Ultrasound-driven ternary conjugation reshapes protein-lipid interfaces and volatile profiles in high-protein egg milk systems

Graphical abstract

Abstract

Heat-induced aggregation, sulfurous off-flavors, and oxidative instability pose challenges for the use of egg white proteins (EWPs) in high-protein beverages. In this study, a ternary conjugated system comprising EWP, milk casein, and xylitol (EMX) was incorporated into skimmed milk powder (SMP) and whole-fat milk powder (WFM) beverages, which were subsequently treated with ultrasound (US; 20 kHz, 300 W). Physicochemical properties, oxidative stability, volatile compounds, in vitro digestion, and storage stability over 30 days at 4 °C were monitored. The results showed that EMX conjugated system exhibited higher solubility (91.04 % vs. 88.54 %) and a 2.4-fold reduction in turbidity compared with skimmed beverages. Ultrasound further decreased particle size and increased molecular flexibility in whole-fat formulations. Additionally, EMX showed the greatest thermal stability (Td 114.17 °C; ΔHd 179.3 J/g), while US treatment improved Td from 102.21 °C to 113.23 °C in SMP. US-treated whole-fat beverages demonstrated lower TBARS values during days 0–14 but higher values than their untreated counterparts on day 30. Protein carbonyl levels remained lowest in US-treated skimmed beverages throughout storage. Altogether, the US increased digestibility, while mitigating dipropyl disulfide and 2-furanmethanethiol sulfurous off-flavors. In conclusion, US-assisted EMX ternary conjugation offers a promising approach for developing the functionality and sensory qualities of protein-enriched dairy model systems.

1. Introduction

The growing global interest in high-protein beverages is spurring efforts to combine complementary protein sources to enhance nutritional and functional quality. Among these, egg white proteins (EWPs) and milk proteins offer a promising synergy for protein-rich beverages due to their high nutritive value and distinct functional properties [1]. Milk proteins (80 % casein, 20 % whey) are well-known for their complete amino acids profile and digestibility [2], [3]. EWPs, comprising ovalbumin, ovotransferrin, ovomucoid, and lysozyme, face several challenges [4]. These proteins exhibit low solubility and readily aggregate or form gels upon heat treatment like pasteurization [5]. Additionally, thermal processing generates sulfur-containing volatiles that impart undesirable “eggy” off-flavors. These stability and sensory issues severely limit the use of high levels of egg white in beverages, underscoring the need for strategies to improve thermal stability, solubility, and organoleptic acceptability of egg milk protein systems [6], [7].

Dual protein interactions have recently gained attention for enhancing functional properties and thermophysical stability. In this respect, Salama et al. [8] incorporated EWPs and milk casein (MCN) under different environmental conditions, including pH, mixing ratio, and ionic strength, to overcome phase transitions and sedimentation. Analogously, the heat stability of hemp protein was enhanced by the chaperone-like action and steric hindrance of MCN, which prevented protein aggregation during heating [9]. EWPs and milk protein complexation were also used to modulate the functionality and organoleptic properties of yogurt [10]. Thus, dual protein interactions are crucial for improving functionality and thermal stability. However, EWPs-MCN concentrations above 3.0% induce heat-driven aggregation, limiting their use in protein-rich beverages [11].

Altogether, ternary conjugated systems with a third component like saccharides optimize protein-based hydrocolloids, enhancing functionality, thermal stability, and flavor masking beyond what binary systems achieve. Xylitol (Xyl), a low-calorie sugar alcohol, has emerged as a promising alternative to sucrose, which is linked to health concerns such as obesity, diabetes, and cardiovascular disease. In addition to its sweetening function, Xyl also enhanced thermal stability and alleviated off-flavors in egg protein-based beverages through a preferential hydration/exclusion mechanism [12], [13], [14]. Trehalose, a disaccharide, helped conjugate whey and lentil proteins to form a complex with improved properties [15]. This function caused structural changes, boosting solubility and digestibility. Besides, conjugating Xyl with an EWP-MCN complex via high-intensity ultrasound (US) enhanced the thermal resistance and function of proteins [11]. These reports suggested that the third component can play multiple roles, including covalent or non-covalent binding to proteins, altering their conformation, providing steric stabilization, or even scavenging reactive compounds.

Despite advancements in dual protein systems and ternary conjugates, there is a significant knowledge gap regarding their application in real high-protein beverages. Previous research demonstrating improvements in solubility, functional properties, and heat stability primarily used simplified models or isolated protein solutions under controlled conditions [16], [17], [18]. However, these systems do not fully represent real food matrices containing proteins, lipids, and minerals. There is insufficient data on the performance of ternary conjugates in real complex food matrices, such as egg milk drinks containing both skimmed and whole milk, which undergo thermal processing and need to remain shelf-stable. In particular, its structural stability and aggregation behavior under processing conditions remain unclear. Interestingly, this ternary system could be enhanced by US treatment (300 W, 20 kHz), which has been reported to modify volatile compounds in thermally processed EWPs, including reductions in fishy and sulfur related flavor fingerprints such as trimethylamine and 3-butenenitrile [14]. In protein hydrocolloid systems, Kong et al. [19] and Zhang et al. [20] noted that US treatment of soy proteins broke apart intermolecular bonds (e.g. disrupting disulfide linkages and hydrophobic clusters), thereby raising the denaturation temperature and thermal stability of the protein solution. In a recent study, ultrasonication of egg protein mixtures was reported to decrease the perception of sulfurous aromas and increase fruity and sweet notes. This effect was attributed to structural modifications and radical-mediated reactions [14]. In particular, integrating a sugar alcohol such as Xyl into a protein conjugate via US is a novel approach for addressing Maillard reaction-related issues [11].

This study developed a ternary conjugate of EWP, MCN, and Xyl via high-intensity US and integrated it into high-protein milk beverages. The research evaluated how the EWP-MCN-Xyl (EMX) conjugate affected protein solubility to prevent sedimentation, enhanced thermal and oxidative stability to resist aggregation, and influenced protein-lipid interfacial behavior with milk fat. Additionally, it investigated aroma profiles and off-flavor modulation, with a focus on sulfur-containing volatiles in egg milk model systems. This innovative system offers a promising approach to addressing stability and off-flavor issues in protein-rich beverages. It also provides a strong scientific foundation for developing future protein beverages with improved functional performance, nutritional quality, and consumer appeal.

2. Materials and methods

2.1. Chemicals and materials

Xylitol, potassium bromide (KBr), and Nile red (NR) were purchased from BioFroxx-Sai Guo Biotechnology Co., Ltd. (Guangzhou, China). Casein with a purity of 92% was provided by Yuanye Bio-Technology Co. (Shanghai, China). Skimmed and whole-fat milk powders were purchased from JBX Pty Ltd. (Maxigenes® brand, Sydney, Australia). In addition, 2-thiobarbituric acid (TBA), 2,4-dinitrophenylhydrazine (DNPH), guanidine hydrochloride (GuHCl), fluorescein isothiocyanate isomer I (FITC), O-phthalaldehyde (OPA), and 1,1,3,3-tetraethoxypropane (TEP) were obtained from Sigma-Aldrich (St. Louis, MO, USA). Trypsin (activity 1:10000), Pepsin (activity 1:12000), and pancreatic (activity 1:8000) enzymes, as well as bile salts, were purchased from Nanning Pangbo Biological Engineering Co., Ltd. (Nanning, China). All reagents and chemicals used were of analytical quality or HPLC grade. Fresh eggs were purchased from the Charoen Pokphand Foods Group Co., Ltd.​ (Wuhan, China).

2.2. Preparation of ternary conjugated system (EMX)

The ternary conjugated system was prepared in detail according to the study of Salama et al. [11], with slight modifications. In a word, EWP (8% w/v) was dispersed in deionized water, stirred at 25 °C for 4 h, then refrigerated overnight for complete hydration. Similarly, MCN (8% w/v) was solubilized in deionized water with the pH adjusted to 11.0. After that, the EWP solution was mixed with the MCN dispersion at a 2:1 ratio at pH 7.0. Xyl was then added at a concentration of 2% (w/v) to the mixture, and the pH of the ternary conjugated system formed was readjusted to 7.0 with gentle stirring (M 85-2, Sile Instrument, China) for 2 h.

2.3. Preparation of beverages

The EMX ternary conjugated system from Section 2.2 was used as the base formulation for beverage preparation. Skimmed milk powder (SMP) and whole-fat milk powder (WFM) were separately solubilized in deionized water to a protein concentration of 8% (w/v). Subsequently, the EMX system was combined with either SMP or WFM dispersions at a 1:1 volume ratio (EMX:SMP = 1:1 and EMX:WFM = 1:1), the pH was set to 7.0, and the mixtures were gently stirred for 1 h, as described by Lyu et al. [1], with some modifications. The resulting formulations were designated as SMP and WFM, respectively. The EMX ternary system without the addition of SMP or WFM served as the reference control. The mixtures were sonicated using an US probe (20 kHz, 300 W, SCIENTZ-IID, Ningbo, China) for 10 min, with 5 s on and 3 s off pulses. During ultrasonication, the sample container was continuously immersed in an ice-water bath to minimize the temperature increase caused by acoustic cavitation. After that, the samples were heated at 90 °C for 30 min in a controlled water bath (HH-4, Guohua Electric, China) and then cooled to room temperature. US-treated SMP and WFM formulations were denoted as U-SMP and U-WFM, respectively, while the corresponding untreated formulations were labeled SMP and WFM. After heating and cooling, each formulation was divided into two portions. Fresh liquid samples were used for colloidal, spectroscopic, microstructural, rheological, storage, oxidative stability, in vitro digestion, volatile profile, and sensory analyses. Separate aliquots were freeze-dried using an ALPHA1-2 LD Plus freeze dryer (Martin Christ GmbH, Germany) and utilized for XRD, FTIR, and DSC measurements.

