Tailoring soy protein isolate–alginate double emulsion composite gel microcapsules for cheese-like texture and probiotic protection in plant-based cheese analogues

Highlights

• RSM mapped the SPI-SA matrix to balance probiotic protection and cheese-like texture.
• SA enhanced mechanical strength via robust Ca²⁺ cross-linking.
• SPI-SA matrices provided superior protection in simulated gastrointestinal fluids.

1 Introduction

The global demand for plant-based alternatives has largely grown in the last decade, due to the rise of concerns regarding sustainability and animal welfare (McClements and Grossmann, 2021; Paul et al., 2020). For example, non-dairy cheese alternatives have become one of the major items in the plant-based food market, which can also serve as a lactose-free product for lactose-intolerant customers (Alehosseini et al., 2025; Kamath et al., 2022; Leal et al., 2025; Sözeri Atik and Huppertz, 2025). However, while recent advancements have successfully mimicked the sensory attributes of traditional dairy products, these plant-based analogs usually lack functional benefits, particularly the presence of beneficial live microorganisms, which are known for various health benefits to support the gut health and wellbeing of the host body (Ringel et al., 2012). Fortifying plant-based products with probiotics represents a critical step toward upgrading these products from sensory substitutes to functional, health-promoting alternatives.

The incorporation of viable probiotics into plant-based matrices, however, presents significant challenges. Probiotic cells are highly sensitive to environmental stresses during mechanical and thermal processing, cold chains, and the harsh acidic conditions of the human gastrointestinal tract (Ayyash et al., 2021; 2021; Kathiriya et al., 2023; Li et al., 2024). There are a few commercial non-dairy analogs and lab-scale samples reported in a recent review and studies (Alehosseini et al., 2025; Naeem et al., 2024; Sharma et al., 2018); however, the addition of probiotics directly into non-dairy alternatives does not ensure the viability of probiotics during processing, storage, retailing, and gastrointestinal (GI) delivery. Microencapsulation has emerged as a highly effective strategy to protect these sensitive microorganisms from these severe environments (Frakolaki et al., 2021; Yao et al., 2020). Among microencapsulation technologies, an emulsion-based microcapsule system has gained significant attention due to the scalability, small particle size, high production rate, and high protectability (Chen and Chen, 2007). In our previous study, we have advanced the system by combining external gelation via W/O/W/O emulsion state to form gelled W/O/W double emulsion microcapsules (Udo et al., 2025a). The use of a W/O/W double emulsion, combined with a protein-polysaccharide composite matrix and Ca²⁺ cross-linking, provides a sophisticated multi-layer barrier including shell surface and multi-nuclear matrices. Soy protein isolate (SPI) offers strong surface activity and superior pH-buffering capacity, and sodium alginate (SA) ensures robust structural cross-linking, which are widely used and documented for use as wall materials. In our previous article, we have reported that this novel microencapsulation system effectively protects the entrapped probiotics from the simulated gastrointestinal juices and cold storage environments (Udo et al., 2025b).

Beyond their biological protection capability, the textural and morphological characteristics of the SPI-SA microcapsule system are highly tunable depending on wall material composition, suggesting potential compatibility with various semi-solid plant-based food matrices, such as semi-hard cheese products. However, despite extensive research on microencapsulation systems, systematic investigation of how wall material formulation simultaneously governs macroscopic textural attributes, such as firmness and stickiness, and biological protection of probiotics remains limited. Understanding these relationships is essential for designing microcapsules that meet both the structural and functional requirements of complex food applications.

Therefore, building on the W/O/W double emulsion microencapsulation system that was successfully established in our previous study (Udo et al., 2025a), the objective of this study was to explore the formulation of the cross-linked SPI-SA wall materials to tailor the cheese-like textural characteristics of the gelled microcapsules while preserving probiotic protection, so that they can serve as a new probiotic-fortified building block for plant-based cheese analogues. To investigate the impacts on functional and textural properties of microcapsules, a 3² full factorial design with center-point replication and Response Surface Methodology (RSM) were employed, which are widely adapted for the optimization of various microencapsulation technologies (Gholivand et al., 2024; Jalalizand and Goli, 2021; Premjit and Mitra, 2021). This study ultimately provides a systematic understanding of how SPI and SA wall material composition governs both the structural integrity and biological performance of W/O/W gelled microcapsules, providing a basis for the future development of probiotic-fortified plant-based cheese analogues.

2 Materials and methods

2.1 Materials

All ingredients used for microencapsulation were of food-grade and Generally Recognized as Safe (GRAS) status. Soy protein isolate (SPI) was obtained from BulkSupplements (Henderson, NV, United States of America), with a declared protein content of 25 g per 30 g serving and containing no added ingredients other than soy protein isolate. Sodium alginate (SA) was obtained from bioWORLD (Ohio, United States of America). For the microbiological procedures, freeze-dried Lactobacillus rhamnosus GG (LRGG) was sourced from Culturelle® (Amerifit Inc., United States of America). Peptone was purchased from Becton, Dickinson and Company (New Jersey, United States of America), and MRS broth and agar were purchased from Alpha Biosciences (Maryland, United States of America). Digestive enzymes for simulated digestion (pepsin, pancreatin, lipase, and bile extract) and all other chemicals not listed above were acquired from Sigma-Aldrich (Missouri, United States of America). All chemical reagents were of analytical grade. To ensure microbiological integrity, all glassware, media, and solutions were rigorously sterilized via autoclaving before use.

