Extraction method-dependent structure-function relationships of Astragali complanati semen protein: Implications for its application as a novel plant protein ingredient

Abstract

Astragali complanati semen is an underutilized protein-rich plant seed resource. Therefore, the effects of different extraction methods, i.e., ammonium sulfate precipitation (ASP), alkaline extraction (AE), and salt extraction (SE) on the yield, structure, and functional properties of Astragali complanati semen protein (ACSP) were investigated to evaluate its potential for food applications. The findings revealed that AE yielded a protein with a flexible structure, relatively stable conformation, high-purity protein (82.71 ± 0.41%), small particle size (239.50 ± 9.01 nm), and a mild odor. Its foaming capacity exceeded 100% at various pH levels, and its solubility was 87.81 ± 2.84% at pH 11. The least gelling concentration (14%) of AE protein was comparable to that of soy protein, exhibiting 88.54 ± 0.59% DPPH radical scavenging capacity at 2.5 mg/mL, and 52.60 ± 0.83% in vitro protein digestibility. In contrast, ASP exhibited superior foaming ability under acidic conditions, whereas SE yielded a higher amount of protein, with excellent oil-holding (4.91 ± 0.04 g/g) and emulsifying activity. Notably, the solubility of AE protein decreased by approximately 4% after 4 weeks of storage at −20 °C, with low levels of lipid and protein oxidation, demonstrating excellent storage stability. This study provides a theoretical foundation for the potential application of ACSP as a novel plant protein ingredient in the food industry.

Graphical abstract

Highlights

• •ACSP can be used as a novel protein in food products.
• •ASP has remarkable foaming capacity under acidic conditions.
• •AE exhibits superior functional properties and antioxidant capacity.
• •SE shows excellent oil-holding capacity and emulsifying properties.

1. Introduction

As the global demand for protein continues to grow, traditional animal-based proteins are no longer sufficient to meet current requirements because of their limited supply and high cost (Shang et al., 2026). Plant-based proteins have attracted widespread attention because they are inexpensive, highly available, sustainably produced, and can meet protein requirements while reducing environmental and health concerns (Li et al., 2026). Currently, plant proteins are extensively used to produce functional and value-added products, including dietary supplements, health foods, and pharmaceuticals (Olatunde et al., 2023). Consumption of plant proteins is associated with numerous health benefits, such as antioxidant and antibacterial effects, along with the regulation of appetite and lipid metabolism (Zhang, Wang, Cai, et al., 2023). However, the limited functional characteristics of plant proteins continue to pose a challenge to their application in the food sector. For example, peanut protein has not been fully utilized because of its poor functional properties, which result from its natural spherical structure and extraction conditions (Boukid, 2022). In addition, plant proteins may possess undesirable sensory attributes such as “bitterness,” “grassiness,” and “greenness,” which can affect consumer acceptance (Tang et al., 2024). Plant protein powders tend to aggregate and lose functional properties during storage, which affects protein quality (Zhang et al., 2025). Addressing these challenges necessitates investigating novel and underutilized plant protein resources, along with comprehensive studies on the effects of extraction conditions on the functional, structural, sensory characteristics, and storage stability of proteins.

Knowledge of the functional characteristics of plant proteins impacted by the extraction method is crucial for their successful application in the food industry (Dhiman et al., 2023). Plant protein extraction methods are classified as “wet” or “dry”. The “wet” method uses a solution to dissolve the protein, whereas the “dry” method primarily involves grinding and air or electrostatic classification (Sundaram et al., 2024). The most commonly used method for industrial food production involves alkaline extraction and acid precipitation, a wet processing technique that depends on the pH-dependent solubility of proteins (Li et al., 2026). Although this method produces high-purity protein, it results in a relatively low yield, and the extensive use of water and chemical reagents during extraction significantly affects the environment (Farooq & Yu, 2026). The salt extraction method is also a common wet-processing method, which separates proteins based on the “salting-in” principle, while preserving their natural structure and minimizing protein denaturation (Patel et al., 2025). The “salting-in” process involves using low-concentration salt ions to inhibit electrostatic interactions among proteins, thereby enhancing protein solubilization and potentially increasing extraction yields (Yang et al., 2023). Conversely, the ammonium sulfate precipitation method is based on the salting-out phenomenon, which operates under conditions of very high ionic strength, where the disruption of electrostatic and hydrophobic interactions reduces protein solubility and promotes aggregation. Although this method effectively precipitates protein components, it often results in contaminants being precipitated along with the target proteins (Jiang et al., 2021). Therefore, the choice of extraction method significantly influences protein recovery and preservation of structural integrity.

Seeds are the primary source of protein, accounting for approximately 31.3% of the total plant protein (Ravindran et al., 2024). However, plant seeds are commonly discarded as waste, fuel, or animal feed, and their potential as nutritional supplements is often overlooked (Marques et al., 2025). Extracting plant protein from agricultural and industrial byproducts and waste can not only meet the increasing protein demand but also ensure the full utilization of plant resources (Zongo et al., 2025). Astragali complanati semen (ACS) refers to the mature, dried seeds of the leguminous plant Astragalus complanatus R. Br. The seeds are kidney-shaped, smooth, light brown with a mild aroma and taste, and have a slight legume flavor (Xie et al., 2024). For millennia, ACS has been used in Asia as a traditional food ingredient, commonly brewed as tea or added to soups. ACS is rich in nutrients such as proteins, carbohydrates, and minerals. Additionally, ACS contains a variety of bioactive compounds, including terpenes, flavonoids, and fatty acids (Chen, Zhang, & Li, 2025). These compounds offer various health benefits, including anti-fatigue, lipid-lowering, and hepatoprotective effects (Fan et al., 2017). Based on these bioactivities, ACS-related health foods have been developed and produced industrially, with more than ten varieties currently registered and available in China, such as health wines for weight loss and health teas supporting immune function. Despite being a legume seed, research on the protein components of ACS has been limited. Our previous study revealed that the four protein fractions obtained from ACS exhibited varying degrees of antioxidant capacity, with the glutelin fraction also showing good functional properties (Cao et al., 2024). However, the extraction process involves multiple steps, protein recovery is low, and the cost of production on an industrial scale is high. Therefore, to enhance the industrial production and use of ACSP, this study compared the effects of the alkaline extraction, salt extraction, and ammonium sulfate precipitation methods on the structural, functional properties, sensory quality, antioxidant activity, and storage stability of ACSP. This research aims to elucidate the structure-function relationship and provide a scientific basis for the application of ACSP in plant-based foods, functional ingredients, and health foods, while also laying the foundation for the high-value and sustainable utilization of ACS as a plant seed resource.

2. Materials and methods

2.1. Extraction of ACSP

The ACS was pulverized using a grinder (1000Y, Bote Electric Co., Ltd., Zhejiang, China), and the resulting powder was sieved through a 0.125 mm mesh. Defatting was performed twice using petroleum ether for 12 h. The defatted ACS powder was air-dried in a fume hood for 24 h and stored at −20 °C for future use. ACSP was extracted from defatted ACS powder using three extraction methods. The extraction processes are shown in Fig. 1 and are detailed in the Supplementary Materials.

Fig. 1.

Procedure for preparing ACSP using different extraction methods.