2.4. Gross composition of beverages

The proximate composition of the samples, including moisture, protein, fat, carbohydrate, and ash contents, was analyzed using a FOSS scan analyzer (NIRS^TM DS2500 F NIR, Hillerød, Denmark). Before measurement, the instrument was calibrated using the reference model. Afterward, 50 g of the homogenized sample was loaded into the sample cup and scanned in reflectance mode over a wavelength range of 850-2500 nm using the beverage model. Data acquisition and analysis were carried out using Mosaic^TM software (FOSS Analytical, Denmark). The total solids content of each formulation was estimated as the sum of protein, fat, carbohydrate, and ash. All measurements were recorded in triplicate [21].

2.5. Solubility and turbidity measurements

The solubility and turbidity of the treatments were quantified according to Salama et al. [11], with minor changes. Samples were diluted to 0.5 mg/mL with 10 mM phosphate buffer (pH 7.0), mixed well, and centrifuged at 10,000 × g using a Sigma 3-30K centrifuge (California, USA) under cooling for 30 min at 4 °C. Afterward, the supernatant was collected, and soluble protein content was measured according to Lowry et al. [22]. The beverages’ solubility was determined computationally as follows:

$$\text{Solubility (\%) =}\frac{\text{Protein content in supernatant}}{\text{Protein content in original sample}}\times 100$$ (1)

Turbidity was measured at 600 nm using a Nanodrop 2000C UV/Vis spectrophotometer (Thermo Fisher Scientific, MA, USA). Phosphate buffer was used to set transmittance (TR) to 100%, and turbidity was expressed as 100 - TR%.

2.6. Molecular flexibility measurement

Proteins’ flexibility was determined based on the method of Ouyang et al. [23]. Briefly, 250 μL of trypsin (1 mg/mL) prepared in 50 mM Tris-HCl at pH 8.0 was mixed with 4 mL of the sample (1 mg/mL) and incubated at 38 °C for 5 min with mild shaking (THZ-82, Guohua Electric, China). Afterward, the enzymatic reaction was terminated by the addition of 4 mL of 5% (w/v) trichloroacetic acid (TCA). Following centrifugation at 5000 × g for 30 min at 4 °C, the resulting supernatant was carefully collected, and its absorbance was measured at 280 nm using a UV spectrophotometer (Thermo Fisher Scientific, MA, USA) in triplicate.

2.7. Particle size and zeta potential measurements

The size distribution and ζ-potential of the samples were analyzed by Nano Zetasizer (Malvern, Worcestershire, UK). Samples were placed in disposable cells for size analysis and in DTS 1070 cells for zeta measurements, both at a concentration of 0.1%. The temperature was set at 25 °C, with 60 s equilibration and 12 cycles per run. All measurements were performed three times.

2.8. Color measurement

The color of the samples was achieved using a colorimeter (CR-400, Minolta Co., Tokyo, Japan). It was evaluated using lightness (L*), red-green (a*), and yellow-blue (b*) values, with three repetitions for each [24].

2.9. Second-derivative UV spectroscopy measurement

Second-derivative UV spectra were employed to reveal detailed alterations in amino acid surroundings. Samples were diluted to 0.2 mg/mL in 10 mM phosphate buffer (pH 7.0) and scanned from 240 nm to 340 nm at 25 °C using a UV spectrophotometer (Thermo Fisher Scientific, MA, USA) with a 1 cm quartz cuvette.

2.10. Intrinsic fluorescence and synchronization measurements

Samples were made at a concentration of 0.05% (v/v) in phosphate buffer (10 mM, pH 7.0). Fluorescence measurements were done using a Shimadzu F-4600 spectrofluorometer (Tokyo, Japan) according to Miao et al. [25], with minor modifications. The excitation wavelength was adjusted at 280 nm and 295 nm. And the emission was recorded between 280 nm and 450 nm. The slit width for both was set to 5 nm, and the photomultiplier voltage was 400 V with a scanning speed of 1200 nm/min. For synchronized fluorescence, both excitation and emission wavelengths were scanned simultaneously, maintaining a constant wavelength difference (Δλ) = (λ_emission - λ_excitation) = 15 nm (signal sensitive to the tyrosine microenvironment) and 60 nm (signal sensitive to tryptophan and the hydrophobic microenvironment).

2.11. X-ray diffraction (XRD) measurement

X-ray diffraction was used to analyze the crystalline structure of the treatments, employing a Bruker D8 Advance diffractometer (Odelzhausen, Germany). The measurements were carried out at an operating voltage of 40 kV and a current of 40 mA, using Cu Kα radiation with a wavelength of 1.5406 Å. And scans were recorded over a scattering angle (2θ) range of 5°- 60° at a speed of 5°/min with a 0.02° step size [26].

2.12. FTIR spectroscopy measurement

According to the methodology of Salama et al. [11], powdered samples were mixed with KBr at a 1:100 ratio, ground, and pressed into transparent tablets. FTIR spectra were recorded using a spectrophotometer (IS10, Nicolet, USA) over the range 4000-500 cm⁻¹, with 64 scans per sample at 4 cm⁻¹ resolution. The spectral data collected from three replicates were analyzed using OMNIC software (V8.0, Thermo, USA).

2.13. Confocal laser scanning microscope (CLSM) measurement

The microstructure and distribution of proteins and lipids in beverage samples were examined using a FV12000MPE CLSM (Olympus Corporation, Tokyo, Japan) following Li et al. [27], with adjustments. Proteins were stained with FITC, while lipids were stained with NR. A stock solution of FITC (1 mg/mL) and NR (0.5 mg/mL) was prepared in ethanol at a 1:1 ratio. Next, 20 μL of dye was added to the samples (1 mL), and the mixture was placed in the dark for 30 min. After staining, 10 µL of the sample was placed on a glass slide, covered with a coverslip, and imaged using an oil-immersion objective lens at 100× magnification. Excitation wavelengths were set at 488 nm for FITC and 543 nm for NR.

2.14. Differential scanning calorimetry (DSC) measurement

DSC (CC 300, Netzsch, Germany) was used to evaluate the thermal stability of the treatments, based on the study of Salama et al. [11]. In a word, 15 mg of the sample was sealed in a DSC crucible and heated from 20 °C to 200 °C at 10 °C/min under a nitrogen flow of 50 mL/min. An empty crucible served as the reference. Denaturation temperature and enthalpy were analyzed using TA Universal Analysis software (Delaware, USA).

2.15. Rheological measurements

The rheological properties of the samples were assessed by an AR2000ex rheometer (DHR-3, TA Instruments, USA) according to the methodology of Yang et al. [28], with slight adjustments. All measurements were carried out at 25 °C using a 40 mm parallel-plate steel with a 1000 μm gap. The viscosity was measured over a shear rate range of 0.1-100 s⁻¹ using a steady shear test. The linear viscoelastic region (LVR) was carried out using an amplitude sweep test at 1 Hz over a strain range of 0.1-100%. Afterward, the frequency sweep test from 0.1 Hz to 100 Hz was performed at 0.5% fixed strain within the LVR. All tests were performed in triplicate.

2.16. Multiple light scattering measurement

The kinetic stability of the treatments was evaluated using a Turbiscan® TOWER analyzer (Formulaction, Toulouse, France) following the method of Salama et al. [11], with modifications. A 25 mL sample was placed in glass tubes and analyzed at 25 °C for 12 h using pulsed near-infrared light (880 nm). Stability parameters included particle diameter (Dm), delta transmission (ΔTr), delta backscattering (ΔBS), and turbiscan stability index (TSI).

2.17. Storage stability of beverages

To evaluate physical storage stability, samples were stored under refrigerated (4 °C) and room temperature (25 ± 2 °C) conditions for 30 days after the addition of 0.02% (w/v) sodium azide to inhibit microorganisms’ growth. Visual appearance was recorded on days 0, 7, 14, and 30 to monitor macroscopic changes under both storage conditions. Particle size distribution and oxidative stability were tracked over time during refrigerated storage (4 °C) for 30 days. Before each measurement, samples were shaken to ensure homogeneity.