2.2 Cell culture

LRGG was cultured for microencapsulation as described in (Udo et al., 2025a). Briefly, the freeze-dried LRGG were revitalized in sterile 0.1% peptone solution at 37 °C for 18 h, followed by sub-culturing on MRS agar at 37 °C for 48 h to obtain LRGG colonies. Bacterial cells (approximately 10 mL) were inoculated into MRS broth and then incubated at 37 °C for 22 h. Cultured cells were harvested and washed by centrifugation at 4,400 × g using sterile peptone water to obtain free LRGG suspension (9–10 Log CFU/mL). The suspension was stored at 4 °C until use.

2.3 Microencapsulation of LRGG

Microencapsulation was performed following a previously published paper with slight modification (Figure 1) (Udo et al., 2025a). To create the LRGG microcapsules, a primary W/O emulsion was first formed by homogenizing a probiotic suspension with soybean oil (1% v/v Tween 80) at a volume ratio of 4:6 (cell suspension to oil) and 10,000 rpm for 30 s. This was then transitioned into a W/O/W double emulsion by mixing with an aqueous phase containing Soy Protein Isolate (SPI) and Sodium Alginate (SA) at a 1:3 ratio, followed by homogenization at 10,000 rpm for 5 min. No pH adjustment was applied to the SPI or SA solutions, both of which were prepared in deionized water without buffering. The formulation of SPI and SA is summarized in Table 1. The resulting emulsion was dispersed in soybean oil and cross-linked by same volume of 0.1 mol/L CaCl₂ at 800 rpm for 30 min. The CaCl₂ concentration was held constant across all formulations so that the effect of the wall material composition could be evaluated independently of the cross-linking agent, and was in excess relative to the alginate carboxylate groups at all SA levels examined. Under these conditions the aqueous CaCl₂ contacts the dispersed W/O/W droplets at the oil–water interface generated by stirring, allowing Ca²⁺ to reach the alginate-containing outer aqueous phase and induce external gelation. Finally, after settling at 4 °C for an additional 30 min for further gelation, the separated oil phase was decanted, and the microcapsules were isolated via vacuum filtration.

TABLE 1. Experimental design matrix of SPI-SA probiotic microcapsules for the response surface methodology.

The asterisks (*) represent the center point of the factorial design matrix with three repetitions.

2.4 Microscopic observations

The morphology and internal structure of the freeze-dried SPI-SA probiotic microcapsules were characterized using optical light microscopy, confocal laser scanning microscopy (CLSM), and scanning electron microscopy (SEM).

2.4.1 Optical light microscopy

The general morphology of the microcapsules was observed using an optical light microscope (Nikon Eclipse E600, Nikon, Tokyo, Japan) as described in our previous article (Udo et al., 2025a). Each sample was loaded onto a glass slide with a few drops of distilled water and covered with a coverslip. Images were captured at ×10 and ×20 magnifications.

2.4.2 Confocal laser scanning microscopy

The oil phase, the outer-phase SPI, and the probiotic cells were labeled with Coumarin 6, Fast Green FCF, and Hoechst 33342, corresponding to excitation/emission maxima of 488/505, 633/698, and 350/461 nm, respectively. Coumarin 6 was first dissolved in ethanol (5 mg/mL), and 180 μL of this stock was incorporated into 10 mL of soybean oil. Fast Green FCF was prepared in distilled water (10 mg/mL), of which 0.1 mL was added to 10 mL of the SPI solution. For cell labeling, a bacterial suspension in peptone water (9 Log CFU/mL, 50 mL) was treated with 50 μL of Hoechst 33342 (10 mg/mL) for 15 min, after which residual dye was removed by three washing cycles (centrifugation at 4400 ×g, 5 min, peptone water). The labeled components were subsequently assembled into microcapsules as described above. Imaging was carried out on a Zeiss LSM 710 system (Carl Zeiss, Oberkochen, Germany) fitted with a ×20 dry objective. Blue, green, and pink fluorescence signals were acquired sequentially under 405 nm diode, argon, and HeNe2 laser excitation, and a stack of roughly ten optical sections (2 μm each) was reconstructed into a single projected image using the Z-stack function.

2.4.3 Scanning electron microscopy

The surface morphology of the microcapsules was examined using scanning electron microscopy (SEM; Hitachi SU3900, Hitachi, Tokyo, Japan). Freeze-dried microcapsule samples were mounted on the stage and observed under variable pressure mode at an accelerating voltage of 20.0 kV and a chamber pressure of 30 Pa, without gold coating.

2.5 FTIR spectroscopy

FTIR spectra were recorded on a Bruker Alpha II spectrometer (Bruker, Germany) in transmission mode, following the procedure described previously. Spectra of SPI, SA, the non-cross-linked SPI-SA mixture, and the Ca²⁺ cross-linked microcapsules were collected over 4000–400 cm^-1 at a resolution of 4 cm^-1, and were baseline-corrected and normalized prior to peak assignment.

