Impact of soybean genotype diversity on the structure and gelation properties of soy proteins

Abstract

Understanding the genetic basis of soy protein gelation performance differences is essential for developing clean-label, high-functionality ingredients. Here, we systematically compared the structural and gelation behavior of soy protein isolates from 20 genetically diverse soybean genotypes. Protein yield, subunit composition, sulfhydryl content, surface hydrophobicity, intrinsic fluorescence, secondary and thermal structures (FTIR, DSC), and gelation behavior (flow and rheology) were evaluated under standardized conditions. Glycinin-dominant isolates showed compact, thermally stable structures but formed weak gels, whereas genotypes with balanced 11S/7S profiles and higher conformational accessibility formed elastic gels with superior water-holding capacity. Highly aggregated isolates displayed high pre-gel viscosity yet poor gel strength, confirming that viscosity alone does not predict gelation. This side-by-side analysis links secondary structure fractions and reactive group exposure to gel viscoelasticity, providing a comprehensive genotype–function dataset for soy proteins. These findings establish a mechanistic framework for breeding and ingredient selection toward texture-focused, plant-based food formulations.

Graphical abstract

Highlights

• •Soy genetics affected protein functionality.
• •Protein viscosity measurements did not necessarily predict gel strength.
• •Balanced 11S/7S ratios yielded elastic gels with high water-holding capacity.
• •Glycinin-rich proteins formed firm but rigid and less extensible gel networks.
• •Excessive aggregation limited gel strength despite high pre-gel viscosity.

1. Introduction

Plant-based proteins are increasingly explored to support a sustainable, nutritious, and functional food supply. Among them, soybean is the most widely consumed and extensively studied source, characterized by high protein (∼40%) and oil (∼18%) contents, a complete amino acid profile, and well-established global production systems (Dilawari et al., 2022). Soy proteins are widely used in traditional products, including tofu and soymilk, as well as in modern plant-based alternatives such as yogurt, cheese, and meat analogues. Many of these applications rely on gelation, a key techno-functional property that remains central to the development of soy-based ingredients (Khatib, Aramouni, Herald, & Boyer, 2002; Yang et al., 2024).

Protein functionality is governed by structure across multiple hierarchical levels. Secondary and tertiary structures modulate this process by controlling conformational flexibility and exposure of reactive sites. A more flexible, loosely ordered structure typically permits greater unfolding and more extensive network formation (Yan, Xu, Zhang, & Li, 2021). These interactions underlie protein gelation: upon heating, proteins unfold and expose these groups, allowing network cross-links to form. Quaternary structure (multi-subunit assembly) also plays a role. Soybean storage proteins β-conglycinin (7S) and glycinin (11S) exemplify this. Utsumi and Kinsella (1985) showed that 7S and 11S form gels through different dominant forces. 7S forms a gel primarily via hydrogen-bonded networks, whereas 11S relies more on disulfide linkages and electrostatic interactions (Utsumi & Kinsella, 1985). Thus, differences in amino acid composition, folding, and subunit organization can lead to markedly different gelation outcomes (Yan et al., 2021).

Previous genotype-focused studies have often correlated gel quality primarily with bulk 11S/7S ratios or a limited set of structural descriptors. However, heat-induced gelation is a multistep, nonlinear process involving denaturation, aggregation, and percolated network formation governed by protein–protein and protein–water interactions (Ma & Chen, 2023). Importantly, subunit identity and molecular accessibility can substantially modulate gel outcomes. For instance, α'-subunit deficiency in β-conglycinin has been shown to increase gel strength and water-holding capacity and to produce a denser gel network, accompanied by changes in surface hydrophobicity and disulfide/hydrophobic interactions (Fu et al., 2023). Genotype plays a crucial role in determining the protein profiles, amino acid composition, and abundance of amino acid groups containing reactive groups like sulfhydryls, which mediate the protein-protein interactions necessary for gel formation (Kunowska & Stelzl, 2022).

Despite the importance of protein functionality, most commercial soybean breeding has historically prioritized yield, agronomic performance, and seed composition. While these traits remain essential, there has been relatively little emphasis on improving protein functionality. As a result, the genetic base of modern U.S. cultivars remains relatively narrow, limiting opportunities to improve functionality-related traits such as gelation (Singer et al., 2023). Thus, researchers are turning to diverse sources of genotypes, such as the USDA Soybean Germplasm Collection, to develop lines with improved functional profiles (Qi, Venkateshan, Mo, Zhang, & Sun, 2011; Yang et al., 2024).

Building on this rationale, the present study applies a standardized extraction–gelation protocol to 20 genetically diverse soybean genotypes, ensuring that observed differences in protein structure and gelation behavior arise from genotype-conditioned molecular phenotypes rather than processing variability. We integrate complementary structural and functional analyses, including subunit profiling, thiol chemistry, conformational accessibility, thermal stability, aggregation state, RVA, and rheology, to mechanistically explain genotype-dependent divergence in gel network formation. This approach establishes an application-oriented classification of elastic, firm, and weak-gelling protein types and supports a genotype → molecular phenotype → gel functionality framework, in which genetic diversity manifests as differences in subunit composition, aggregation propensity, molecular accessibility, and thermal behavior that ultimately govern gel network architecture.

2. Materials and methods

2.1. Materials

Twenty soybean genotypes were obtained from the USDA Soybean Germplasm Collection via the Soybean and Nitrogen Fixation Research Unit (Raleigh, NC, USA): five released cultivars and fifteen advanced breeding lines (Table 1). These lines include both domestic and exotic pedigrees, selected to cover wide ranges in protein (38–47% DW), oil (18–25%) (Momen, Sawant, Fallen, & Girard, 2025), and trait categories (e.g., high protein, high oleic, low phytate, large seed; Table S1). Although this represents a fraction of the >20,000 soybean accessions in the Collection, they capture diverse genetic and compositional backgrounds and provide a representative view of the functional variability underpinning gelation. All other chemicals used were analytical grade, obtained from Thermo Fisher Scientific or Sigma-Aldrich.

Table 1.

Breeding line, pedigree, and physicochemical characteristics of 20 soybean protein isolates (SP-1 to SP-20)¹.

Sample	Breeding Line	Mass Recovery (%)	Protein Recovery (%)	Fluorescence maxima (nm)	Diameter size (nm)2
SP-1	N6202	31.2 ± 2.4 a	63.4 ± 4.6 b	331 ± 0.5 e	566 ± 45 fg
SP-2	USDA-N6003LP	26.9 ± 2.5 b	65.4 ± 5.8 b	332 ± 0.1d	470 ± 13 hi
SP-3	N11–7477	22.1 ± 1.7 c	49.5 ± 4.8 de	331 ± 1.4 de	758 ± 47 c
SP-4	N14–7017	26.4 ± 5.3 ab	58.7 ± 11.2 cd	331 ± 0.7 e	958 ± 55 a
SP-5	N16–9124	27.2 ± 2.2 b	69.8 ± 5.5 ab	330 ± 0.5 f	869 ± 30 b
SP-6	USDA-N7007	18.5 ± 1.8 de	44.3 ± 4.3 de	336 ± 0.2 a	279 ± 11 jk
SP-7	N16–9211	17.4 ± 5.3 cde	36.4 ± 10.6 f	333 ± 1.4 cd	338 ± 5 j
SP-8	N16–9924	24.8 ± 1.9 bc	58.0 ± 4.4 cd	335 ± 2.1 ab	613 ± 33 ef
SP-9	N16–10044	23.1 ± 3.2 bc	46.4 ± 5.7 de	332 ± 0.7 d	340 ± 15 j
SP-10	N16–10075	32.4 ± 3.9 a	75.9 ± 10.5 a	334 ± 1.4 bc	433 ± 6 i
SP-11	N16–1286	21.4 ± 1.1 cd	50.9 ± 1.7 d	332 ± 2.1 cd	690 ± 56 cd
SP-12	N17–2535	27.2 ± 5.0 ab	56.4 ± 10 cd	335 ± 0.4 b	243 ± 14 k
SP-13	N17–30715	25.7 ± 2.9 bc	65.5 ± 8.3 ab	332 ± 0.7 d	914 ± 47 ab
SP-14	N17–31531	22.5 ± 0.3 c	53.9 ± 6.6 cd	332 ± 1.4 cd	411 ± 12 i
SP-15	N17–31805	26.9 ± 6.1 ab	61.7 ± 13.8 bc	333 ± 0.7 c	710 ± 1 cd
SP-16	N19–0318	27.2 ± 0.7 b	61.6 ± 2.0 c	333 ± 1.4 bc	522 ± 19 gh
SP-17	STPR14–547	20.9 ± 3.0 de	48.6 ± 7.0 de	335 ± 1.4 ab	319 ± 2 j
SP-18	NC-Raleigh	16.2 ± 1.0 e	38.3 ± 3.4 f	335 ± 1.4 ab	652 ± 28 de
SP-19	TC12FASDSZ-9	22.2 ± 0.8 c	52.0 ± 2.8 d	333 ± 1.4 bc	413 ± 17 i
SP-20	USDA-N5001	26.1 ± 1.3 b	60.5 ± 3.7 c	331 ± 1.3 cde	750 ± 49 c

¹Data are presented as mean ± SD (n = 3). Within each column, values sharing different lowercase letters are significantly different (p < 0.05).