2.2. Protein quality

2.2.1. Yield and protein content

The Kjeldahl method was used to determine the nitrogen content of the ACSP (Pokorski et al., 2025). Briefly, the total protein content was calculated by multiplying the nitrogen content by 6.25. The extraction yield and protein yield were calculated using the following equations:

$$\text{Extraction yield}\left(\%\right)=\frac{\text{Weight of protein product}\left(\mathrm{g}\right)}{\text{Weight of powder used for extraction}\left(\mathrm{g}\right)}\times 100$$ (1)

$$\text{Protein yield}\left(\%\right)=\frac{\text{The weight of protein in ACSP}\left(\mathrm{g}\right)}{\text{The weight of protein in}\mathrm{ACS}\text{powder}\left(\mathrm{g}\right)}\times 100$$ (2)

2.2.2. Amino acid composition analysis

Amino acid composition was determined using a previously described method (Cao et al., 2025), with minor modifications. The protein sample and 6.2 mol/L hydrochloric acid solution were sea led in a glass tube, into which nitrogen gas was introduced. This mixture was hydrolyzed at 100 °C for 24 h. After complete hydrolysis, the mixture was filtered through a 0.45 μm microporous filter. Sodium citrate buffer was then added to dissolve the solid residue completely, facilitating deacidification. Prior to injection, the samples were filtered through a 0.22 μm microporous filter, and the amino acid composition of the samples was analyzed in detail using an automated amino acid analyzer (Biochrom 30⁺; Biochrom Ltd., Cambridge, UK).

2.3. Sensory properties

2.3.1. Color

Images of the protein samples were captured under identical lighting conditions. A colorimeter (TS70, Sanenshi Technology Co., Ltd., Shenzhen, China) was used to measure the lightness (L*), redness (a*), and yellowness (b*) of the samples, following the method described previously (Teixeira et al., 2026).

2.3.2. Electronic nose (E-nose)

E-nose detection was performed according to a previously described method (Zhang et al., 2025). The e-nose (CNose-18, Shanghai Baosheng Technology Co., Ltd., China) was equipped with 18 gas sensors. Volatile compounds were analyzed by combining sensor array responses with pattern recognition technology. The protein samples were placed in headspace vials and allowed to equilibrate for 30 min. The target compounds for each of the 18 sensors (S1-S18) are listed in Supplementary Material Table S1.

2.3.3. Sensory evaluation

Sensory evaluation was performed following the method (Mortenson et al., 2008), with minor modifications. A panel of 10 trained assessors (seven women and three men, aged between 23 and 29) conducted the sensory evaluation using quantitative descriptive analysis (QDA). The panel evaluated the protein samples for aroma, taste, and mouthfeel, creating a standardized list of attributes (Table S2). The intensity of each attribute was rated on a 10 cm continuous scale, with 0 indicating the lowest intensity and 10 the highest intensity. For aroma assessment, 1 g of each protein sample was placed in opaque glass vials, labeled with random three-digit numbers, and presented randomly to the evaluators for scoring based on their smell. For taste and mouthfeel evaluation, 15 mL of protein solution (5%, w/v) was placed in an opaque cup. The evaluators held the solution in their mouth for approximately 5 s before spitting it out without swallowing. The sensory profiles of the samples were depicted using a radar chart, and the relationships between the samples and sensory attributes were statistically analyzed using principal component analysis (PCA). The protocol was conducted in accordance with the ethical principles of the World Medical Association (Declaration of Helsinki) for experiments involving human participants. The foods tested were safe and hygienic, and participation was completely voluntary. This study involved only noninvasive sensory testing and, therefore, did not require ethical review. All data were collected anonymously.

2.4. Molecular and structural characteristics

2.4.1. Particle size measurement

Protein particle size was measured using a nanoparticle size analyzer (NS-90Z Plus, OMEC Instruments Co., Ltd., Zhuhai, China). Protein sample solutions were prepared at 1 mg/mL.

2.4.2. Sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE)

Sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) was conducted according to a previously described method (Yue et al., 2025), with minor modifications. The samples were mixed with the loading buffer containing dithiothreitol (DTT) and inactivated at 100 °C for 5 min. Next, 10 μL of the protein markers (10–180 kDa) and protein samples (10 mg/mL) were placed into the loading wells. Electrophoresis was performed using a PowerPac Basic power supply (Bio-Rad Laboratories, Inc., Hercules, USA) at constant voltages of 70 V and 145 V. The gel was stained with Coomassie Blue R250 for 40 min after electrophoresis. Subsequently, the decolorization step was performed. The iBright™ Imaging System (FL1500, Thermo Fisher Scientific Inc., USA) was used to capture images. The protein molecular weights were calculated using the Gel Molecular Weight Analyzer App within OriginPro 2024b software (OriginLab Corporation, USA).

2.4.3. Fourier transform infrared spectroscopy (FTIR)

Fourier transform infrared spectroscopy (FTIR) of the samples was performed using a spectrometer (Nicolet iS50, Thermo Fisher Scientific Inc., USA). The spectra were recorded in the wavelength range of 4000–400 cm^-1.

2.4.4. Circular dichroism (CD)

Circular dichroism (CD) analysis was performed following a previously described method (Kwaambwa & Maikokera, 2008), with slight modifications. The CD spectra of the samples were recorded using a CD spectrometer (MOS-500, Bio-Logic Scientific Instruments, Claix, France). Measurements were taken using a 1-mm path length quartz cuvette covering a scanning range of 190–260 nm. The secondary structure of the samples was analyzed using the CONTIN-LL algorithm on the DichroWeb website (http://dichroweb.cryst.bbk.ac.uk).

2.4.5. X-ray diffraction(XRD)

X-ray diffraction (XRD) was performed according to a published method (Sadaf et al., 2024), with minor modifications. The XRD patterns of the proteins were recorded using an X-ray diffractometer (D8 ADVANCE, Bruker, Germany). Copper was used as the target material, and the scanning range was 5°–90°. The crystallinity index (%) was calculated by analyzing the peak fitting using OriginPro 2024b software (OriginLab Corporation, USA) to determine the peak area ratio.

2.4.6. Scanning electron microscopy (SEM)

The surface morphology of the samples was examined according to a previously described method (Jiang et al., 2026), with minor modifications, using a scanning electron microscope (SEM, Sigma 360, Zeiss, Germany). A 10 mg sample was placed on a specimen holder and coated with a gold layer using an ion sputtering system. The spectra were recorded at magnifications of 200× and 2,000×.

2.5. Functional properties

2.5.1. Solubility and ζ-potential

Solubility was assessed following a previously reported method (Mir et al., 2019), with minor modifications. A 10 mg/mL sample solution was adjusted to different pH values (pH 3–11) using 1 M NaOH or HCl. After centrifugation, the supernatant was collected. Solubility was calculated by dividing the protein content in the supernatant by the total protein content in the sample. The ζ-potential of the samples was measured using a zeta potential analyzer (NS-90Z Plus, Zhuhai OMEC Instruments Co., Ltd., China).

2.5.2. Water holding capacity (WHC) and oil holding capacity (OHC)

The WHC and OHC were determined according to the previous method (Hadnadev et al., 2018), with minor modifications. A 100 mg sample was combined with 10 mL ultrapure water or oil in a centrifuge tube. After vortexing for 10 min, the mixture was centrifuged at 8,000 × g for 30 min and the supernatant was discarded. The WHC and OHC were calculated using the following formulas:

$$\mathrm{WHC}\text{or}\mathrm{OHC}\left(\mathrm{g}/\mathrm{g}\right)=\frac{{\mathrm{W}}_2-{\mathrm{W}}_1}{{\mathrm{W}}_0}$$ (3)

W₀ is the weight of the sample, W₁ is the total weight of the sample with the centrifuge tube, and W₂ is the weight of the centrifuge tube with the sample precipitated after removal of water or oil.