2.6 Particle size measurements

Particle size distribution was determined by laser diffraction using an LS 13 320 analyzer equipped with a Universal Liquid Module (Beckman Coulter, CA, United States of America), covering a measurement range of 0.375–2000 µm. Samples were dispersed in deionized water and measured for 90 s at an obscuration of approximately 20%. Size distributions were calculated using the Fraunhofer optical model (rf780 d); as this model does not require the refractive index of the sample to be specified, no refractive index assumptions were made. Samples were sonicated using the built-in sonicator of the module throughout the measurement. The 10, 50 (median), and 90 percentiles of the particles were obtained as D _*10*, D _*50*, and D _*90* in addition to the mean. Measurements were conducted in analytical triplicate. The distribution of particle sizes was calculated as SPAN using the equation as follows:

SPAN = D 90 − D 10 D 50

2.7 Moisture contents

The moisture content of the microcapsules was determined gravimetrically. Microcapsules were held at 4 °C after collection before the initial weight was recorded. Approximately 1 g of the wet microcapsules was placed into aluminum dishes and dried in a hot air oven at 105 °C for 3 h. The moisture content was calculated as the percentage of weight loss relative to the initial weight of the sample, using the following equation:where W initial is the weight of the wet microcapsules, and W dried is the weight of the microcapsules after drying. All measurements were performed in analytical triplicate.

Moisture % = W initial − W dried W initial × 100

2.8 Encapsulation efficiency

The encapsulation efficiency (EE) of LRGG was determined by comparing the viable bacterial counts in the initial cell suspension and in the final microcapsules, as described previously (Premjit and Mitra, 2021; Udo et al., 2025a; Zhou et al., 2024). To release the encapsulated bacteria, 1 g of microcapsules was suspended in 9 mL of a 1% (w/v) sterile sodium citrate solution at pH 6.0 and gently shaken at room temperature on an orbital shaker for 30 min until complete dissolution. After dissolution, the suspension was serially diluted and plated onto MRS agar, followed by incubation at 37 °C for 48 h before counting the colony-forming units (CFU). The EE (%) was calculated using the following equation:where N is the number of viable cells released from the microcapsules (CFU/g), and N₀ is the total number of viable cells added (CFU/mL). Following the convention adopted in previous studies using this expression, EE here describes the proportion of the logarithmic viable count retained in the microcapsules relative to that of the initial cell suspension, and is not the absolute fraction of encapsulated cells.

E E % = log ⁡ N log ⁡ N 0 × 100

2.9 Texture analysis

Texture analysis of the microcapsules was performed using a TA-XT2i universal testing machine (Stable Micro Systems Ltd., Godalming, UK) equipped with a 25-kg load cell. The measurement was conducted using a TTC Spreadability Rig (HDP/SR) mounted on a Heavy Duty Platform (HDP/90). Specifically, a TA-2A 90-degree acrylic cone probe was used as the upper male fixture, working in conjunction with matching female sample holders.

Approximately 15 g of samples were packed into the female sample holders and gently pressed down to eliminate any visible air pockets and excess sample on the surface, leaving a flat test area. The TA-2A cone probe was then driven into the sample in compression mode with a test speed of 3.0 mm/s, a post-test speed of 10.0 mm/s, and a penetration distance of 23 mm. Data acquisition rate was set at 200 pps. From the obtained force-time curve, the maximum positive force during the downward stroke was recorded as the firmness (g) of the sample, while the maximum negative force observed during probe withdrawal was recorded as stickiness (g). Firmness and stickiness are reported in grams (g), as recorded by the texture analyzer, and these values correspond to the force required, expressed as the equivalent mass. A single measurement was performed for each experimental run, while the pure experimental error was estimated from the three independent center point replicates.

2.10 CIE Lab* measurements

The color of the microcapsules was measured using a HunterLab MiniScan EZ Spectrophotometer (Hunter Associates Laboratory, VA, United States of America) with a 45/0 geometry and a Large Area View (LAV) port. The CIE L ^*** a ^*** b ^*** color scale was evaluated under illuminant A and a 10° standard observer angle. L ^*** indicates lightness/darkness, a ^*** indicates redness/greenness, and b ^*** indicates yellowness/blueness. The measurements were performed in analytical triplicate by filling microcapsule samples in a cup to ensure a uniform surface.

To further characterize the optical properties, the chroma (C ^***), hue angle (h), and whiteness index (WI) were calculated using the following equations:

C * = a 2 + b 2

h = tan − 1 b * / a *

WI = 100 − 100 − L * 2 + a * 2 + b * 2

The C ^*** value represents the color intensity or saturation, while h defines the color nuance angles. The WI was used to assess the proximity of the sample color to ideal white.

2.11 In vitro digestion study

The survivability of microencapsulated LRGG was evaluated as described in previous papers with a slight modification (Udo et al., 2025a; Udo et al., 2025b). Briefly, 2 g of samples were suspended in 18 mL of simulated gastric fluid (SGF, pH 2.5) containing 5 mg/mL pepsin and 6 mg/mL mucin. The mixture was incubated in a shaking bath at 37 °C and 150 rpm for 2 h. Subsequently, 10 mL of the resulting chyme was mixed with 10 mL of simulated intestinal fluid (SIF, pH 8.1) containing 18 mg/mL pancreatin, 3 mg/mL lipase, and 60 mg/mL bile extract. The intestinal digestion was conducted for an additional 4 h at 37 °C. Viable cell counts (CFU/g) were enumerated at 0 (initial count), 1, 2, 4, and 6 h using the plate count method on MRS agar.

2.12 Storability in cold storage

The storability of LRGG was investigated during cold storage as described in a previously published article with a slight modification (Udo et al., 2025a). Cold storage stability in the present study refers to the retention of probiotic viability during refrigerated storage. Briefly, the microencapsulated LRGG were stored at 4 °C for 2 months, a timeframe relevant to refrigerated food applications. The number of viable cells was monitored at 1-month intervals using serial dilutions and the standard plate count method.