²Particle size of aqueous samples determined using a Zetasizer Nano ZS.

2.2. Protein extraction and yield determination

Defatted flour was prepared by extracting soybean flour with n-hexane (1:10 w/v, 2×, 2 h total) at room temperature, followed by solvent evaporation under a fume hood overnight. Protein extraction was conducted by adding defatted flour to deionized water (1:10 w/v) at pH 8.5 (1 M NaOH), stirring for 2 h, then centrifuging (10,000 ×g, 30 min, 4 °C). The supernatant was adjusted to pH 4.5 (1 M HCl) to precipitate proteins, then centrifuged (6000 ×g, 20 min, 4 °C). Pellets were washed twice with DI water, neutralized to pH 7.0, and freeze-dried to obtain soy protein isolates (SPIs) (Momen et al., 2025). The protein content of the initial defatted flour and of the freeze-dried SPI was measured by the Kjeldahl nitrogen determination (AOAC 920.87), using a nitrogen-to-protein conversion factor of 6.25 (Jung et al., 2003). All genotypes were processed using identical alkaline extraction and isoelectric precipitation conditions to minimize processing-induced variability.

Because alkaline extraction can bias subunit composition via differential solubility (Wittmann, Duarte, Ayub, & Rossi, 2024), extraction yield and protein recovery were calculated based on the study of Q. Guo, Lin, He, and Zheng (2020):

$$\text{Extraction yield}\left(\text{Mass recovery},\%\right)=\left(\mathrm{W}₂/\mathrm{W}₁\right)\times 100$$ (1)

$$\text{Protein recovery}\left(\%\right)=\left(\mathrm{W}₃/\mathrm{W}₄\right)\times 100$$ (2)

where W₁ represents the weight of the defatted soybean flour, W₂ is the weight of the freeze-dried SPI, W₃ is the amount of protein recovered in the extracted SPI, and W₄ is the total protein content in the original defatted flours.

2.3. Protein characterization

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

Gel electrophoresis profiles of defatted flours (SF-1–SF-20) and SPI powders (SP-1–SP-20) were assessed under reducing and non-reducing conditions. Flours (∼1 mm) were suspended in 10 mM phosphate buffer (pH 7.0, 10 mg/mL) and hydrated overnight at 4 °C. Samples were diluted to 2 mg/mL, mixed 1:1 with 2× Laemmli buffer with or without 5% β-mercaptoethanol, heated for 5 min at 90 °C, and 10 μL (∼20 μg protein) was loaded per lane on 4–12% TGX gels (Bio-Rad). A pre-stained MW ladder (Bio-Rad) was included. All samples were analyzed under identical conditions to ensure uniform ionic strength and that observed results were not electrophoretic artifacts.

Electrophoresis was performed at 30 mA. Gels were stained with Coomassie Brilliant Blue R-250 and destained with methanol/acetic acid. Images were captured using a ChemiDoc system. Densitometry (ImageJ, NIH, Bethesda, MD, USA, http://rsb.info.nih.gov/ij) employed rolling-ball background subtraction (Din et al., 2021). Band groups were quantified with their relative abundance normalized to lane intensity.

Quantification of 11S/7S ratios was performed only on reducing SDS-PAGE gels (β-mercaptoethanol) to ensure dissociation of disulfide-linked polymers into resolved 7S and 11S subunits. The 11S fraction was calculated as the summed density of glycinin acidic and basic subunits, and the 7S fraction as the summed density of α', α, and β subunits. High-molecular-weight material retained near the stacking interface was not included in the 11S/7S calculation.

2.3.2. Free and total sulfhydryl content

The free –SH and total –SH contents of SPIs were quantified using Ellman's reagent (DTNB) (Beveridge et al., 1974; Zhu et al., 2020). A stock SPI solution (10 mg/mL) was diluted to 5 mg/mL in Tris-glycine buffer (86 mM Tris, 90 mM glycine, 4 mM EDTA, pH 8.0) for free –SH or in Tris-glycine buffer containing 8 M urea for total –SH analysis. 20 μL of DTNB reagent (4 mg/mL) was added to 2 mL samples, and absorbance at 412 nm was measured after incubation for 15 min at 25 °C in the dark. Sulfhydryl content was calculated using Eq. (3):

$$\mathrm{SH}\left(\text{μmol}/\mathrm{g}\right)=\frac{\left(73.53\times A_{412}\times \mathrm{D}\right)}{\mathrm{C}}$$ (3)

where A₄₁₂ is absorbance, D is dilution factor, and C is protein concentration (mg/mL).

2.3.3. Surface hydrophobicity

Surface hydrophobicity (H₀) was assessed using the ANS (8-anilino-1-naphthalenesulfonic acid) fluorescent probe. SPI solutions (0.01–0.1 mg/mL) were incubated with 8 mM ANS in 10 mM phosphate buffer (pH 7.0). Fluorescence intensity was measured (excitation: 390 nm, emission: 470 nm) using a PTI QuantaMaster spectrofluorometer (HORIBA Scientific, Edison, NJ, USA). H₀ was obtained from the slope of fluorescence intensity versus protein concentration.

2.3.4. Proteolytic susceptibility (conformational accessibility)

Proteolytic susceptibility was used as an indirect index of conformational accessibility (flexibility) in the hydrated, pre-heated state. Limited proteolysis preferentially targets locally flexible regions; thus, higher peptide release reflects greater accessibility of cleavage sites rather than the extent of unfolding during the subsequent heating step (Fontana et al., 2004).

SPI powders were dispersed in 0.1 M Tris-HCl buffer (pH 8.0) at 1.0 mg/mL. Trypsin (0.1%) was added at a 1:15 substrate-to-enzyme ratio and incubated at 37 °C for 10 min (Zhu et al., 2020). The reaction was terminated with 5% TCA, the sample centrifuged at 10,000 ×g for 10 min, and the supernatant peptide concentration quantified by Bradford assay. A relatively short digestion time (10 min) was chosen to assess initial proteolytic susceptibility before extensive protein breakdown occurs (Zhu et al., 2020). This limited trypsin treatment allows more structurally ‘open’ proteins to be preferentially cleaved, while more compact proteins remain largely intact – thereby highlighting differences in conformational accessibility among samples.

Proteolytic susceptibility was expressed as the % of solubilized peptides relative to total protein content.

$$\text{Solubilized peptides}\%=\frac{\text{Supernatant protein}}{\text{Total}}\times 100$$ (4)

2.3.5. Fluorescence spectroscopy

Tryptophan fluorescence was measured to assess protein conformational changes. SPI powders were hydrated overnight in 10 mM phosphate buffer (pH 7.0) at 0.05 mg/mL. Fluorescence spectra were recorded from 300 to 500 nm (excitation at 280 nm) using a PTI QuantaMaster spectrofluorometer with excitation and emission slits set at 5 nm. All readings were performed in quartz cuvettes at room temperature.