2.5.3. Foaming capacity (FC) and foam stability (FS)

Foaming capacity (FC) and foam stability (FS) were measured according to the previously described method (Higa et al., 2024), with some modifications. A 10 mg/mL sample solution was adjusted to different pH values (pH 3–11) using 1 M NaOH or HCl. The mixture was homogenized for 2 min using a homogenizer (HR-500DG, HUXI Corporation, Shanghai, China). The following formula was used to determine FC and FS:

$$\mathrm{FC}\left(\%\right)=\frac{{\mathrm{V}}_1-{\mathrm{V}}_0}{{\mathrm{V}}_0}\times 100$$ (4)

$$\mathrm{FS}\left(\%\right)=\frac{{\mathrm{V}}_2-{\mathrm{V}}_0}{{\mathrm{V}}_1-{\mathrm{V}}_0}\times 100$$ (5)

V₀ and V₁ are the volumes before and after foaming, respectively; V₂ is the volume after 30 min of foaming.

2.5.4. Emulsifying activity index (EAI) and emulsion stability index (ESI)

The emulsifying activity index (EAI) and emulsion stability index (ESI) were measured according to a previously described method (Ye et al., 2024), with slight modifications. Soybean oil (5 mL) and the sample solution (15 mL) were mixed and homogenized for 3 min. The emulsion formed was pipetted out from the mixture (50 μL) at 0 and 10 min after emulsification and diluted with 0.1 % SDS to 5 mL. The absorbance was measured at 500 nm using a UV spectrophotometer (UV-3600i Plus, Shimadzu, Japan). The EAI and ESI were calculated using the following equations:

$$\mathrm{EAI}\left(\mathrm{m}2/\mathrm{g}\right)=\frac{2\times 2.303\times {\mathrm{A}}_0\times \mathrm{D}}{\mathrm{C}\times \left(1-\mathrm{\phi }\right)\times {10}^4}$$ (6)

$$\mathrm{ESI}\left(\%\right)=\frac{{\mathrm{A}}_0}{{\mathrm{A}}_0-{\mathrm{A}}_{10}}\times \mathrm{t}$$ (7)

A₀ and A₁₀ are the absorbances measured at 0 and 10 min after homogenization; D is the dilution factor (100); C is the sample (mg/mL); φ is the oil volume fraction (0.25); t is time (10 min).

2.5.5. Least gelling concentration (LGC)

The least gelling concentration (LGC) was determined according to a previously described method (Shen et al., 2025). Briefly, the samples were prepared at concentrations ranging from 10% to 20% (w/w) and heated in an oven at 100 °C for 30 min. Subsequently, the samples were incubated overnight at 4 °C. To measure the LGC of the protein, the sample tubes were inverted, and the samples were classified as gels if no flow was observed within 30 s after inversion.

2.6. In vitro protein digestibility(IVPD)

The IVPD was performed according to the previously described method (You et al., 2010), with minor modifications. This test was conducted using the INFOGEST method, with the detailed protocol described in Method 1.3 in the Supplementary Materials.

2.7. Antioxidant activities

DPPH and ABTS⁺ radical scavenging capacities and reducing power were determined using the previously described method (Sharifi et al., 2025), with a slight modification. The detailed method is described in Method 1.4 of the Supplementary Materials.

2.8. Storage stability measurement

Storage stability was assessed according to the previously described method (Zhang, Wang, Chu, et al., 2023), with minor modifications. Protein samples were stored in sealed centrifuge tubes at −20 °C for 0, 1, 2, and 4 weeks. The degree of lipid oxidation during this storage was determined using the thiobarbituric acid (TBARS) method and measured using the Malondialdehyde (MDA) Content Assay Kit (Beijing Solarbio Science & Technology Co., Ltd.). Protein oxidation was determined by measuring the carbonyl group content using the Protein Carbonyl Content Assay Kit (Beijing Solarbio Science & Technology Co., Ltd.). The solubility of the samples during storage was determined by the method described in 2.5.1.

2.9. Statistical analysis

Each experiment was conducted with a minimum of three repetitions. Data are presented as mean ± standard deviation. To identify statistically significant differences (p < 0.05) among the samples, a one-way analysis of variance (ANOVA) was performed, followed by Duncan’s multiple range test using IBM SPSS Statistics 20 (SPSS, Chicago, USA).

3. Results and discussion

3.1. Protein content and yield of ACSP from different extraction methods

The yield and protein content of the ACSP obtained from different extraction methods are shown in Fig. 2A. The protein contents of ASP, AE, and SE were 30.42 ± 0.09%, 82.71 ± 0.41%, and 44.41 ± 1.33%, respectively. The highest protein content in AE may be attributed to the alkaline extraction condition, which disrupts disulfide bonds and facilitates the ionization of acidic and neutral amino acids, thus increasing protein solubility (Bainto-Ancheta et al., 2025). The protein yields for ASP, AE, and SE were 10.78 ± 1.06%, 18.85 ± 0.74%, and 29.71 ± 0.38%, respectively. The extraction yields for ASP, AE, and SE were 5.37 ± 0.65%, 3.44 ± 0.08%, 10.12 ± 0.35%, respectively (Table S3). The highest extraction yield from SE aligns with previous findings on Trichosanthes kirilowii seed protein, demonstrating that salt extraction achieves a greater yield than other extraction methods (Peng et al., 2024). The conditions for salt extraction were mild, leading to minimal protein loss and thus resulting in high extraction yield and protein recovery (Yang et al., 2021). These findings indicate that the three extraction methods significantly differ in yield, protein recovery, and purity. SE achieves the highest yield and recovery, whereas AE has the highest protein purity.

Fig. 2.

(A) Protein yield and content of ACSP obtained by different extraction methods, (B) Color, (C) Electronic nose detection, (D) PCA analysis utilizing electronic nose data, (E) Sensory profile, (F) Biplot of the sensory profile. Values are expressed as the means ± standard deviation, and different letters denote statistically significant differences (p < 0.05). ASP, ammonium sulfate precipitation; AE, alkaline extraction; SE, salt extraction. *smell; **taste; ***mouthfeel.

3.2. Amino acid composition of ACSP obtained by different extraction methods

Table 1 presents the amino acid composition of the protein samples. ACSP comprises 17 amino acids, including 7 essential amino acids (EAA). Tryptophan was not detected in the sample, likely owing to its degradation under hydrolytic conditions, and a similar phenomenon was observed in safflower seed meal protein (Yang et al., 2026). The total amino acid (TAA) and EAA contents were ranked as follows: AE > SE > ASP. Among all protein samples, Glu was the most abundant, followed by Asp and Arg. Previous studies have reported that Asp and Glu are common flavor-enhancing amino acids (Guo et al., 2025), suggesting that ACSP has potential as an umami ingredient in food products. AE demonstrated the highest EAA content, with relatively high levels of Leu, Phe, and His, which are key precursors of bioactive peptides (Valdez-Arana et al., 2025). Both AE and SE met the recommended EAA intake values recommended by the FAO for older adults (de Oliveira et al., 2025) (Table 1). Furthermore, the EAA/TAA ratio of AE was 28.19%, which aligns with the WHO/FAO/UNU recommended daily protein intake of 0.66 g/kg body weight for adults (Gorissen et al., 2018), suggesting that ACSP is a high-potential new dietary protein source.