2.13 Experimental design and statistical analysis

A 3² full factorial design with center-point replication was chosen for the response surface methodology (RSM) (Table 1). Two independent variables were SPI concentration (X _*1*, 3.0%–6.0% w/w) and SA concentration (X _*2*, 0.5%–2.0% w/w). The experimental design consisted of 11 runs in total, performed in random order, comprising a 3 × 3 = 9 runs plus two additional replications at the center point (SPI 4.5%, Alginate 1.0%) to estimate the pure experimental error and assess the lack of fit. The three center point runs (Runs 5, 6 and 7) were prepared as separate batches and therefore constitute independent production replicates. Particle size, moisture content, encapsulation efficiency and color were measured in analytical triplicate on each batch, whereas texture was measured once per run.

The experimental data were fitted to a second-order polynomial response surfacewhere Y represents the predicted response variables, including EE, moisture content, texture profile, CIE, storability, and survival of LRGG; β _*0* is the intercept; β _*1* and β _*2* are the linear coefficients; β _*12* is the interaction coefficient; β _*11* and β _*22* are the quadratic coefficients.

Y = β 0 + β 1 X 1 + β 2 X 2 + β 12 X 1 X 2 + β 11 X 1 2 + β 22 X 2 2

All statistical analyses and response surface modeling were performed using R software (version 4.2.1) with the rsm, and psych packages. Stepwise regression analysis was employed to eliminate non-significant terms (p > 0.05) and simplify the models where appropriate, based on Akaike Information Criterion (AIC). Stepwise selection was performed on grouped first-order, interaction and pure quadratic terms, thereby maintaining model hierarchy. The statistical significance of the models and individual regression coefficients was evaluated via Analysis of Variance (ANOVA) at a 95% confidence level (p < 0.05). The adequacy and predictive accuracy of the fitted models were verified by evaluating the coefficient of determination (R ²), adjusted R ², and the lack-of-fit test. Viable counts below the detection limit of the plate count method (<2 Log CFU/g) were assigned the value of the detection limit (2.0 Log CFU/g) for regression analysis.

3 Results and discussion

3.1 Model fitting and statistical analysis

The matrix of the 3² full factorial design with center-point replication and the corresponding experimental responses for the nine evaluated parameters of the probiotic microcapsules are presented in Table 2. To evaluate the adequacy and predictive accuracy of the response surface models, an analysis of variance (ANOVA) was performed, and the statistical parameters are also summarized in Table 3. The main text of this study focuses selectively on the most critical nine key performance indicators (KPI) for the physical and functional evaluation of the microcapsules, which could be described by RSM models. Namely, encapsulation efficiency (EE), moisture content, firmness, stickiness, median particle size (D _*50*), SPAN, gastrointestinal survival (2 h and 6 h), and cold storage stability (2 months) are discussed in detail, as these parameters directly dictate the structural integrity and applicability of the microcapsules to non-dairy analogs. Detailed response surface plots and supplementary data for the secondary parameters, such as color parameters and the rest of particle size, gastrointestinal survival, and storability parameters, are provided in the Supplementary Material (Supplementary Tables S1–S3; Supplementary Figures S1–S3).

TABLE 2. Experimental design and observed responses of microcapsules under different SPI and SA concentrations.

Asterisks (*) represent that the viable cell counts are lower than the detection limit (<2 Log CFU/g). SPI: soy protein isolate; SA: sodium alginate; EE: encapsulation efficiency; D₅₀: Median particle size. Gastric (2 h), Intestinal (6 h), and Storage (2 m) values are presented as viable counts (log CFU/g). Runs 5–7 are replicate runs at the center point (SPI, 4.5%, SA, 1.0%).

TABLE 3. Summary of ANOVA for the fitted response surface models after AIC-based stepwise term selection.

Significance levels: *p < 0.05, p < 0.01, *p < 0.001. ¹Terms retained in the final model with p < 0.10. Terms with p < 0.05 are marked with asterisks as indicated above; terms without asterisks were retained as marginally influential and are not considered statistically significant.

The developed models exhibited high coefficients of determination (R ²), ranging from 0.695 to 0.990. To assess predictive reliability rather than goodness of fit alone, predicted R ² values were calculated from the PRESS statistic for each model (Table 3). On this basis the models fall into two groups. For cold storage stability at 2 months, firmness, median particle size, stickiness and intestinal survival at 6 h, the predicted R ² ranged from 0.596 to 0.931 and the lack-of-fit tests were non-significant, indicating that these equations describe and predict the responses adequately within the investigated design space. For moisture content, SPAN, encapsulation efficiency and gastric survival at 2 h, the predicted R ² was 0.409 or lower, and the lack-of-fit test was significant for gastric survival at 2 h. These Supplementary Material four models are therefore presented as descriptive representations of the observed trends and are not used for prediction. The model validity, regression equations, stationary points, and RSM plots of each KPI will be further discussed in the following sections.