2.18. Oxidative stability of beverages

Lipid oxidation was determined using the thiobarbituric acid reactive substances (TBARS) method according to Nasiru et al. [29], with minor changes. Briefly, 1 mL of the sample was mixed with 2 mL of 20 mM TBA reagent (prepared in 15% TCA with 0.25 N HCl). The mixture was vortexed (MX-F, Sci-LOGEX, China) and incubated in a water bath at 95 °C for 15 min away from light to allow the formation of the pink chromogen. After cooling, the mixture was centrifuged at 5000 × g for 10 min at 4 °C to remove precipitates. The absorbance of the resulting supernatant was recorded at 532 nm using a UV/Vis spectrophotometer, and the data were expressed as mg malondialdehyde (MDA)/kg sample using a standard curve for TEP.

Concurrently, protein oxidation was assessed by measuring carbonyl content in the samples, as described by Hu et al. [30], with slight changes. Samples were first adjusted to a protein concentration of 5 mg/mL. In the process, 0.5 mL of the sample was mixed with 0.5 mL of a 10 mM DNPH reagent (dissolved in 2 M HCl) and incubated in the dark at 40 °C for 1 h. A corresponding blank was prepared using 2 M HCl in place of DNPH. The reaction was stopped by adding 3 mL of 20% TCA, followed by centrifugation at 10,000 × g for 10 min at 4 °C. The supernatant was discarded, and the pellets were washed three times with a 1:1 ethanol-ethyl acetate solution (1 mL) to remove unreacted DNPH. The final pellets were dissolved in 1 mL of 6 M GuHCl (prepared in 20 mM potassium phosphate buffer, pH 6.5), and incubated at 37 °C for 20 min until completely dissolved. The absorbance was measured at 370 nm using a UV/Vis spectrophotometer. The carbonyl content was calculated using an extinction coefficient of 22,000 M⁻¹ cm⁻¹ and was expressed as μmol carbonyl/g protein. Each sample was repeated three times.

2.19. In vitro digestibility of beverages

The digestibility of the treatments was characterized by quantifying the degree of hydrolysis (DH) following the protocol described by Khalifa et al. [31], with changes. In a word, 10 mL of the beverage was subjected to sequential gastric and intestinal digestion. For the gastric phase, the sample was mixed with 6 mL of simulated gastric liquid (7 mM KH₂PO₄ and 0.2% NaCl, pH 3.0), 1.6 mL of pepsin solution (3000 U/mL), 4 μL of CaCl₂ (0.3 M), and 0.396 mL of ultrapure water, giving a total volume of 18 mL. The pH was adjusted to 3.0 using 1 M HCl, and the mixture was incubated at 37 °C in a shaking incubator (THZ-82, Guohua Electric, China) at 120 rpm for 2 h to simulate gastric digestion. After that, the pH was increased to 7.0 using 1 M NaOH to inactivate pepsin. For the intestinal phase, the gastric digest was subsequently mixed with 7.7 mL of simulated intestinal liquid (7 mM KH₂PO₄ and 0.2% NaCl, pH 7.0), 3.5 mL of pancreatin solution (800 U/mL), 1.75 mL of bile salt solution (160 mM), 28 μL of CaCl₂ (0.3 M), and 14 mL of ultra-pure water. The pH was maintained at 7.0 using 1 M NaOH, and the system was then incubated at 37 °C for 2 h to mimic small intestinal digestion. The enzymatic reaction was then terminated by heating the system at 95 °C for 5 min. The mixtures were cooled and centrifuged at 5000 × g for 10 min at 4 °C. Further, 400 μL of the supernatant was added to 4 mL of OPA reagent (0.04 g OPA dissolved in a mixture of 1 mL methanol, 25 mL 0.01 M sodium tetraborate decahydrate, 2.5 mL 20% SDS, and 0.1 mL β-mercaptoethanol up to 50 mL with deionized water), vortexed, and incubated at 25 °C for 2 min. The absorbance was recorded at 340 nm using a UV/Vis spectrophotometer. An L-serine standard curve was prepared from a 10 mM stock solution, and the calibration curve exhibited excellent linearity (R² > 0.99). The concentration of free amino groups released during digestion was calculated from the standard curve and expressed as the number of hydrolyzed peptide bonds (mmol/g). The DH was then calculated according to this equation:

$$\mathrm{D}\mathrm{H}(\%)=\frac{\text{H}}{{\mathrm{H}}_{\mathrm{t}\mathrm{o}\mathrm{t}}}\times 100$$ (2)

Where $\text{H}$ denotes the number of hydrolyzed peptide bonds, and ${\mathrm{H}}_{\mathrm{t}\mathrm{o}\mathrm{t}}$ stands for the total number of peptide bonds in the protein (8.2 mmol/g).

2.20. Volatile profiles of beverages by GC-IMS

The identification and semi-quantification of volatiles were performed using a FlavourSpec® 25 gas chromatography-ion mobility spectrometry (GC-IMS, G.A.S, Dortmund, Germany) based on the method described by Sun et al. [32]. Semi-quantitative analysis was based on signal intensity values (a.u.) obtained through the HeightAboveAreaMin evaluation method. In brief, a 20 mL headspace glass vial was filled with 5 mL of the beverage, then sealed and secured with an aluminum cap. To encourage volatile release, the vials were incubated at 60 °C for 20 min. The analysis time was set to 41 min with an injected sample size of 500 µL. Nitrogen was employed as the carrier gas with a purity of ≥ 99.99%. The gas flow was initially set at 2 mL/min for 2 min, then gradually increased to 30, 100, and 150 mL/min according to the programmed analysis conditions. The identification was accomplished by matching the obtained spectra against the internal reference library (GC-IMS, NIST2020 RI DB-5, Dortmund, Germany), while retention indices were determined using a homologous series of n-alkanes from C4 to C9. To ensure reproducibility, each sample was done in triplicate.

2.21. Sensory evaluation of beverages

Sensory evaluation was conducted following Lyu et al. [1], where 20 mL of the beverage was served in randomly coded disposable cups and presented in a balanced order to a trained panel of 10 members (5 men and 5 women) from the College of Food Science, Huazhong Agricultural University. Panelists rated color (10), texture (10), taste (10), flavor (20), and overall acceptability (50). Each sample was evaluated three times. The sensory evaluation protocol was reviewed and approved by the Ethics Committee of Huazhong Agricultural University, under approval ID HZAUHU-2025-0107. All panelists provided informed consent prior to participation.

2.22. Data processing and statistical analysis

All analyses were performed three times, and the data are reported as mean ± standard deviation. The results were statistically analyzed using SPSS version 19.0 (IBM Corporation, New York, USA), with significance considered at p < 0.05. One-way analysis of variance (ANOVA) followed by Tukey's post hoc multiple comparison test was used to evaluate significant differences among samples. Storage-related measurements were analyzed by comparing samples at each storage interval under the same statistical criteria. Figures and graphs were created using OriginPro 2025 (OriginLab Software, Northampton, USA). Volatile organic compounds analysis (VOCal) software version 0.4.12 (Gesellschaft für Analytische Sensorsysteme mbH, FlavourSpec® 25, Germany) enabled the application of both the Gallery-sort and Reporter modules, allowing a clear comparison of signal patterns and aroma profiles across the beverages. Multivariate statistical analysis was conducted using SIMCA software version 18.0.1 (Umetrics, Malmö, Sweden).

3. Results and discussion

This study investigates the development of protein-rich egg milk beverages using ultrasound-assisted EMX ternary conjugates with SMP and WFM, highlighting how tailored protein-milk fat interactions can significantly affect thermal and oxidative stability, as well as nutritional quality and digestibility. In this work, EMX is considered a model conjugate system for elucidating protein-protein and protein-Xyl interactions, while the beverage formulations represent application systems incorporating these conjugates into a complex dairy matrix. By integrating advanced analyses of structural, functional, thermodynamic, and storage stability, this work establishes a comprehensive framework for producing heat-stable and high-protein dairy beverages. This approach advances the understanding of protein-lipid systems and highlights their potential for developing functional beverages with improved stability and nutritional value.

3.1. Proximate composition analysis

The gross chemical composition of beverages, particularly protein, fat, lactose, and added sweeteners, fundamentally dictates their functionality, sensory properties, and stability throughout processing and storage [33], [34]. As shown in Table 1, there were significant differences between the samples in their chemical composition and colloidal behavior (p < 0.05). Compared with the beverage formulations, the EMX ternary conjugated system demonstrated higher moisture content and lower ash content. Adding milk powders significantly increased (p < 0.05) protein, fat, and carbohydrate contents compared to the EMX reference system. This distinction reflected the simpler composition of the conjugate system relative to the more complex beverage matrix. Altogether, total solids were highest in whole-fat milk formulations, followed by skimmed milk, and lowest in EMX. The higher solids likely enhanced network formation, viscosity, and stability. Similar results for fermented egg milk beverage were observed [1]. In model systems, ultrasonication of casein-whey powders with added calcium modified water activity and protein interactions, enhancing hydration, mixing, and stability of dried milk formulations [34].

Table 1.

Chemical composition and colloidal behavior of beverage samples.