Table 1.

Amino acid profiles of ACSP.

Amino acids (g/100 g)	ASP	AE	SE	FAO recommendationa (2011)
Aspartic Acid (Asp)	2.28 ± 0.04c	6.21 ± 0.24a	3.41 ± 0.06b
Threonine (Thr)⁎	1.29 ± 0.02c	2.15 ± 0.05a	1.61 ± 0.02b	0.25
Serine (Ser)	1.08 ± 0.04c	3.33 ± 0.08a	1.56 ± 0.02b
Glutamic acid (Glu)	4.77 ± 0.06c	16.73 ± 0.20a	8.06 ± 0.14b
Glycine (Gly)	1.27 ± 0.02c	2.99 ± 0.02a	1.83 ± 0.01b
Alanine (Ala)	0.89 ± 0.02c	1.95 ± 0.11a	1.06 ± 0.01b
Cystine (Cys)	0.92 ± 0.02c	1.04 ± 0.03b	1.82 ± 0.03a
Valine (Val)⁎	0.83 ± 0.01c	2.19 ± 0.03a	1.24 ± 0.07b	0.40
Methionine (Met)⁎	0.19 ± 0.004c	0.51 ± 0.01a	0.33 ± 0.01b	0.23
Isoleucine (Ile)⁎	0.77 ± 0.02c	2.03 ± 0.01a	1.44 ± 0.02b	0.30
Leucine (Leu)⁎	0.98 ± 0.03c	3.62 ± 0.07a	1.51 ± 0.01b	0.61
Tyrosine (Tyr)	0.83 ± 0.02c	1.89 ± 0.02a	1.42 ± 0.07b
Phenylalanine (Phe)⁎	0.75 ± 0.01c	3.15 ± 0.04a	1.32 ± 0.04b	0.41
Histidine (His)⁎	0.77 ± 0.01c	2.38 ± 0.01a	1.35 ± 0.03b	0.16
Lysine (Lys)⁎	1.39 ± 0.04c	2.23 ± 0.02b	2.82 ± 0.09a	0.48
Arginine (Arg)	2.11 ± 0.03c	8.74 ± 0.25a	4.02 ± 0.18b
Proline (Pro)	1.01 ± 0.05c	3.63 ± 0.04a	2.52 ± 0.05b
Total amino acids	22.13 ± 0.23c	64.78 ± 0.92a	37.33 ± 0.64b

Values are presented as the means ± standard deviation, and different letters denote statistically significant differences (p < 0.05).

^⁎Essential amino acids.

^aOlder children, adolescents, and adults.

3.3. Sensory properties of ACSP obtained by different extraction methods

3.3.1. Color

The color of the protein is a crucial marker for assessing the visual quality of the product, as it directly affects sensory appeal and consumer acceptance (Deen et al., 2025). Fig. 2B presents images and color metrics of the ACSP obtained using different extraction methods. ASP (left) appeared light brown, AE (middle) was a grayish-brown powder, and SE (right) was brownish-yellow. The extraction method notably affected the lightness (L*) of all protein samples. AE exhibited the darkest color, with the lowest L* value (33.23 ± 0.04) and relatively low a* (−1.07 ± 0.02) and b* (7.53 ± 0.01) values. Previous studies have shown that silflower protein isolate obtained by alkaline extraction was also grayish-brown in color (Teng et al., 2026). This browning might result from the oxidation of polyphenols under alkaline conditions and the incomplete removal of pigments during protein precipitation at the pI, causing darkening of the protein (Carrillo et al., 2024). Compared with AE and SE, ASP had a higher L* value and a lighter color, indicating a lower pigment content (Zhang et al., 2017), resulting in a lighter color. SE exhibited a higher b* value (13.97 ± 0.11) and appeared dark brownish-yellow. These findings imply the significant influence of the extraction method on the color parameters of ACSP, thereby affecting its visual characteristics.

3.3.2. E-nose

The E-nose can mimic human olfaction and rapidly analyze the flavor characteristics of samples (Chen, Dai, et al., 2025). Fig. 2C shows the responses of the 18 sensors to volatile compounds in the samples. A positive correlation between the amount of volatile compounds present in the samples and their response values on the sensors. Overall, the radar chart profiles of all samples displayed distinct characteristics, with ASP and SE showing higher response values across most sensors owing to the elevated volatile compound content in these samples and richer aromatic composition (Zhang, Wu, Zhang, Thakur, et al., 2025). The overall sensor response values for AE were lower, indicating that the alkaline extraction method can effectively improve the flavor characteristics of ACSP. Similar findings have been reported in earlier studies (Zhang et al., 2024).

The PCA data in Fig. 2D show that PC1 and PC2 accounted for 93.27% of the variance. The protein samples were distinctly separated, consistent with the findings from the radar chart. Further analysis of the sensor response values revealed that ASP had stronger correlations with sensors S1, S6, S8, S12, S14, S15, and S18, suggesting higher levels of aldehydes and ketones, short-chain alkanes, alcohols, and carbon-containing compounds. SE exhibited stronger correlations with sensors S4, S5, S6, S8, S12, S14, and S18, indicating increased levels of sulfides, nitrogenous compounds, acetones, short-chain alkanes, and alcohols. The higher levels of sulfides, alcohols, and alkanes in ASP and SE may be attributed to the minor oxidation of methionine and lipids (Zhang et al., 2025). Consequently, AE displayed the lowest overall response value, and the total area of the radar plot decreased significantly, suggesting that AE had the lowest volatile content. This reduction could be because alkaline conditions cause many odor-causing compounds to chemically transform or degrade (Modupalli et al., 2024), thereby significantly lowering volatile organic compounds and contributing to a cleaner and more neutral aroma profile.

3.3.3. Sensory evaluation

The radar chart and PCA biplot illustrating the sensory evaluation of protein samples are shown in Fig. 2E and F. The PCA biplot showing the relationship between protein samples and sensory attributes indicates that the combined contribution of PC1 (82.7%) and PC2 (17.3%) is 100%. ASP is positioned on the positive side of PC1 and shows a positive association with astringent, salty, and nutty attributes. Green and grainy are located on the negative side of PC2 and are positively associated with SE. Both ASP and SE exhibited strong overall aromatic profiles, with relatively high scores in spicy, green, and grainy attributes ranging from 2.54 to 6.39. These aromatic traits might be associated with volatile compounds such as aldehydes and ketones (Fig. 2C) (Shen et al., 2026). The astringent and salty characteristics of ASP could be attributed to residual polyphenols and salts from the extraction process (Ammam et al., 2025). In contrast, AE is positioned on the negative side of PC1 and is positively associated with beany and sweet attributes. AE exhibited relatively low overall sensory attribute scores, ranging from 0.18 to 3.22. Evaluators consistently noted that AE had a mild aroma with a subtle beany note, an almost flavorless mouthfeel, and a sweet aftertaste. It has been reported that lipoxygenases facilitate the production of lipid peroxides and volatile compounds, including aldehydes and alcohols. Alkaline extraction conditions can effectively suppress lipoxygenase activity (Lakshmipathy et al., 2025). Additionally, after isoelectric precipitation, carbonyl-containing carbohydrates can be removed from the protein (Cui et al., 2020), potentially reducing the development of off-flavors. These findings are consistent with the E-nose results, where AE exhibited the lowest overall sensor responses.

