the study of Chen et al. [67], the anthocyanin retention rate in double emulsions was 69.96% after being heated at 50 °C for 30 min, whereas it remained at 94.03% ± 0.69% even after 6 h at 50 °C in this study. This exceptional thermal stability can be attributed to the stable multi-protective structure of the W₁/O/W₂ emulsions. Primarily, with anthocyanins encapsulated within the W₁ phase, the surrounding oil phase acted as a physical thermal barrier, directly impeding the inward conduction of external heat. More critically, the oil–water interface stabilized by Ultra CNCs formed a compact interfacial film, which effectively retarded the diffusion of water molecules, thus suppressing key degradation pathways such as anthocyanin hydration and glycosidic bond hydrolysis. Additionally, this stabilization can minimize bioactive integrity loss during pasteurization.

This study demonstrates the considerable potential of Ultra CNC-stabilized W₁/O/W₂ emulsions for practical applications in the food industry. The extended half-life of anthocyanins in the emulsion system indicates its potential for developing thermally stable functional foods (e.g., fortified dairy products, meal replacement shakes) that can maintain their health-promoting properties and natural color during processing and storage.

4. Conclusions

This study developed an acidic W₁/O/W₂ emulsion system for anthocyanin encapsulation, centered on nanoparticle stabilization. The findings demonstrated that the highest TAC of ‘France’ Prunus domestica L. peel was obtained using 80% (v/v) methanol as the solvent, combined with ultrasonic extraction. Crucially, ultrasonic treatment induced a significant nanostructural transformation of CNCs, effectively enhancing their hydrophobicity and interfacial activity. This improvement was fundamental to forming a stable W₁/O/W₂ emulsion with markedly reduced droplet size, higher ζ-potential, and superior encapsulation efficiency compared to emulsions stabilized with untreated CNCs. It is noteworthy that the W₁/O/W₂ emulsion stabilized by Ultra CNCs remained stable after seven days of storage at 37 °C and effectively prevented the degradation of anthocyanins. Furthermore, the emulsion system provided exceptional protection to anthocyanins at high temperatures compared to the free-form, drastically extending their thermal stability. The robust encapsulation capability offers a promising strategy for designing targeted nutrient delivery systems, potentially enhancing the bioavailability of bioactive compounds. In addition, the use of biomass-derived CNCs aligns with consumer demand for clean-label ingredients. Future work may focus on developing advanced delivery systems stabilized by modified CNCs, with applications in the food and pharmaceutical industries.