3.2 Structural characterization of microcapsules

First, the visual appearance of the SPI-SA probiotic microcapsules under different formulations is shown in Figure 2. Overall, the microcapsules exhibited a spread or crumbly appearance depending on the concentrations of SPI and SA. It was clearly indicated that the more SPI and SA were added, the coarser, more agglomerated microcapsule textures were found in the samples. This agglomeration may be due to the increased particle size distribution and sticky texture. As the SA and SPI levels increased, the flowability of the emulsion decreased, which led to the large droplet size and, thereafter, large particle size in the end products (Table 2). Furthermore, the strong cross-linking of the alginate network, combined with the protein matrix of SPI, likely promoted the fusion of microcapsules into denser, clustered macroscopic structures. This transition of textures between spreadable and crumbly textures is highly desirable for the target applications, as it may span a range of structural and textural characteristics comparable to those of crumbly and spreadable non-dairy alternatives.

3.2.1 Microscopic observations

The structural organization of the SPI-SA microcapsules was examined by different imaging techniques (Figure 3). Optical light microscopy (Figure 3a) revealed irregularly shaped, agglomerated particles with a broad size distribution, consistent with the median particle size determined by laser diffraction (Table 2). The particles exhibited a dense appearance with entrapped W/O/W emulsion, rather than discrete spherical beads.

CLSM imaging (Figure 3b) resolved the multi-phase structure within the microcapsules. The green fluorescence of Coumarin 6, which labeled the oil phase, was distributed as discrete droplets throughout the particle interior, confirming that the inner W/O emulsion was successfully entrapped within the outer matrix. Notably, the blue fluorescence of Hoechst 33342, corresponding to LRGG cells, overlapped with the green signal to produce a turquoise appearance, indicating that the probiotic cells were localized in close association with the oil droplets. This co-localization is consistent with the encapsulation of LRGG within the internal aqueous droplets. The pink signal of Fast Green FCF, corresponding to SPI, extended across the continuous phase and surrounded the oil-cell domains, indicating that the protein was distributed both at the oil-water interfaces and within the surrounding alginate network. These observations confirm the multi-nuclear architecture characteristic of a W/O/W double emulsion embedded in a Ca²⁺ cross-linked gel matrix.

SEM observation (Figure 3c) further showed a rough, irregular surface topography with visible folds and cavities, which likely arose from the collapse of the hydrated gel network during freeze-drying. Such porous surface morphology has been commonly reported for calcium-alginate based microparticles (Coluccino et al., 2016). Structurally, this organization, a continuous SPI–alginate protein–polysaccharide matrix enclosing discrete entrapped oil droplets, resembles the microstructure described for natural and plant-based cheeses, in which a continuous protein network encloses dispersed fat globules. It should be noted that SEM was performed on freeze-dried particles, and the observed surface morphology may therefore differ appreciably from the hydrated structure that governs the textural behavior of the microcapsules.

3.2.2 FTIR spectroscopy

The molecular interactions within the SPI-SA microcapsule system were investigated by FTIR spectroscopy, comparing the spectra of pure SPI, pure SA, their physical mixture (prior to cross-linking), and the Ca²⁺ cross-linked W/O/W microcapsules (Figure 4).

Pure SPI exhibited its characteristic amide bands, with the amide I band (C=O stretching) at 1638 cm^–1 and the amide II band (N–H bending and C–N stretching of the peptide backbone) at 1521 cm^–1. Pure SA displayed the asymmetric and symmetric stretching vibrations of its carboxylate (COO⁻) groups at 1597 and 1410 cm^–1, respectively, along with the C–O–C glycosidic band at 1027 cm^–1 (Alesaeidi et al., 2022; Li et al., 2016; Zhang et al., 2023). The clearest spectral indication of ionic cross-linking was observed in the symmetric COO⁻ stretching region. This band appeared at 1410 cm^–1 in pure SA and at 1401 cm^–1 in the physical mixture, remaining close to that of native alginate. In contrast, it shifted to 1418 cm^–1 in the Ca²⁺ cross-linked microcapsules, representing a shift of +17 cm^–1 relative to the physical mixture. Because this shift was absent in the non-cross-linked mixture, it is consistent with coordination of Ca²⁺ ions by the carboxylate groups of alginate rather than with simple physical blending of the two biopolymers. Similar high-wavenumber shifts of the symmetric COO⁻band upon Ca²⁺ cross-linking of alginate have been reported previously (Urquiza et al., 2011). We note that the asymmetric COO⁻stretching region (1580–1630 cm^–1) overlapped with the amide I band of SPI in the microcapsule spectra and was therefore not used as a primary indicator of cross-linking to avoid ambiguous assignment. Accordingly, the symmetric COO⁻shift, which lies in a spectral region free from protein amide contributions, was adopted as a marker consistent with Ca²⁺ coordination.

Moreover, the amide II band provided a further indication of protein–polysaccharide interaction. This band remained unchanged at 1521 cm^–1 in both pure SPI and the physical mixture, but it shifted to 1545 cm^–1 (+24 cm^–1) in the Ca²⁺ cross-linked microcapsules. This is consistent with a change in the microenvironment of the peptide backbone following incorporation of SPI into the calcium–alginate network (Cao et al., 2022).

3.3 Morphological and physical properties

The macroscopic observations are quantitatively supported by the median particle size (D _*50*) and the moisture content analysis. As shown in Table 2, the median particle size of the microcapsules ranged widely from 98.91 to 334.20 µm. According to the ANOVA results (Tables 3, 4), the SA concentration exhibited a dominant positive effect on the particle size (p < 0.001), with an excellent determination coefficient (R ² = 0.983). The response surface plot (Figure 5a) illustrates that the increase in SA dramatically enlarged the microcapsule size, while the effect of SPI was relatively minor. This increase in size at higher SA concentrations is likely related to the thickening effect of the alginate as a more viscous aqueous phase is expected to resist droplet breakup during homogenization and to yield larger droplets, which are then stabilized by Ca²⁺ cross-linking. Other studies have also reported a similar trend the size of alginate particles is proportional to the concentration (Anani et al., 2022).