3.4. Structural properties of ACSP obtained by different extraction methods

3.4.1. Particle size

Particle size and the PDI are crucial metrics for evaluating the physical stability and dispersibility of proteins (Karabulut et al., 2025). Fig. 3A shows the particle size and PDI of the protein samples. AE exhibited the smallest average particle size (239.50 ± 9.01 nm) and PDI (0.26 ± 0.04). Under alkaline extraction conditions, the structure of ACSP partially unfolds, and the hydrogen bonds and/or electrostatic interactions within the structure are disrupted, leading to a loss of order and smaller particle sizes (Jeganathan et al., 2024). Previous studies have shown that smaller protein particle sizes correlate with enhanced solubility and gastrointestinal mobility (Guevara-Zambrano et al., 2024). Consequently, AE might exhibit superior functional properties and bioavailability. Conversely, SE showed the largest particle size (695.97 ± 40.75 nm) and PDI (0.91 ± 0.03). While salt extraction preserves the natural conformation of proteins, the protein molecules aggregate into larger particles through interactions with salt ions or the formation of salt bridges (Ajayi et al., 2025). The particle size (560.80 ± 4.78 nm) and PDI (0.54 ± 0.04) of ASP were intermediate. This can be explained by the presence of salt ions enhancing competition among protein molecules for binding to water molecules, causing the proteins to aggregate primarily through hydrophobic interactions (Edward et al., 2023), thereby resulting in moderate particle sizes. Overall, different extraction methods led to unique protein aggregation patterns, intermolecular interactions, and conformational shifts, resulting in differences in particle size across different protein samples.

Fig. 3.

Molecular and structural characteristics of ACSP obtained by different extraction methods: (A) Particle size, (B) SDS-PAGE (lane a for molecular weight marker, lanes b, c, and d represent ASP, AE, and SE, respectively), (C) FT-IR spectra, (D) Protein secondary structure, (E) CD spectra, (F) X-ray Diffraction. ASP, ammonium sulfate precipitation; AE, alkaline extraction; SE, salt extraction

3.4.2. SDS-PAGE

Fig. 3B shows the effect of the extraction method on the SDS-PAGE profile of ACSP, alongside the migration patterns of protein markers with known molecular weights (10–180 kDa). All protein samples exhibited multiple bands within the 13–97 kDa molecular weight range, which is similar to that of yellow pea proteins (Lao et al., 2023). AE exhibited the highest number of protein bands, comprising 11 polypeptide chains, while ASP and SE showed similar band patterns, each with 9 polypeptide chains. CD spectroscopy indicated that the α-helix content in the AE sample was lower than that in the ASP and SE samples, suggesting protein structural unfolding (Zheng et al., 2020). AE showed new distinct bands around 22 and 37 kDa, indicating partial degradation of high-molecular-weight proteins in ACSP. Moreover, broad and intense diffuse bands were noted at approximately 20 kDa in AE, signifying a high content of low-molecular-weight proteins. All samples exhibited five bright bands near 18, 20, 39, 48, and 63 kDa.

3.4.3. FTIR

Fig. 3C shows the FTIR spectra of the protein samples. All protein samples showed a broad and intense absorption peak near the amide band at 3410 cm^-1, which corresponds to the intramolecular O-H stretching vibration (Kantakul et al., 2024). In the 2900-3000 cm^-1 range, the absorption peaks correspond to C-H bonds (Goormaghtigh et al., 2006). Notably, SE displayed an absorption peak near 2930 cm^-1, while ASP and AE exhibited weak absorption peaks near 2920 cm^-1 and 2910 cm^-1, respectively, both exhibiting significant blue shifts. This indicates that the extraction method affects the C-H stretching vibrations of the protein, thereby altering its hydrophobic microenvironment (Kim et al., 2026).

All protein samples showed a distinct peak at approximately 1650 cm^-1, which was attributed to the C=O stretching vibration of the peptide bond, signifying the presence of ordered structures (Zhu et al., 2024). Additionally, the absorption peak near 1540 cm^-1 corresponds to the N-H (Liu et al., 2024). Compared to the other samples, AE exhibited a pronounced peak near 1410 cm^-1, likely resulting from the symmetric stretching vibration of the carboxylate group (–COO^-) and may be related to the higher solubility of AE. The carboxylate group being highly hydrophilic and anionic, generates electrostatic repulsion that can inhibit protein aggregation or precipitation (Diaz-Espinoza et al., 2017). Moreover, the peak observed near 1240 cm^-1 corresponds to N-H bending vibrations (Li et al., 2024), while the absorption peak near 618 cm^-1 corresponds to the C-H bending vibration, likely caused by the amino acid side chains (Patil et al., 2024). These findings indicate that the higher-order conformation of ACSP obtained using different extraction methods was altered, thereby impacting its functional properties.

3.4.4. CD

The CD spectra of the protein samples are shown in Fig. 3D. ASP and SE showed positive molar ellipticity peaks near 194 nm and negative peaks at approximately 213 nm and 209 nm, respectively, which correspond to β-sheets. Moreover, SE exhibited a negative peak near 194 nm, and the peak with reduced molar ellipticity appearing near 213 nm corresponds to a disordered structure (Jeganathan et al., 2024). AE displayed a positive molar ellipticity peak at approximately 198 nm and a negative peak at near 222 nm, which are associated with the α-helix structure. Notably, AE also exhibited a strong negative molar circularity peak near 237 nm, which is likely due to the presence of regular π-π stacking (Wang et al., 2017), suggesting the presence of β-sheets or aromatic amino acid residues (Ranjbar & Gill, 2009).

Fig. 3E presents the secondary structure composition of the protein samples. Overall, ASP and SE primarily consisted of α-helices and β-turns. In contrast, the secondary structure of AE was dominated by β-sheets (39.36 ± 0.35%) and random coils (28.51 ± 0.59%), with significantly lower percentages of α-helices (13.69 ± 0.29%) and β-turns (18.44 ± 0.05%) compared to ASP and SE, while the percentages of β-sheets and random coils were relatively high. The increase in α-helices enhances structural rigidity, whereas a decrease in α-helices leads to increased molecular disorder (Günal-Köroglu et al., 2023). Under alkaline extraction conditions, the ordered structure of ACSP (such as the α-helix) was disrupted, hydrogen bonds were broken, allowing the exposed polypeptide chains to rearrange through intermolecular hydrogen bonding, partially forming stable β-sheet structures (Kong et al., 2026). β-sheets are relatively extended and ordered structures (Ma et al., 2025). A greater proportion of β-sheets may enhance hydrogen bonding between protein molecules, thereby promoting the formation of a more stable gel network (Kazemi-Taskooh & Varidi, 2023). Additionally, the increased content of random coils indicated that the internal structure of AE became loose and more flexible, potentially facilitating the adsorption of protein molecules at interfaces and enhancing protein digestion (Othmeni et al., 2025). In contrast, ASP and SE exhibited higher levels of α-helices and β-turns, resulting in more compact and rigid structures. Furthermore, the higher proportion of β-turns may help anchor the protein effectively at the oil-water interface and enhance emulsion stability (Kim et al., 2024). These findings suggest that the secondary structures in protein samples influence their functional properties.

