Lactiplantibacillus plantarum SLpl116 attenuates OVA-induced food allergy with ecological restoration of the gut microbiota and immune rebalancing

Abstract

Gut dysbiosis is increasingly recognized as a key contributor to food allergy, yet probiotic strains capable of restoring allergic microbiota and rebalancing host immunity remain limited. Here, we identified Lactiplantibacillus plantarum SLpl116 through a multi-criteria screening pipeline integrating anti-allergic activity, safety, and processing stability, and evaluated its efficacy in a prophylactic ovalbumin (OVA)-induced murine food allergy model. SLpl116 significantly attenuated allergic symptoms, including diarrhea and hypothermia, and suppressed serum IgE, IgG1, OVA-specific immunoglobulins, and mucosal mast cell protease-1. It was also associated with suppression of Th2-related responses and enhancement of systemic Th1-associated signaling, indicating restoration of Th1/Th2 immune balance. Microbiome analysis showed that SLpl116 was associated with ecological restoration of the dysbiotic gut community, including suppression of allergy-associated taxa such as Alistipes finegoldii and Bacteroides and enrichment of beneficial commensals, particularly Lachnospiraceae. Correlation analysis supported an association between microbial reconfiguration and immune rebalancing, while PICRUSt2-based functional prediction suggested enriched butyrate-associated metabolic potential in the effective strain groups. Comparative genome-informed analysis further indicated that SLpl116 possessed distinctive phenotype-linked features, providing a plausible molecular rationale for its favorable phenotype. Together, these findings identify SLpl116 as a promising strain-level probiotic candidate associated with direct immune rebalancing and microbiome-associated ecological restoration.

Introduction

Food allergy is characterized as an adverse immune reaction to ingested food proteins and has emerged as a significant public health concern with a rapidly increasing prevalence worldwide^1,2. The pathophysiology of this condition is primarily driven by a disruption in immune homeostasis, specifically an excessive type 2 helper T (Th2) cell response³. Upon exposure to allergens, activated Th2 cells secrete pro-inflammatory cytokines, including interleukin (IL)-4, IL-5, and IL-13. These cytokines stimulate B cells to undergo isotype switching and produce allergen-specific immunoglobulin E (IgE), which binds to high-affinity receptors on mast cells⁴. Subsequent exposure to the same allergen triggers mast cell degranulation, releasing inflammatory mediators such as histamine, leukotrienes, and prostaglandins⁵. These mediators cause symptoms ranging from mild gastrointestinal disturbances and urticaria to severe, life-threatening anaphylaxis⁶. Consequently, suppressing this Th2-skewed hyper-responsiveness and enhancing type 1 helper T (Th1) immunity to restore the Th1/Th2 balance is considered a key immunological strategy for the prevention and treatment of food allergies⁷.

Currently, there are no definitive standardized curative therapies for individuals with a history of food allergies². The primary management strategy relies heavily on the strict avoidance of causative allergens in daily life, which can be socially isolating and difficult to maintain⁸. Pharmacological interventions, such as antihistamines, are commonly prescribed to alleviate symptoms upon accidental exposure⁹. However, their use is limited due to side effects such as dizziness, fatigue, and drowsiness. While oral immunotherapy has been developed as an allergen-specific desensitization program to induce tolerance, the treatment process is often cumbersome, time-consuming, and carries a risk of adverse reactions¹⁰. Moreover, its effects frequently do not persist without continuous maintenance dosing, meaning it generally fails to induce long-term sustained unresponsiveness. In parallel with the rising prevalence of allergies, emerging evidence suggests that gut dysbiosis, an imbalance in the intestinal microbial community, is closely linked to the failure of immune tolerance¹¹. Allergic individuals often exhibit a reduction in beneficial commensals and an increase in opportunistic pathogens, such as Bacteroides, which are associated with immune-mediated inflammatory conditions¹². These microbial shifts can compromise the intestinal barrier and promote systemic allergic sensitization¹³. Therefore, novel therapeutic approaches that go beyond symptomatic relief to fundamentally restore the gut ecosystem and immune regulatory capacity are urgently needed.

Probiotics, defined as live microorganisms that confer health benefits when consumed in adequate amounts, have recently gained attention for their potential to manage allergic diseases¹⁴. They have been shown to modulate intestinal immune responses, strengthen the mucosal barrier, and restore the Th1/Th2 balance¹⁵. Numerous studies have reported the anti-allergic effects of specific probiotic strains. For instance, supplementation with Lactobacillus rhamnosus GG has been shown to increase the abundance of butyrate-producing bacteria such as Faecalibacterium prausnitzii and Roseburia, which upregulate tight junction proteins and reduce allergen translocation¹⁶. Yet, such studies predominantly rely on well-established commercial strains, often lacking rigorous comparative evaluations to identify the most effective candidates among diverse indigenous isolates¹⁷. Similarly, while Bifidobacterium longum subsp. infantis B6MNI has been observed to alleviate allergic symptoms in murine models by significantly reducing serum IgE and immunoglobulin G1 (IgG1) levels and suppressing Th2-associated cytokines (IL-4, IL-5, and IL-13) in splenocytes¹⁸, these investigations often focus on phenotypic outcomes without fully elucidating the ecological mechanisms, specifically, how the strain selectively suppresses dysbiosis-driving pathobionts. Furthermore, although multi-strain probiotic mixtures have been reported to induce mucosal CD103⁺ dendritic cells, which promote regulatory T cell (T_reg) differentiation and reshape the gut microbiota by increasing beneficial phyla such as Verrucomicrobia¹⁹, strains demonstrating such high bioactivity often lack the rigorous safety and stability verification required for industrial application prior to in vivo assessment²⁰.

To address these cumulative limitations, this study employed a comprehensive multi-criteria screening strategy. Four distinct probiotic candidates meeting rigorous criteria for processing stability and safety were selected to ensure their industrial viability alongside therapeutic potential. Subsequently, to validate their anti-allergic efficacy in vivo, a comparative evaluation was performed using a robust ovalbumin (OVA)-induced murine food allergy model, characterized by systemic sensitization (intraperitoneal injection, i.p.) followed by intragastric challenge (intragastric gavage, i.g.) to mimic severe clinical symptoms. Through a direct head-to-head comparison of these pre-selected strains in this rigorous model, Lactiplantibacillus plantarum SLpl116 was identified as the most promising strain-level candidate within the present comparative framework. Furthermore, beyond phenotypic alleviation, the present study investigated the microbiome- and immune-associated features linked to the efficacy of this strain, particularly its association with reversal of gut dysbiosis and immune rebalancing, to propose a practical framework for probiotic-based prevention of food allergy.

Results

Screening and selection of probiotic candidates with high anti-allergic potential

To identify potential probiotic candidates with anti-allergic properties, a total of 30 bacterial strains were isolated from Korean traditional fermented foods and human feces (Supplementary Table S1). As a primary screening step, the inhibitory effect of these 30 isolates on β-hexosaminidase release, a key marker of mast cell degranulation, was evaluated in IgE-sensitized RBL-2H3 cells. L. plantarum CJLP133, a commercial strain approved by the Ministry of Food and Drug Safety in Korea for immune modulation, was included as a positive control to benchmark the efficacy of the isolates²¹. Based on the assay results, eight strains exhibiting degranulation inhibition rates comparable to or higher than that of CJLP133 (59.23%) were selected for secondary evaluation (Fig. 1A).

Fig. 1. Screening of anti-allergic probiotic strains.

A Primary screening of 30 bacterial isolates assessing their inhibitory effects on β-hexosaminidase release in DNP-IgE/BSA-stimulated RBL-2H3 cells. The degranulation rate of the stimulated positive control (PC) was set to 100%. The eight strains selected for secondary evaluation are indicated by bold borders. B–D Secondary evaluation of eight selected strains on the transcriptional regulation of pro-inflammatory cytokines. Relative mRNA expression levels of (B) IL-4, (C) IL-13, and (D) TNF-α were quantified using qRT-PCR. In panels B–D, strains exhibiting lower mRNA expression levels than the reference strain CJLP133 are indicated by bold borders. Data are presented as mean ± SEM of independent experiments (n = 3-4). Statistical significance is indicated as follows: n.s, not significant; *p < 0.05; **p < 0.01 compared to the CJLP133 reference group (in A) or the PC group (in B–D). PC, DNP-IgE/BSA-stimulated cells.

Subsequently, the anti-allergic efficacy of these eight selected strains was further assessed by measuring their ability to suppress the mRNA expression of pro-inflammatory cytokines (IL-4, IL-13, and TNF-α) in DNP-IgE/BSA-stimulated RBL-2H3 cells (Table 1). While exposure to DNP-IgE/BSA significantly upregulated cytokine expression, treatment with the probiotic strains effectively downregulated these levels. Among the tested strains, Lactobacillus acidophilus SLac002, L. gasseri SLga015, Lacticaseibacillus paracasei SLpa017, and L. plantarum SLpl116 demonstrated the most potent inhibitory effects, showing efficacy comparable to or exceeding that of the control strain CJLP133 (Fig. 1B–D). Consequently, these four distinct species were selected as the final candidates for further characterization.

Table 1.