%	Treatments
	EMX	SMP	U-SMP	WFM	U-WFM
Moisture	90.38 ± 1.13a	82.66 ± 1.23b	82.42 ± 1.72b	72.50 ± 1.41c	72.78 ± 1.12c
Protein	7.23 ± 0.11c	8.39 ± 0.30ab	8.48 ± 0.25a	8.25 ± 0.26ab	7.92 ± 0.10b
Fat	0.00 ± 0.00b	0.14 ± 0.004b	0.12 ± 0.001b	7.71 ± 0.34a	7.46 ± 0.09a
Carbohydrate	1.47 ± 0.04c	7.49 ± 0.30b	7.58 ± 0.19b	10.62 ± 0.83a	10.71 ± 0.54a
Ash	0.51 ± 0.03b	1.48 ± 0.06a	1.53 ± 0.03a	1.56 ± 0.05a	1.51 ± 0.03a
Solubility	91.04 ± 0.62bc	88.54 ± 0.41d	93.19 ± 0.28a	90.02 ± 0.33c	91.44 ± 0.03b
Turbidity	1.23 ± 0.26d	2.95 ± 0.77c	0.92 ± 0.22d	90.05 ± 0.04a	83.65 ± 0.09b

All data represent the mean of triplicates ± standard deviation. Different letters in the same row indicate significant differences (p < 0.05).

3.2. Colloidal behavior analysis

The highest turbidity was observed in WFM samples compared to the other treatments (Table 1). This was due to the presence of milk fat globules, which act as light-scattering centers in the beverage matrix, thus increasing turbidity [35]. In contrast, SMP lacked significant fat content, reducing the number of scattering centers and thus resulting in clearer solutions. Concurrently, ultrasonication treatment remarkably increased solubility and decreased turbidity (p < 0.05). Where U-SMP had the highest solubility (93.19%) than SMP (88.54%). The same trend was noted for U-WFM versus WFM (91.44% vs. 90.02%). Additionally, turbidity of U-SMP was decreased by 3.2 times than SMP, while in U-WFM was reduced by 1.07-fold than WFM. This enhancement in turbidity was attributed to the acoustic cavitation generated by ultrasound (US), which created microbubbles that collapse violently, producing localized shear forces capable of breaking protein aggregates and fat globule clusters into smaller and more uniformly dispersed particles [36]. These findings agreed with Salama et al. [11], who indicated that US significantly increased solubility and decreased turbidity of ternary conjugates after heating, reflecting improved thermal stabilization. A similar result for skimmed milk model’s turbidity containing dual β-lactoglobulin-casein complex was observed [37].

3.3. Molecular flexibility analysis

A higher absorbance indicated greater peptide release, reflecting increased molecular flexibility. As depicted in Fig. 1a, the EMX conjugated system exhibited the highest flexibility, followed by US-treated beverage samples (p < 0.05). This indicated that US-assisted conjugation promoted structural modifications and increased molecular mobility. However, this reflects the behavior of the conjugate system and should be distinguished from beverage performance, which was influenced by the complex food matrix. This result was in agreement with the findings of Ouyang et al. [23] for rice glutelin-EWP-polyphenols ternary conjugates. US-induced cavitation disrupted weak intramolecular forces, while shear stretching of polypeptides destabilized rigid structures, enhancing protein flexibility [38]. Additionally, Qian et al. [39] elucidated that US loosened the protein’s compact structure via acoustic cavitation, reduced structural rigidity, and exposed more cleavage sites. This structural flexibility explained the greater peptide release observed as higher absorbance.

Fig. 1.

Molecular flexibility (a); particle size and PDI (b); zeta potential (c); and particle size distribution (d) of beverage samples. Different letters indicate significant differences (p < 0.05). EMX, SMP, and WFM refer to EWP-MCN-Xyl ternary conjugated system, skimmed milk beverages, and full fat milk beverages, respectively. U-SMP and U-WFM denote the same skimmed and full fat beverages treated with ultrasonication for 10 min.

3.4. Particle size and ζ-potential analysis

Fig. 1b shows the changes in particle size and polydispersity index (PDI) for the treatments. The EMX conjugated system exhibited the lowest particle size compared to beverage samples (p < 0.05). This was due to the enhanced interaction between Xyl and EWP-MCN dual protein via non-covalent interactions, aligning with those of Salama et al. [11]. Additionally, US significantly decreased particle size in WFM, with no significant changes observed in SMP (p > 0.05). Cavitation during ultrasonication reduced particle size by disrupting particles and dispersing aggregates [40], [41]. Meanwhile, the PDI varied significantly across all treatments (p < 0.05). WFM demonstrated the highest particle size and PDI compared to all other treatments. This suggested that the presence of more fat globules greatly interacted with proteins and other components, promoting the formation of larger clusters, which naturally increased the PDI.

As shown in Fig. 1c, no significant differences in ζ-potential among treatments were noted. While US reduced particle size, it did not alter the overall ζ-potentials, suggesting that it mostly affected physical dispersion rather than modifying protein charge properties. This indicated that the colloidal stability of the beverage formulations was primarily governed by other factors, such as particle size or steric interactions, rather than changes in surface charge. This agreed with a similar result observed in a skimmed milk model containing β-lactoglobulin-casein dual complex [37]. The size distribution of EMX and US-treated beverage samples exhibited a shift toward smaller diameters than those of untreated ones, suggesting that US influenced the size distribution (Fig. 1d). The decreased diameters in EMX conjugates system were attributed to protein-polyol interactions, which aligned with Zhao et al. [42], who indicated that Xyl incorporation into dual systems significantly reduced their diameters. The same action was also obtained for US-treated samples, aligning with the data for US-assisted ternary conjugates [11].

3.5. Color analysis

The EMX system exhibited the lowest values of L*, a*, and b* among all samples (Table 2). The ternary system appeared darker and less colorful due to its role as a model conjugate system and the absence of natural white and yellow pigments provided by milk powders. Incorporation of milk powders increased brightness (L*) and enhanced a* and b* values, giving a more visually appealing, creamy, and familiar milk-like appearance. In addition, WFM demonstrated the highest b* values (yellowness), which was consistent with the natural color of milk fat. This was due to WFM containing carotenoids and fat-soluble pigments that contributed yellow hues, while SMP had less fat and thus lower b* values [3]. Meanwhile, the negative values of a* indicated a slight greenish hue rather than red. US treatment produced a greenish hue in SMP, suggesting minor structural changes in proteins or increased dispersion of conjugates. This was possibly due to the exposure of protein-Xyl conjugates affecting light absorption/reflection. Conversely, U-WFM shifted greenish hue toward red. This meant that US improved the dispersion of fat and protein components, leading to better light scattering and a masking of any green tint, shifting a* toward neutral/red. This was consistent with Carrillo-Lopez et al. [43], who reported that US altered milk color by affecting fat content and homogenization, influencing L*, a*, and b* values.

Table 2.

Color characteristics of beverage samples.

Treatment	L*	a*	b*
EMX	35.01333 ± 0.93821c	−1.17667 ± 0.13051a	−0.13333 ± 0.60136c
SMP	67.66000 ± 0.58924b	−4.47667 ± 0.54455b	2.5400 ± 0.2623b
U-SMP	70.75001 ± 0.53563a	−4.78667 ± 1.02632b	2.50333 ± 0.49339b
WFM	67.18333 ± 0.07095b	−0.5410 ± 0.09539a	8.67667 ± 0.10116a
U-WFM	71.38667 ± 1.24821a	0.1003 ± 0.71084a	8.7800 ± 0.04359a

All data represent the mean of triplicates ± standard deviation. Different letters in the same column indicate significant differences (p < 0.05).

3.6. Second-derivative UV spectra analysis

As displayed in Fig. 2a, WFM and U-WFM demonstrated higher absorbance than other treatments, indicating protein structural changes and enhanced aromatic residue exposure, together with light scattering and milk fat-protein interactions. Meanwhile, US-treated samples showed lower peak intensity compared to untreated ones. This was due to US-induced conformational rearrangements in beverages, showing reduced near-UV absorbance due to the exposure of aromatic residues to the aqueous phase [44]. This was in accordance with Jun et al. [45], who noticed US-induced structural alterations in EWP. Similar results for US-assisted ternary conjugates prior to and post-heating were observed [11].

Fig. 2.

UV spectra (a); SD-UV spectra (b); intrinsic fluorescence at 280 nm (c); intrinsic fluorescence at 295 nm (d); synchronized fluorescence spectra with Δλ of 15 nm (e); synchronized fluorescence spectra with Δλ of 60 nm (f) of beverage samples.

Due to the overlapping absorption bands of tryptophan (Trp), tyrosine (Tyr), and phenylalanine (Phe), their individual contributions cannot be clearly distinguished using UV spectra alone. Second-derivative UV (SD-UV) enhances signal separation, with Trp showing peaks at 290-300 nm, Tyr at 280-285 nm, and Phe at 250-260 nm [46]. As demonstrated in Fig. 2b, the SD-UV spectra of all samples exhibited pronounced differences in peak positions and intensities, reflecting alterations in the microenvironment of the chromophores. These spectral variations were particularly pronounced in WFM and U-WFM samples. This was attributed to the high amount of fat, which promoted closer protein-lipid associations and enhanced structural rearrangements of ternary conjugates. Furthermore, US treatment intensified these effects, likely due to cavitation-induced unfolding and reaggregation, which increased the exposure and polarity shifts of Trp and Tyr. Consequently, the more evident peak shifts in WFM and U-WFM samples indicated a higher degree of conformational alteration and complex structural dynamics compared with EMX, SMP, and U-SMP samples. Similar findings for whey protein isolates with phosphoric acid were obtained [47].