3.4.5. SEM

Fig. 4 shows the surface morphology of the protein samples. The ASP surface exhibited irregular, flake-like structures of different sizes. This is attributed to the interactions between proteins during the freeze-drying process, and surface erosion caused by prolonged freeze-drying (Akyüz & Ersus, 2026). Under high-magnification microscopy, the ASP surface appeared uneven, with small particles and strip-like formations. Under low magnification, AE showed relatively fine, irregular flake-like and rod-like structures with cross-linking arms. These fine fragmented structures may result from partial denaturation of the protein under alkaline extraction conditions, leading to the unfolding of the secondary structure, breaking hydrogen bonds and forming smaller fragments, thereby reducing the particle size (Yin et al., 2025). Particle size analysis also revealed that AE had the smallest average particle size (239.50 ± 9.01 nm). At 2,000× magnification, the AE surface appeared smooth with some wrinkles, indicating high purity (Li, Li, Khedr, & Du, 2026). The SE surface exhibited a porous, lamellar structure; this network-like configuration enhances the penetration of water or oil, resulting in superior OHC/WHC. At 2,000× magnification, the SE surface appeared rough, with micro-particles and spherical protrusions.

Fig. 4.

SEM images of ACSP obtained by different extraction methods

3.4.6. XRD

XRD patterns are commonly used to assess the presence of crystalline structures in proteins. Sharp and narrow peaks in the pattern typically signify the presence of a crystalline structure, whereas broad, diffuse peaks suggest an amorphous structure (Senthilkumar et al., 2026). Fig. 3F shows the XRD patterns of the samples. Both AE and SE displayed broad peaks at 2θ ≈ 20°, indicating that they are primarily amorphous. However, AE also exhibited several narrow, sharp peaks between 25° and 80°, suggesting a small amount of crystalline structure. However, ASP exhibited numerous diffraction peaks with varying intensities from 15° to 60°, indicating a high level of crystallinity.

The crystallinity of the protein samples was as follows: ASP (69.89%) > AE (18.93%) > SE (9.66%). AE and SE exhibited the typical amorphous structure of plant proteins, which enhances hydration and solubility compared to that of crystalline proteins (Singh et al., 2026). Structural flexibility is crucial for the functional properties of proteins, and amorphous structures provide this flexibility (Ghasemi et al., 2025). The structure of AE is rich in β-sheet structures, which confer relatively high conformational stability (Li et al., 2026). These structural features form the basis of its functional properties.

3.5. Functional properties of ACSP obtained by different extraction methods

3.5.1. Solubility and ζ-potential

Protein solubility directly affects its use as a food ingredient in liquid products (Yao et al., 2023). The solubility of all protein samples gradually increased with rising pH, with the lowest solubility occurring at pH 5, forming a characteristic U-shaped solubility curve. Similar pH-solubility patterns have been reported in earlier studies on bitter melon seed proteins (Horax et al., 2011). Protein samples exhibited the lowest solubility at pH 5 (near pI), where protein-water interactions weakened, prompting protein aggregation and reduced solubility. At pH 5, the solubility of ASP, AE, and SE decreased to 22.39 ± 1.72%, 12.31 ± 2.84%, and 19.79 ± 1.80%, respectively. Subsequently, the protein surface charge increased as the pH shifted away from the pI, leading to greater electrostatic repulsion and enhanced interactions between the protein and water (Li et al., 2022), thereby enhancing solubility. Among all protein samples, AE exhibited the highest solubility at pH 3, 7, 9, and 11, peaking at 87.81 ± 2.84% at pH 11. The AE structure rich in random coils experienced a decrease in α-helices and protein unfolding under alkaline extraction conditions, which exposed hydrophilic groups and enhanced interactions with water (Adiamo et al., 2023), thereby increasing its solubility. The superior solubility of AE offers distinct advantages for applications such as plant-based protein beverages.

The stability of proteins plays a crucial role in food processing by preventing their aggregation and precipitation. Proteins with exposed charged groups can enhance the stability of protein solutions by increasing the electrostatic repulsion between particles, thereby preventing aggregation (Kelany et al., 2024). The ζ-potentials of the protein samples across the pH range of 3–11 are shown in Fig. 5B. The ζ-potential values of all samples gradually increased with increasing pH values. The pH values corresponding to the ζ-potential of 0 for ASP, AE, and SE were 3.52, 3.95, and 4.56, respectively, aligning with the pI of the proteins. The ζ-potentials of ASP, AE, and SE at pH 7 were −17.29 ± 0.72, −22.47 ± 0.69, and −19.44 ± 1.09 mV, respectively. AE exhibited the highest ζ-potential values across all pH levels, reaching −31.72 ± 1.53 mV at pH 11. This could be attributed to the use of alkaline solvents during extraction, which unfolds the protein structure, exposing more hydrophobic groups (Hu et al., 2023), thereby enhancing electrostatic repulsion and reducing protein aggregation and precipitation. These results suggest that AE demonstrates superior solubility across a range of pH conditions, offering greater benefits in food processing.

Fig. 5.

(A) Solubility and (B) ζ-potential of ACSP obtained using different extraction methods at various pH levels.

3.5.2. FC and FS

FC is the foaming capacity of a protein to generate bubbles, whereas FS refers to the duration for which these bubbles persist (Bebartta et al., 2025). Fig. 6A and B show the FC and FS of the protein samples across a pH range of 3 to 11. Under acidic conditions, ASP exhibited higher FC values, specifically 159.08 ± 3.42% and 141.79 ± 3.24% at pH 3 and 5, respectively. At pH values near the pI, the net charge of ASP was relatively low, reducing electrostatic repulsion among proteins. This reduction facilitated aggregation through hydrophobic interactions and enabled swift adsorption at the air-water interface (Foo et al., 2026), thereby increasing its FC. AE showed higher values at pH 7, 9 and 11, and FC increased gradually with rising pH, reaching a maximum of 156.15 ± 2.28% at pH 11. AE’s FC value exceeded 100% under various pH conditions, surpassing previously reported values for sunflower protein (Huang et al., 2024). Alkaline extraction partial unfolded AE, exposing hydrophilic and hydrophobic groups and increasing random coil content, which enhanced protein flexibility (Huerta-Rodriguez et al., 2025). This flexibility allowed proteins to rapidly adsorb at the gas-water interface (Katekhong & Phuangjit, 2024). Simultaneously, AE’s smaller particle size and higher solubility enhanced foam dispersibility and supported the formation of a continuous interfacial film (Jiang et al., 2026), aiding bubble formation.

Fig. 6.

Functional properties of the protein samples at various pH levels: (A) FC, (B) FS, (C) EAI, (D) ESI. Values are expressed as mean ± standard deviation, with different letters indicating statistically significant differences (p < 0.05) at the same pH value.

