P. acidilactici XZ31 was isolated from Chinese traditional sourdough as previously described (Fu et al., 2020b). P. acidilactici XZ31 was cultured in fresh de Man, Rogosa, and Sharpe broth at 38°C for 24 h. The bacteria pellets were collected by centrifugation at 6,000 g for 10 min, freeze-dried, and then re-suspended in phosphate-buffered saline (PBS) for oral administration in the mouse model.

All experiments were approved by the Beijing Municipal Science and Technology Commission of China. Female (3 weeks old) specific pathogen-free BALB/c mice were purchased from Sipeifu Biotechnology (Beijing, China) and housed in individual cages under controlled conditions of temperature (23 ± 3°C), humidity (50 ± 10%), and light (12/12-h light/dark cycle). The mice were maintained on AIN-93M diet (wheat-free) before and during sensitization.

In vivo sensitization to wheat gluten was performed by using the protocol described by Caminero et al. with modifications (Caminero et al., 2019). The protocol is illustrated in Figure 1A and Supplementary Table S1 . In brief, the mice were randomly divided into the following groups (n = 10): gluten-sensitized and challenged (WP), P. acidilactici XZ31-treated (Pa), non-immunized negative control (Ctrl), and group injected with adjuvant (Ad). Gluten from wheat (Shanghai Yuanye Bio-Technology) was prepared as described previously (Galipeau et al., 2011) with modifications. Briefly, gluten was dissolved in artificial gastric juice (Yuanye Co., Ltd., Shanghai, China; content: 0.2 N HCl, pH 1.5) for 2 h in 37°C water bath with pepsin (Sigma-Aldrich, 2,240 U/ml). Subsequently, the mixture was adjusted to pH 7.5 by the addition of 0.5 M NaHCO₃. Trypsin (Sigma-Aldrich, 125 U/ml) was added, followed by shaking the mixture at 37°C for 1 h. Pepsin–trypsin (PT)-digested gluten was freeze-dried and stored at −20°C. The lypohilized samples were dissolved in PBS buffer (pH 7.0–7.2) before use, and the protein concentration was quantified by Bradford method (Akagawa et al., 2007).

Each mouse in the WP and Pa groups was sensitized by an intraperitoneal injection of 150 μl PT-digested gluten together with 50 μl aluminum adjuvant (Sigma-Aldrich). The total amounts of gluten were 100 μg on days 21, 28, 35, and 42, followed by a booster challenge with 600 μg gluten on day 50. The Ctrl group received 200 μl sterile PBS, and the Ad group received an equal amount of PBS containing 50 μl aluminum at each sensitization and challenge point. The Pa group was orally administered with a bacterial suspension (5 × 10 ⁸ to 6 × 10⁸ colony-forming units per mouse) once per day for 22 days from day 28 to 49, while the other groups received the same volume of sterile PBS. The mice were bled from the retro-orbital plexus on day 51. Fresh blood samples were incubated at room temperature for 30 min and centrifuged at 3,000 rpm for 10 min. The upper serum was taken and stored at -80°C. All mice were sacrificed on day 51, and organ and intestinal samples were collected for further analysis.

The clinical symptoms of the mouse were observed within 60 min after an intraperitoneal challenge. The symptoms were evaluated using a standardized scoring system as detailed in Table 1 (Fu et al., 2020a). Rectal temperature was measured before the gluten challenge and again at 30 and 60 min after the intraperitoneal challenge.

Gluten-specific IgE and IgG2a in the serum were measured by enzyme-linked immunosorbent (ELISA) assay as previously described (Caminero et al., 2019). Briefly, 100 μl of diluted proteins (5 μg/ml in 50 mM carbonate–bicarbonate buffer, pH 9.6) was coated onto a microplate and incubated overnight at 4°C. The plates were blocked with bovine serum albumin (1% w/v in 0.02 M Tris-buffered saline) for 2 h at 37°C, followed by washing three times with Tris-buffered saline-Tween 20, and incubated for 2 hours with 100 μl of mouse serum (diluted to 1:5 for IgE measurement and 1:200 for IgG2a). After the washing cycles, the plates were incubated with HRP-conjugated goat anti-mouse IgE or IgG2a. The binding was assessed with a tetramethylbenzidine substrate kit. The results were measured at 450 nm using a microplate reader (Thermo Fisher Multiskan FC).

At sacrifice, the duodenum and colon were collected in 4% paraformaldehyde. The tissues were processed, paraffin-embedded, and stained with hematoxylin and eosin (H&E) for evaluation of tissue morphology. Mast cells and goblet cells were determined on paraffin-embedded sections after staining with toluidine blue and periodic acid–Schiff (PAS), respectively. Histological scoring was performed according to a scoring system described by Fu et al. (2018; 2020a) and by Ding and Wen (2018) with modifications ( Table 2 ). The number of mast cells was quantified in at least five mice per group using ImageJ software.

Spleen from the mouse was excised and kept in cold RPMI 1640 medium containing 10% fetal bovine serum. Single-cell suspensions were prepared and re-stimulated with Cell Stimulation Cocktail for 6 h (eBioscience, USA). T help (Th) cells were stained with FITC-conjugated anti-CD4, APC-conjugated anti-IFN-γ, and PE-conjugated anti-IL-4 (eBioscience, USA) according to the manufacturer’s instructions. The stained cell samples were detected using BD LSRII flow cytometry. FACS Diva8.0 software was used to analyze the data.

Fecal samples were collected under sterile conditions and stored at −80°C until processing for DNA isolation. Total genome DNA from samples was extracted using the cationic detergent cetyl-trimethylammonium bromide method (Stewart and Via, 1993). DNA purity was monitored on 1% agarose gels (Sambrook, 2008). According to the concentration, DNA was diluted to 1 ng/μl using sterile water. 16S rRNA genes of distinct regions (16S V3 and V4) were amplified using specific primers with barcodes. The amplified products were mixed in equidensity ratios and purified with Qiagen Gel Extraction Kit (Qiagen, Germany). Sequencing libraries were generated using TruSeq^® DNA PCR-Free Sample Preparation Kit (Illumina, USA) following the manufacturer’s recommendations. The library concentrations and quality were checked using Qubit@ 2.0 Fluorometer (Thermo Scientific) and Agilent Bioanalyzer 2100 system, respectively. Lastly, the library was sequenced on Illumina NovaSeq platform. The original data was spliced and filtered out to get high-quality-tag sequences. The optimized sequences, with a limit of 3% distance, were gathered into operational taxonomic units (OTU) (Edgar, 2013), and the representative sequences were aligned against Silva database (http://www.arb-silva.de). Community richness and the diversity of each sample were determined by alpha diversity using the index of Chao and Shannon, which were calculated with QIIME (Version 1.7.0) and displayed with R software (version 2.15.3). Beta diversity analysis was calculated by QIIME software (version 1.9.1) and used as a comparative analysis of microbial communities in different samples.

Data are expressed as mean ± standard error of mean for each group. As noted, the statistical analysis was performed with Graph Prism using ANOVA test. P <0.05 was considered statistically significant.

We firstly examined the clinical allergenic response ( Figures 1B, C ). After the gluten intraperitoneal challenge, the WP group showed the highest anaphylaxis score and a persistent decrease of rectal temperature within 60 min, providing preliminary evidence of the validity of the model. The anaphylaxis score was significantly lower in the Pa group than in the WP group, and the rectal temperature recovered slightly after 60 min. The fluctuation of the rectal temperature was small in the Ctrl and Ad groups, and no obvious clinical reaction was observed.

Previous studies have shown that food allergy is always accompanied by gastrointestinal manifestations. We performed histopathological analyses of blinded tissue sections for the duodenum and colon from each group. H&E and PAS staining were performed to observe the intestinal morphology. As shown in Table 2 , we examined the villus height, crypt depth, mucosal degeneration, and goblet cell distribution. The colonic infiltration of immune cells was also evaluated. Villus height and crypt depth correlated with the pathological score and demonstrated shedding of intestinal villi in tissues with greater pathological changes. Representative H&E staining results are shown in Figures 2A , 3A . In the Ctrl and Ad groups, the duodenum showed a well-differentiated brush border structure. There was no significant difference in the histological score of the duodenum between the Ctrl group, Ad group, and Pa group ( Figure 2C ). In the WP group, gluten caused the duodenum villi to become shorter and disordered ( Figures 2A, B , black box). Dysplastic glands invaded through the muscularis mucosa into the submucosa and tunica muscularis ( Figure 2A , red arrow). Reduced crypt depth and extensive inflammatory infiltration were also observed in the colon of the mice in the WP group ( Figure 3B , black box). In the Ctrl, Ad, and Pa groups, a large number of goblet cells were distributed evenly in the duodenum and colon ( Figures 2B , 3B , black arrow). The gut sections were also stained with toluidine blue for mast cells. As we can see in Figures 4A, B , a higher number of mast cells and extensive inflammatory cell infiltration ( Figure 4B , black box) were observed mainly in the colon of the WP group. The inflammatory cell infiltration in the Pa group was less than that in the WP group. These results further demonstrated the role of P. acidilactici XZ31 in alleviating gluten-induced allergic responses.

To explore the mechanism of P. acidilactici XZ31 in alleviating an allergenic reaction, we further characterize T cell differentiation in the spleen (anti-mouse IFN-γ for Th1 polarization and anti-mouse IL-4 for Th2 polarization). The IL-4 and IFN-γ production of splenic CD4 T cells is shown in Supplementary Figure S1 . The production of IFN-γ was significantly lower in the WP group compared with that in the Ctrl group, and the production of IL-4 was higher and of IFN-γ was lower in the WP group than in the Pa group ( Figures 5C, D ). Food allergy is commonly a Th2-dependent disease and mediated by IgE, and the imbalance of Th1 and Th2 is an effective indicator of food allergy (Lee et al., 2018). As shown in Figure 5E , the Th1/Th2 ratio was significantly lower in the WP group. By contrast, we found that the ratio of Th1/Th2 increased significantly in the Pa group, indicating that the Th1/Th2 balance drifted to Th1 after P. acidilactici XZ31 administration. As shown in Figures 5A, B , the levels of IgG2a and IgE were lower in the Ctrl and Ad groups. Gluten challenge resulted in a significant increase in gluten-specific IgE. The application of P. acidilactici XZ31 significantly reduced the concentration of gluten-specific IgE. In contrast, the production of gluten-specific IgG2a was promoted by P. acidilactici XZ31 compared to the WP group. These results show that P. acidilactici XZ31 regulated the Th1/Th2 immune balance toward Th1 polarization, which subsequently induced a reduction in gluten-specific IgE production.

It is increasingly evident that intestinal bacteria are critical for regulating allergic responses to dietary antigens (Nermes et al., 2013; Iweala and Nagler, 2019). Gut microbial community composition was assessed to test the effect of the intervention of P. acidilactici XZ31 on the intestinal microbiota homeostasis. The number of effective tags was more than 74% of the paired-end reads in all samples ( Supplementary Table S2 ). The sequences that showed more than 97% similarity were assigned to the same OTU. The OTU coverage (≥0.999) indicated that a lower abundance of OTU has also been covered, which could accurately reveal the gut bacterial community structure of these samples ( Supplementary Figure S2 ). At the phylum level ( Figure 6A ), Bacteroidota and Firmicutes are predominant in all groups. The proportion of Bacteroidota decreased slightly in the gluten sensitization group, while the proportion of Firmicutes increased compared with the Ad and Ctrl groups. The intake of P. acidilactici XZ31 reduced the ratio of Firmicutes/Bacteroidota in the intestinal microbiota. At the family level ( Figure 6B ), the abundance of Akkermansiaceae and Erysipelotrichaceae was most affected by gluten and the P. acidilactici XZ31 treatment. In short, the abundance of Akkermansiaceae and Lactobacillaceae in the Pa group was higher than that in the WP group. The increase in Caulobacteraceae and Erysipelotrichaceae after gluten sensitization was balanced by the administration of P. acidilactici XZ31. Additionally, the administration of P. acidilactici XZ31 favored an increase of buryrate-producing bacteria, such as Faecalibacterium prausnitzii and Collinsella aerofaciens ( Supplementary Figures S3 , S4 ).

The bacterial community diversity and richness were assessed using the Shannon index and Chao index, respectively. The larger Shannon and Chao index values indicate higher diversity and richness of the sample. As shown in Figures 6C, D , there was no significant difference in community richness between the different groups. The P. acidilactici XZ31 administration suppressed gluten-induced decline in intestinal microbiota diversity.

Linear regression analysis was performed to reveal the correlation between bacterial community in the gut and allergenic reaction. The results ( Figures 6E, F ) showed that the concentration of gluten-specific IgE antibody was positively related to the relative abundance of Erysipelotrichaceae and Firmicutes. Similar results were also found in Pearson analysis, of which the increase of gluten-specific IgE is accompanied by an increase in the relative abundance of Erysipelotrichaceae and Firmicutes (Pearson correlation analysis: r = 0.6608, P = 0.0039; r = 0.5365, P = 0.0264).

Beta diversity was used as a comparative analysis of microbial communities in different samples. Principal coordinate analysis (PCoA) revealed the correlation between gluten sensitization, Pa administration, and microbiota composition ( Figure 7 ). Points spaced closer together are more similar in microbiota composition (Minchin, 1987). The WP group was far from the Pa group, implying a lower similarity in bacterial composition between them. Interestingly, the separation with the Ctrl and Ad groups increased in the Pa group compared to the WP group. The present results demonstrate that the administration of P. acidilactici XZ3 can diminish gastrointestinal inflammation and allergic responses and increase the abundance of beneficial microbiota in the gut. We speculate that P. acidilactici XZ3 may exhibit more beneficial effects in different types of intestinal homeostasis.