2.3.6. FTIR spectroscopy

Spectra were collected on a Nicolet iS50 with diamond ATR (32 scans, 2 cm⁻¹, 4000–500 cm⁻¹) at ambient temperature. A fresh background was collected between samples. Amide-I (1700–1600 cm⁻¹) was baseline-corrected (rubberband) and smoothed (Savitzky–Golay, 9 points). Regarding deconvolution, Gaussian curve-fitting assigned peaks to parallel β-sheet (∼1630–1633 cm⁻¹), antiparallel β-sheet (∼1616–1620 and 1680–1690 cm⁻¹), α-helix/random (∼1648–1656 cm⁻¹), β-turns (∼1660–1670 cm⁻¹). Areas were normalized to total amide-I (Devkota, Kyriakopoulou, Bergia, & Dhital, 2023).

2.3.7. Differential scanning calorimetry (DSC)

Thermal denaturation properties of SPIs were analyzed using a DSC Q200 calorimeter (TA Instruments, New Castle, DE, USA) calibrated with indium and sapphire standards. 10 ± 1 mg of protein suspension (17% w/v in deionized water) was sealed in T_zero hermetic aluminum pans, with an empty pan used as a reference. Each sample was scanned from 20 to 150 °C at a heating rate of 10 °C/min under a nitrogen purge (50 mL/min). Baseline correction was performed using the instrument's universal analysis software. The onset (Tₒ), peak denaturation (T_d), and endset (Tₑ) temperatures were determined from the thermograms, and the enthalpy change (ΔH, J/g protein) was calculated by integrating the peak area after baseline subtraction (S. Li, Wei, Fang, Zhang, & Zhang, 2014; Yue et al., 2021). Each genotype was analyzed in duplicate (n = 2).

2.3.8. Nanoscale surface morphology and particle size

The nanoscale morphology of SPIs was examined using a BioScope Catalyst AFM (Bruker, Santa Barbara, CA, USA) operated in tapping mode. Five representative SPIs were selected after initial compositional and solution-state screening (SDS–PAGE patterns, particle size distribution, and conformational accessibility assays) to span a range of apparent aggregation and dispersion behaviors. The selected SPIs represented a range of physicochemical properties: SP-4 and SP-5 had compact, aggregated morphologies; SP-7 had a unique protein profile; SP-10 had intermediate conformational accessibility; and SP-18 had a highly aggregated state and high thermal stability. AFM was then used to qualitatively visualize deposited nano-scale features, and DSC was used to compare thermal transitions under identical conditions.

Protein solutions (2 mg/mL in PBS) were prepared, and a 2 μL drop was placed on freshly cleaved mica and air-dried at room temperature to form thin deposits. Imaging was performed using a silicon nitride tip (triangular cantilever, nominal spring constant 0.4 N/m) under default tapping parameters with a scan size of 2 × 2 μm and a line rate of 1 Hz. Images were processed using NanoScope Analysis v1.40 (Bruker) for surface roughness and height distribution. Because images were acquired from dried films, possible drying-induced aggregation artifacts are acknowledged when interpreting surface features.

Additionally, the particle size of the five selected SPIs was measured by dynamic light scattering (Zetasizer Nano ZS; Malvern Instruments, UK). Backscatter at 173° was recorded at 25 °C using disposable cuvettes. Samples (2 mg/mL in PBS) were gently vortexed, centrifuged (5000 ×g, 5 min) before analysis. The Z-average and polydispersity index were taken from triplicate runs (12 scans/run).

2.4. Gelation analyses

2.4.1. Flow behavior of pre-gel soy protein solutions

The SPI powder (10% w/v) was dissolved in 10 mM phosphate buffer (pH 7.0). The viscosity of the SPI solutions was measured using an Ares G2–101 rheometer (TA Instruments, New Castle, DE, USA) with a 25 mm diameter parallel plate geometry and a 1 mm gap at 25 °C. The shear rate was increased from 0.01 to 200 s⁻¹, and the resulting shear stress versus shear rate data were fitted to the Power Law model, Eq. (5):

$$\mathrm{\tau }=\mathrm{K}.\mathrm{\gamma }\dot{}{}^{\mathrm{n}}$$ (5)

where τ represents shear stress (Pa), γ˙ is the shear rate (s⁻¹), K is the consistency coefficient (Pa·sⁿ), and n is the flow behavior index.

In addition to steady shear flow tests, a pasting test was performed on the 10% solution using a Rapid Visco Analyzer (RVA 4800; PerkinElmer, USA). The RVA program was: 0–2 min at 25 °C; 2–7 min ramping to 95 °C; 7–17 min holding at 95 °C; 17–22 min cooling to 25 °C; and 22–27 min holding at 25 °C. The paddle speed was set to 960 rpm for 10 s, then reduced to 160 rpm. This method enabled assessment of thermal pasting characteristics of the SPIs, complementing the rheological analysis.

2.4.2. Gel preparation and least gelation concentration screening

The least gelation concentration (LGC) was determined following a standardized tube inversion method. This binary gel/no-gel criterion provides a clear threshold for gel formation (Alabi & Amonsou, 2024). It is a qualitative measure and thus does not quantify gel strength or capture subtle differences in weakly gelled samples near the threshold. In this method, SPI powders were dispersed in 10 mM phosphate buffer (pH 7.0) at concentrations ranging from 5 to 10% (w/v) and stirred for 2 h at ambient temperature. The dispersions were transferred into 10 mL glass vials with tight-fitting caps and heated in a water bath at 90 °C for 30 min, followed by rapid ice bath cooling. The samples were then stored at 4 °C for 48 h to allow complete network stabilization (Fang, Chen, & Rao, 2023; Webb, Dogan, Li, & Alavi, 2023). Gelation was evaluated with a binary inversion criterion (no visible flow upon 180° inversion for 30 s). All genotypes formed self-supporting gels at 10% (w/v), which was selected as the standard concentration for subsequent rheological and water-holding capacity analyses. All genotypes were subjected to an identical heating/cooling program to minimize processing variability and enable attribution of functional differences to intrinsic genotype-conditioned protein properties, consistent with recommendations that gel properties depend on both intrinsic composition and thermal history (C. Wu, Hua, Chen, Kong, & Zhang, 2017).

2.4.3. Gel rheological measurements

The heat-induced gelation behavior and viscoelastic properties of SPI solutions were characterized using a Discovery DHR-2 hybrid rheometer (TA Instruments, New Castle, DE, USA) with a 20 mm parallel-plate geometry and a 1000 μm gap. Freshly prepared 10% (w/v) SPI dispersions at pH 7.0 were loaded directly onto the Peltier-controlled plate. Excess material was trimmed, and paraffin oil was applied around the perimeter to prevent moisture loss. Samples were allowed to equilibrate on the plate for 3–5 min before testing. Amplitude sweeps (0.01–10% strain at 1 Hz) were performed to identify the linear viscoelastic region (LVR). All subsequent measurements were conducted at 0.01% strain, which fell within the LVR for all genotypes.

The temperature-dependent gelation profile was obtained by a temperature sweep from 25 to 90 °C (5 °C/min), holding for 15 min at 90 °C, and cooling back to 25 °C at the same rate. The storage modulus (G′) and loss modulus (G″) were continuously recorded at 1 Hz to monitor network formation and stabilization (Abdi Kordlar, 2022). After gelation, frequency-sweep tests (0.01–10 Hz, 0.01% strain, 25 °C) were performed to evaluate the frequency dependence and mechanical strength of the gels (Andlinger & Kulozik, 2023; Momen, Rodrigue, & Aider, 2023).

2.4.4. Water-holding capacity (WHC)

To measure the WHC, SPI gels were prepared as described in Section 2.4.2 directly in 5 mL centrifuge tubes (Momen et al., 2023). After gelation, the samples were stored at 4 °C for 48 h to allow full network stabilization. The gels were then weighed. After centrifuging (1000 ×g, 10 min), the expelled water was removed using a pipette, and the gels were immediately reweighed. WHC was calculated using the following equation, which reflects the structural integrity and water-binding efficiency of the gel matrix formed by different SPI:

$$\mathrm{WHC}\left(\%\right)={\mathrm{W}}_{\mathrm{r}}/{\mathrm{W}}_{\mathrm{t}}\times 100$$ (6)

where W_t = gel mass before spin; W_r = gel mass after spin.