TABLE 4. Regression equations and stationary points for the fitted response surface models after AIC-based stepwise term selection.

X_*1*: Soy Protein Isolate (%); X _*2*: Sodium Alginate (%). Equations are expressed in uncoded (actual) units. Stationary points located outside the investigated ranges of SPI (3.0%–6.0%) and SA (0.5%–2.0%) represent mathematical extrapolations of the fitted models.

In addition to the average particle size, the SPAN value was evaluated to determine the uniformity of the size distribution, as presented in Table 2. The SPAN values ranged from 1.55 to 2.19, indicating a relatively narrow and consistent size distribution across all formulations. According to the ANOVA results and regression models (Tables 3, 4), the SPAN was significantly affected by SA concentration (p = 0.004) and the SPI × SA interaction (p = 0.002), with an R ² of 0.907. The significant interaction indicates that the effect of SPI on size uniformity depends on the SA level. The RSM plot (Figure 5b) shows that the lowest SPAN values were observed at moderate concentrations of both polymers. This suggests that an optimal balance between SPI and SA is essential to maintain a stable emulsion viscosity, which in turn ensures a uniform droplet breakup during the extrusion process. From a sensory perspective, a narrow size distribution is highly desirable for non-dairy applications, as it prevents a “gritty” mouthfeel and contributes to a smooth, homogeneous texture similar to traditional dairy products (Emon et al., 2025). However, no sensory evaluation was conducted in the present study, and the oral perception of these particles remains to be assessed.

Regarding the moisture content, the samples ranged from 40.0% to 64.9% (Table 2). The regression model for moisture content was significant (p < 0.05), and the response surface plot (Figure 5c) revealed a saddle-shaped trend. This suggests a complex interaction in which the SPI and SA concentrations balance the matrix’s water-holding capacity. These values are typical of hydrated gel systems and should not be interpreted in the same terms as those of dried probiotic powders, in which low moisture content is the primary means of extending shelf life by suppressing metabolic activity during storage (Chen et al., 2023). Wet microcapsules follow a different route to stability, and have been reported to maintain probiotic viability for two to 3 months under refrigeration, considerably longer than free cells (Bustamante et al., 2025; Teymoori et al., 2024; Xu et al., 2026). Within the range examined here, moisture content therefore reflects differences in matrix water-holding capacity rather than a direct indicator of storage stability. Water activity, which would be more informative in this respect, was not determined in the present study.

On the other hand, the concentrations of SPI and SA did not exhibit a substantial impact on the color parameters, including CIE Lab,* Chroma (C), Hue Angle, and Whiteness Index (WI) (Supplementary Tables S1–S3). Lightness (L*) and yellowness (b**) are considered critical attributes for characterizing samples as non-dairy alternatives. For example, the obtained Lab* values are quite similar to feta-type cheese samples previously reported (Hammam et al., 2021). However, these parameters could not be significantly explained by the regression models (p > 0.05, with no significant terms). This lack of sensitivity to polymer concentration suggests that other factors, such as particle size distribution and the light-scattering effects of the internal oil droplets, may play a more dominant role in determining the optical properties. Similarly, Chroma and WI did not yield significant regression models (p > 0.05), indicating that desired color values cannot be precisely tuned by varying SPI and SA levels alone within the tested range. Nevertheless, the WI values remained relatively high and stable, ranging from 81.51 to 86.18 across all samples. This inherent whiteness demonstrates that the microcapsules possess a favorable neutral base, implying the potential to easily modify the final product color through the incorporation of other functional ingredients or natural colorants. For reference, natural cow-milk cheeses have been reported at approximately L* = 82.6, a* = 0.3, and b* = 17.4, and commercial plant-based block-style cheese analogues span L* = 79.8–93.6 (Ali et al., 2025; Milovanovic et al., 2020); the high lightness and whiteness index of the present microcapsules therefore fall within the range reported for white, fresh-type cheeses and plant-based analogues, supporting their suitability as a neutral base for plant-based cheese products.

3.4 Encapsulation efficiency

The ability of the wall materials to efficiently entrap the probiotic cells is a critical factor for successful microencapsulation. As shown in Table 2, the encapsulation efficiency (EE) of LRGG within the SPI-SA microcapsules was remarkably high across all experimental runs, ranging from 93.65% to 98.59%. The close values of EE were also found in other emulsion-based probiotic microcapsules (Sun et al., 2024; Zhou et al., 2024). The overall regression model for EE did not reach statistical significance (p = 0.087, R ² = 0.695), indicating that EE was largely insensitive to the wall material composition within the tested range. Within the model, only the linear term of SA was significant (p = 0.040), whereas neither the linear nor the quadratic term of SPI contributed significantly (p > 0.30). EE was therefore treated as a secondary indicator rather than a response that could be reliably modelled by RSM, and the corresponding response surface (Figure 5d) is presented for descriptive purposes only.