3.7. Intrinsic and synchronized fluorescence analysis

In Fig. 2c, SMP showed the highest fluorescence intensity, followed by U-SMP and EMX at 280 nm. This was due to the high exposure of chromophores, which increased fluorescence intensity. Additionally, the lower intensity observed in U-SMP compared to SMP could be attributed to US-induced partial unfolding or aggregation, leading to some quenching of exposed aromatic residues. In contrast, WFM and U-WFM had the lowest intensity with redshifted, probably due to the hydrophobic interactions between proteins and fat globules, which buried aromatic residues and reduced fluorescence emission. Similar results for rice starch-polyphenol-whey protein isolate ternary conjugates [48] and β-lactoglobulin and mono- or di-glucoside anthocyanins [31] were observed.

At 295 nm, where fluorescence primarily originates from Trp residues, the intensity of WFM and U-WFM indicated that Trp residues were more exposed or in a more hydrophobic environment compared to the others (Fig. 2d). This was attributed to fat creating a hydrophobic microenvironment that enhanced Trp fluorescence by reducing water-induced quenching. A slight blue shift was noted in U-WFM, suggesting that US altered the protein microenvironment (Fig. 2d). Conversely, SMP, U-SMP, and EMX showed lower Trp fluorescence, likely due to a more polar environment or tighter protein-protein/polyol interactions. Kong et al. [12] indicted that the conjugation of whey protein with MCN and Xyl altered the protein’s tertiary structure, potentially burying Trp residues within the conjugate. This burial reduced the exposure of Trp to the solvent, decreasing fluorescence intensity. Moreover, US induced structural changes that altered the microenvironment of Trp residues, affecting their fluorescence characteristics [11]. Analogously, Li et al. [49] found blue shifts of rice bran protein-chitosan complexes treated by US.

At Δλ of 15 nm or 60 nm, synchronous fluorescence spectroscopy primarily tracks Tyr and Trp fluorescence, offering insights into protein-ligand interactions [25], [50]. As shown in Fig. 2e and f, SMP exhibited the highest fluorescence intensity, followed by U-SMP and EMX. In contrast, WFM and U-WFM showed the lowest fluorescence intensity. This trend suggested that the protein environment and matrix composition strongly influenced samples’ intrinsic fluorescence. The higher intensity in SMP reflected a greater exposure of aromatic amino acid residues due to its lower fat content, which reduced quenching effects (Fig. 2e and f). In addition, US treatment reduced the intensity in U-SMP with a blue shift, possibly indicating partial conformational changes and surrounding microenvironment. Concurrently, the lower intensity in WFM and U-WFM could be attributed to the higher lipid content, which shielded or quenched protein fluorophores, limiting their fluorescence emission. Similar findings were noted for β-lactoglobulin-resveratrol dual system complex [25]

3.8. XRD analysis

Fig. 3a revealed clear structural differences among the treatments. EMX exhibited a relatively strong and narrow diffraction peak (intensity 946 at 20.83°), consistent with a higher degree of ordering or larger coherent crystalline domains in the conjugated system. In contrast, SMP showed a lower intensity and a broader peak (648 at 22.60°). These differences arose from their fundamentally different system compositions. While U-SMP exhibited a slight reduction in intensity (640) combined with a shift to higher 2θ (24.52°) and notable peak broadening. This pattern indicated structural disruption by US, resulting in smaller crystallites and/or an increased amorphous fraction. Additionally, WFM showed the highest native intensity (1175 at 20.37°), implying substantial crystalline ordering (likely from fat/lactose domains). While US treatment in U-WFM reduced that ordering (843 at 20.85°) and produced a modest peak shift. This reduction caused by US indicated partial disruption or fragmentation of the crystalline domains [51]. In summary, EMX showed higher crystalline order than SMP and U-SMP, while WFM exhibited the greatest native crystallinity due to fat and lactose domains, aligning with the data in Table 1. US reduced ordering in both SMP and WFM, and this was stronger in SMP. Overall, EMX conjugation helped maintain ordered structures, whereas US disrupted crystallinity, influencing stability and mouthfeel.

Fig. 3.

XRD patterns (a); FTIR spectra (b); and CLSM micrographs (c) of beverage samples. Proteins were dyed with FITC (green color), and milk fat was stained with NR (red color). The scale bar = 20 μm.

3.9. FTIR analysis

As illustrated in Fig. 3b, WFM and U-WFM displayed pronounced shifts in FTIR spectra, suggesting that the fundamental backbone structures largely changed. All treatments exhibited absorption bands between 3100 and 3500 cm⁻¹, primarily corresponding to O–H and N–H stretching vibrations, which reflected hydrogen bonding in proteins and associated water molecules [28]. The broader bands observed in EMX and SMP indicated a higher degree of hydrogen bonding interactions or greater structural flexibility, likely due to more extensive protein-sugar interactions or water incorporation. This result was in accordance with Salama et al. [11], who reported that ternary system stabilization was governed by hydrogen bonding. Conversely, the narrower bands in the WFM suggested fewer hydrogen bonds or a more rigid and ordered protein network, possibly due to tighter protein packing. The absorption peak around 2928 cm⁻¹ corresponded to C–H stretching vibrations, which were primarily associated with lipid or hydrophobic protein regions [26]. Its pronounced presence in WFM and U-WFM reflected the higher lipid content, whereas the slight peak in U-SMP might indicate minor exposure of hydrophobic groups due to US-induced structural loosening. Meanwhile, the absence signal in SMP and EMX suggested less accessible hydrophobic regions, consistent with their fat-free content or tighter protein structures limiting the C–H vibrational signal.

All samples displayed peaks at around 1650 cm⁻¹ (amide I), 1540 cm⁻¹ (amide II), and 1242 cm⁻¹ (amide III), demonstrating that the protein backbone structure was largely retained [52]. While slight shifts suggested the effects of lipid-protein conjugation, US treatment, and milk powder formulation on secondary structure and intermolecular interactions. Similar shifts in these bands were observed for lentil-whey protein-trehalose ternary conjugates [44]. Interestingly, the absorption peak around 1075 cm⁻¹, attributed to C–O stretching of sugars like lactose and Xyl [53]. This was prominent in WFM, U-WFM, and U-SMP, reflecting accessible sugar residues, while its weakness in EMX and SMP suggested lower exposure due to tighter protein-sugar interactions or more compact protein networks.

3.10. CLSM microstructure analysis

Milk fat was stained red, while proteins were labeled green (Fig. 3c). EMX, SMP, and U-SMP samples exhibited a dark background in the NR channel, indicating the absence of visible fat globules and confirming their fat-free composition. In contrast, WFM samples displayed numerous large oil droplets, reflecting the presence of milk fat and incomplete integration of the lipid phase. US led to a remarkable reduction in droplets size, producing a more homogeneous and finely dispersed fat-protein network in U-WFM groups. Furthermore, the merged images clearly revealed that US facilitated closer association between fat droplets and protein matrices, suggesting improved interfacial stabilization and emulsification efficiency [54]. A similar case was observed in the FITC channel for SMP groups (Fig. 3c). Concurrently, EMX exhibited the smallest and most uniform protein particles among all treatments, which agreed with the particle size analysis results. Comparable microstructural improvements have been reported in US-treated dairy and protein-based emulsions, in which cavitation-induced shear and turbulence reduced particle clustering and facilitated compact and stable matrices formation [20], [43], [55].

3.11. DSC thermal stability analysis

The thermodynamic profiles of all beverages are presented in Fig. 4a and Table 3. Each sample exhibited a shift toward lower peak temperatures compared to EMX, indicating marked alterations in thermal stability following formulation and matrix modification. As also seen in Table 3, EMX exhibited the highest denaturation temperature (Td=114.17 °C) and enthalpy change (ΔHd=179.3 J/g), indicating superior thermal stabilization at the molecular level, independent of the effects of complex beverage matrices. This result was aligned with the enhanced heat stability of the dual protein system [5]. On the contrary, WFM showed the lowest Td (96.44 °C) and ΔHd (91.87 J/g), suggesting a lower thermal resistance. Remarkably, U-SMP displayed a higher Td (113.23 °C) than SMP (102.21 °C). Likewise, U-WFM showed an increased Td (100.60 °C) compared to WFM (96.44 °C), reflecting the stabilizing effect of US on protein structures and improved thermal stability. Yousefi and Abbasi [40] indicated that the US enhanced whey protein thermal stability by disrupting peptide chains, disulfide bonds, and hydrophobic interactions. Higher ΔHd indicated stronger molecular interactions and structural stability, while lower values reflected weaker thermal resistance [11]. This suggested weaker interactions in WFM, likely due to fat interference with protein-protein conjugation. The same trend for ΔHd was observed in EWP microgel-polysaccharide dual complexation [28].