2.5. Statistical analysis

All measurements were performed in triplicate, unless otherwise indicated. Data were analyzed by one-way analysis of variance (ANOVA) and are presented as mean ± standard deviation. Duncan's multiple range test (p < 0.05) was used to assess significant differences among means, except for FTIR data, which were analyzed using Tukey's post-hoc test. All analyses were performed using IBM SPSS software (version 16).

3. Results and discussion

3.1. Extraction yield and protein recovery

The 20 soybean genotypes exhibited pronounced differences in extractability under identical alkaline–isoelectric conditions, indicating genotype-dependent variation in protein solubility and recovery behavior (Table 1). Extraction yield – i.e., the percentage of dry soy protein isolate (SPI) mass relative to the initial defatted flour – ranged from 16.2% (SP-18) to 32.4% (SP-10; mean 24.31%). Protein recovery – calculated as the percentage of total protein retained in the isolate relative to the starting protein content of the flour – ranged from ∼36% (SP-7 and SP-18, respectively) to 75.9% (SP-10; mean 55.9%). SP-10 consistently achieved the highest extraction, with mass and protein recovery exceeding 30% and 75%, respectively.

Higher-yield lines likely contained globulin fractions more readily solubilized at pH 8.5 and efficiently precipitated at pH 4.5 (e.g., β-conglycinin-rich) (Pavlicevic, Tomic, Djonlagic, Stanojevic, & Vucelic Radovic, 2018; Qi et al., 2011), whereas low-yield lines likely contained more compact, hydrophobic, or aggregated proteins that resisted extraction (Khatib et al., 2002; Liu et al., 2008). Because all genotypes were processed under identical alkaline-isoelectric conditions, observed differences reflect inherent solubility or compositional variation rather than methodological bias.

3.2. Physicochemical characterization of protein powders

3.2.1. Subunit composition and SDS–PAGE profiles

SDS–PAGE profiles revealed pronounced genotype-dependent variation in storage protein composition and aggregation behavior (Fig. 1A–B). Defatted flour extracts displayed the full complement of β-conglycinin (7S; α', α, β) and glycinin (11S; acidic and basic) subunits, with occasional lipoxygenase (LOX) bands near 95 kDa (Fig. 1A), consistent with previous reports (Momen et al., 2025; Yang et al., 2024). In contrast, SPI profiles diverged considerably among genotypes (Fig. 1B). Several genotypes (SP-6, SP-7, SP-11, SP-16, SP-17) exhibited diminished β-conglycinin bands, whereas SP-1, SP-10, SP-15, and SP-18–20 retained more balanced 7S and 11S intensities. Distinct high-molecular-weight (HMW) smearing in SP-4 and SP-5, and faint streaking in SP-18, suggested partial aggregation during alkaline extraction or drying (Wang, Hill, Campbell, & O'Connor, 2022; Yang et al., 2024).

Fig. 1.

Subunit composition and 11S/7S ratios of soybean proteins from 20 genotypes. (A) Reducing SDS–PAGE of defatted soybean flours. (B) Reducing SDS–PAGE of corresponding soy protein isolates (SPIs). Representative non-reducing profiles are shown in Supplementary Fig. S1. (C) Quantified 11S/7S ratios (mean ± SD, n = 3); different lowercase letters indicate significant differences among genotypes (p < 0.05). “SP-x” sample codes correspond to the genotypes listed in Table 1.

Under non-reducing conditions (Fig. S1), all SPI samples exhibited prominent HMW smears accumulating near the stacking interface (>150–250 kDa), indicating the presence of very large aggregates that failed to migrate into the resolving gel. Upon reduction with β-mercaptoethanol (Fig. 1B), these HMW species largely disappeared and resolved into expected 7S and 11S subunits, consistent with disulfide-mediated aggregation reported for alkali-processed or stored soy isolates (Guo et al., 2020. Y. Yang et al. (2024) also proved that conventional isolation of plant proteins (extreme pH or thermal treatments) can induce protein denaturation and aggregation, whereas milder processing preserves a more native state with fewer large aggregates.

In SP-4 and SP-5, some residual HMW smear persisted even after reduction (Fig. 1B), suggesting that a minor fraction of aggregates may be stabilized by non-reducible covalent cross-links. Possible mechanisms include dityrosine formation or Maillard-type glycation, both known to generate HMW polymers resistant to reducing agents (Liu et al., 2008; Momen et al., 2023). For example, oxidative treatment caused proteins to form aggregates >245 kDa that remained stuck at the top of a 10% gel despite reducing conditions (Ke & Huang, 2016). Liu et al. (2008) observed that soy protein isolates tend to form disulfide-mediated aggregates during storage, especially under conditions that promote oxidation, leading to decreased solubility and HMW species.

Lipoxygenase bands (LOX; ∼94–97 kDa) decreased or disappeared in several isolates (e.g., SP-7, SP-17), consistent with genotype-dependent differences in LOX abundance or partial removal during isoelectric precipitation. Reduced LOX content is associated with lower formation of lipid oxidation derived off-flavor compounds such as hexanal and nonanal, suggesting potential sensory implications (X. Li et al., 2023). LOX content in soybean products can vary due to processing and genotype, and isolate lacking LOX can improve sensory profiles.

Densitometric evaluation of the reducing SDS-PAGE gels quantitatively supported the visual band patterns observed in Fig. 1B, revealing substantial variation in the relative proportions of the two major soy storage proteins, glycinin (11S) and β-conglycinin (7S), among the genotypes tested (Fig. 1C). The calculated 11S/7S ratios spanned from a low of 0.68 in SP-16 to a high of 1.29 in SP-4, indicating genotype-dependent differences in storage protein composition. Specifically, genotypes SP-4, SP-5, SP-8, and SP-18 exhibited 11S/7S values greater than 1.1, consistent with a predominance of glycinin over β-conglycinin. In contrast, SP-11, SP-12, SP-16, and SP-17 showed ratios below 0.75, reflecting a relatively higher proportion of β-conglycinin (7S) subunits. These quantitative ratios not only confirm the gel profiles but also align with previous studies demonstrating that the 11S/7S ratio can vary widely among soybean varieties and is influenced by genetic background under standard extraction conditions (Hou et al., 2025). Isoelectric precipitation and alkaline solubilization can influence subunit recovery and aggregation behavior (J. Yang, Zamani, Liang, & Chen, 2021). However, because all genotypes were processed identically, relative differences among isolates reflect intrinsic compositional and molecular characteristics rather than procedural variability (Verfaillie, Janssen, Van Royen, & Wouters, 2023).

3.2.2. Sulfhydryl content and surface hydrophobicity

Free –SH content ranged from 1.17 to 5.87 μmol g⁻¹ (lowest SP-18, highest SP-4), while total –SH ranged from 14.9 to 19.7 μmol g⁻¹ (lowest SP-5, highest SP-4; Fig. 2A). Surface hydrophobicity (H₀) also varied widely, with SP-7 and SP-13 among the highest (∼270 × 10³ a.u.) and SP-19 the lowest (∼130 × 10³ a.u.; Fig. 2B). The high free and total –SH contents of SP-4, combined with visible aggregation on SDS–PAGE (Fig. 1B), suggest it has accessible cysteines that promote intermolecular disulfide bonding despite partial aggregation (M. Wu et al., 2019; Zhang et al., 2024). SP-5 exhibited low total –SH and low H₀, pointing to a compact, tightly folded conformation with limited exposure of reactive groups, consistent with its pronounced aggregation (Liu et al., 2008). SP-18 also showed low free –SH and low H₀, supporting a globular, compact state with limited rearrangement potential.

Fig. 2.

Structural reactivity and conformational flexibility of soy protein isolates (SPIs) from 20 soybean genotypes. (A) Free and total sulfhydryl (–SH) group content, (B) Surface hydrophobicity (H₀) measured by ANS binding fluorescence, and (C) Proteolytic susceptibility to trypsin digestion as a proxy for tertiary structure flexibility and conformational accessibility. Data represent mean ± SD (n = 3). Different letters above bars indicate significant differences among genotypes (p < 0.05). “SP-x” codes refer to Table 1.