At pH 3, the FS values of ASP, AE and SE were 63.89 ± 1.20%, 60.89 ± 3.27% and 63.32 ± 3.86%, respectively, with no significant differences (p > 0.05). AE displayed the highest FS at pH 5 and 7, with values of 81.01 ± 0.82% and 89.10 ± 4.19%, respectively. AE exhibited higher FS under both acidic and neutral pH conditions. This can be attributed to the higher β-sheet content in AE resulted from protein unfolding during the extraction process, which results in an adsorption film at the air-water interface with a greater elastic modulus (Dur et al., 2025), thereby enhancing FS. At pH levels of 9–11, all proteins exhibited reduced FS, with values of all protein samples remaining below 50%. This reduction may be attributed to the loosening of protein structures in alkaline conditions, which limited molecular interactions between protein molecules and weakens interfacial films (Cao et al., 2024), thereby reducing FS. ASP’s high FC under acidic conditions makes it suitable as a foaming ingredient in products such as beverages and mousses. AE demonstrated strong FC across all pH conditions and exhibited excellent FS in acidic and neutral conditions, making it more suitable for neutral to slightly acidic food systems, such as plant-based yogurt and baked goods.

3.5.3. EAI and ESI

Plant proteins are extensively researched as emulsifiers because they rapidly adsorb onto oil droplets, thereby reducing interfacial tension and prevent flocculation through strong steric hindrance and electrostatic repulsion (Hadidi et al., 2023). Fig. 6C and D show the EAI and ESI of the protein samples at pH 3 to 11. Overall, the EAI and ESI of the protein samples showed a pH-dependent pattern similar to their solubility curves. At the pI, the solubility of proteins was very low, impeding their diffusion to the water-oil interface and preventing the formation of a film around the oil droplets in an O/W emulsion (Illingworth et al., 2024). As the pH moved away from the pI, both EAI and ESI gradually increased. SE exhibited the highest EAI (47.41 ± 3.70 m²/g) and ESI (94.02 ± 2.95%) at pH 11. SE’s superior emulsifying properties at most pH levels could be attributed to its higher surface hydrophobicity, enabling rapid adsorption at the oil-water interface and decreases interfacial tension (He & Zhou, 2026). The EAI values of ASP at pH 5 and 7 were 10.58 ± 1.12 m²/g and 17.07 ± 0.92 m²/g, respectively, while its ESI (53.01 ± 2.22%) was higher at pH 3. Both ASP and SE had high α-helix and β-turn contents, and this compact conformation may contribute to the formation of a relatively dense, ordered interfacial film following adsorption at the interface (Yan et al., 2026). In contrast, AE exhibited a relatively high EAI (23.47 ± 2.95 m²/g) only at pH 9. This phenomenon may be attributed to its high WHC (3.21 ± 0.17 g/g), fewer hydrophobic clusters, and high solubility, which may limit its interaction with the lipid phase and reduce its anchoring capability at the oil-water interface (Mundi & Aluko, 2012). SE’s excellent emulsifying properties make it suitable for foods requiring emulsification, such as meat products and mayonnaise.

3.5.4. WHC and OHC

The WHC and OHC of protein samples are directly related to the texture and flavor retention of food (Hosseini et al., 2025). WHC of proteins refers to the water-holding capacity of the food matrix, while OHC refers to their ability to hold oil (Sharma et al., 2023). The WHC values of ASP, AE, and SE were 0.93 ± 0.09 g/g, 3.21 ± 0.17 g/g, and 2.74 ± 0.08 g/g, respectively (Fig. 7A). The WHC of AE surpassed that of pea protein obtained through alkaline extraction (2.4–2.6 g/g) (Stone et al., 2015). AE exhibited excellent WHC, potentially because the alkaline extraction process exposes water-binding regions on the protein’s side chains, coupled with AE’s abundance of polar amino acids(such as Arg, Glu and Asp) (Table 1). The side chains of these amino acid residues possess stronger hydrogen bonding and electrostatic interactions with water molecules (Bagiyal et al., 2026). Simultaneously, the smaller particle size of AE increases its surface area, providing more water-binding sites and thereby enhancing the overall hydrophilicity of the protein. Moreover, previous studies have reported that WHC of proteins extracted using alkaline methods is superior to other methods such as salt extraction (Shao et al., 2024). The high WHC of AE is more suitable for plant-based dairy products, baked goods, and other foods that require moisture retention.

Fig. 7.

(A) Oil and water holding capacity, (B) LGC, (C) IVPD of the protein samples. Values are means ± standard deviation, and different letters denote statistically significant differences (p < 0.05).

The OHC values of ASP, AE, and SE were 3.10 ± 0.14 g/g, 3.76 ± 0.06 g/g, and 4.91 ± 0.04 g/g, respectively. SE showed the highest OHC, surpassing that of the chickpea protein isolates (3.4–4.4 g/g) (Onder et al., 2023). Previous studies on bamboo shoot proteins have also observed that the OHC of the salt-extracted proteins is higher than that from other methods such as the alkaline extraction (Xiao et al., 2024). The higher OHC of SE may be attributed to more hydrophobic regions and nonpolar groups on its protein surface, which facilitate oil binding and interaction (Ge et al., 2021). Moreover, the higher OHC can form a denser interfacial film around the oil droplets, reducing coalescence, potentially improving emulsion stability (Badjona & Dubey, 2026). The high OHC of SE makes it suitable for products that require the incorporation of fats and oils, such as emulsified foods and meat analogues.

3.5.5. LGC

The LGC values for the protein samples are listed in Fig. 7B. Compared to ASP (20%) and SE (18%), AE showed the lowest LGC (14%), demonstrating strong gel-forming ability. The LGC of AE (14%) was comparable to that of soy protein isolate (Brishti et al., 2017). The higher β-sheet content (46.36 ± 0.35%) in AE indicated enhanced intermolecular hydrogen bonding interactions, leading to the formation of a more organized gel matrix (Zhou & Yang, 2019). The higher proportion of random coil structures reflected greater conformational flexibility. This balanced structural composition contributed to the formation of a network during the thermogelation process (Yu, Chitrakar, et al., 2026). Compared with ASP and SE, AE had the highest solubility and the smallest average particle size, which further led to greater protein-protein interactions upon heating (Wang et al., 2020), resulting in a denser, more tightly packed gel network. These findings suggest that AE may be suitable for food products requiring thickening and gel formation, as well as for nanogel materials.

3.5.6. In vitro protein digestibility

The digestibility of protein is a crucial measure of its nutritional value. A higher IVPD signifies that the protein is hydrolyzed to release more small-molecule peptides, facilitating easier absorption and utilization by the body (Zhang et al., 2022). The IVPD values of protein samples are shown in Fig. 7C. Overall, the IVPD of the protein samples continued to increase as the digestion time extended. In the gastric digestion phase (0-120 min), all protein samples showed a rapid increase in IVPD. This initial increase could be attributed to the cleavage of peptide bonds by gastric enzymes, which reduced molecular weight and loosened the peptide chains, thereby enhancing hydrolysis (Zhao et al., 2025). ASP gradually surpassed SE after 60 min of gastric digestion. At the end of gastric digestion (120 min), the IVPD values for ASP, AE, and SE were 25.24 ± 2.02%, 34.91 ± 0.69%, and 23.18 ± 1.07%, respectively. With the addition of trypsin, the IVPD of all proteins continued to increase during the intestinal digestion phase (120-240 min). At the end of digestion (240 min), the IVPD of AE was 52.60 ± 0.83%, surpassing that of ASP (28.10 ± 1.44%) and SE (34.49 ± 1.06%). AE’s high digestibility might be attributed to the reduction in α-helix content under alkaline extraction conditions, resulting in a more flexible structure (More et al., 2026). Additionally, the smaller particle size, higher solubility, and larger surface charge of AE collectively enhanced interactions between proteins and enzymes (Tanyitiku et al., 2026). Conversely, the relatively low digestibility of ASP and SE might be related to their rigid structures (Beaubier et al., 2025), which hinder enzymatic hydrolysis. These findings suggest that the superior digestibility of AE implies improved digestion and bioavailability in vivo, making it more promising for applications such as nutritional supplements and functional food ingredients.