The consistently high EE values, even at lower concentrations of wall materials, demonstrate the robust protective capacity of the SPI-SA matrix. This high performance can be attributed to the synergistic interaction between the emulsifying properties of SPI and the rapid gelation of SA. SPI likely stabilized the probiotic cells within the oil droplets of the emulsion (Guo et al., 2025), while the subsequent Ca²⁺-induced cross-linking of the SA network formed a secondary physical barrier that effectively prevented the leakage of the bacterial cells during the hardening process. Other studies also demonstrated that the encapsulation efficiency of SPI-SA microcapsules showed an increase in EE when a higher SA concentration was used, possibly due to the dense cross-linking network formed in the microcapsules (Hadzieva et al., 2017; Soliman et al., 2013). This entrapment capability is highly advantageous for non-dairy applications, as it ensures that the intended dosage of probiotics remains within the food matrix during processing, thereby guaranteeing the functional value of the final product.

3.5 Textural properties

The textural characteristics of probiotic microcapsules, specifically firmness and stickiness, are essential in determining their suitability for integration into plant-based food matrices. The force-time curves obtained from the compression tests (Figure 6) provided a clear profile of these attributes. As summarized in Table 2, the firmness of the microcapsules varied significantly across the experimental design, ranging from 113.3 g to 14610.2 g. Because firmness spanned more than two orders of magnitude across the design space, a second-order model fitted to the untransformed data showed a significant lack of fit (p = 0.022) despite a high coefficient of determination (R ² = 0.964). Firmness was therefore modelled on a log₁₀ scale. All five terms of the resulting second-order model were significant (p = 0.001–0.032), and the model gave R ² = 0.987 with a non-significant lack of fit (p = 0.098) and a predicted R ² of 0.914 (Tables 3, 4). The residuals were within ±0.24 log units across the entire design space. It should be noted that the pure error estimated from the three center point replicates was small (0.042 log units), which renders the lack-of-fit test sensitive to modest model deviations even for a well-fitting model.

The response surface plot (Figure 5e) demonstrates a log-linear increase in firmness as the SA concentration rose from 0.5% to 2.0%, consistent with the formation of a more robust and densely cross-linked calcium alginate hydrogel network at higher polymer densities. SPI also contributed significantly to firmness (p = 0.008), and its presence within the alginate matrix likely provided a filler effect reinforcing the overall structure. The negative SPI × SA interaction term (p = 0.013) indicates that the two polymers do not act additively: the increase in firmness with SA concentration was less pronounced at high SPI levels, suggesting that the protein matrix partly moderates the stiffening effect of the calcium–alginate network.

Regarding stickiness, the values ranged from −106.2 g to −2370.6 g. Unlike firmness, the stickiness was significantly influenced by both SPI and SA concentrations (p < 0.05). Higher concentrations of both polymers led to an increase in the negative force required to withdraw the probe, indicating a more cohesive and adhesive matrix (Figure 5f). This increased stickiness at higher SPI levels can be attributed to the surface-active nature of soy proteins around the capsules.

This broad range of firmness and stickiness is highly strategic for the product development of plant-based dairy foods. For example, formulations with lower firmness are ideal for creating spreadable textures, whereas those with high firmness can provide the structural integrity required for firmer semi-solid products. For reference, previous studies have reported the hardness of Gouda cheese (1636 g), pulse-based cheese analogs (1883–2903 g), commercial vegan cheese analogs (5787 g), and goat and sheep milk cheeses (about 3000–5000 g) (Alkhalaileh, 2024; Ferawati et al., 2021). Notably, commercial plant-based cheese analogue slices have been reported at firmness values of roughly 2.8–3.8 N (approximately 285–390 g) (Ali et al., 2025), whereas the block-type analogues and natural cheeses cited above fall in the ∼1,600–5,800 g range. Benchmarking our formulations against these reference values indicates that intermediate SA levels (approximately 1.0%–1.5%) yield firmness overlapping the semi-hard and sliceable cheese range, while the lowest SA levels approach the spreadable end, allowing the optimal formulation to be selected according to the targeted cheese texture.

Furthermore, it is widely known that the moisture and stickiness of hydrocolloid-structured plant-based dairy foods are generally higher than those of their conventional counterparts due to the use of hydrocolloids for structuring (Ali et al., 2025). As highlighted in a recent systematic review by Rune et al. (2026), the absence of a continuous protein network in many plant-based products necessitates a heavy reliance on hydrocolloids, such as alginates and gums, to provide structural integrity. The strong water-binding capacity of these complex polysaccharide networks typically results in higher adhesiveness and gumminess. Therefore, the pronounced stickiness of the SPI-SA composite may be directly attributed to the hydration and water-retention of the alginate matrix. Importantly, this property provides the binding force required for the microcapsules to cohere into a unified, macroscopic mass.

It was also indicated that the firmness and stickiness are strongly correlated, as shown in the correlation matrix table (Supplementary Figure S4). While some degree of stickiness is necessary for the microcapsules to bind within a semi-solid mass, excessive stickiness must be balanced to ensure the final product remains easy to handle and does not adhere unpleasantly to the palate. Overall, it was confirmed that the formulation of SPI and SA wall materials significantly impacts the textural profile of microcapsules, which enables the textural properties of the microcapsules to be adjusted for incorporation into plant-based food products.

Nevertheless, two limitations of the present textural characterization should be noted. Rheological parameters such as the storage and loss moduli or yield stress, and additional texture-profile attributes, were not determined, and these would be required for a full description of a cheese-like characteristics. Further textural profile analysis and comparison with cheese products should be considered for future applications.