Notably, SP-7 showed low free –SH yet high H₀. A plausible explanation is that β-conglycinin deficiency and glycinin dominance yields a compact core with limited thiol exposure but exposes hydrophobic patches upon partial rearrangement (Maruyama et al., 1999), increasing H₀. This pattern was not universal among glycinin-rich lines (e.g., SP-5, SP-18), indicating that subunit ratio, aggregation state, and local packing jointly govern surface exposure. In contrast, SP-4 and SP-10 paired accessible thiols with moderate–high hydrophobicity, a combination expected to favor disulfide and hydrophobic crosslinking during gelation.

3.2.3. Proteolytic susceptibility and conformational exposure

Trypsin susceptibility differed among genotypes, from ≤47% (e.g., SPs 5, 18, 19) to 75% (SP-17), with SP-10 also relatively high (60%; Fig. 2C). Broadly, lines with higher trypsin susceptibility showed greater surface exposure (free –SH/H₀) and less persistent high MW aggregation, while protease-resistant lines appeared compact (Kato, Komatsu, Fujimoto, & Kobayashi, 1985; Kato & Nakai, 1980; Zhu et al., 2020). Because limited proteolysis targets flexible regions, higher susceptibility suggests a lower barrier to rearrangement and exposure of reactive patches during heating (Fontana et al., 2004). This helps explain why, within a moderate range, more accessible proteins may form more connected networks under a fixed heating program, whereas overly aggregated or overly stable proteins may incorporate less into the network and yield weaker gels (J. M. S. Renkema, 2001).

3.2.4. Nanoscale morphology and particle size (AFM and DLS)

AFM imaging of selected SPI samples (SPs 4, 5, 7, 10, 18) was performed on air-dried deposits on mica and revealed clear genotype-dependent differences in nanoscale surface topology (Fig. 5A). To relate these observations to hydrated dispersions, we compared AFM trends with independent solution-state DLS measurements (Fig. 5B, Table 1) rather than inferring solution particle size directly from dried AFM features (Ruggeri et al., 2018).

Fig. 5.

(A) Atomic force microscopy (AFM) images of five representative SPI samples (SP-4, SP-5, SP-7, SP-10, SP-18) captured in tapping mode at 20 μm × 20 μm scan size. (B) Average particle size (nm) of corresponding SPIs determined by dynamic light scattering (DLS) (mean ± SD, n = 3). Different letters indicate significant differences among genotypes (p < 0.05). “SP-x” codes refer to Table 1.

AFM images of SP-4 and SP-5 showed rough, densely packed surfaces with large, irregular clusters (> 800 nm, Fig. 5), concurrent with their large Z-average values measured in solution by DLS (Table 1) and their aggregation signatures on SDS-PAGE (Fig. 1B). SP-10 displayed a smoother surface with loosely clustered nanoaggregates (433 nm), indicating moderate conformational accessibility and surface reactivity favorable for gelation. SP-7 showed more uniformly distributed globular deposits; this qualitative morphology is consistent with its smaller DLS Z-average (338 nm) and comparatively lower aggregation signatures. SP-18 exhibited heterogeneous deposited clusters: while drying may contribute to the appearance of these features, the trend is broadly consistent with its larger DLS Z-average (652 nm), supporting a higher aggregation tendency under our standardized conditions.

3.3. Analysis of secondary structure by FTIR

Deconvolution of amide I revealed parallel β-sheets (42.6–45.7%) as the dominant motif across genotypes, followed by α-helix/random coil (21.8–26.7%), antiparallel β-sheets (13.3–18.7%), and β-turns (13.9–16.4%) (Fig. 3A–B; Table S2). SP-4 exhibited the highest antiparallel β-sheet content (18.7%), while SP-5 and SP-13 were lower (15.4% and 13.3%, respectively) with slightly elevated coil/helix fractions. SP-4, with the highest antiparallel β-sheet content, also showed aggregation in SDS–PAGE (Fig. 1B) and low protease susceptibility (Fig. 2C), pointing to a compact, hydrogen-bonded structure. Similarly, SP-5 exhibited aggregation in SDS–PAGE and low protease susceptibility, despite having a lower antiparallel β-sheet content, suggesting that aggregation is driven by different molecular interactions.

Fig. 3.

Secondary-structure composition of soy protein isolates (SPIs) from 20 genetically diverse soybean genotypes determined by FTIR amide I band deconvolution. (A) Relative abundance of anti-parallel β-sheets (orange), parallel β-sheets (green), α-helix and random coil (blue), and β-turns (pink) for all SPIs. (B) Detailed comparison for five representative genotypes (SP-4, SP-5, SP-7, SP-10, SP-18). Data represent mean ± SD (n = 3); different letters within each structural category indicate significant differences (p < 0.05). “SP-x” codes are listed in Table 1. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

In contrast, SP-10, featuring a lower antiparallel β-sheet and higher unordered content (Fig. 3A–B), displayed higher enzymatic susceptibility and increased surface reactivity (Fig. 2A–C), suggesting a more accessible protein structure. This is consistent with prior research demonstrating that higher unordered content (coil, helix) and reduced β-sheet organization correlate with enhanced conformational flexibility and improved functional responsiveness (Carbonaro, Maselli, & Nucara, 2012; Zhu et al., 2020). Interestingly, SP-7, a glycinin-dominant genotype lacking β-conglycinin subunits (Fig. 1B), showed a relatively balanced secondary structure composition. This suggests that while subunit composition may influence physicochemical traits, tertiary and quaternary structures – such as molecular packing, aggregation states, or solvation interfaces – likely drive functional divergence (Q. Tang, Roos, & Miao, 2024).

Although the FTIR deconvolution indicates only modest differences in the fractional secondary structures among genotypes, these values represent averaged backbone motifs and can mask large differences in tertiary packing, aggregation pathway, and network incorporation. Heat-induced protein gelation proceeds through denaturation → aggregation → network formation, and small changes in structural order can shift the balance of attractive vs repulsive interactions, the exposure of hydrophobic and thiol groups, and ultimately the fraction of protein incorporated into the load-bearing network (Ma & Chen, 2023). Notably, prior FTIR–texture analyses have reported strong correlations between β-sheet enrichment and gel hardness and between random-coil enrichment and reduced hardness, supporting that even moderate shifts in secondary structure can correspond to substantial macroscopic differences when they reflect underlying changes in network density and topology (Zheng, Regenstein, Zhou, Mokhtar, & Wang, 2023).

3.4. Thermal stability of soy protein isolates

Differential scanning calorimetry (DSC) revealed significant genotypic variation in thermal stability (Table 2). Representative thermograms exhibited a single broad endotherm (Fig. S2), indicating overlapping 7S and 11S transitions under our concentrated conditions. Onset temperatures (T_o) ranged from 95.0 °C (SP-4) to 105.0 °C (SPs 1, 3, 5), while peak denaturation temperatures (T_d) varied from 100.4 °C (SP-4) to 119.6 °C (SP-5). Enthalpy changes (ΔH) displayed a marked scatter, with low values (∼3–7 J/g) in SPs 1, 2, 4, 6, and others, compared to notably higher values (∼8–17 J/g) in SPs 3, 7, 10, 18, and 19, showing major differences in protein packing and intramolecular interactions among genotypes (Table 2).

Table 2.

DSC thermal parameters of soybean protein isolates (SP-1 to SP-20).

