Endotoxin-free, desalted, and phosphorothioated class-B CpG-ODNs [MsST (11)] were synthesized by GeneDesign, Inc. (Osaka, Japan). MsST was reconstituted in endotoxin-free water and sterilized by passage through a 0.22-μm pore microfilter. The sequence of MsST was as follows: 5′-CAGGACGTTGTATCACTGAA-3′.

MsST encapsulated in carbonate apatite-based particles was prepared using a modification of a previously established method (10, 12, 13). First, 1 M CaCl₂ (900 μL) was mixed with 30 mL of inorganic solution (44 mM NaHCO₃, 0.9 mM NaH₂PO₄, 25 mM D-glucose, pH 7.5) containing 6 mg MsST, and the mixture was allowed to stand for 60 min at 37°C. The suspension then was centrifuged at 700 g for 5 min, and the pellet was washed and freeze-dried. The control particles (Cap) were prepared without ODN by the same procedure. Finally, the pellet (1 mg) was blended with a commercial mouse powder feed (1 g, MF, Oriental Yeast, Tokyo, Japan) to prepare ODNcap-feed (ODNcap-F) or Cap-feed (Cap-F).

All experimental procedures were carried out in accordance with the Regulations for Animal Experimentation of Shinshu University, and the animal protocol was approved by the Committee for Animal Experiments of Shinshu University (No. 280063).

Female BALB/c mice (4 weeks of age) were purchased from Japan SLC (Shizuoka, Japan), housed under temperature- and light-controlled conditions, and fed a standard diet (MF, Oriental Yeast) and sterile water ad libitum. After preliminary housing for 2 weeks, we established the following four experimental groups (n = 6/experiment): non-treatment (NT), standard diet-feed (Ctrl-F), Cap-F, and ODNcap-F. Three independent experiments (Exp. 1 to 3) were performed.

The experimental protocol is shown schematically in Figure 1 . Briefly, mice were fed the standard or customized feeds freely for 70 days. An OVA-induced allergic airway inflammation model was used by modification of a previously established method (14). Mice were immunized on Days 42, 49, and 56 with OVA (Sigma-Aldrich, St. Louis, MO, USA) by intraperitoneal injection of 1 μg OVA adsorbed to 2 mg of aluminum oxide gel. From Days 63 to 69, mice were exposed (once daily for 30 min/day) to an aerosol of 2% OVA in phosphate-buffered saline; exposure employed an ultrasonic nebulizer (NE-U17; OMRON, Kyoto, Japan). On Day 70, 24 h after the last OVA inhalation, mice were subjected to different procedures as follows: in Exp. 1 and 2, mice were euthanized to collect serum, feces, bronchoalveolar lavage fluid (BALF), and lungs; in Exp. 3, mice were tracheostomized under anesthesia to measure the lung mechanics.

Levels of OVA-specific IgG₁ and cytokines in serum and BALF were quantified using commercial enzyme-linked immunosorbent assay (ELISA) kits (IgG₁: Cayman Chemical, Ann Arbor, MI, USA; cytokines: eBioscience, San Diego, CA, USA) according to the manufacturers’ instructions. Quantification of OVA-specific IgG_2a was performed according to a previously established procedure (15).

Total RNA was isolated from the lungs with TRIzol Reagent (Life Technologies, Carlsbad, CA, USA), and then was reverse-transcribed using PrimeScript RT Master Mix (Takara Bio, Shiga, Japan) according to the manufacturers’ instructions. Real-time quantitative PCR was performed with a Thermal Cycler Dice Real Time System II (TaKaRa Bio) using TB Green Premix Ex Taq II (TaKaRa Bio) as described previously (16). The optimized primers for target-specific amplification were purchased from Perfect Real Time Support System (TaKaRa Bio), except for Il22ra1- (17), Reg3g- (18), and Extl3-specific (17) primers that were synthesized by Integrated DNA Technologies (Coralville, IA, USA).

The number of BALF cells was counted by trypan blue staining. We omitted the NT group from this analysis because the total number of BALF cells in that group was excessively low (< 1×10⁵ cells; Figure 2A ). BALF cells were fixed in 4% paraformaldehyde. The fixed cells (1×10⁵ cells/tube) then were stained with fluorescein isothiocyanate (FITC)-conjugated rat anti-mouse CD11b (1/1,000; 101205; BioLegend, San Diego, CA, USA), FITC-conjugated rat anti-mouse Ly6G (1/100; 127606; BioLegend), Alexa Fluor 488-conjugated rat anti-mouse CD3 (1/100; 100210; BioLegend), Alexa Fluor 488-conjugated rat anti-mouse CD19 (1/100; 115521; BioLegend), Alexa Fluor 488-conjugated mouse anti-mouse NK1.1 (1/100; 108718; BioLegend), phycobiliprotein (PE)-conjugated rat anti-mouse SiglecF (1/100; 552126; BD Biosciences, Franklin Lakes, NJ, USA), PE/Cyanine5-conjugated rat anti-mouse CD11b (1/100; 101210; BioLegend), and/or peridinin-chlorophyll-protein-conjugated hamster anti-mouse CD11c (1/200; 117326; BioLegend) antibodies for 60 min at 4°C. The cells then were washed, and the numbers of eosinophils (CD11b⁺ CD11c^– SiglecF⁺), neutrophils (CD11b⁺ Ly6G⁺), lymphocytes (CD3⁺ CD19⁺ NK1.1⁺), and monocytes (CD11c⁺) were determined using a flow cytometer (FACSCalibur, BD Biosciences). All data were analyzed using FlowJo software (ver. 10.3; FlowJo, LLC, Ashland, OR, USA).

Lungs were fixed in 10% buffered formalin, embedded in paraffin wax, and sectioned (5 μm) according to standard procedures. Hematoxylin and eosin (HE) or periodic acid-Schiff (PAS) staining of the sections was performed according to standard procedures (10, 19). Histopathologic evaluation was performed on the HE-stained sections. PAS-stained sections were used to assess goblet cell hyperplasia. A ranking system (0: none, 1: rare, 2: mild, 3: moderate, 4: intermediate between moderate and gross, 5: gross, and 6: very gross) was employed to score the PAS-positive area around the bronchioles. Immunohistochemistry was performed to detect pulmonary vein cardiomyocytes (PVCs). The deparaffinized sections were subjected to antigen retrieval by microwave irradiation in 10 mM citrate buffer for 10 min. Endogenous alkaline phosphatases and non-specific absorption of antibodies were blocked with Endogenous Peroxidase and Alkaline Phosphatase Blocking Solution (Vector Laboratories, Burlingame, CA, USA) and Blocking One Histo (Nacalai Tesque), respectively. The sections then were reacted with mouse anti-cardiac muscle troponin T (cTnT) (1/40; MA5-12960; Thermo Fisher Scientific) or mouse isotype control (1/1,000; 31903; Thermo Fisher Scientific) antibodies overnight at 4°C. The reaction was detected with ImmPRESS-AP Horse Anti-Mouse IgG Polymer Reagent (Vector Laboratories) and alkaline phosphatase substrate solution (Vector Laboratories) according to the manufacturer’s instructions. All sections were observed by light microscopy (BZ-X800; KEYENCE, Osaka, Japan).

Mice were tracheostomized under anesthesia and connected to a ventilator (flexiVent; SCIREQ, Montreal, Canada) to assess responsiveness to methacholine according to a previously established procedure (20). Mice were then exposed to nebulized methacholine (0, 3, 10, and 30 mg/mL), and respiratory system resistance (Rrs) values were monitored after each methacholine challenge using the snapshot technique. All data were analyzed using the flexiWare software (ver. 7.6).

Total RNA was isolated from the lungs of eight mice (two mice/group). Transcriptome profiling based on DNA microarray analysis was performed by Macrogen Japan (Kyoto, Japan) using the SurePrint G3 Mouse Gene Expression 8x60 K platform (Agilent, Santa Clara, CA, USA). The microarray results were extracted using Agilent Feature Extraction software (ver. 11.0) (Agilent), and then log-transformed and normalized using the quantile method. A total of 32,312 probes were identified in the filtered data. The statistical significance of the expression data was determined using the local-pooled-error test and fold change. The false discovery rate (FDR) was controlled by adjusting the P values using the Benjamini-Hochberg algorithm. All data analyses were conducted using R software (ver. 3.2.1) (21).

BALF samples were boiled in an equal volume of 2× sample buffer (Wako, Osaka, Japan) for 5 min. A 20-μL aliquot was resolved by sodium dodecyl sulfate-polyacrylamide gel (15% polyacrylamide) electrophoresis. The resolved bands were transferred from the gel onto a polyvinylidene difluoride membrane (GE Healthcare, Amersham, UK). The membrane was blocked with 3% skim milk for 60 min at room temperature. The blots then were reacted with rabbit anti-regenerating islet-derived protein 3 gamma (Reg3γ) antibody (1/500; SAB4301004; Sigma-Aldrich) overnight at 4°C, and were further incubated with horseradish peroxidase-conjugated donkey anti-rabbit IgG antibody (1/5,000; 406401; BioLegend) for 60 min at room temperature. The reaction was detected with a lumino-image analyzer (ImageQuant LAS 500; GE Healthcare) with ECL Prime Western Blotting Detection Reagent (GE Healthcare). Densitometric analysis of the signals was performed with ImageJ software (ver. 1.53a) (22).

DNA was isolated from fecal samples using the QIAamp Fast DNA Stool Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer’s instructions. The 16S V3–V4 region of the fecal DNA was amplified as described previously (23, 24). The amplicon was barcoded and enriched by PCR using the Nextera XT Index Kit (Illumina, San Diego, CA, USA), and then purified and quantified. A pool containing equivalent quantities of the barcoded V3–V4 amplicons from each sample was sequenced (by TaKaRa Bio) using a MiSeq instrument (Illumina) to generate paired-end (2×250 bp) Illumina-sequencing data. The QIIME2 pipeline (ver. 2019.1) (25) was employed to process and analyze the data. The demultiplexed raw sequences were filtered by trimming primer sequences, removing low-quality and chimeric reads, joining paired-end reads, and identifying and enumerating amplicon sequence variants (ASVs) using the q2-dada2 plugin. The resultant data were aligned to build a phylogenetic tree using the q2-phylogeny plugin. The q2-diversity plugin was used to compute and statistically analyze the α and β diversity metrics of the aligned data. ASVs were referred to a trained classifier (gg-13-8-99-nb-classifier.qza) using the q2-feature-classifier plugin to assign taxonomy. The denoised data contained 3,991,059 high-quality reads from 48 samples (median: 87,864 reads/sample). The data of three samples that were identified as having excessively low read numbers were abrogated. The data were then rarefied to 55,682 reads per sample in the diversity analyses.

The statistical analyses of the microarray data and the diversity metrics of the fecal microbiome are described in the subsections titled “Lung transcriptome profiling” and “16S ribosomal RNA gene amplicon sequencing”, respectively. Other statistical analyses were performed using a statistical software package (GraphPad Prism 8; GraphPad Software, San Diego, CA, USA). Two-tailed one-way analysis of variance followed by Tukey’s tests or the Kruskal–Wallis test with Dunn’s multiple comparisons was used to determine the significance of the differences. Differences between groups in the lung mechanics data were analyzed using a mixed effects model for repeated measures; the P values were adjusted using an FDR method. Correlations between the relative abundance of fecal bacteria and OVA-specific IgG₁ levels were identified using Spearman’s correlation. Differences at P < 0.05 were considered significant.

Allergen-induced airway inflammation represents an important pathology in allergic asthma and is mediated primarily by the Th2 immune response (1). Therefore, we investigated the effect of free feeding of ODNcap-F on the prevention of OVA-induced airway inflammation in mice. The body weight and feed consumption of each mouse were comparable among all groups over the course of the study ( Supplementary Figure 1 ). Signatures of Th2-mediated inflammation in BALF, lungs, and serum were analyzed on Day 70. The total number of BALF cells was increased similarly in Ctrl-F and Cap-F compared to the non-treated healthy control group (NT) ( Figure 2A ). This increase was suppressed significantly in ODNcap-F ( Figure 2A ). Flow cytometric analysis of BALF cells showed a significant decrease in the number of eosinophils, neutrophils, and lymphocytes in ODNcap-F compared to Ctrl-F and Cap-F ( Figure 2B ). Th2 cytokines (interleukin [IL]-4, IL-5, and IL-13) in BALF accumulated to significantly higher levels in Ctrl-F and Cap-F compared to NT, whereas the levels of these cytokines were comparable between NT and ODNcap-F ( Figure 2C ). In histopathological analysis with HE-stained lung sections, similar patterns of massive infiltration of inflammatory cells around the bronchi were observed in Ctrl-F and Cap-F, but were suppressed in ODNcap-F ( Figure 2D ). OVA-specific immunoglobulin G₁ (IgG₁) was detected in the serum and BALF of OVA-sensitized/challenged mice, while the level of these antibodies was significantly reduced in ODNcap-F compared to Ctrl-F and Cap-F ( Figure 2E ). Collectively, these results showed that free feeding of ODNcap-F prophylactically suppresses OVA-induced Th2 responses and inflammation.

Previous reports have demonstrated that CpG-ODNs strongly induce the Th1 immune response to counteract to the Th2-biased response (26–28). Therefore, we measured levels of OVA-specific IgG_2a in the serum and interferon-gamma (IFNγ) in BALF, as well as the expression of IFNγ-encoding (Ifng) mRNA in the lungs, as markers of Th1-mediated responses. Unexpectedly, the serum level of OVA-specific IgG_2a was comparable among OVA-sensitized/challenged groups ( Supplementary Figure 2A ). The BALF level of IFNγ was under the detection limit in all mice (data not shown). Ifng expression in the lungs was significantly upregulated in Ctrl-F and Cap-F compared with NT, whereas the Ifng transcript level in ODNcap-F resembled the baseline level ( Supplementary Figure 2B ). The levels in BALF of IL-10, a key mediator of the regulatory-type immune response (29), were comparable among all groups ( Figure 2C ). These results suggested that the attenuation of allergic airway inflammation by ODNcap-F feeding may not be involved in inducing the protective Th1- and IL-10-mediated counter-immunity.

Airway remodeling accompanied by goblet cell hyperplasia in airway epithelia also contributes to the pathogenesis of asthma (1). We observed PAS-stained lung sections to evaluate mucin production from goblet cells in the bronchi ( Figure 3 ). PAS-positive cells in the bronchial epithelia were prominent in Ctrl-F and Cap-F, but were significantly decreased in ODNcap-F ( Figure 3 ). These results showed that free feeding of ODNcap-F attenuates the development of goblet cell hyperplasia and mucus hyperproduction in the airways of OVA-sensitized/challenged mice.

To investigate the effect of free feeding of ODNcap-F on AHR, we evaluated airway responsiveness to methacholine challenge using the forced oscillation technique. The values of respiratory system resistance (Rrs) ( Figure 4 ) in Ctrl-F were increased by methacholine challenges, and these effects were dose dependent. The values of Rrs in Cap-F were comparable to those in Ctrl-F. The Rrs values in ODNcap-F remained at baseline even with methacholine challenge. These results showed that free feeding of ODNcap-F attenuates OVA-induced AHR.

To elucidate the suppressive mechanism(s) of action of ODNcap-F in OVA-induced airway insults, we analyzed the lung transcriptome using DNA microarray analysis. Although this analysis was conducted with few replicates (two mice/group), we identified 108 genes that were differentially expressed among the OVA-sensitized/challenged groups ( Figure 5A ). Significantly upregulated and downregulated genes are visualized in Figures 5B, C , respectively. Interestingly, gene ontology analysis showed that genes related to cardiac muscle contraction (GO:0060048; Myh6, Myl4, Myl1, Tcap, Tnni3, Tnnt2, Ttn, Nppa, and Tnnt3) were significantly upregulated in ODNcap-F compared to Ctrl-F, Cap-F, or both. Further analyses revealed that genes preferentially or highly expressed in cardiac muscle [specifically, Myl7 (30), Myom2 (31), Sln (32), Pgam2 (33), Cox8b (34), Myh7b (35), and Fabp3 (36)] also were upregulated in ODNcap-F ( Figures 5A, B, D ). In contrast, the levels of all of these transcripts were decreased in Ctrl-F compared to NT ( Figure 5D ). Among the genes exhibiting ODNcap-F-specific upregulation, we identified Reg3g, which encodes the antimicrobial peptide Reg3γ ( Figure 5A ).

Cardiomyocytes, components of cardiac muscle, are known to be present in the lungs, specifically along the large pulmonary veins (37). Using quantitative PCR, we confirmed that mRNA expression of cardiomyocyte-specific markers (specifically, Tnnt2, Myh6, and Myl7) in the lungs was significantly suppressed in Ctrl-F and Cap-F compared to NT, whereas the expression of these genes in ODNcap-F was comparable to that in NT ( Figure 6A ). Lung sections were stained for cTnT to confirm the presence of pulmonary vein cardiomyocytes (PVCs). In NT, staining for cTnT constituted a dense ring around the large pulmonary vein ( Figure 6B and Supplementary Figure 3 ). Interestingly, a disjunct area in the cTnT-positive ring was observed occasionally in Ctrl-F as well as Cap-F, suggesting allergic injury of PVCs ( Figure 6B ). Notably, such clearly injured areas were not observed in ODNcap-F ( Figure 6B ). These results suggested that free feeding of ODNcap-F inhibits allergic injury of PVCs.

In lungs, Reg3γ is expressed in airway epithelial cells (AECs) responsive to methicillin-resistant Staphylococcus aureus infection or IL-22 (17, 38). Interestingly, a previous study demonstrated that Reg3γ induced by IL-22 inhibits allergic airway inflammation in asthma models (17). To confirm that free feeding of ODNcap-F promotes Reg3γ expression in the lung, we analyzed Reg3g expression in the lungs and secretion of Reg3γ into BALF. The Reg3g transcript level was significantly upregulated in ODNcap-F compared to that in other groups ( Figure 7A ). Secretion of Reg3γ into BALF also was increased in ODNcap-F ( Figure 7B ). We next analyzed the lung mRNA expression of genes encoding IL-22 signaling-related proteins, which would be expected to contribute to the induction of Reg3γ production. The expression of Il22ra1, which encodes a functional IL-22 receptor (IL-22R1) (39), was significantly impaired in Ctrl-F and Cap-F, but was restored in ODNcap-F ( Figure 7C ). The expression levels of Il22 and Il22ra2, which encode a natural antagonist of IL-22 signaling (IL-22 binding protein) (39), were comparable among the allergen-sensitized/challenged groups ( Figure 7C ). We also found that the expression of Extl3, which encodes a putative receptor for Reg3γ (exostatin-like 3; EXTL3) (40), was significantly downregulated in Ctrl-F and Cap-F compared to NT, whereas Extl3 expression was comparable between NT and ODNcap-F ( Figure 7C ). Elsewhere, it has been reported that IL-22R1 and EXTL3 are preferentially expressed on AECs in the lungs (17). Collectively, these results suggested that free feeding of ODNcap-F regulates the mechanisms for the induction, secretion, and action of Reg3γ in AECs to suppress allergic airway inflammation.

To evaluate the effects of ODNcap-F on the gut microbiota, next-generation sequencing data for fecal 16S rRNA gene amplicons were analyzed. α diversity indices were comparable among all groups ( Supplementary Figure 4 ). Principal coordinate analysis ( Figure 8A ) and permutational multivariate analysis of variance ( Figure 8B ) based on weighted UniFrac distances revealed that the fecal microbiome structure of ODNcap-F was significantly different from those of NT and Ctrl-F. Changes in the abundances of Firmicutes and Bacteroidetes (the two most-abundant phyla) resulted in a significant decrease in the ratio of Bacteroidetes to Firmicutes in ODNcap-F compared to those of other groups ( Figures 8C, D, F ). Other phyla with minor abundances were also altered in the gut microbiome of ODNcap-F ( Figure 8E ). In deeper analysis within Firmicutes and Bacteroidetes phyla, the proportions of some bacterial groups, such as the S24-7 family and Clostridium genus, in ODNcap-F were significantly different from those in other groups ( Figure 8G ). To identify specific gut bacteria related to allergic inflammation, Spearman’s correlations between the proportions of each bacterial group and the amounts of OVA-specific IgG₁ in serum and BALF were analyzed. We found 27 bacteria groups that significantly correlated with OVA-specific IgG₁ levels in serum, BALF, or both ( Figure 8H ).