3.6 Gastrointestinal digestions and cold storage

The survivability of free LRGG has been reported previously under the same digestion and storage conditions (Udo et al., 2025a); the present study therefore focuses on how the wall material composition modulates this protection, and differences reported here are among encapsulated formulations. According to the ANOVA (Tables 3, 4) and response surface plots (Figure 5g), the regression model for gastric survival at 2 h was marginally significant (p = 0.050, R ² = 0.835), although the lack-of-fit test was significant (p = 0.014). Within these limits, the response surface indicates that the first-order effects of SPI and SA were significant (p = 0.031), whereas the interaction term did not reach significance at p < 0.05 (p = 0.066).

By the end of 6 h of the combined gastric and intestinal digestion, the survival ranged from the detection limit (<2 Log CFU/g) to 7.11 Log CFU/g, which could be well explained by the linear effects of both SPI and SA (p = 0.002, R ² = 0.867). The regression model (Tables 3, 4) and response surface plot (Figure 5h) also indicate the significant impacts of SPI and SA concentrations for the delivery of LRGG in the intestinal site. This is possibly due to the combination of the high buffering capacity of SPI and the structural framework of SA. It should be noted, however, that viability at the end of digestion varied considerably across the design space, and only the formulations combining the highest SPI and SA levels approached the therapeutic threshold of 10⁷ CFU/g commonly cited for probiotic products (Ouwehand, 2017). The remaining formulations fell below this level, and intestinal delivery in the present system should therefore be understood as strongly formulation-dependent rather than uniformly achieved.

The long-term viability of probiotics during refrigerated storage (4 °C) is also a critical quality parameter to ensure an adequate viable number of LRGG in the products. By the end of 2 months of refrigerated storage, the viability dropped into the range of 6.96 and 7.93 Log CFU/g (Table 2). A previous study demonstrated a similar result that the SA-SPI emulsion microcapsules extended the shelf-life of Lacticaseibacillus paracasei ProSci-92 at 4 °C compared to free cells (Peng et al., 2025). Another study also reported that cross-linked alginate particles increased the storability of encapsulated probiotics compared to free cells (Yeung et al., 2016).

The ANOVA for storage stability revealed highly robust models for both 1 month (p < 0.001, R ² = 0.882) (Supplementary Table S2) and 2 months (p < 0.001, R ² = 0.990) (Tables 3, 4). Notably, the 2-month stability model exhibited a strong fit to the experimental data, underscoring the effectiveness of the SPI-SA microcapsules in maintaining cell viability over time. By the end of 2 months of the cold storage, the survivability was significantly impacted by both the linear and quadratic terms of SPI and SA (p < 0.01 and p < 0.001, respectively). This is supported by Pearson’s correlation coefficient between moisture content and storability at 2 months, which exhibited a moderate positive correlation of r = 0.58 (Supplementary Figure S4).

At 2 months, the highest concentration of SPI showed a slight decrease in storability, probably due to the osmotic stress and water restriction caused by the dense structure, which requires cells to consume more energy to balance their turgor and the external osmotic osmolarity (Rascón et al., 2018). In general, the wet microcapsules are less advantageous in terms of the shelf-life of probiotics compared to the dried particles due to the continuation of metabolic activities (Bustos et al., 2025; Dianawati et al., 2016). That being said, the viable counts remained in the range of 6.96–7.93 Log CFU/g after 2 months, that is, at or marginally below the threshold of 7 Log CFU/g commonly cited for probiotic products, indicating reasonable long-term stability of the microcapsule system (Ouwehand, 2017).

4 Conclusions

This study successfully developed double emulsion-based probiotic microcapsules using soy protein isolate (SPI) and sodium alginate (SA) containing L. rhamnosus GG for probiotic delivery in plant-based food applications. Using response surface methodology, it was found that the microcapsules’ structural and functional properties could be controlled by the individual concentrations and interactions among the wall materials. SA concentration was the primary determinant of the physical and textural characteristics; higher SA levels led to a substantial increase in median particle size and to a log-linear increase in firmness, attributable to the formation of a calcium–alginate cross-linked network. The combination of SPI and SA was associated with consistently high encapsulation efficiencies and with protection of the probiotic cells during simulated gastric digestion, and the structural integrity provided by both polymers supported cellular viability over 2 months of cold storage.

Taken together, the SPI–SA composite system provides a systematically tunable range of textural and functional properties, and the present system is regarded as a probiotic-loaded structural ingredient for the future development of plant-based cheese analogues. Its evaluation in an actual cheese analogue, including meltability, sliceability and sensory acceptance, together with the incorporation of other bioactive compounds, remains to be undertaken. Future studies will investigate the incorporation of these microcapsules into formulated plant-based food products under realistic processing conditions.

Generative AI statement

The author(s) declared that generative AI was used in the creation of this manuscript. During the preparation of this work, the authors used Claude Opus 5 to improve the language, grammar, and readability of the text. After using it, the authors reviewed and edited the content as needed and take full responsibility for the content of the publication.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/frfst.2026.1972949/full#supplementary-material

Supplementary material Supplementary material at the source

Footnotes

Edited by: Tomás García-Cayuela, Food and Biotech Lab, Mexico

Reviewed by: Tejaswi Boyapati, South Dakota State University, United States

Shery M Varghese, Mar Athanasios College for Advanced Studies Tiruvalla (MACFAST), India