Genotype1	To (°C)2	Td (°C)	Te (°C)	ΔH (J/g)
SP-1	105.0 ± 1.2 a	118.8 ± 1.5 a	125.0 ± 1.0 a	3.1 ± 0.2 c
SP-2	100.0 ± 0.8 b	108.1 ± 1.3 c	115.0 ± 1.2 b	3.0 ± 0.3 c
SP-3	105.0 ± 1.1 a	118.3 ± 1.2 a	125.0 ± 1.5 a	9.5 ± 0.6 a
SP-4	95.0 ± 2.0 c	100.4 ± 1.8 f	112.0 ± 2.1 d	5.8 ± 0.4 b
SP-5	105.0 ± 1.0 a	119.6 ± 1.4 a	125.0 ± 1.5 a	5.4 ± 0.1 e
SP-6	105.0 ± 1.3 a	118.5 ± 1.2 a	125.0 ± 1.6 a	6.5 ± 0.5 b
SP-7	105.0 ± 1.1 a	118.5 ± 1.4 a	119.6 ± 1.3 c	10.3 ± 0.8 a
SP-8	104.0 ± 1 a	117.9 ± 1.5 a	123.1 ± 1.7 a	7.2 ± 0.5 b
SP-9	104.5 ± 1 a	118.8 ± 1.3 a	119.7 ± 1.4 c	6.4 ± 0.4 b
SP-10	104.8 ± 1.2 a	118.2 ± 1.4 a	123.0 ± 1.6 a	8.0 ± 0.6 a
SP-11	103.0 ± 1.4 ab	117.5 ± 1.5 a	122.2 ± 1.8 a	6.7 ± 0.5 b
SP-12	103.5 ± 1.1 ab	117.8 ± 1.3 a	122.5 ± 1.6 a	6.9 ± 0.5 b
SP-13	104.2 ± 1.3 ab	118.0 ± 1.2 a	122.8 ± 1.7 a	7.1 ± 0.6 b
SP-14	104.0 ± 1.0 ab	118.1 ± 1.4 a	123.0 ± 1.5 a	7.3 ± 0.5 b
SP-15	104.5 ± 1.2 a	118.3 ± 1.3 a	123.2 ± 1.6 a	7.5 ± 0.5 b
SP-16	100.0 ± 1.1 b	118.5 ± 1.4 a	119.6 ± 1.5 c	6.5 ± 0.5 b
SP-17	103.8 ± 1.0 ab	118.0 ± 1.2 a	123.0 ± 1.6 a	6.8 ± 0.6 b
SP-18	101.0 ± 1.3 b	107.2 ± 1.8 d	114.9 ± 1.7 c	16.7 ± 4.1 a
SP-19	101.4 ± 1.2 b	112.6 ± 1.7 d	115.3 ± 1.6 c	9.2 ± 1.9 a
SP-20	100.8 ± 1.4 b	106.2 ± 1.9 d	115.9 ± 1.8 c	10.4 ± 1.5 a

¹Genotypes (SP-1 to SP-20) correspond to the soybean lines listed in Table 1.

²Data are presented as mean ± SD (n = 2). Within each column, values sharing different lowercase letters are significantly different (p < 0.05).

SP-4 exhibited the lowest T_o and T_d with a moderate ΔH (5.82 J/g), consistent with its partially unfolded and aggregation-prone state (Section 3.2). These thermal features support its capacity for intermolecular crosslinking during gelation due to exposed reactive sites (Zhong & Sun, 2000). In contrast, SP-5 and SP-7 displayed higher T_d and moderate-to-high ΔH values (∼10 J/g), indicating more tightly folded, thermally stable protein matrices. This aligns with their glycinin-rich composition (Fig. 1C) and rigid secondary structures (Fig. 3), which contribute to stability but may restrict reactivity during functional processing (Mo, Zhong, Wang, & Sun, 2006; J. Renkema, Knabben, & Van Vliet, 2001).

SP-18 had the highest denaturation enthalpy (ΔH ≈ 16.8 ± 4.1 J/g), indicating strong cooperative unfolding and dense structural packing. Similar high ΔH values under low-moisture conditions have been linked to aggregate-rich, less soluble soy protein conformations. For example, Kitabatake, Tahara, and Doi (1990) reported that as moisture decreased from 94% to 11%, the denaturation temperature rose sharply (reaching ∼180 °C) due to tighter protein packing and stronger intermolecular interactions. The elevated ΔH and unfolding resistance align with SP-18's compact, aggregate-rich structure (Fig. 5, Table 2) and low protease susceptibility (Fig. 2C). In this case, the high thermal stability reflects a rigid molecular arrangement that hinders conformational rearrangement, which may reduce its functional adaptability (C.-H. Tang, Li, Wang, & Yang, 2007).

We acknowledge that DSC measurements were conducted in duplicate (n = 2), which is a limitation in terms of statistical robustness. However, each genotype's two thermograms were highly consistent; the standard deviation for T_d was typically around 1 °C or less. The clear differences observed between genotypes (such as the markedly higher ΔH of SP-18) were reproducible even with two runs and aligned with our other analytical results (e.g., SP-18's pronounced aggregation and rigid gel texture supported its high ΔH).

3.5. Flow behavior and gelation properties

Because all samples were subjected to the same gelation conditions, the substantial variation in gelation behavior observed arose from genotype-dependent differences in the molecular state of the proteins prior to thermal gelation. These differences reflect intrinsic structural and interaction properties that influence how proteins unfold and aggregate during heat-induced gel formation (Yuqi Zhang, Sharan, Rinnan, & Orlien, 2021).

3.5.1. Least gelation concentration (LGC)

LGC varied by genotype (Fig. 4). All twenty isolates formed visible, self-standing gels at 10% (w/v), confirming that 10% exceeds the LGC for each line. At 8–9%, gel integrity diverged: only SPs 1, 5, 6, 7, 8, 9, 10, 14, 15, and 20 maintained cohesive structures. At 7%, several isolates (e.g., SPs 1, 5, 6, 8, 9, 10, 14, 15, 16, 18, 19, and 20) formed viscous, semi-solid matrices but not self-supporting gels. All samples remained fluid at <7% protein concentration.

Fig. 4.

Least Gelation Concentration (LGC) screening of soy protein isolates across 5% to 10% (w/v) protein concentrations using tube inversion; a non-flowing gel indicated successful network formation. Highlights in the 8% row indicate key genotypic differences in gelation capacity at sub-threshold concentrations. Solid-line boxes denote stable gels, dashed-line boxes denote partial or viscous gels, and dotted-line boxes indicate no gel formation. “SP-x” codes are listed in Table 1.

Because all samples formed self-supporting gel networks at 10% (w/v), this concentration was selected for subsequent rheological and water-holding capacity analyses to ensure complete gel formation across all lines. Although extraction yield differed among genotypes, identical isolation and re-dispersion protocols ensured comparable protein concentrations and ionic conditions in gelation tests, thereby minimizing compositional bias. This allows observed differences in gel strength and network elasticity to be attributed primarily to intrinsic protein structural differences rather than to extraction efficiency.

3.5.2. Flow behavior of pre-gel solutions

Steady-shear data for 10% (w/v) SPI dispersions fit the power law τ = K · γ˙ⁿ with R² > 0.99 for all genotypes (Table 3). Most isolates were pseudoplastic (n < 1), consistent with shear-induced disentanglement of protein aggregates. They spanned 0.69 to 1.07, indicating microstructural diversity. SP-18 uniquely exhibited n > 1, suggesting mild dilatant or hydrocluster behavior at this solids level. Brief pH shift (7.0 → 8.0) or 50–100 mM NaCl lowered K but did not improve gel G′ upon heating, consistent with early aggregation limiting rearrangement rather than ionic screening. SP-18 also had the highest K (7.70), a large Z-avg. (653 nm; Table 1), high ΔH (Table 2), and AFM heterogeneity (Fig. 5), indicating strong inter−/intra-molecular interactions and low conformational flexibility (Zheng et al., 2023).

Table 3.

Flow behavior (consistency index K and flow behavior index n), dynamic rheological properties (storage modulus G', loss modulus G", and tan δ), and water-holding capacity (WHC) of heat-induced soy protein isolate gels prepared from 20 genetically diverse soybean genotypes.