Fig. 4.

DSC thermograms (a); strain sweeps (b); frequency sweeps (c); and flow sweeps (d) of beverage samples.

Table 3.

Thermodynamic properties of beverage samples.

Treatment	ΔTd (°C)	Td (°C)	ΔHd (J/g)
EMX	38.00–––147.74	114.17	179.3
SMP	41.72–––136.18	102.21	132.0
U-SMP	48.58 – 160.90	113.23	169.1
WFM	44.22–––119.85	96.44	91.87
U-WFM	44.20–––132.63	100.60	92.76

3.12. Rheological properties analysis

The viscoelastic properties were assessed using strain, frequency, and flow sweeps. As shown in Fig. 4b, all treatments had storage modulus (G') values greater than loss modulus (G''), demonstrating predominantly elastic or gel-like properties across formulations [56]. This suggested a weak network formed by protein-protein and protein-milk fat interactions, maintaining structural integrity under small deformations. WFM and U-WFM showed the lowest initial G' values, indicating a weaker elastic network. The higher fat content likely acted as a plasticizer, disrupting protein interactions, while US only partially offset this softening at low strains.

The frequency sweep results indicated that all treatments exhibited G' > G'' across the measured range, confirming a predominantly elastic behavior (Fig. 4c). Similar trends were observed in US-assisted dual complexation of rice flour and soy protein isolate [57]. Both G' and G'' elevated with frequency, showing weak networks with some viscoelasticity. Notably, US-treated samples showed higher G', suggesting enhanced network interactions due to reduced particle size and improved protein complexation. This aligned with Guo et al. [58], who found that US and pH-shifting improved the stability of protein beverages. As depicted in Fig. 4d, the flow sweep test revealed shear-thinning behavior, as viscosity reduced with raising shear rate, a typical feature of protein-polysaccharide networks [28]. WFM and U-WFM samples had higher viscosity due to fat globules, which enhanced cross-linking. This was consistent with Janahar et al. [59], who demonstrated that US treatment increased the viscosity of the pea protein-dairy model, particularly in higher-fat systems. Overall, viscoelastic analyses confirmed that US treatment effectively improved protein-fat interactions and structural integrity.

3.13. Multiple light scattering analysis

The kinetic physical stability of the treatments is illustrated in Fig. 5. Among all formulations, the EMX sample exhibited the highest TSI, indicating system instability (Fig. 5a). As EMX is a model conjugate system without added milk powders, its higher optical clarity makes the Turbiscan measurement more sensitive to subtle changes in light transmission and backscattering. This increased sensitivity means that even small particle migrations or slight structural rearrangements are more readily detected, resulting in a higher TSI value. Conversely, formulations containing milk solids and fat were more opaque, which could mask minor structural changes and led to lower TSI values, even if particle movement occurred. Similar trends were reported in the β-lactoglobulin-casein dual system during UHT milk processing, where complexation increased TSI [37].

Fig. 5.

Changes in turbiscan stability index (TSI) (a); transmission (ΔTr) (b); backscattering (ΔBS) (c); and particle diameter (Dm) (d) of beverage samples.

The increase in ΔTr correlated with a decrease in particle concentrations, indicating stability. EMX showed the highest ΔTr and lowest ΔBS, explaining the discrepancies in TSI values (Fig. 5b and c). The other treatments showed lowest and stable ΔTr over the 12 h of tracking time (Fig. 5b). This was due to the higher contents of milk solids and fat. Furthermore, ΔBS profiles revealed that WFM exhibited the highest signal intensity, followed by U-WFM, whereas SMP and U-SMP showed comparatively lower values (Fig. 5c). This pattern reflected the greater light scattering caused by the higher fat content in WFM compared to SMP. Concurrently, US treatment notably reduced ΔBS in both WFM and SMP formulations. This was due to improved dispersion and reduced particle aggregation, which was consistent with previous observations in US-treated soybean milk [60]. These data agreed with the turbidity results in Table 1.

As displayed in Fig. 5d, EMX system showed the lowest Dm compared to the systems containing milk powders, confirming its stability. U-WFM and U-SMP had smaller Dm than their untreated counterparts, aligning with the results of particle size. Salama et al. [11] found that US-assisted ternary conjugation reduced TSI and Dm due to hydrogen bonding and hydrophobic interactions. Abou-Elsoud et al. [54] elucidated that pectin-low density lipoprotein dual complexation assisted by US and high-pressure homogenization significantly affected Dm and TSI values by controlling the protein-polysaccharide mixing ratio and the physical treatment applied.

3.14. Storage stability analysis

Fig. 6 displays the visual appearance and particle size of beverages during 30 days of storage. At refrigerated temperature, all formulated systems exhibited excellent stability without signs of sedimentation or phase separation throughout the storage period (Fig. 6a). This indicated that low-temperature conditions effectively suppressed aggregation and destabilization mechanisms. This finding was consistent with earlier reports that refrigerated storage slowed protein-protein interactions and fat destabilization, thereby enhancing the colloidal stability of protein beverages [61].

Fig. 6.

Visual appearance at 4 °C (a); at ambient temperature (b); changes in particle size (c); and particle size distribution on day 30 (d) of beverage samples during 30 days of storage at 4 °C. Different letters indicate significant differences (p < 0.05).

At ambient temperature, all samples exhibited stability for the first 7 days, but noticeable changes emerged with prolonged storage. By day 14, SMP developed a slight phase-separated layer, which intensified by day 30 (Fig. 6b). In contrast, other formulations maintained visual stability throughout the storage period. The phase separation in SMP was primarily due to protein aggregation, which exacerbated by higher molecular mobility at room temperature. This observation aligned with Bolling et al. [62], who indicated that skimmed milk powders, lacking fat globules, were more susceptible to aggregation and exhibited less emulsion stability compared to fat-containing systems. The high fat content in WFM formulations enhanced stability through interfacial mechanisms, with fat globules acting as fillers in the colloidal network, diminishing protein aggregation [63].

Additionally, US improved protein conformation and resulted in finer particles, leading to a more uniform colloidal network that resisted sedimentation. Particle size tracking revealed that WFM and U-WFM had the largest sizes during storage due to fat globules (Fig. 6c). Although particle size decreased slightly at 14 days, it increased significantly by day 30, indicating gradual aggregation attributed to the presence of fat globules contributing to the larger droplet size (p < 0.05). US-treated beverages maintained smaller particle sizes than their untreated counterparts (p < 0.05), confirming the effectiveness of US in enhancing stability and protein solubility [20]. This agreed with the particle size distribution of the treatments at 30 days of storage (Fig. 6d). Carrillo-Lopez et al. [43] reported that US reduced particle growth and delayed destabilization in dairy and protein beverages during storage.

3.15. Oxidative stability analysis

The oxidative stability of the beverages was monitored over 30 days of storage at 4 °C, as illustrated in Fig. 7a and b. For lipid oxidation, no detectable levels of lipid oxidation were observed in EMX, SMP, and U-SMP samples. This was attributed to their negligible fat content, thereby minimizing the substrates available for oxidative reactions (Fig. 7a). This aligned with the data in Table 1.

Fig. 7.

Contents of malondialdehyde (MDA) (a); and protein carbonyl (b) in beverage samples during 30 days of storage at 4 °C. The in vitro digestion simulation layout (c); and degree of hydrolysis (DH) (d) of beverage samples. Different letters indicate significant differences (p < 0.05).

In contrast, both WFM and U-WFM exhibited gradual increases in lipid oxidation during storage, consistent with their higher fat levels, which make them more susceptible to peroxidation. Notably, lipid oxidation in WFM was higher than in U-WFM during the early storage period (0-14 days). At 30 days, U-WFM showed a significant increase (0.6522 mg MDA/kg) compared to WFM (0.4784 mg MDA/kg). This suggested that while US initially inhibited oxidative processes, prolonged storage might accelerated oxidation, potentially due to US-induced changes in lipid-protein interactions and radical generation. Similar trends have been reported in milk and dairy products, where US treatment was shown to delay lipid oxidation at early stages but promote it during extended storage, depending on the intensity and exposure conditions [64], [65]. Analogously, Abbas et al. [66] indicated that the oxidative stability in milk was significantly enhanced by mild cold plasma treatment in the early storage period (7 days). These findings highlighted that US processing might provide short-term protection against lipid oxidation in full fat beverages, but the increased TBARS values in U-WFM on day 30 indicated a potential trade-off over the long term, emphasizing the need for optimization.