Primer sequences for qRT-PCR analysis of cytokine mRNA expression

Gene		Sequence (5’ to 3’)		Annealing Temp. (°C)	Reference
RBL-2H3 (Rat)	GAPDH	Fa	TTC ACC ACC ATG GAG AAG GC	58.0	75
		Ra	GGC ATG GAC TGT GGT CAT GA
	IL-4	F	TCC ACG GAT GTA ACG ACA GC	58.0	76
		R	TCA TTC ACG GTG CAG CTT CT
	IL-13	F	TCA TTC ACG GTG CAG CTT CT	58.0	76
		R	AGC GGA AAA GTT GCT TGG AG
	TNF-α	F	ATG GGC TCC CTC TCA TCA GT	58.0	75
		R	GAA ATG GCA AAT CGG CTG AC
BALB/c mouse	β-actin	F	CTA AGG CCA ACC GTG AAA G	58.0	77
		R	ACC AGA GGC ATA CAG GGA CA
	IL-4	F	TGT TCT CAT GGA GCT GCA GA	58.0	78
		R	GGA CTC ATT CAT GGT GCA GC
	IL-10	F	CTT ACT GAC TGG CAT GAG GAT CA	58.0	79
		R	GCA GCT CTA GGA GCA TGT GG
	IL-17	F	TCC CTC TGT GAT CTG GGA AG	58.0	80
		R	AGC ATC TTC TCG ACC CTG AA
	IFN-γ	F	GCC AAG TTT GAG GTC AAC AAC	58.0	81
		R	CCG AAT CAG CAG CGA CTC

^aF forward primer; R reverse primer.

Safety assessment and stability under simulated gastrointestinal and processing conditions

Assessment of safety profiles according to the authoritative guidelines of EFSA²² and FAO/WHO²³ confirmed the suitability of the selected strains (L. acidophilus SLac002, L. gasseri SLga015, L. paracasei SLpa017, and L. plantarum SLpl116) for human consumption. Rigorous verification of safety, particularly the absence of transmissible antibiotic resistance and virulence factors, is a fundamental prerequisite for probiotic candidates to preclude potential health risks such as horizontal gene transfer to intestinal pathogens. Antibiotic susceptibility testing revealed that all four candidates were susceptible to nine clinically relevant antibiotics, with MIC values falling below the established cut-off limits (Supplementary Table S2). Phenotypic evaluations further demonstrated the absence of hemolytic activity (γ-hemolysis) and bile salt hydrolase (BSH) activity, ensuring no risk of red blood cell lysis or toxic secondary metabolite production (Supplementary Figs. S1 and S2). Notably, metabolic analysis indicated that these strains predominantly produced L-lactate rather than D-lactate under the in vitro conditions tested, supporting their phenotypic safety characterization (Supplementary Fig. S3). Additionally, cytotoxicity assays on Caco-2 cells confirmed that exposure to high bacterial concentrations (10⁹ CFU/mL) did not induce significant cell death (Supplementary Fig. S4). Furthermore, whole-genome bioinformatic analysis of the SLac002, SLga015, SLpa017, and SLpl116 strains was performed to evaluate their safety at the genomic level. These analyses, covering general genomic features, virulence factors, antibiotic resistance profiles, and PHASTEST-based prophage profiling, supported the overall safety assessment of all strains as potential probiotics and are detailed in the supplementary data (Fig. S5, Tables S3, and S4). Although prophage regions were predicted in SLpl116, no known virulence factors or acquired/transferable antibiotic resistance genes were identified within these predicted prophage regions. In addition, a comparative genome-informed summary of phenotype-linked features, including predicted surface-associated proteins, bacteriocin/secondary metabolite-related loci, and carbohydrate-active enzymes, is provided in Supplementary Table S5.

Survival through the harsh conditions of the gastrointestinal tract is a fundamental requirement for probiotic efficacy. Evaluation under simulated gastric and bile conditions demonstrated that all candidate strains possessed robust resistance to sequential digestive stress (Fig. 2A). Among the tested isolates, L. plantarum SLpl116 exhibited the highest survival rate of 82.5%, significantly outperforming the other candidates, which suggests the highest capacity among the four strains to reach the intestine in a viable state.

Fig. 2. Probiotic characterization of selected strains regarding gastrointestinal stability and adhesion properties.

A Survival rates of the four selected strains under simulated gastrointestinal conditions (sequential exposure to gastric acid and bile salts). B Adhesion ability of the selected strains to Caco-2 intestinal epithelial cells. Background adhesion to the plastic surface was assessed using blank wells without Caco-2 cells and subtracted from the final adhesion values. C Heat tolerance of the selected strains after incubation at 45 °C for 3 h. Data are presented as mean ± SEM of independent experiments (n = 3-4). Different lowercase letters (a–d) above the bars indicate significant differences between strains (p < 0.05) determined by one-way ANOVA followed by Duncan’s multiple range test.

Beyond survival, the ability to adhere to the intestinal epithelium may contribute to persistence potential and host interaction. Adhesion assays using Caco-2 cells revealed that all four strains exhibited strong attachment capabilities (Fig. 2B). Specifically, L. plantarum SLpl116 demonstrated a marked adhesion rate of 73.9%, which may support enhanced persistence potential and interaction with mucosal immune cells, although in vivo colonization was not directly assessed in the present study.

Finally, thermal stability was assessed to determine the suitability of these strains for industrial processing and storage. Exposure to thermal stress at 45 °C for 3 h resulted in viability retention of over 90% for all strains, indicating resilience against conditions commonly encountered during fermentation or product formulation (Fig. 2C). Collectively, these findings indicate that L. plantarum SLpl116 combines a verified safety profile with favorable gastrointestinal stability and adhesion-related properties, supporting its potential as a promising functional food ingredient. However, direct in vivo persistence or colonization of the administered strains was not assessed in this study.

Comparative evaluation of clinical symptom alleviation in a food allergy mouse model

To evaluate the in vivo efficacy of the selected probiotic strains, the severity of allergic symptoms was assessed in an OVA-induced food allergy murine model. The experimental design and treatment information were shown in Fig. 3A and Table 2. Terminal body weight measured at sacrifice showed some group differences, with significantly lower values in the SLpl116+OVA and SLpa017+OVA groups compared with the OVA group (Fig. 3B). However, this pattern was not uniformly observed across all probiotic-treated groups and, because only endpoint values were available, it was interpreted cautiously. Following the oral challenge with OVA, mice in the OVA group exhibited characteristic signs of anaphylaxis and allergic reactions, validating the successful establishment of the model. Specifically, the OVA group showed a marked increase in the occurrence of diarrhea and higher diarrheal symptom scores compared to the PBS group (Fig. 3C and 3D). However, oral administration of the selected probiotic strains significantly attenuated these symptoms. Overall, all four tested strains showed a tendency toward symptom improvement, whereas SLac002 and SLpl116 showed the most consistent and statistically supported reductions. Among the probiotics-OVA groups, the SLpl116+OVA group exhibited the most significant reduction in both diarrhea incidence and symptom scores compared to the OVA group (p < 0.05).

Fig. 3. Experimental design and alleviation of clinical allergic symptoms by selected probiotic strains in an OVA-induced food allergy murine model.

A After two-week adaptation and quarantine period, probiotics were administered for six weeks, and on day 7 and 21, 50 μg OVA and 1 mg of Alum were injected. samples were collected on day 0 and day 21. For challenge, OVA were gavaged (50 mg/mouse) every other day from day 35 to day 43. Blood, spleen and colon were collected on day 44. OVA, ovalbumin; Alum, aluminum hydroxide gel. B Terminal body weight measured at sacrifice on day 43. C Diarrhea occurrence monitored during the OVA challenge period. D Blinded diarrheal symptom scores evaluated after the final OVA challenge (Day 43). E Changes in rectal temperature, calculated as the difference between measurements taken at 0 min and 60 min to assess anaphylactic shock severity. Data are presented as mean ± SEM (n = 8 mice per group). Asterisks indicate statistical significance compared with the OVA control group by one-way ANOVA followed by Dunnett’s multiple comparison test (*p < 0.05; **p < 0.01).

Table 2.

Description of experimental groups

Group	Description	Oral administrationa	Sensitization (i.p.)b	Challenge (i.g.)c
		Day 1 ~ 43	Day 7, Day 21	Day 35-43
PBS	Normal control	PBS	PBS	PBS
OVAd	OVA-induced Food allergy control	PBS	OVA (50 μg) + Alum (1 mg)	OVA (50 mg)
SLac002 + OVA	Probiotic-OVA groups	Lactobacillus acidophilus	OVA (50 μg) + Alum (1 mg)	OVA (50 mg)
SLga015 + OVA	Probiotic-OVA groups	Lactobacillus gasseri	OVA (50 μg) + Alum (1 mg)	OVA (50 mg)
SLpa017 + OVA	Probiotic-OVA groups	Lacticaseibacillus paracasei	OVA (50 μg) + Alum (1 mg)	OVA (50 mg)
SLpl116 + OVA	Probiotic-OVA groups	Lactiplantibacillus plantarum	OVA (50 μg) + Alum (1 mg)	OVA (50 mg)

^aProbiotics were administered at a dose of 10⁹ CFU/mouse/day for 43 days.

^bIntraperitoneal injection, 0.2 mL of saline solution containing 50 μg of OVA and 1 mg of Alum (aluminum hydroxide gel).

^cIntragastric gavage, every other day with 0.2 mL of saline solution containing 50 mg of OVA from day 35 to 43.

^dOvalbumin.