Sample1	Flow consistency index (K = mPa.s)2	Flow behavior index (n, dimensionless)	Storage modulus (G')	Loss modulus (G")	Tan δ	WHC (%)
SP-1	1.81 ± 0.10 c	0.71 ± 0.02 bc	74 ± 3 k	24 ± 3 i	0.18 ± 0.01 c	94 ± 0.9 ab
SP-2	2.01 ± 0.09 a	0.89 ± 0.03 a	78 ± 4 k	28 ± 6 i	0.21 ± 0.01 b	96 ± 0.5 ab
SP-3	1.77 ± 0.08 c	0.81 ± 0.02 b	112 ± 24 k	35 ± 3 i	0.27 ± 0.02 a	84 ± 0.7 c
SP-4	1.59 ± 0.07 d	0.70 ± 0.01 bc	2085 ± 134 c	464 ± 16 c	0.20 ± 0.01 bc	96 ± 1.5 ab
SP-5	1.55 ± 0.06 d	0.70 ± 0.02 bc	344 ± 69 i	73 ± 5 h	0.21 ± 0.03 b	80 ± 5.6 cd
SP-6	1.89 ± 0.11 b	0.86 ± 0.03 ab	454 ± 20 h	94 ± 6 fg	0.20 ± 0.02 bc	86 ± 2.9 c
SP-7	2.05 ± 0.12 a	0.90 ± 0.02 a	2049 ± 174 c	437 ± 77 c	0.19 ± 0.01 bc	96 ± 1.0 ab
SP-8	1.66 ± 0.05 d	0.69 ± 0.01 c	1620 ± 167 d	302 ± 26 d	0.21 ± 0.03 b	96 ± 2.2 ab
SP-9	1.70 ± 0.08 c	0.71 ± 0.01 bc	140 ± 47 k	38 ± 7 i	0.27 ± 0.04 a	97 ± 1.9 ab
SP-10	1.72 ± 0.07 c	0.72 ± 0.02 bc	2502 ± 25 i	384 ± 58 cd	0.18 ± 0.01 c	99 ± 0.1 a
SP-11	1.99 ± 0.10 b	0.88 ± 0.02 ab	1089 ± 86 e	221 ± 22 e	0.20 ± 0.01 bc	94 ± 1.2 ab
SP-12	1.87 ± 0.09 b	0.85 ± 0.02 ab	385 ± 41 hi	98 ± 16 fg	0.25 ± 0.01 a	97 ± 1.1 ab
SP-13	1.71 ± 0.06 c	0.80 ± 0.01 b	2686 ± 67 b	601 ± 61 b	0.22 ± 0.03 b	97 ± 0.8 ab
SP-14	1.83 ± 0.07 b	0.84 ± 0.02 ab	2411 ± 58 bc	492 ± 71c	0.20 ± 0.04 b	88 ± 2.7 bc
SP-15	1.96 ± 0.08 b	0.84 ± 0.02 ab	3647 ± 310 a	1059 ± 38 a	0.20 ± 0.02 b	97 ± 1.2 ab
SP-16	2.05 ± 0.10 a	0.89 ± 0.02 a	938 ± 59 ef	213 ± 25 e	0.22 ± 0.02 b	97 ± 0.9 ab
SP-17	2.06 ± 0.10 a	0.90 ± 0.02 a	353 ± 98 h	98 ± 7 fg	0.27 ± 0.01a	98 ± 0.9 ab
SP-18	7.70 ± 0.20 e	1.07 ± 0.03 d	626 ± 36 g	135 ± 14 f	0.29 ± 0.04 a	82 ± 4.3 bc
SP-19	1.59 ± 0.05 d	0.75 ± 0.02 bc	573 ± 24 h	224 ± 41 e	0.21 ± 0.01 b	93 ± 0.9 b
SP-20	1.75 ± 0.07 c	0.80 ± 0.02 b	500 ± 78 h	110 ± 21 f	0.22 ± 0.01 b	96 ± 0.6 ab

¹Genotypes (SP-1 to SP-20) correspond to the soybean lines listed in Table 1.

²Data are presented as mean ± SD (n = 3). Within each column, values sharing different lowercase letters are significantly different (p < 0.05).

SP-4 and SP-5 showed moderate K (∼1.6) and strong shear-thinning (n ≈ 0.61–0.66; Table 3) alongside large particles (958 and 869 nm, respectively; Table 1), consistent with partially aggregated, densely packed suspensions. SP-7 displayed K ≈ 2.05 with n ≈ 0.90 (near-Newtonian; Table 3), and the smallest particles (∼338 nm; Table 1), indicating a compact, weakly structured dispersion, aligned with its glycinin-dominant profile. Previous studies have shown that β-conglycinin contributes to flexible, extended conformations that enhance gel elasticity, while glycinin-rich proteins typically form weaker, less cohesive gels (Mo et al., 2006; J. Renkema et al., 2001). SP-10 was intermediate (K = 3.54; n = 0.78 ±; Z-avg ≈ 434 nm; Table 1, Table 3) and showed loosely clustered nanoaggregates by AFM (Fig. 5A), consistent with moderate conformational flexibility favorable for controlled gelation.

To further assess thermal responses during processing, RVA (Rapid Visco Analyzer) profiles were obtained for all samples (Fig. S3), but are discussed here for five representative genotypes: SPs 4, 5, 7, 10, and 18 (Fig. 6A). Peak viscosities were ranked in the following order: SP-5 > SP-4 > SP-18 > SP-10 > SP-7. These RVA results corroborate established observations that protein peak viscosity reflects folding state, aggregation, and swelling capacity under heat-shear conditions, particularly for plant proteins (Onwulata, Tunick, & Thomas-Gahring, 2014). SP-5 displayed the highest peak viscosity, likely due to pre-existing aggregates that swell under heat and shear. In contrast, SP-7's minimal viscosity increase reflects limited swelling and low flexibility, driven by a rigid protein matrix lacking subunits for reconfiguration. These results demonstrate that the flow behavior of pre-gel soy protein solutions is closely tied to particle size, conformational flexibility, and protein subunit composition. Highly aggregated, rigid proteins (e.g., SP-18, SP-5) exhibit high consistency and limited flow, while more flexible and dispersed systems (e.g., SP-10) display lower viscosity but enhanced reactivity, critical for forming structured thermal gels.

Fig. 6.

Thermal and viscoelastic properties of heat-induced soy protein gels from five representative genotypes (SP-4, SP-5, SP-7, SP-10, SP-18); (A) RVA pasting profiles showing viscosity development during heating (25 → 95 °C), holding, and cooling phases. (B) Frequency sweep analysis of gels at 25 °C, showing storage modulus (G′) and loss modulus (G″) across a frequency range of 0.01–10 Hz. (C) Complex viscosity (η*) and overlaid G′/G″ values (bottom right) across frequencies, highlighting differences in gel network strength and resistance to deformation. All measurements were performed in triplicate (n = 3). “SP-x” codes refer to Table 1.

3.5.3. Rheological characterization of heat-induced gels

3.5.3.1. Elastic modulus and frequency dependence

Small-amplitude oscillatory shear (0.01% strain, within LVR; 0.01–10 Hz) confirmed solid-like behavior (G′ > G″) for all gels (Table 3). The highest G′ values occurred for SP-15 (3647 Pa) and SP-13 (2686 Pa), indicating dense, cohesive networks, whereas SPs 1, 2, 3, and 9 were < 150 Pa. tan δ values were low (0.18–0.29), reflecting elastic dominance with modest between-genotype differences.

Representative viscoelastic spectra (Fig. 6B) illustrate the range: SPs 10, 4, and 7 exhibited high G′ (2502, 2085, 2049 Pa, respectively) with low tan δ (0.18–0.20) and weak frequency dependence: signatures of robust, cross-linked networks. Moderate pre-aggregation in SP-4 did not hinder gelation. Rather, partially soluble aggregates likely integrated into the network and increased mesh density, provided reactive groups remained accessible. Their superior gel strength is likely attributed to sufficient protein unfolding during heating, which facilitates exposure of reactive moieties, such as sulfhydryl and hydrophobic groups. Exposure of these moieties promotes disulfide bonding and hydrophobic interactions, thereby enabling the formation of a robust, 3-D network (Ma & Chen, 2023).

In contrast, SP-5 and SP-18 had lower G′ (<600 Pa) and stronger frequency dependence, indicating weaker, more heterogeneous matrices. For SP-18, the combination of high K, n > 1, and large particles (Table 1, Table 3) is consistent with early-stage aggregation that restricts rearrangement and network fusion during heating. This finding aligned with prior findings where excessive protein aggregation impairs dynamic crosslinking and gel elasticity, resulting in brittle or discontinuous networks (Brodkorb, Croguennec, Bouhallab, & Kehoe, 2016).