During a 30-day storage period, protein oxidation, measured by carbonyl content, showed notable differences (p < 0.05) among various treatments (Fig. 7b). At 0-14 days, the EMX system had significantly higher carbonyl levels than U-WFM and WFM beverage formulations, indicating greater early oxidative susceptibility in the simplified conjugate system compared to the complex beverage matrices. Conversely, SMP and U-SMP demonstrated the lowest carbonyl contents throughout this period, suggesting better oxidative stability due to lower fat content. This aligned with findings from Scheidegger et al. [67], showing SMP’s reduced carbonyl content versus whole milk powder. By the end of the storage period, the carbonyl levels in WFM, U-WFM, and SMP increased but without significant differences among them, implying progressive protein degradation over time. U-SMP consistently maintained lower carbonyl levels than SMP, indicating that US modulated protein oxidative stability. This might result from cavitation-enhanced protein interactions with MCN and Xyl, which shielded oxidation-sensitive residues and improved radical scavenging activity, consistent with Salama et al. [11]. On the contrary, U-WFM had higher carbonyl levels than WFM, highlighting the pro-oxidative effect of fat under US treatment, reinforcing the concept that fat content influences oxidative stress in milk powders. Zhao et al. [68] reported that whole milk showed higher carbonyl levels and glycation products than skimmed milk during thermal treatment, demonstrating the pro-oxidative role of fat. Of note, oxidative stability was evaluated using TBARS and carbonyl content as partial indicators, but a broader set of markers would better capture the oxidation mechanisms in such complex matrices.

3.16. In vitro digestion analysis

The digestive behavior of the model conjugate system and beverage samples was evaluated by simulating the in vitro gastric and intestinal digestion phases as represented in Fig. 7c. The results showed that US significantly increased DH (p < 0.05) in both SMP and WFM beverages compared with their untreated counterparts (Fig. 7d). This improvement was attributed to the US-induced conformational modification of protein structures, which promoted partial unfolding and exposure of enzymatic cleavage sites, particularly hydrophobic residues and peptide bonds buried within the native protein matrix [24]. Moreover, the cavitation effect generated by US disrupted protein aggregates, decreased particle size, and improved surface area accessibility, thereby facilitating enzyme-substrate interactions and accelerating proteolysis [69]. As also shown in Fig. 7d, EMX conjugated system demonstrated the lowest DH values (p < 0.05) among all beverage formulations. This could be attributed to the limited accessibility of enzymatic sites due to strong intermolecular interactions and aggregate formation between EWP and MCN in the absence of US treatment. The presence of Xyl might also contribute to stabilizing hydrogen bonds within the protein network, increasing conformational rigidity and thus reducing enzymatic susceptibility [50]. This was consistent with that of Meng et al. [48], who indicated that rice starch-polyphenol-whey protein isolate ternary conjugates had the lowest digestibility as a result of the suppressive impact of polyphenols on enzyme activity. Similar observations were reported by Lorieau et al. [70] for dairy products, and by Khalifa et al. [31] for β-lactoglobulin and mono- or di-glucoside anthocyanins on in vitro digestion changes.

3.17. Volatile profiles analysis

The volatile compounds (VCs) observed in the different treatments were identified and semi-quantified using GC-IMS. As represented in Fig. 8a, the 3D topography of the samples consisted of the Z-axis that referred to the signal intensities of ions detected, and the Y-axis, which indicated the retention time (Rt) of the compound took to travel through the GC column. The X-axis denoted the drift time (Dt) of ions through the IMS under the electric field. Higher peaks correspond to VCs present in greater quantities [32]. The 3D plot demonstrated that the EMX conjugated system was the highest signal intensities compared with the beverage formulations. This indicated higher amounts of VCs in EMX than others. This was attributed to differences in matrix complexity and volatile-matrix interactions between the conjugate system and the beverage systems, which influenced volatile retention and release. In addition, these signals in U-SMP and U-WFM decreased, indicating that US significantly affected VCs contents.

Fig. 8.

3D topography (a); 2D topography map (b); 2D topographic subtraction map (c) of volatile compounds for beverage samples.

As shown in Fig. 8b, the 3D plot was scaled down to create a 2D topography map for beverage samples. Multiple signals were found in all samples within the Rt range of 100 to 800 s. EMX revealed higher signal intensities than the other treatments. Meanwhile, these signals notably decreased in WFM and U-WFM. This suggested that milk fat significantly modulated VCs in beverages. The data from the 2D map agreed with the 3D topography plot. Concurrently, Fig. 8c displays the 2D topographic subtraction map of the treatments. EMX was used as a reference plot by subtracting its spectrum from the others to highlight the differences. VCs concentrations in SMP, U-SMP, WFM, and U-WFM are similar to those in EMX, indicated in white. Concentrations above or below EMX are noted in red and blue, respectively. The concentrations of VCs increase or decrease depending on the intensity of red and blue colors [32], [71]. The 2D topographic subtraction plot demonstrated significant differences in all beverage formulations compared to the reference plot (Fig. 8c). This indicated that VCs in beverages are mainly based on the milk type and the physical treatment used. This reflected that milk fat with US altered protein molecular interactions, resulting in binding some VCs via hydrophobic interactions, aligning with signals in 2D and 3D topographic maps. These results were supported by Tan et al. [72], who concluded that increased fat content enhanced both volatile profiles and flavor attributes. Besides, milk fat has been shown to intensify volatile formation, with higher levels of aldehydes and esters detected in whole milk compared to skimmed milk. This was mainly due to lipid oxidation pathways linked to fatty acids such as C16:0 and C18:2 [73]. This highlighted that fat-rich systems in the presence of US generated more pronounced volatile changes.

3.18. Volatile fingerprints analysis

Fig. 9a presents VCs’ fingerprints of the model conjugate system and beverage samples. In the plot, red areas represent regions of higher VCs abundance, whereas white areas correspond to lower concentrations. The fingerprint profiles showed that the contents of dipropyl disulfide, 2-furanmethanethiol, 5-Hepten-2-one, 6-methyl-, 1-hexanal, 4-methylthiazole, and 2,6-dimethyl pyrazine were notably higher in the EMX conjugated system compared to the beverage samples (p < 0.05). Conversely, WFM and U-WFM had higher contents of 2-Pentanol, 2-Butanone, 3-hydroxy-, Butanal, 2,3-Butanedione, n-pentanal, 2,5-Dimethylfuran, Pentyl formate, and Butyl acetate than others. These results agreed with the data in Table 4 (p < 0.05). Compared to EMX, SMP and U-SMP beverages decreased dipropyl disulfide by 1.62 and 1.57 times, while WFM and U-WFM reduced the compound by 7.78 and 8.38-fold, respectively. Likewise, 2-furanmethanethiol decreased significantly (p < 0.05) in SMP and U-SMP (1.32 and 1.30 a.u.), while in WFM and U-WFM, the values were 0.89 and 0.69, respectively, compared to 1.65 a.u. in EMX (Table 4). More importantly, these two VCs (dipropyl disulfide and 2-furanmethanethiol) exhibited a sulfur and onion odor, according to the odor description in Table S1. Thus, these two odors contributed to the off-flavors in EWP within the EMX conjugated system [7]. WFM and U-WFM beverages exhibited higher amounts of volatiles with butter, sweet, and fruity odor descriptions, this came at the expense of sulfur aromas (Table S1). This reflected that the capability of milk fat in modulating volatile profiles and alleviating undesirable flavors.

Fig. 9.

Fingerprints (a); and PCA biplot (b) of volatile compounds distribution across beverage samples. Sensory evaluation scores (c) of beverage samples. Color, texture, and taste were rated out of 10 points each, while flavor received 20 points. Overall acceptability was rated out of 50.

Table 4.

Volatile compounds detected in beverage samples using GC-IMS.