Rectal temperature changes, calculated as the difference between measurements taken at 0 and 60 min post-challenge, were assessed as a key indicator of systemic anaphylaxis. Following the final OVA challenge, the OVA group exhibited a severe drop in body temperature, recording a mean difference of approximately 3.38 °C, which signifies acute anaphylactic shock. In contrast, oral administration of the selected probiotic strains effectively attenuated this hypothermic response, limiting the temperature reduction to a range of 0.86 °C to 1.76 °C (Fig. 3E). Notably, among the probiotics-OVA groups, L. plantarum SLpl116 demonstrated the greatest protective effect against OVA-induced hypothermia. Although all four tested strains showed a tendency to attenuate allergic symptoms, SLac002 and SLpl116 exhibited the most consistent and statistically supported improvements, with SLpl116 significantly alleviating, but not completely normalizing, both diarrhea and hypothermia in this food allergy model.

Suppression of systemic allergic responses and mast cell activation

To determine whether the alleviation of clinical symptoms by probiotic administration was accompanied by the suppression of systemic allergic responses, serum levels of total and allergen-specific immunoglobulins were analyzed. The production of IgE and IgG1, which are hallmarks of a Th2-mediated allergic response, was significantly elevated in the OVA group compared to the PBS group. However, oral administration of the selected probiotic strains effectively downregulated these levels. In particular, the SLpl116+OVA group exhibited a significant reduction in both total IgE and IgG1, as well as OVA-specific IgE and OVA-specific IgG1 levels (p < 0.05), indicating a potent suppression of systemic sensitization and humoral immune responses (Fig. 4A–D).

Fig. 4. Effect of selected probiotic strains on serum antibody levels and mast cell activation markers in OVA-sensitized mice.

Serum samples were collected on Day 43 after the final OVA challenge to quantify systemic allergic responses. A Total serum IgE, (B) IgG1, (C) OVA-specific IgE, (D) OVA-specific IgG1 levels, and (E) Serum mucosal mast cell protease-1 (MCPT-1) levels, a marker of systemic mast cell degranulation. Data are presented as mean ± SEM (n = 8 mice per group). Asterisks indicate statistical significance compared with the OVA control group by one-way ANOVA followed by Dunnett’s multiple comparison test (*p < 0.05; **p < 0.01). PBS, naïve control; OVA, OVA-sensitized/challenged control.

Crucially, to further investigate the impact of L. plantarum SLpl116 on the immediate hypersensitivity reaction in the gut mucosa, the MCPT-1 was measured from the serum. MCPT-1 is a specific marker released into the systemic circulation upon the degranulation of mucosal mast cells, directly reflecting the severity of intestinal allergic inflammation. Following the OVA challenge, the serum MCPT-1 level in the OVA group was dramatically increased, confirming intense mucosal mast cell activation. Notably, treatment with L. plantarum SLpl116 significantly attenuated this increase, suppressing MCPT-1 release more effectively than other strains (Fig. 4E). This marked reduction in MCPT-1, together with the observed decrease in diarrhea scores, supports the interpretation that L. plantarum SLpl116 attenuates food allergy symptoms in association with suppression of mucosal mast cell-related allergic responses in the intestinal tissue.

Modulation of Th1/Th2 immune balance in lymphoid and intestinal tissues

To elucidate the immunomodulatory mechanism underlying the alleviation of allergic symptoms, the profiles of Th1, Th2, and Th17 cytokines were analyzed in both systemic lymphoid organs (spleen) and local intestinal tissues (colon). Food allergy is typically characterized by a skewed Th2-dominant immune response²⁴. Consistent with this, the OVA group exhibited a significant elevation in Th2-associated cytokines (IL-4) and the Th17 cytokine (IL-17), alongside a reduction in the Th1 cytokine (IFN-γ), compared to the PBS group. In splenocytes, oral administration of the probiotic strains differentially modulated this cytokine imbalance. All probiotics-OVA groups showed a significant suppression of IL-4 and IL-17 secretion compared to the OVA group (p < 0.01) (Fig. 5A-5C). However, regarding Th1 immunity, L. plantarum SLpl116 was the only strain among the four candidate strains to significantly upregulate IFN-γ production (p < 0.05), whereas other strains did not show a significant increase (Fig. 5D). Consequently, L. plantarum SLpl116 exhibited the highest restoration of the Th1/Th2 balance among the strains, as evidenced by the highest IFN-γ/IL-4 ratio among the probiotics-OVA groups (Fig. 5E).

Fig. 5. Modulation of Th1/Th2 immune balance in splenocytes and colon tissues by selected probiotic strains.

A–E Cytokine production in splenocytes. Splenocytes were isolated on day 36 and stimulated with OVA (100 μg/mL) for 72 h. Levels of (A) IL-4, (B) IL-10, (C) IL-17, and (D) IFN-γ were quantified by ELISA. E The IFN-γ/IL-4 ratio in splenocytes. F–H Cytokine levels in colon tissue homogenates. F IL-4 and G IFN-γ levels were normalized to total protein. H The IFN-γ/IL-4 ratio in colon tissues. Data are presented as mean ± SEM. Asterisks indicate statistical significance compared with the OVA control group by one-way ANOVA followed by Dunnett’s multiple comparison test (*p < 0.05; **p < 0.01). PBS, naïve control; OVA, OVA-sensitized/challenged control.

In the colon tissues, which represent the local site of allergic inflammation, a similar immunomodulatory trend was observed. All probiotic strains effectively downregulated the local expression of IL-4 (p < 0.01) (Fig. 5F). Although no significant differences were observed in local IFN-γ levels among the probiotics-OVA groups (Fig. 5G), the suppression of IL-4 was sufficient to significantly recover the IFN-γ/IL-4 ratio in the colon (Fig. 5H). Collectively, these findings suggest that while all selected strains possess anti-allergic potential by suppressing Th2 responses, L. plantarum SLpl116 showed the strongest systemic Th1-associated immune shift among the four candidate strains, thereby providing a more comprehensive restoration of the Th1/Th2 immune balance.

Restoration of gut microbiota composition and reversal of dysbiosis

16S rRNA gene sequencing of fecal samples was conducted to investigate the mechanistic link between the alleviation of allergic symptoms and alterations in the gut microbiota. PCoA revealed that food allergy induction caused a significant shift in the overall gut microbiota structure, separating the OVA group from the PBS group (Fig. 6A). The gut microbiota of the groups administered with L. plantarum SLpl116 and L. acidophilus SLac002 clustered closely with that of the PBS group, indicating a structural recovery to a healthy state. In contrast, the SLpa017+OVA group remained clustered near the OVA group, suggesting a limited capacity to restore microbial balance (Fig. 6A). This strain-dependent restorative effect was further confirmed at the phylum level (Fig. 6B). The OVA group exhibited a severe dysbiosis characterized by a reduced Firmicutes and Bacteroidetes ratio (27.0:69.7). Both SLpl116+OVA (47.2:49.0) and SLac002+OVA (50.5:45.5) groups effectively restored this ratio to levels comparable to the PBS group (46.8:49.3), whereas SLpa017+OVA group failed to reverse the imbalance (21.5:76.0). Consistent with the phylum-level alterations, the taxonomic composition at the family and genus levels further corroborated a strain-dependent restoration of the gut microbiota (Supplementary Fig. S6). The microbial structure of the SLpa017+OVA group retained a dysbiotic pattern highly similar to that of the OVA group. In contrast, the SLpl116+OVA and SLac002+OVA groups exhibited a distinct shift toward a healthy equilibrium, resembling the composition of the PBS group. These findings suggest that the therapeutic efficacy of the effective strains is linked to their ability to promote ecological restoration of the gut ecosystem across multiple taxonomic levels.

Fig. 6. Modulation of gut microbiota composition by probiotic strains.

A PCoA plot based on 16S rRNA gene sequencing data, showing the β-diversity and structural separation of gut microbiota among groups. B Taxonomic composition at the phylum level. Relative abundance of key bacterial taxa identified as biomarkers by LEfSe analysis: the genus Bacteroides (C), A. finegoldii (D), the families Lachnospiraceae (E) and Ruminococcaceae (F). Data are presented as mean ± SEM. Different letters indicate significant differences at p < 0.05. G Co-occurrence network analysis of gut microbiota at the family level. Each node represents a bacterial family, and the node size reflects its relative abundance. Nodes are colored according to their modularity class, indicating clusters of taxa with similar co-occurrence patterns. Edges represent significant Spearman’s correlations ( | r | > 0.6, p < 0.05). Blue edges denote positive correlations (co-occurrence), while red edges denote negative correlations (mutual exclusion). The thickness of each edge corresponds to the strength of the correlation coefficient (r).

To pinpoint specific bacterial taxa driving the allergic dysbiosis, Linear Discriminant Analysis Effect Size (LEfSe) was performed to identify key biomarkers that significantly differed between the PBS and OVA groups (Supplementary Fig. S7). Based on this selection, the modulation of these representative taxa was further analyzed (Fig. 6C–F). The genus Bacteroides, which exhibited a dysbiosis-associated expansion to 36.1% in the OVA group, was most potently suppressed by L. acidophilus SLac002 (6.7%), followed by L. plantarum SLpl116 (14.9%), while L. paracasei SLpa017 (36.1%) showed no significant suppressive effect (Fig. 6C). Notably, the species-level analysis highlighted Alistipes finegoldii as an allergy-associated expanded taxon in the OVA group²⁵. While the relative abundance of A. finegoldii markedly increased in the OVA group, treatment with L. plantarum SLpl116 and L. acidophilus SLac002 significantly reduced this species to levels comparable to the PBS group (Fig. 6D), indicating that the effective strains suppressed a dysbiosis-associated taxon enriched under allergic conditions.

Conversely, beneficial taxa associated with the maintenance of intestinal homeostasis showed distinct enrichment patterns. The family Lachnospiraceae was most prominently enriched in the SLpl116+OVA group (33.4%), followed by SLac002+OVA group (29.2%), compared to the OVA group (15.2%) (Fig. 6E), and Ruminococcaceae was also lowered in the OVA group (6.2%) but was effectively restored by L. acidophilus SLac002 (12.2%) to levels comparable to the PBS group (12.7%, Fig. 6F). These findings highlight that the mice in both SLpl116+OVA and SLac002+OVA groups effectively reverse allergy-induced dysbiosis. However, their protective mechanisms involve a dual action: suppression of allergy-associated dysbiosis taxa, including Bacteroides and A. finegoldii, and enrichment of beneficial commensal families, particularly Lachnospiraceae in the L. plantarum SLpl116 group.

To further elucidate ecological interactions within the gut microbiota, a family-level co-occurrence network analysis was performed (Fig. 6G). This analysis revealed a distinct module centered on Lachnospiraceae, which showed strong positive correlations (thick blue edges) with Oscillospiraceae, Butyricicoccaceae, and Ruminococcaceae. These taxa formed a coherent cluster, indicating synchronized abundance patterns among families associated with the restored microbial community. In contrast, this module displayed clear negative correlations (red edges) with Bacteroidaceae, reflecting mutual exclusion between taxa associated with contrasting ecological states. Collectively, these findings indicate that the probiotic effect was not restricted to isolated compositional changes, but was accompanied by coordinated guild-level reorganization of the gut microbiota. This co-occurrence structure provides a network-based basis for ecological restoration, suggesting a shift from a dysbiosis-associated community configuration toward a more homeostatic microbial architecture.

Correlation between gut microbiota composition and allergic parameters

Spearman’s correlation analysis was conducted to assess statistical associations between gut microbiota composition and host immune responses. The resulting heatmap visualization revealed distinct correlation patterns corresponding to the amelioration of food allergy symptoms (Fig. 7). In particular, the positive correlations of A. finegoldii with diarrhea score, Th2-associated cytokines, and serum IgE-related parameters support its interpretation as a context-specific candidate pro-allergic taxon in this OVA-induced food allergy model. Specifically, both Bacteroides and A. finegoldii exhibited significant positive correlation with the Diarrhea Score (p < 0.001) and splenic pro-inflammatory cytokines (p < 0.001). Furthermore, robust positive correlations were observed between these dysbiosis-associated taxa and systemic sensitization markers, including serum IgE and OVA-specific IgE (p < 0.01). These statistical associations indicate that the dysbiosis-associated expansion of Bacteroides is closely linked to the induction of Th2-skewed inflammation and clinical symptom severity.

Fig. 7. Correlation analysis between gut microbiota and allergic parameters.

Heatmap showing Spearman’s correlation coefficients between the relative abundance of key bacterial taxa (based on LEfSe data) and clinical or immunological parameters. Red squares indicate positive correlations, while blue squares indicate negative correlations. The intensity of the color represents the strength of the correlation (r). Statistical significance is indicated by asterisks (*p < 0.05; **p < 0.01; ***p < 0.001).

In parallel, the bacterial taxa that were significantly enriched following probiotic administration displayed distinct protective profiles. Notably, the family Lachnospiraceae, which was uniquely boosted by L. plantarum SLpl116, and the family Ruminococcaceae, which was effectively restored by L. acidophilus SLac002, showed significant negative correlations with serum IgE and MCPT-1 levels. This negative correlation suggests a potential role for these taxa in suppressing mast cell degranulation and systemic sensitization. Most importantly, a marked positive correlation was identified between these beneficial phylotypes and the splenic IFN-γ/IL-4 ratio (p < 0.001), a key indicator of immune homeostasis. Collectively, these findings support an association between strain-specific bacterial modulation, particularly the enrichment of Lachnospiraceae by L. plantarum SLpl116 and Ruminococcaceae by L. acidophilus SLac002, and restoration of the Th1/Th2 balance accompanied by alleviation of food allergy.

Discussion

The rising prevalence of food allergies necessitates therapeutic strategies that go beyond symptomatic relief to address the underlying immune dysregulation^2,26. In this study, the selected probiotic strains showed overall anti-allergic tendencies, and L. plantarum SLpl116 together with L. acidophilus SLac002 exhibited the most consistent protection against OVA-induced allergic symptoms. Among these candidates, SLpl116 was associated with restoration of the Th1/Th2 immune balance together with marked restructuring of the gut microbiota. Rather than reflecting a single mode of action, these findings suggest that the superior phenotype of SLpl116 among the four candidates is best interpreted through multiple complementary perspectives, including direct immune rebalancing, strain-specific ecological restoration of the gut microbiota, and genome-informed features potentially linked to host interaction and ecological fitness. The following discussion therefore considers how these immunological, ecological, and strain-specific properties may collectively explain the anti-allergic phenotype observed in the present model.

In terms of direct immunomodulation, the primary immunological effect underlying the clinical efficacy of L. plantarum SLpl116 was the suppression of allergen-specific Th2 responses together with enhancement of systemic Th1-associated signaling, resulting in restoration of the Th1/Th2 balance. The attenuation of anaphylactic symptoms, including diarrhea and hypothermia, was closely mirrored by a significant reduction in serum MCPT-1, a direct marker of mucosal mast cell degranulation²⁷. Together, these findings suggest that L. plantarum SLpl116 attenuated the terminal effector phase of the allergic response. Upstream of this, the strain profoundly downregulated allergen-specific IgE and IgG1 production (Fig. 4), which are critical for mast cell sensitization⁵. Crucially, our comparative cytokine profiling in both systemic (spleen) and local (colon) tissues revealed a consistent restoration of immune homeostasis. The suppression of type 2 immunity (IL-4 and IL-17) was accompanied by a distinct upregulation of IFN-γ, a Th1 cytokine known to cross-regulate and inhibit Th2 differentiation (Fig. 5)²⁸. Notably, this Th1-associated shift was observed not only systemically but also within the intestinal mucosa, distinguishing L. plantarum SLpl116 from other candidate strains that primarily suppressed Th2 cytokines. This pattern supports the interpretation that enhancement of Th1-associated immune rebalancing may be an important contributor to the superior ability of SLpl116 to control local hypersensitivity in the present model.

In addition to this direct immune rebalancing, our data also support a second, microbiome-associated layer of immunomodulation, in which ecological restoration of the gut microbiota was accompanied by immune homeostasis. This microbiome-associated arm should be interpreted as a model-supported and literature-informed framework, rather than as direct evidence of causality between specific microbial changes and immune regulation in this study. Within this framework, a key implication of the present study is that probiotic efficacy was strongly strain-dependent at the ecological level. The effective strains, particularly L. plantarum SLpl116 and L. acidophilus SLac002, were associated with restoration of the dysbiotic microbiota toward a more homeostatic configuration. These findings suggest that successful in vivo probiotic activity may depend not only on intrinsic immunomodulatory properties, but also on the ecological competence of a strain to reorganize the resident microbial community.

This strain-dependent ecological divergence was most clearly illustrated by L. paracasei SLpa017, which served as a comparatively less effective in vivo strain rather than a merely ineffective strain. Although SLpa017 displayed favorable anti-inflammatory activity during in vitro screening, its in vivo effects were weaker and less consistent than those of L. plantarum SLpl116, rather than being entirely absent. In particular, the SLpa017-treated group remained clustered near the OVA group in the PCoA and failed to restore the Firmicutes/Bacteroidetes ratio, indicating that in vitro bioactivity alone was insufficient to predict in vivo efficacy. This divergence highlights an important limitation of conventional probiotic screening approaches, which often emphasize intrinsic immunomodulatory activity without adequately considering the ecological competence of a strain within the resident gut microbiota. Previous studies have frequently focused on the direct effects of probiotic cell components or secreted metabolites on host immune responses^29,30. However, the present results suggest that successful in vivo anti-allergic efficacy may require more than intrinsic anti-inflammatory potential and may depend on the ability of a strain to reorganize the resident microbial community toward a more homeostatic configuration³¹. Because the present study did not directly measure strain persistence or colonization in vivo, the mechanistic basis for the lower in vivo performance of SLpa017 cannot be conclusively determined. Nevertheless, its more limited microbiota restoration, together with the absence of a distinct Th1-promoting effect comparable to that of SLpl116, supports the interpretation that broader immune rebalancing and ecological remodeling are key determinants of in vivo anti-allergic efficacy.

In this context, comparative genome-informed analysis provided an additional plausible molecular rationale for the superior phenotype of SLpl116 (Supplementary Table S5). Because the four candidate strains belong to different species, these genomic comparisons were interpreted conservatively and were not used to assign the phenotype of SLpl116 to any single causal gene. Nevertheless, SLpl116 showed a distinctive combination of phenotype-relevant features, including an expanded repertoire of predicted surface-associated proteins, a plantaricin-related locus together with broader secondary metabolite-associated regions, and the broadest high-confidence CAZyme profile among the four strains. Collectively, these genome-informed features are consistent with enhanced host interaction, persistence potential, and metabolic flexibility, and may therefore partly explain the superior gastrointestinal survival, adhesion, and microbiota-remodeling properties of SLpl116. However, these observations should be regarded as a supportive molecular rationale rather than direct proof of the responsible mechanism.

Beyond these phenotype-linked genomic features, the prophage prediction results also provided an additional safety-relevant perspective. Recent large-scale analyses have shown that prophages are highly prevalent among human gut-derived bacteria and probiotic-associated lactobacilli, including L. plantarum, indicating that prophage carriage itself is not unusual for this species^32,33. Nevertheless, prophages may contribute to bacterial adaptation and can also serve as potential vehicles for mobile genetic elements, including antibiotic resistance or virulence-related genes. Therefore, in the present study, the predicted prophage regions in SLpl116 were interpreted together with genome-based screening for resistance- and virulence-associated features. Although prophage regions were detected in SLpl116, no known virulence factors or acquired/transferable antibiotic resistance genes were identified within these regions, supporting the view that prophage detection did not alter the overall genomic safety interpretation of this strain. However, experimental evaluation of prophage inducibility or transferability would be required for a more definitive assessment.

Consistent with this strain-dependent ecological restoration, species-level analysis further identified A. finegoldii as a prominent taxon associated with allergic dysbiosis in the present model. Importantly, however, the genus Alistipes is functionally heterogeneous and has been reported to exert both detrimental and beneficial effects depending on the host and disease context³⁴. Recent work has even suggested a beneficial association of A. finegoldii in the setting of cancer immunotherapy³⁵. Therefore, A. finegoldii should not be regarded as a universally pathogenic taxon. Rather, in the present OVA-induced food allergy model, the marked expansion of A. finegoldii occurred in parallel with the increase in Bacteroides, and both taxa showed strong positive correlations with IgE and IL-4. Because A. finegoldii and Bacteroides are Gram-negative bacteria that carry lipopolysaccharides (LPS), their expansion may be consistent with a pro-inflammatory microbial configuration relevant to allergic inflammation^12,36. This interpretation is further supported by previous reports linking Alistipes species, including A. finegoldii, to inflammatory or pathobiont-like behavior in other disease settings, such as colitis-associated tumorigenesis, obesity-associated colorectal carcinogenesis, and bacteremia in patients with gastrointestinal pathology^37–39. Taken together, these findings support the interpretation of A. finegoldii as a context-specific allergy-associated dysbiosis taxon in the present model, rather than as a uniformly pathogenic member of the genus. Accordingly, the ability of L. plantarum SLpl116 and L. acidophilus SLac002 to suppress these allergy-associated dysbiosis taxa may partly explain their protective effects in this setting.

Beyond suppression of dysbiosis-associated taxa, the findings also support a process of ecological restoration, in which the administered probiotics reshape the gut ecosystem toward a more homeostatic configuration by promoting endogenous beneficial bacteria. A striking observation in this study was the robust enrichment of the family Lachnospiraceae (33.4%) in the SLpl116+OVA group and the restoration of Ruminococcaceae (12.2%) in the SLac002+OVA group. Reinforcing this ecological restoration, our network analysis revealed that these taxa do not exist in isolation but form a distinct co-occurrence module with Oscillospiraceae and Butyricicoccaceae (Fig. 6G). This clustering indicates a coordinated recovery of a specific bacterial guild rather than independent changes in single taxa. Importantly, the positive association among Lachnospiraceae, Ruminococcaceae, Oscillospiraceae, and Butyricicoccaceae, together with their opposing relationship with Bacteroidaceae, suggests a transition from a dysbiosis-associated community configuration toward a more homeostatic microbial architecture. Thus, the network analysis provides a structural basis for ecological restoration by showing that beneficial taxa recovered as an interconnected consortium rather than as isolated compositional shifts. However, because this analysis was based on 16S-derived co-occurrence patterns, it should be interpreted as supportive evidence for community architecture rather than as direct proof of microbial interactions or perturbation dynamics.

The immunological relevance of this restored consortium may extend beyond general diversity. Prior studies have shown that members of Clostridia clusters XIVa and IV, including many taxa within Lachnospiraceae- and Ruminococcaceae-related Firmicutes, are associated with colonic Foxp3⁺ regulatory T-cell accumulation^40,41. In addition, microbiota-derived short-chain fatty acids (SCFAs), particularly butyrate, have been reported to promote T_reg differentiation and homeostasis through epigenetic and receptor-mediated pathways⁴². In the present study, this pathway was not directly validated at the cellular or metabolic level; therefore, we interpret it as a literature-supported mechanism that may contribute to immune homeostasis downstream of microbiota restoration, rather than as a directly demonstrated effect of SLpl116 treatment. Consistent with this literature-based interpretation, PICRUSt2-based functional prediction suggested increased abundance of several butyrate-associated KEGG Orthologs in the effective strain groups, particularly SLpl116, compared with the OVA group (Supplementary Fig. S8). In particular, K00634 and K00929, which are consistent with terminal butyrate biosynthetic steps, showed higher predicted abundance in SLpl116 and SLac002 than in the OVA group, supporting the possibility of enhanced butyrate-associated metabolic potential. Concurrently, the enrichment of Oscillospiraceae is significant given its role in maintaining the mucus barrier and its recognized depletion in inflammatory conditions, serving as a marker of gut homeostasis^43,44. Furthermore, Butyricicoccaceae has been accompanied by butyrate production near the mucosa and may contribute to epithelial integrity and local immune homeostasis^45,46. Although direct metabolic quantification was not performed, the synchronous expansion of this beneficial consortium, together with the predicted enrichment of butyrate-associated functional pathways, provides supportive evidence for a microbiota-associated mechanism that may contribute to the observed immunological improvements. These findings suggest that the anti-allergic efficacy of L. plantarum SLpl116 is associated not only with the administered strain itself but also with its ability to restore an endogenous bacterial guild linked to immune homeostasis, potentially contributing to a more tolerogenic intestinal environment.

Crucially, the observed restoration of immune homeostasis was not confined to the local colonic environment but extended systemically to secondary lymphoid organs, specifically the spleen. The correlation analysis provides statistical support for an association between microbial reconfiguration and restoration of the Th1/Th2 balance. The strong inverse correlations between enriched Lachnospiraceae and allergic markers (serum IgE and MCPT-1), together with their positive correlation with the IFN-γ/IL-4 ratio, suggest that ecological restoration of the gut microbiota may contribute to the observed immunological improvements (Fig. 7). More specifically, the simultaneous reduction of allergy-associated dysbiosis taxa, including Bacteroides and A. finegoldii, alongside enrichment of beneficial commensals such as Lachnospiraceae, supports a model in which probiotic treatment shifts the gut ecosystem toward a less pro-inflammatory and more homeostatic configuration^12,36,47. Although these relationships remain correlative, they support an integrated microbiome–immune framework in which ecological restoration is associated with systemic immune rebalancing in the present model (Fig. 8).

Fig. 8. Schematic diagram illustrating the proposed prophylactic mechanism of L. plantarum SLpl116 in alleviating food allergy (adapted from Kanagaratham et al.⁸²).

The protective action of L. plantarum SLpl116 is driven by a dual mechanism: Direct Immunomodulation (A) and Microbiome-associated Immunomodulation (B). The strain was associated with enhancement of Th1-associated signaling (I) and, in parallel, with ecological restoration of the gut microbiota, including enrichment of Lachnospiraceae-related commensals linked in the literature to regulatory immune pathways and putative SCFA-mediated immune homeostasis (II). Together, these complementary changes were associated with a suppressed Th2 response (III), reduced allergen-specific humoral immunity including serum IgE and IgG1 (IV), and reduced mast cell degranulation upon allergen challenge (V). Collectively, these coordinated effects were associated with restoration of the splenic Th1/Th2 balance (VI) and attenuation of allergic symptoms.

However, this study has several limitations that should be considered. Because fecal microbiota profiling was performed only at a single terminal time point, the present data are more appropriately interpreted as ecological restoration rather than ecological resilience. In addition, the study was conducted in a single experimental setting, an OVA-induced food allergy model using female BALB/c mice with intraperitoneal sensitization, a prophylactic administration schedule, and a single dosing regimen, which limits assessment of the optimal dose, timing, therapeutic window, and broader generalizability to other models, host backgrounds, sexes, and humans. Body weight was also available only as a terminal endpoint measurement, and, therefore, its biological significance could not be fully resolved. Mechanistically, although the data support an association between microbiota restoration and immune rebalancing, direct validation of Foxp3⁺ T_reg induction and other immune-cell changes was not performed, and causality between microbiota reconfiguration and immune protection was not formally tested. Likewise, the relative contributions of direct strain-specific immunomodulation, microbiota-mediated effects, and in vivo persistence or colonization of the administered strains remain unresolved. Functional interpretation was based on 16S profiling and predicted butyrate-associated pathways, without direct quantification of luminal or systemic metabolites such as SCFAs and LPS-related signals. Therefore, future studies using higher-resolution metagenomics, targeted metabolomics, immune-cell profiling, microbiota-transfer or depletion approaches, and direct comparison with well-characterized reference probiotic strains will be needed to better define the mechanism, translational value, and strain specificity of SLpl116. Furthermore, the present findings should not be generalized to all L. plantarum strains or to probiotics more broadly.