3.6. Antioxidant activity of ACSP obtained by different extraction methods

Currently, researchers are focused on exploring new protein sources from plants, with significant interest in the antioxidant characteristics of these proteins (Kadam et al., 2022). Considering the complexity and diversity of oxidative stress mechanisms, this study assessed the antioxidant potential of protein samples using three distinct chemical antioxidant assay methods. The radical scavenging and reducing capacities of all protein samples increased with increasing concentration, compared with Vc. AE exhibited the most potent DPPH radical scavenging ability (Fig. 8A) across all concentration ranges. ASP showed higher DPPH scavenging activity than SE between 0.5 and 1.5 mg/mL, while SE exhibited superior scavenging activity compared to ASP at concentrations surpassing 1.5 mg/mL, reaching 73.36 ± 1.07% at 2.5 mg/mL. AE exhibited the highest DPPH radical scavenging capacity (88.54 ± 0.59%) at a concentration of 2.5 mg/mL. In the ABTS⁺ radical scavenging assay (Fig. 8B), the order of ABTS⁺ radical scavenging capacity within the 0.5–2.5 mg/mL range was AE > ASP > SE. AE also exhibited the strongest reducing capacity (Fig. 8C). ASP only exceeded SE in reducing capacity at 0.5 mg/mL. AE exhibited its highest reducing capacity (2.94 ± 0.09) at 2.5 mg/mL.

Fig. 8.

Antioxidant activity of ACSP obtained by different extraction methods: (A) DPPH radical scavenging capacity, (B) ABTS⁺ radical scavenging capacity, (C) Reducing power. Storage stability of ACSP obtained using different extraction methods during storage for 0–4 weeks: (D) MDA content, (E) Carbonyl content, (F) Solubility. Values are means ± standard deviation, with different letters indicating statistically significant differences (p < 0.05) at the same concentration.

Overall, AE exhibited superior antioxidant activity among all samples. This is attributable to the rich content of antioxidant amino acids such as Lys, Leu, and Asp (Ge et al., 2025), as well as the relatively high levels of hydrophobic amino acids including Met and Pro. Previous studies have shown that these antioxidant and hydrophobic amino acids contribute to free radical scavenging (Feng et al., 2022). Notably, the hydrophobic groups exposed during alkaline extraction of AE provide binding sites for phenolic compounds, enabling the co-extracted phenolics to bind to proteins through noncovalent interactions such as hydrophobic interactions and hydrogen bonds, thereby forming protein-phenolic complexes (Bejarano-Luján & Netto, 2010). The residual phenolic compounds also enhanced its antioxidant activity (Yu, Zhang, et al., 2026). The excellent antioxidant activity of AE makes it a promising natural antioxidant for use in functional food ingredients.

3.7. Storage stability of ACSP obtained by different extraction methods

The storage stability of ACSP obtained using different extraction methods was preliminarily evaluated by measuring the extent of protein and lipid oxidation and changes in solubility of protein samples stored at −20 °C for 0–4 weeks (Fig. 8D-F). Throughout the storage period, the protein samples exhibited relatively low MDA and carbonyl levels, suggesting minimal oxidation. The MDA content of all protein samples gradually increased with prolonged storage. After 4 weeks of storage, the MDA contents of ASP, AE, and SE were 0.1152 ± 0.0018 nmol/mg, 0.0755 ± 0.0026 nmol/mg, and 0.1018 ± 0.0070 nmol/mg, respectively. ASP exhibited a higher MDA level, indicating greater susceptibility to lipid oxidation during storage. The relatively high levels of lipid oxidation in ASP and SE may affect protein oxidation, as the free radicals generated by lipid oxidation might interact to form protein-derived free radicals, thus enhancing protein oxidation (Xiong et al., 2025).

Amino acids with NH or NH₂ groups on their side chains are particularly prone to oxidation. During the oxidation of proteins, these groups are converted into carbonyl groups (Siddique et al., 2016). The carbonyl content in protein samples gradually increased with prolonged storage. Compared to the rapid increase observed in ASP and SE, the carbonyl content in AE increased slowly and eventually stabilized. After 4 weeks of storage, the carbonyl content of AE (0.36 ± 0.08 μmol/g) was significantly lower than that of ASP (2.19 ± 0.20 μmol/g) and SE (1.84 ± 0.14 μmol/g).

During storage from 0 to 4 weeks, the solubility of the protein samples decreased to varying degrees. The solubility of ASP, AE, and SE decreased from initial values of 52.61 ± 1.09%, 55.11 ± 3.16%, and 45.61 ± 2.49% to 46.92 ± 0.87%, 51.32 ± 1.12%, and 39.21 ± 1.82%, respectively. The overall reduction in AE solubility was comparatively minor, with a decrease of less than 4%, possibly because its protein oxidation level was relatively low. Reduced oxidation can limit protein aggregation and structural changes (Xiong & Guo, 2021), thus maintaining functional properties more effectively. These results indicate that during storage, AE maintained low levels of both carbonyl and MDA and its solubility decrease was minimal, suggesting superior storage stability of AE.

4. Conclusion

In summary, different extraction methods significantly affected the extraction yield, sensory, structural and functional properties, and storage stability of ACSP. Alkaline extraction led to partial unfolding of the protein, reduced particle size, and resulted in a more flexible conformation, which enhanced its interfacial characteristics. Additionally, the relatively high content of β-sheet structures contributed to maintaining conformational stability and improved gel-forming properties. The exposure of amino acid residues resulting from the structural unfolding enhanced the functional properties of the protein while contributing to its excellent antioxidant capacity. Sensory evaluations showed that electronic nose response values and sensory scores were lower, and the aroma and taste were milder, suggesting that alkaline extraction enhanced the sensory quality of ACSP. Despite producing high-purity protein with excellent functionality, alkaline extraction had a low yield and generated large volumes of acidic and alkaline wastewater, adversely affecting the environment. Future research should investigate eco-friendly extraction techniques (such as ultrasonic or enzyme-assisted extraction) to improve yield while maintaining functional properties and minimizing the use of chemical reagents. In contrast, the salt extraction method offered a relatively high extraction yield, but lower purity due to the co-extraction of non-protein components. Nevertheless, it exhibited excellent oil-holding capacity and emulsifying properties. These findings indicate that ACSP, as a novel plant protein resource, shows considerable potential for practical applications. However, future research should aim to optimize the extraction process to balance quality and yield, thereby facilitating its use in the food industry.

CRediT authorship contribution statement

Guoshi Cao: Writing – original draft. Jiahao Liu: Validation. Yue Chen: Methodology. Yuxin Shi: Data curation. Lingrong Kong: Investigation. Yanan Sun: Resources. Meng Xu: Visualization. Hongyin Zhang: Funding acquisition. Mingming Yan: Writing – review & editing.

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

Supplementary material 1

Data availability

Data will be made available on request.