SP-7 (glycinin-dominant) nonetheless formed a firm gel (high G′, low tan δ; Table 3), indicating that disulfide-stabilized glycinin assemblies can support a compact, rigid network even in the absence of β-conglycinin (Z. Huang et al., 2022; J. Renkema et al., 2001). These observations can indicate that the presence of specific subunits and their ratio and aggregation behavior critically influence the rheological properties of soy protein gels.

3.5.3.2. Complex viscosity trends and network strength

Across genotypes, complex viscosity (η*) decreased with frequency (shear-thinning), but the magnitude and slope varied (Fig. 6C; Table 3). SP-10 and SP-4 exhibited the highest η* values and flatter η* slopes, matching their elevated G′ and low tan δ and suggesting a well-connected structure that resists deformation across a range of timescales (Xia et al., 2022). SP-5 and SP-18 showed low η* with steeper declines, reflecting transient networks with greater rearrangement under oscillatory shear. SP-18's poor η* performance, despite its high pre-gel consistency (K = 7.70, Table 3), further highlights the disconnect between pre-gel viscosity and final gel integrity, a result likely driven by its extensive pre-aggregation and poor network fusion during heating (Qi et al., 2011).

SP-7 was intermediate, with relatively stable η* across frequency, indicating a dense, rigid but less extensible network. The glycinin-dominant composition of SP-7 likely contributes to this compact network, which, while firm, may lack structural flexibility and viscoelastic balance due to the absence of β-conglycinin subunits (Z. Huang et al., 2022; Xia, Siu, & Sagis, 2021). Together, η* trends mirror the G′ and tan δ hierarchy and underscore the role of genotype-specific structure (thiol exposure, conformational accessibility, aggregation state) in governing network connectivity.

3.5.3.3. Water holding capacity (WHC) and gel cohesiveness

WHC varied markedly, from 80% (SP-5) to 99% (SP-10, Table 3). Most isolates exceeded 90%, indicating efficient water entrapment. Consistent with a densely cross-linked and elastic network with small pores that resist water expulsion, SP-10 exhibited high G′, high η*, and maximal WHC (Table 3; Fig. 6C). In contrast, SP-5 and SP-18 exhibited lower WHC (80–82%), aligning with their weaker, heterogeneous gels and greater aggregation, which likely reduced network uniformity and hydration (Zhu et al., 2020). SP-7 showed moderately high WHC, indicating that its compact, fine-stranded network retained water effectively despite reduced extensibility (Z. Huang et al., 2022; Xia et al., 2021). Because isolates were prepared identically and showed similar residual ash (<5%) (Momen et al., 2025), WHC differences primarily reflect protein network topology and hydration rather than non-protein constituents.

Interestingly, one genotype (SP-9) exhibited very low gel elasticity (G′ ≈ 140 Pa, among the weakest gels) yet an extremely high WHC (∼97%, one of the highest). This indicates that water retention in the gel matrix is not solely governed by stiffness (elastic modulus). SP-9's soft gel could still form a very fine, cohesive network that traps water in its microstructure. Previous studies have noted that β-conglycinin-rich proteins tend to form softer, finer gels that hold water well, whereas glycinin yields firmer but more porous gels prone to water loss (Y. Huang et al., 2025). SP-9 may parallel those observations – its network, while weak, is dense enough at the microscopic level to prevent water expulsion under the gentle centrifugation used for WHC measurement. This example underscores that maximum WHC does not always require a high G′; the structural arrangement of proteins and the nature of protein-water interactions are equally important factors.

In summary, pre-gel flow and gel rheology collectively demonstrated that genotype governs network formation through the interplay of subunit composition (11S/7S ratio, Fig. 1), conformational accessibility (free –SH, H₀, protease susceptibility, Fig. 2), and aggregation state (AFM, DLS, Fig. 5). Genotypes with balanced 11S/7S and greater molecular accessibility (e.g., SP-10, SP-4) unfolded and cross-linked efficiently, forming cohesive, elastic gels with high WHC. In contrast, the glycinin-dominant SP-7 produced firm but less extensible gels, while highly aggregated isolates such as SP-5 and SP-18 exhibited high pre-gel viscosity yet weak, brittle networks, confirming that viscosity alone does not predict gel strength when early aggregation restricts network fusion.

Secondary structure trends further supported these genotype effects. A higher fraction of unordered structures (α-helix/random coil, Fig. 3) was generally associated with greater conformational flexibility and higher gel elasticity, whereas elevated antiparallel β-sheet content corresponded to more compact, rigid matrices and lower WHC (Table 3). These results align with earlier reports showing that moderate structural disorder enhances segmental mobility and intermolecular bonding during heat-induced gelation (Carbonaro et al., 2012; Fu et al., 2023; Pavlicevic et al., 2018). Thus, subtle shifts toward less-ordered conformations appear to facilitate unfolding and crosslinking, improving both viscoelastic strength and water retention.

Thermal behavior also paralleled gel performance. Genotypes with intermediate ΔH values produced stronger, more elastic gels, while those with very high ΔH (e.g., SP-18, Table 2) formed brittle networks, indicating that excessive structural cooperativity and packing hinder dynamic rearrangement during heating. T_d showed little association with gel strength, reinforcing that thermal stability alone does not ensure network fusion when conformational mobility is limited (Fu et al., 2023). Overall, ΔH reflects structural rigidity but not functionality in isolation, emphasizing the combined role of subunit balance, accessibility, and moderate stability in governing gelation.

Soybean genotype governs gel network formation by encoding distinct molecular phenotypes that emerge under standardized processing. Variations in subunit composition, aggregation propensity, sulfhydryl accessibility, conformational flexibility, and thermal cooperativity represent genotype-conditioned traits that collectively determine how proteins unfold, interact, and percolate into load-bearing gel networks.

4. Limitations and future directions

This study surveyed 20 soybean genotypes, which, while diverse, represent only a fraction of global soybean diversity. Further research on a larger panel of genotypes would broaden understanding of the protein's functional range. Additionally, our analyses were conducted under controlled lab conditions (pH 7 buffer, 10% protein, no added salt or polysaccharides). Future studies could examine how these genotypes perform in actual product matrices (e.g., in tofu or meat analogue formulations) and under various pH or ionic environments. Despite these caveats, the clear genotype-dependent trends we observed suggest genuine underlying differences worthy of further exploration, particularly at the genetic and molecular levels (e.g., sequencing protein-coding genes to pinpoint variants associated with superior functionality).

5. Application outlook

Our findings carry implications for both crop breeding and food formulation. The pronounced differences in gelation performance among genotypes indicate that it is feasible to breed or select soybeans for enhanced functionality. Breeders could use traits like 11S/7S ratio, solubility, or sulfhydryl content as selection criteria or screening proxies for protein quality in addition to yield and seed composition. Exploiting the genetic diversity in germplasm collections (as we have done here) can thus facilitate the development of soybean varieties tailored for specific end-uses (Singer et al., 2023). From the food industry perspective, our “elastic vs. firm vs. weak” gel classification suggests that certain cultivars may be naturally suited for particular product textures. For example, an ‘elastic’ high-WHC isolate could improve the juiciness and springiness of plant-based meat analogues, whereas a ‘firm’ isolate might better lend structure to a tofu or cheese analogue. Manufacturers might even blend protein isolates from different genotypes to achieve optimal texture and water retention in formulations. Importantly, utilizing genotype-selected soy proteins with superior functionality could reduce reliance on additives or extensive processing, aligning with clean-label trends.

CRediT authorship contribution statement

Shima Momen: Writing – original draft, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Sanjana Sawant: Writing – review & editing, Methodology, Investigation, Formal analysis. Lily Lincoln: Writing – review & editing, Formal analysis. Benjamin D. Fallen: Writing – review & editing, Funding acquisition, Conceptualization. Audrey L. Girard: Writing – review & editing, Supervision, Resources, Project administration, Methodology, Funding acquisition, Conceptualization.

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

Data availability

Data will be made available on request.