In conclusion, despite these limitations, the present study shows that, in an OVA-induced food allergy model using female BALB/c mice, the superior efficacy of L. plantarum SLpl116 was associated with suppression of dysbiosis-associated taxa together with enrichment of endogenous protective commensals, accompanied by restoration of immune homeostasis. These findings support a model in which the anti-allergic effect of SLpl116 can be interpreted through complementary layers of direct immune rebalancing and microbiome-associated ecological restoration. Although the microbiome-associated component remains inferential, SLpl116 emerged as the most promising strain-level candidate within the present experimental setting based on its integrated clinical, immunological, ecological, safety/stability, and genome-informed phenotype. Broader validation in additional models, host backgrounds, and sexes will be required to determine the general applicability of these findings.

Methods

Isolation and identification of probiotic strains

Probiotic strains were isolated from various traditional Korean fermented foods, including Kimchi, Makgeolli, and cheese and healthy human feces. Food samples (10 g or 10 mL) were homogenized in 90 mL of 0.1% sterile buffered peptone water using a stomacher (Interscience, France). The homogenates were inoculated into de Man, Rogosa, and Sharpe (MRS) broth (Difco, USA) supplemented with 0.05% (w/v) L-cysteine hydrochloride (Sigma-Aldrich, USA) and incubated anaerobically at 37 °C for 48 h for enrichment. Fecal samples were serially diluted in PBS (pH 7.4; GenDEPOT, USA) and spread onto selective agar plates, including MRS agar, Bifidobacterium iodoacetate medium-25 (BIM-25)⁴⁸, and M17 agar (BD, USA), followed by anaerobic incubation at 37 °C for 48 h. The samples were collected from healthy human individuals. This study was approved by Seoul National University Hospital Institutional Review Board (IRB no. 2206-001-014). Written informed consent was obtained from all participants, according to the institutional guidelines.

Colonies exhibiting distinct morphological characteristics were picked and purified. For genotypic identification, genomic DNA was extracted, and the 16S rRNA gene was amplified using universal primers 27 F and 1492R⁴⁹. The amplicons were sequenced (Bionics, Korea) and analyzed using the NCBI BLASTN database⁵⁰. The isolated and identified strains were stored in 20% glycerol at -80 °C for further screening.

RBL-2H3 cell-based β-hexosaminidase release assay

To select probiotic candidates with anti-allergic potential, the ability of isolated strains to inhibit mast cell degranulation was evaluated using rat basophilic leukemia cells (RBL-2H3; ATCC CRL-2256). RBL-2H3 cells were obtained from the Korean Cell Line Bank (KCLB, Korea) and cultured in Minimum Essential Medium (MEM; Welgene, Korea) supplemented with 10% fetal bovine serum (FBS; GW Vitex, Korea) and 100 μg/mL penicillin-streptomycin (Welgene). Cells were maintained at 37 °C in a humidified atmosphere containing 5% CO₂.

For the degranulation assay, RBL-2H3 cells were seeded in 24-well plates at 2 × 10⁵ cells/well and sensitized overnight with 50 ng/mL anti-2,4,6-dinitrophenyl immunoglobulin E (anti-DNP IgE; Sigma-Aldrich). After washing with PBS, the cells were incubated with or without bacterial strains at a multiplicity of infection (MOI) of 100 for 3 h. Degranulation was then induced by stimulation with 50 ng/mL DNP-BSA (Sigma-Aldrich) for 1 h.

β-Hexosaminidase release was used as an indicator of RBL-2H3 cell degranulation, following a previously reported protocol with slight modifications⁵¹. Briefly, culture supernatants were collected and incubated with 1.3 mg/mL p-nitrophenyl-N-acetyl-β-D-glucopyranoside (Sigma-Aldrich) at 37 °C for 4 h. The reaction was terminated by adding 0.2 M glycine, and absorbance was measured at 405 nm. Total β-hexosaminidase content was determined by complete cell lysis using 0.1% Triton X-100. The inhibition rate of β-hexosaminidase release was calculated using the following equation:

$$\mathrm{Inhibition\; rate}\left(\%\right)=\left[1-\frac{T-C}{A-C}\right]\times 100$$ 1

where C represents β-hexosaminidase release under the non-stimulated condition [DNP-BSA( − )], T represents β-hexosaminidase release from stimulated cells treated with bacterial strains [DNP-BSA(+) + bacterial treatment], and A represents the total β-hexosaminidase content obtained after complete lysis with 0.1% Triton X-100. Based on the inhibition rates, the top eight strains showing the highest anti-degranulation activity were selected for secondary screening.

Analysis of cytokine gene expression in RBL-2H3 cells

The anti-inflammatory potential of the eight selected strains was further evaluated by measuring the mRNA expression levels of pro-inflammatory cytokines in DNP-IgE/BSA-stimulated RBL-2H3 cells. Cells were sensitized with anti-DNP IgE and treated with bacterial strains as described above, followed by stimulation with DNP-BSA. After stimulation, total RNA was extracted from RBL-2H3 cells using TRIzol reagent and the PureLink RNA Mini Kit (Invitrogen) according to the manufacturer’s instructions. RNA concentration and purity were determined using a NanoDrop 2000 spectrophotometer (Thermo Scientific), and cDNA was synthesized using the PrimeScript RT Reagent Kit (TaKaRa, Japan).

Quantitative real-time PCR (qRT-PCR) was performed using SYBR Green Supermix (Bio-Rad, USA) to quantify the expression levels of IL-4, IL-13, and tumor necrosis factor alpha (TNF-α). The PCR conditions were as follows: initial denaturation at 95 °C for 3 min, followed by 40 cycles of 95 °C for 5 s and 58 °C for 30 s. GAPDH was used as the reference gene for normalization, and relative mRNA expression was calculated using the 2^-ΔΔCT method⁵². The primer sequences used for qRT-PCR are listed in Table 1.

Antibiotic susceptibility test

For antibiotic resistance test of selected four strains, experiment was conducted in accordance with the safety assessment guidelines for probiotics provided by MFDS and EFSA²². Minimum Inhibitory Concentration (MIC) tests of nine antibiotics (ampicillin, vancomycin, gentamycin, kanamycin, streptomycin, erythromycin, clindamycin, tetracycline, and chloramphenicol) were performed as described previously⁵³ with slight modification: MIC was determined through E-TEST method. Bacterial strains were cultured in MRSc medium until 600 nm optical density absorbance at 1.0. Then, the concentration of 1.5 × 108 CFU/ml of bacterial cell suspensions were mixed with 30 ml of 0.6% LSM (LAB susceptibility test medium; A mixed formulation of Iso-Sensitest medium (90%; MBcell) and MRS medium (10%; Difco) with or without supplementation with L-cysteine·HCl (Sigma)) agar, and pour onto base agar plates. E-test strips (Biomerieux, France) were placed at the center of each agar plate, and the plates were then incubated anaerobically at 37 °C for 48 h. The lowest portion of the strip where the microorganism did not grow was confirmed as the MIC. In accordance with the cut-off values specified by European Food Safety Authority (EFSA) for each species, probiotics displaying values below the cut-off of all antibiotics were considered acceptable for use.

Hemolytic and bile salt hydrolase activity test

For hemolytic activity, the selected four strains were streaked on tryptic soy agar (BD) supplemented with 5% (v/v) defibrinized sheep blood (MB cell). The plates were incubated anaerobically at 37 °C for 24 h. When observing lysis activity below the colony, only strains displaying γ-hemolysis were considered safe⁵⁴. Staphylococcus epidermidis ATCC 35983, Staphylococcus aureus ATCC 29213, Lactobacillus rhamnosus GG was used as a positive control for α, β, and γ hemolysis, respectively.

Bile salt hydrolase (BSH) activity of selected strains was measured in MRS agar containing 0.5% (w/v) taurodeoxycholic acid (sodium salt) hydrate (Sigma). The plates were incubated anaerobically at 37 °C for 48 h. If strain exhibits BSH activity, a precipitation zone surrounding the colony or white, opaque colony appears.

D/L-lactic acid production test

Four selected strains were cultured in MRSc medium anaerobically at 37 °C for 24 h. Afterward, the cultures were centrifuged at 10,000 × g for 10 min and the supernatant was then filtered using 0.2 μm sterile acrodisc syringe filter (Pall, USA) to obtain cell free supernatant (CFS). D- and L-lactic acid in CFS was quantified using D-Lactate Assay Kit and L-Lactate Assay Kit (Bioassay, USA), and the test was assayed, according to the manufacturer’s protocol. Lacticaseibacillus rhamnosus GG (LGG) was included as a commercially used reference probiotic strain with an established history of food use and safety⁵⁵.

Cytotoxicity test on Caco-2 cells

To confirm toxicity on Caco-2 cells of four selected strains, lactate dehydrogenase (LDH) in the medium was quantified. LDH cytotoxicity assay kit (DOGENBio, Korea) was used, and the test was conducted following the manufacturer’s protocol. Caco-2 cells were seeded at a concentration of 5 × 104 cells/well in 96 well cell culture plate and incubated at 37 °C containing 5% CO₂ for 24 h to establish monolayer. Then, PBS-washed strains were inoculated at Caco-2 cells at a concentration of 10⁷, 10⁸, and 10⁹ CFU/well and incubated at 37 °C containing 5% CO₂ for 24 h. Escherichia coli O157:H7 ATCC 43895, a pathogenic strain, was used as a positive control, while a negative control consisted of medium without any strains. Cytotoxicity was calculated as follows:

$$\mathrm{Cytotoxicity}\left(\%\right)=\left(\frac{{\mathrm{OD}}_{450}\mathrm{for\; experimental\; release}-{\mathrm{OD}}_{450}\mathrm{for\; spontaneous\; release}}{{\mathrm{OD}}_{450}\mathrm{for\; maximum\; release}-{\mathrm{OD}}_{450}\mathrm{for\; spontaneous\; release}}\right)\times 100$$ 2

The spontaneous release represents the amount of LDH derived from untreated Caco-2 cells, while maximum release indicates quantity of LDH released when untreated cells are completely lysed using a lysis solution.

Oxygen tolerance test

The 1% (v/v) of overnight cultures of selected strains were inoculated into 12 well plates containing 2 mL of MRSc medium and incubated at 37 °C for 24 h shaking aerobically at 180 rpm. Optical density absorbance at 600 nm were measured at two-hour intervals from 0 h to 10 h and at 24 h using Microplate reader (Tecan, Switzerland). For anaerobic growth, the inoculated plates were placed in an anaerobic jar with an anaerobic gas pack (Oxoid, United States) and incubated at 37 °C for 24 h with shaking at 180 rpm. Optical density of the broth was then measured at 600 nm using Microplate reader (Tecan).

Heat tolerance test

Overnight cultures of probiotic strains were 1% (v/v) inoculated into 1.7 mL tubes containing 1 μl of MRSc medium and the heat tolerance assay was proceeded in the following sequences: (1) Heat shock under two conditions: at 45 °C for 3 h and 60 °C for 3 h under anaerobic condition. (2) The number of viable cells was measured by counting on MRS agar and was calculated as log CFU/mL. CFU was compared to that of a control incubated at 37 °C for 3 h under anaerobic condition. The percentage of survival rate was calculated as follow:

$$\mathrm{Survival\; rate}\left(\%\right)={(\log \mathrm{CFU}}_{\mathrm{test}}/{\log \mathrm{CFU}}_{\mathrm{control}})\times 100$$ 3

Gastric acid and bile acid resistance test

To determine intestinal viability of isolates, the simulated gastric acid (125 mM NaCl, 7 mM KCl, 45 mM NaHCO₃, and pepsin 3 g/l at a final pH 2.5 with HCl) and bile salt fluids (0.1% pancreatin (w/v), 0.3% bile (w/v) at a final pH 8.0 with NaOH) were used described previously⁵⁶. The experiment was performed by following previously reported procedures, with slight modifications⁵⁷. Four selected probiotics were cultured anaerobically in MRSc medium to 600 nm optical density absorbance of 1.0 by spectrophotometer (Thermo Scientific) and cells were harvested by centrifugation at 13,000 rpm for 2 min. The harvested cells were washed twice with 1x PBS buffer (GenDEPOT) and resuspended in simulated gastric acid and incubated anaerobically at 37 °C. After 3 h of incubation, cells were harvested by centrifugation at 13,000 rpm for 2 min, washed twice with 1 x PBS buffer and resuspended in bile salt fluids. Then, suspensions were incubated anaerobically at 37 °C. After 3 h of incubation, the cells were collected and washed twice with 1 x PBS buffer. 100 μl of serially diluted samples with 1x PBS buffer were spread on MRSc medium plates supplemented with 1.8% Bacto agar (BD), followed by anaerobic incubation at 37 °C for 24 h. In plates where approximately 200 to 300 colonies were observed per plate, the bacterial survival rate was determined by counting colony-forming units (CFUs). At this time, the control is performed with 1x PBS buffer instead of gastric acid and bile salt juice under same process. Separate experiments were performed three times. The percentage of survival was calculated as follows:

$$\mathrm{Survival\; rate}\left(\%\right)=(\log {\mathrm{CFU}}_{\mathrm{assay}}/\log {\mathrm{CFU}}_{\mathrm{control}})\times 100$$ 4

Caco-2 cell adhesion test

The adhesion of selected strains to Caco-2 cells was evaluated. Caco-2 cells (4 × 10⁵ cells/mL) were seeded in 12-well plates and incubated at 37 °C and 5% CO₂ for 24 h. Four selected strains were cultured in MRS broth at 37 °C for 24 h, then resuspended in DMEM without FBS and penicillin-streptomycin (10⁷ cells/well), and treated with Caco-2 cells. In parallel, blank wells without Caco-2 cells were prepared and treated identically with the bacterial suspensions to assess nonspecific attachment to the plastic surface. After 2 h, the Caco-2 cells were washed twice with PBS to detach non-adherent bacterial strains. The number of viable bacteria was measured by counting on MRS agar and was calculated as log CFU/mL. The final adhesion values were corrected by subtracting the bacterial counts recovered from the blank wells.

Bacterial genomic DNA whole genome sequencing

Selected bacterial strains cells were harvested by centrifugation at 10,000 × g for 5 min after cultivation. The genomic DNA was extracted and purified using the G-spin Total DNA Extraction Kit (iNtRON Biotechnology, Korea), according to manufacturer’s protocol. The DNA was sequenced using MinION (Oxford Nanopore Technologies, UK). Libraries were generated using a Ligation-Sequencing kit (SQK-LSK109) with the Native Barcoding Expansion 1-12 (EXP-NBD104) and 13-24 (EXP-NBD114) kit, according to the manufacturer’s instructions (Oxford Nanopore Technologies). The final library was loaded onto MinION-R9.4.1 flow cell and sequenced with MinKNOW software-v5.4.

For whole-genome sequencing, raw long-read data were first filtered by length and quality using Filtlong v0.2.1 to obtain reads suitable for assembly. Chromosomal assemblies were generated and merged using Trycycler v0.5.4⁵⁸, and the resulting consensus contigs were further polished with long reads using Medaka v1.0.3. Structural annotation was then performed using Prokka v1.14.6. For genome-based safety assessment, antibiotic resistance genes were analyzed using the Comprehensive Antibiotic Resistance Database (CARD) with default settings⁵⁹. Virulence-associated genes and putative horizontally transferred genes were identified using VFDB 2022⁶⁰ and HGTree v2.0⁶¹, respectively, with default parameters. In addition, prophage regions were predicted using PHASTEST⁶², and the results were cross-checked using PhiSpy to support genome-based safety characterization⁶³. To further investigate phenotype-linked genomic features, comparative genome-informed analyses were additionally performed for the four final candidate strains. Predicted subcellular localization of protein-coding sequences was analyzed using PSORTb 3.0, and proteins predicted to localize to the cell wall or extracellular space were used to summarize surface-associated features⁶⁴. Biosynthetic gene clusters and bacteriocin-related loci were identified using antiSMASH 8.0⁶⁵ and BAGEL4⁶⁶, respectively. Carbohydrate-active enzymes (CAZymes) were analyzed using dbCAN3⁶⁷, and only ORFs supported by all three tools (HMMER_dbCAN, DIAMOND_CAZy, and HMMER_dbCAN-sub) were considered high-confidence CAZyme candidates.

Microbial network analysis

Microbial association networks were constructed from relative abundance data using R. To reduce sparsity, taxa with zero counts, prevalence below 20% of samples, mean relative abundance below 0.05%, or zero variance across samples were excluded prior to network inference. Associations were inferred using the semi-parametric rank-based correlation and partial correlation estimation (SPRING) method implemented in the NetCoMi package (version 1.1.0), which estimates sparse conditional association networks accounting for compositionality⁶⁸. User-defined parameters included the use of the approximate rank-based estimation method, 50 repetitions, and an association threshold of |r | ≥ 0.10; all remaining parameters were kept at their default values. The resulting signed association matrix was converted into an undirected weighted graph using igraph package (version 2.2.1), retaining both positive and negative associations. Network layout was computed using a force-directed Fruchterman–Reingold algorithm with edge weights influencing node positioning. Network properties were analyzed using NetCoMi, including community detection by the Louvain algorithm and calculation of node centrality measures (degree, eigenvector centrality, and betweenness). The network was visualized in Gephi (version 0.10.1).

Animals experimental design

Before mouse experiments, Institutional Animal Care and Use Committee (IACUC) of Seoul National University approved the experimental plan (approval number, SNU-240306-6). And then, all experiment on the mice were performed in compliance with the Guide for the Care and Use of Laboratory Animals from the Seoul National University. Forty-eight female specific pathogen free BALB/c mice (4-weeks-old, 18–20 g) were purchased from RaonBio (South Korea). After two weeks of quarantine period, all mice were acclimatized to standard condition with temperature at 23 ± 2 °C and 50 ± 10% humidity under a 12-h light/dark cycle for one week. Feed and water were provided ad libitum. The food allergy mouse model was referenced from a previous publication⁶⁹. The mice were randomly divided into six groups (n = 8 in each group) as follows: one normal control (PBS group), one food allergy control (OVA group), and four probiotics-OVA groups (SLac002+OVA, SLga015+OVA, SLpa017+OVA, and SLpl116+OVA). Each mouse in five OVA-treated groups was injected with 0.2 mL of saline solution containing 50 μg of OVA (Sigma-Aldrich) and 1 mg of aluminum hydroxide gel (InvivoGen, USA) on days 7 and 21 for sensitization. From day 35, each mouse was oral-gavaged every other day with 0.2 mL of saline solution containing 50 mg of OVA until day 43 to challenge. Mice were fasted for 3–4 h before each intragastric challenge to ensure that the OVA antigen could quickly pass through the stomach without being destroyed by gastric acid. The oral gavage was performed using oral gavage needle (3.8 cm, 24 G, SamwooKurex, Korea) attached to 1 mL syringe. The mice in four probiotics-OVA groups were oral-gavaged every day with 0.2 mL (1.0 × 10⁹ CFU/mL) of the fresh live probiotic suspension in PBS from day 0 to day 43. The mice in the PBS and OVA groups were oral-gavaged with 0.2 mL of sterile 1 x PBS buffer, instead of probiotics. Final body weight was measured at the terminal time point immediately before euthanasia on day 43.

Evaluation of diarrhea severity and anaphylactic hypothermia

After the last challenge on day 43, core body temperature was measured using a rectal probe (Testo, USA). To minimize subjective bias, clinical symptom scores were evaluated in a blinded manner. The severity of diarrhea was scored using the following scale, 0: normal (normal stool), 1: minimal (soft stool), 2: slight (a little wet, unshaped stools), 3: moderate (a small amount of wet and unshaped stools with moderate perianal staining of the coat), 4: severe (severe and watery stools with severe perianal staining of the coat). The incidence of each diarrhea score (0 to 4) and the average diarrhea score were used to evaluate the severity of diarrhea²⁷. Rectal temperature was measured using a thermometer at 20-min intervals over a 60-min period, and the temperature changes for the mice in each group were calculated accordingly. Twenty-four hours after the final allergen challenge, mice were humanely euthanized, and biological samples were harvested for analysis.

Measurement of serum immunoglobulins and mast cell protease-1

The blood was collected into 2 mL K₂EDTA tubes (BD) and centrifuged at 3000 × g for 15 min, and then, the serum samples were collected and stored at −80 °C for further analysis. For measurement of IgE, IgG1, and mouse mast cell protease-1 (MCPT-1) levels in the samples, IgE Mouse Uncoated ELISA Kit (Invitrogen), IgG1 Mouse Uncoated ELISA Kit (Invitrogen), and Mouse MCPT-1 Uncoated ELISA kit (Invitrogen) were used, according to the manufacture’s protocols. Concentrations of OVA-specific IgE and OVA-specific IgG1 levels in the serum were measured with Mouse Serum Anti-OVA IgE Antibody Assay Kit (Chondrex, USA) and Mouse Anti-OVA IgG1 Antibody Assay Kit (Chondrex), according to the manufacturer’s protocols.

Analysis of cytokine profiles in spleen and colon

To assess systemic immune responses, splenocytes were isolated and stimulated ex vivo. The spleen was aseptically removed and crushed through a 40 μm cell strainer (BD) into RPMI 1640 medium (Welgene). After lysing red blood cells with ACK lysing buffer (Gibco), the single-cell suspension was washed and resuspended in RPMI 1640 medium supplemented with 10% FBS. The splenocytes were seeded into 24-well plates (5 × 10⁶ cells/well) and stimulated with OVA (100 μg/mL) for 72 h at 37 °C in a 5% CO₂ atmosphere. The culture supernatants were collected and stored at −80 °C until analysis.

To determine cytokine levels of the colon, the colon tissue samples measuring 1 cm in length were placed into ice-cold RIPA buffer (150 mM sodium chloride, 50 mM Tris-HCl, pH 7.4, 1 mM EDTA, 1 mM phenylmethyl-sulfonyl fluoride, 1% Triton X-100, 1% sodium deoxycholic acid, 0.1% sodium dodecyl sulfate, 5 μg/mL of leupeptin; ATTO, Japan) supplemented with protease inhibitor with stainless beads (GeneReach Biotechnology, Taiwan). The colon tissues were homogenized by bead beater (Bead Ruptor Elite, OMNI International, USA) and the colon tissue lysate was centrifuged at 12,000 × g for 10 min at 4 °C. The total protein concentration was determined using a BCA protein assay kit (Pierce, USA).

For protein analysis, the levels of cytokines (IL-4, IL-10, IL-17, and interferon gamma; IFN-γ) in the splenocyte culture supernatants and colon tissue homogenates were quantified using commercial ELISA kits (Invitrogen), according to the manufacturer’s instructions. The cytokine concentrations in the colon were normalized to the total protein content of the tissue samples. To further validate the cytokine profiles at the transcriptional level, qPCR was performed. Total RNA was extracted from the collected splenocyte pellets and colon tissues using TRIzol reagent and the PureLink RNA Mini Kit (Invitrogen) according to the manufacturer’s protocol. Total RNA from all samples was quantified using the NanoDrop 2000 (Thermo Fisher Scientific) and then reverse transcribed into cDNA using the PrimeScript RT Reagent Kit (TaKaRa, Japan). qPCR was conducted using SYBR Green Master Mix (Bio-Rad) on a real-time PCR system (CFX96, Bio-Rad). The relative mRNA expression levels were calculated using the 2^-ΔΔCt method with β-actin serving as the housekeeping gene⁵². The specific primer sequences used for qPCR are listed in Table 1.

Gut microbiota analysis

Fresh fecal samples were collected from all mice on Day 43 (24 h after the final challenge) and immediately stored at −80 °C until processing. Total genomic DNA was extracted from the fecal samples using the QIAamp Fast DNA Stool Mini Kit (Qiagen, Germany) according to the manufacturer’s protocol. The concentration and purity of the extracted DNA were assessed using a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific). All DNA samples were normalized to a concentration of 5 ng/μL using Ultra-Pure Water (Welgene) for downstream applications.

The V3–V4 hypervariable regions of the bacterial 16S rRNA gene were amplified using primers 341 F and 805 R, containing Illumina overhang adapters⁷⁰. The PCR reaction mixture contained KAPA HiFi HotStart Ready Mix (Roche, Switzerland), 15 μM of each primer, and 12.5 ng of template DNA. The PCR amplification was performed under the following thermal cycling conditions: initial denaturation at 95 °C for 3 min, followed by 35 cycles of denaturation at 95 °C for 30 s, annealing at 54 °C for 30 s, and extension at 72 °C for 1 min, with a final extension at 72 °C for 5 min. The PCR amplicons were purified and sequenced on the Illumina NextSeq system (Illumina, USA) using a 2 × 300 bp paired-end sequencing protocol. Library preparation and sequencing were conducted by Sanigen Co., Ltd. (Korea).

The raw sequencing data were processed using the QIIME 2 pipeline⁷¹. Quality control and denoising were performed using DADA2 to generate amplicon sequence variants (ASVs)⁷². For taxonomic classification, representative sequences were aligned against the SILVA database (release 138) using a Naive Bayes classifier, optimized for species-level identification⁷³. Beta diversity (Principal Coordinate Analysis, PCoA) based on Bray-Curtis distance was calculated to assess microbial community structural changes. To identify specific bacterial taxa differentially abundant between groups, Linear Discriminant Analysis Effect Size (LEfSe) was performed with an alpha value of 0.05 for the factorial Kruskal–Wallis test and a threshold of > 4.0 for the logarithmic LDA score.

To elucidate the functional interactions between the gut microbiota and host immune responses, Spearman’s correlation analysis was performed. The analysis examined the associations between the relative abundances of differential bacterial taxa (identified as key biomarkers via LEfSe) and allergic parameters, including serum immunoglobulins (IgE, IgG1), MCPT-1, and cytokine levels in all groups. Consistent with the comprehensive microbiome profiling described above, all statistical calculations and visualizations were implemented using the microeco package⁷⁴ in the R software environment (version 4.3.0). A p-value of < 0.05 was considered statistically significant.

Statistical analysis

Statistical analyses were conducted using the Statistical Package for Social Sciences (SPSS) 26.0 (SPSS Inc, USA) and GraphPad Prism 8.0.1 (GraphPad Software, USA). All data were expressed as the mean and standard error of the mean (SEM). The differences between means were analyzed via one-way analysis of variance (ANOVA), followed by Dunnett’s multiple comparison to determine the significance of differences among the means. Different lowercase letters in the figures indicate significant differences between groups. Also, the Student’s t-test was conducted for the comparison of two groups. A p < 0.05 was considered to assess statistical significance between groups.

Author contributions

Conceptualization: S.J. and J-H.L.; Resources: S.J. and J.Y.; Data curation: S.J., Y-T.K., and J.Y.; Formal analysis: S.J., Y-T.K., and J.Y.; Methodology: S.J. and J.Y.; Validation: S.J. and Y-T.K.; Visualization: S.J. and Y-T.K.; Investigation: Y-T.K. and J.Y.; Funding acquisition: J-H.L.; Project administration: J-H.L.; Supervision: J-H.L.; Writing – original draft: S.J., Y-T.K., J.Y., and J-H.L.; Writing – review & editing: Y-T.K. and J-H.L. All authors reviewed the manuscript.

Data availability

The raw sequencing data generated in this study have been deposited in the NCBI Sequence Read Archive (SRA) under the BioProject accession number PRJNA1422456.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41538-026-00978-2.