Volatile compounds	Peak signal intensity (a.u.)
	EMX	SMP	U-SMP	WFM	U-WFM
dipropyl disulfide	1.09 ± 0.09a	0.67 ± 0.01b	0.69 ± 0.007b	0.14 ± 0.01c	0.13 ± 0.009c
3-methylbutyl butanoate	0.94 ± 0.02a	0.74 ± 0.01b	0.75 ± 0.001b	0.26 ± 0.004c	0.25 ± 0.001c
2-Ethyl-1-hexanol	1.49 ± 0.1a	1.04 ± 0.01b	1.07 ± 0.01b	0.29 ± 0.003c	0.27 ± 0.009c
ethyl E-2-hexenoate	0.63 ± 0.08a	0.22 ± 0.002b	0.24 ± 0.005b	0.02 ± 0.0007c	0.02 ± 0.0007c
5-Hepten-2-one, 6-methyl-	1.74 ± 0.12a	1.28 ± 0.01b	1.32 ± 0.005b	1.03 ± 0.01c	1.02 ± 0.007c
3-(methylsulfanyl)propanol	0.56 ± 0.06a	0.24 ± 0.003b	0.26 ± 0.007b	0.15 ± 0.006c	0.14 ± 0.002c
Hexanoic acid methyl ester	0.92 ± 0.14a	0.51 ± 0.21ab	0.42 ± 0.01bc	0.20 ± 0.19bc	0.09 ± 0.01c
3-Heptanol	0.55 ± 0.07a	0.39 ± 0.05b	0.38 ± 0.01bc	0.30 ± 0.07bc	0.25 ± 0.002c
2-furanmethanethiol	1.65 ± 0.13a	1.32 ± 0.12ab	1.30 ± 0.01ab	0.89 ± 0.36bc	0.69 ± 0.02c
Pentyl formate	1.86 ± 0.08a	1.86 ± 0.10a	1.95 ± 0.008a	1.89 ± 0.05a	1.80 ± 0.03a
Sarin	1.44 ± 0.04a	1.17 ± 0.09ab	1.31 ± 0.31a	0.88 ± 0.54ab	0.41 ± 0.18b
1-hexanal	3.11 ± 0.24a	2.41 ± 0.31ab	2.27 ± 0.02ab	1.72 ± 0.48bc	1.34 ± 0.04c
4-methylthiazole	0.56 ± 0.01a	0.49 ± 0.06a	0.47 ± 0.02a	0.49 ± 0.006a	0.47 ± 0.01a
Butyl acetate	0.51 ± 0.04ab	0.40 ± 0.03b	0.42 ± 0.01b	0.54 ± 0.13ab	0.65 ± 0.01a
2-Pentanone	0.57 ± 0.07b	0.59 ± 0.005b	0.58 ± 0.001b	1.20 ± 0.06a	1.18 ± 0.003a
2,3-Pentadione	0.70 ± 0.08a	0.40 ± 0.01c	0.58 ± 0.01b	0.38 ± 0.01c	0.47 ± 0.01bc
1-Penten-3-one	0.64 ± 0.06a	0.41 ± 0.009c	0.42 ± 0.01c	0.50 ± 0.01b	0.45 ± 0.005bc
2,5-Dimethylfuran	0.36 ± 0.16b	0.28 ± 0.004b	0.27 ± 0.004b	1.05 ± 0.01a	0.99 ± 0.007a
n-pentanal	0.29 ± 0.05b	0.21 ± 0.003c	0.19 ± 0.007c	0.78 ± 0.02a	0.74 ± 0.002a
3-Methyl butanal	1.70 ± 0.21a	1.15 ± 0.03b	1.33 ± 0.04b	1.22 ± 0.02b	1.13 ± 0.02b
methyl propanoate	2.53 ± 0.10a	2.22 ± 0.03c	2.67 ± 0.02a	2.35 ± 0.03bc	2.37 ± 0.02b
acetic acid ethyl ester	1.93 ± 0.09b	2.12 ± 0.02a	2.09 ± 0.007a	2.05 ± 0.003ab	2.03 ± 0.003ab
2,3-Butanedione	1.06 ± 0.17a	1.19 ± 0.03a	1.25 ± 0.04a	1.17 ± 0.01a	1.14 ± 0.01a
Butanal	0.58 ± 0.02c	1.03 ± 0.01b	1.07 ± 0.005b	1.05 ± 0.006b	1.14 ± 0.006a
Tetrahydrofurane	0.41 ± 0.10a	0.24 ± 0.02b	0.23 ± 0.01b	0.26 ± 0.007b	0.23 ± 0.001b
(Z)-4-heptenal	0.03 ± 0.003c	0.03 ± 0.002c	0.04 ± 0.001c	0.90 ± 0.02a	0.84 ± 0.01b
Ethyl pentanoate	0.31 ± 0.04c	0.34 ± 0.007c	0.35 ± 0.002c	1.62 ± 0.02a	1.56 ± 0.01b
2-Butanone, 3-hydroxy-	0.56 ± 0.16a	0.43 ± 0.01a	0.45 ± 0.006a	0.56 ± 0.006a	0.52 ± 0.008a
Propylacetate	1.24 ± 0.22a	1.03 ± 0.02ab	0.97 ± 0.02abc	0.75 ± 0.02bc	0.73 ± 0.01c
2-Pentanol	0.97 ± 0.22b	0.87 ± 0.01b	0.83 ± 0.02b	1.49 ± 0.02a	1.43 ± 0.007a
acetic acid propyl ester	2.23 ± 0.12c	2.39 ± 0.03ab	2.30 ± 0.003abc	2.41 ± 0.004a	2.24 ± 0.006bc
2,6-dimethyl pyrazine	0.53 ± 0.08a	0.40 ± 0.001bc	0.42 ± 0.007b	0.30 ± 0.007cd	0.29 ± 0.004d

All data represent the mean of triplicates ± standard deviation. Different letters in the same row indicate significant differences among treatments (p < 0.05).

As shown in Table 4, U-SMP and U-WFM remarkably decreased (p < 0.05) sulfur aromas and increased fruity ones compared to SMP and WFM. This enhancement by US was mainly due to protein’s structural alterations, which exposed additional reactive sites and promoted stronger interactions with VCs through non-covalent binding [74]. Furthermore, the •OH radicals generated by cavitation covalently interacted with sulfur odorants via Schiff base formation, mitigating sulfur aromas [14]. Similar findings were reported for enhancing the beany flavor and storage stability in soybean milk using 7 min of US [60]. Likewise, glycation of soy protein isolate-dextran conjugates significantly decreased soy protein’s off-flavors by 24.7% [75]. In a nutshell, both milk fat and US treatment modulated the volatile fingerprints of egg milk beverages, with WFM and U-WFM showing higher relative signal intensities of compounds tentatively associated with buttery, fruity, and sweet notes.

To further elucidate the relationship between VCs and sample differentiation, principal component analysis (PCA) was performed. As depicted in Fig. 9b, PCA biplot (R²X [1] = 0.635, R²X [2] = 0.213) explained 84.8% of the total variance, demonstrating a clear separation of volatile profiles among treatments. EMX samples were positioned far to the right, associated with sulfur-containing compounds such as 4-methylthiazole, 2-furanmethanethiol, and dipropyl disulfide. In contrast, U-WFM and WFM beverages clustered closely in the upper left quadrant, strongly correlated with fruity, buttery, and sweet volatiles such as butyl acetate, (Z)-4-heptenal, and 3-hydroxy-2-butanol. SMP and U-SMP are grouped near the center, reflecting their moderately volatile contributions dominated by esters like acetic acid propyl ester and acetic acid ethyl ester. These findings indicated that US and milk fat interactions influenced protein structures and altered volatile profiles of the treatments.

3.19. Sensory evaluation analysis

Organoleptic properties are crucial for the acceptance of protein-enriched beverages, reflecting taste, aroma, texture, and appearance [1]. As illustrated in Fig. 9c, sensory evaluation demonstrated that U-WFM beverage formulations achieved the highest overall acceptability. This was mainly due to superior color and total solids, as well as enhanced flavor and richer taste contributed by milk fat. In contrast, SMP and U-SMP formulations rated higher for texture but lower for taste and overall preference, reflecting a balance between smooth mouthfeel and flavor intensity. Additionally, the EMX model conjugate system consistently received the lowest sensory scores, especially in flavor (9.5/20) and overall acceptance (24/50). This was attributed to its lower total solids content together with the elevated sulfur-related odor detected by GC-IMS. In summary, these results suggested that both fat content and US treatment played critical roles in improving sensory attributes by enhancing visual uniformity, modulating volatile profiles, and balancing mouthfeel. These observations were consistent with previous findings indicating that protein-fat interactions and US processing could contribute to consumer acceptance of protein-enriched beverages [2], [11], [40].

4. Conclusions

This study demonstrates that US-assisted ternary conjugates of EMX act as an advanced hydrocolloid system capable of stabilizing high-protein egg milk beverages. EMX effectively reorganized protein secondary and tertiary structures, strengthened hydrogen-bonded microdomains, and prevented aggregation, resulting in superior solubility (91.04%), reduced particle size, and the highest thermal stability (Td 114.17 °C; ΔHd 179.3 J/g) among all formulations. US further enhanced structural flexibility and refined protein-lipid interfacial assembly, yielding uniform microstructures and improved colloidal stability across 30 days of storage. The combination of milk fat and US significantly modulated volatile profiles and oxidative behavior. US-treated whole-fat beverages showed enhanced oxidative stability in the early stages (0-14 days), but this effect diminished over time, with higher TBARS values in U-WFM than WFM on day 30. These formulations also promoted the retention of fruity and buttery esters while dramatically alleviating sulfurous off-flavors, reducing dipropyl disulfide by up to 8.38-fold and 2-furanmethanethiol by 2.39 times. Additionally, US treatment enhanced digestibility, reflecting improved enzyme accessibility and protein unfolding. These results were translated into enhanced sensory and functional quality of the final egg milk beverages. This study presents a promising approach using US-assisted ternary conjugation to improve the stability and functionality of protein-rich beverage systems. However, the small sensory panel, tentative volatile identification, and short storage period remain limitations that should be addressed in future studies.

CRediT authorship contribution statement

Mohamed Salama: Writing – review & editing, Writing – original draft, Visualization, Validation, Software, Methodology, Formal analysis, Data curation, Conceptualization. Mostafa Gouda: Writing – review & editing, Validation. Mahmoud Abou-Elsoud: Formal analysis. Hamdy Zahran: Software. Dewei Shu: Writing – review & editing. Long Sheng: Writing – review & editing, Validation. Ammar AL-Farga: Investigation, Writing – review & editing. Zhaoxia Cai: Writing – review & editing, Supervision, Project administration, Funding acquisition.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following are the Supplementary data to this article: