The AD mouse model was generated via ovalbumin (OVA) sensitization as previously described, with some modifications.24 Briefly, 6-week-old female mice were anesthetized and their backs were shaved. Skin injury was then induced by repeated tape stripping, after which 100 µg of OVA (grade V; Sigma-Aldrich, St Louis, MO, USA) in 100 µL of phosphate-buffered saline (PBS), or PBS alone, was applied to a sterile patch and attached to the tape-stripped area with an adhesive tape. Treatment was repeated once during the 1-week sensitization period, followed by a 2-week resting period before a repeat of the sensitization steps over 1 week. The mice were euthanized and samples were collected 24 hours after the last treatment (Supplementary Fig. S1).

Female BALB/c mice weighing 16-20 g (3 weeks old) were purchased from Orient Bio Inc. (Seongnam, Korea) and treated in accordance with the guidelines of the Institutional Animal Care and Use Committee of the Asan Medical Center and Ulsan University College of Medicine (identification code No. 2014-02-182).

For the induction of gut dysbiosis, mice were treated with an antibiotic cocktail in their drinking water for 2 weeks before primary sensitization with ampicillin (1 g/L; Ratiopharm, Ulm, Germany), vancomycin (500 mg/L; Cell Pharm, Hannover, Germany), ciprofloxacin (200 mg/L; Bayer Vital, Leverkusen, Germany), imipenem (250 mg/L; MSD, Haar, Germany), and metronidazole (1 g/L; Fresenius, Bad Homburg, Germany).25

To investigate the effect of fecal microbiota transplantation on gut dysbiosis, fecal samples were collected directly from untreated, age-matched, healthy donor mice. Fresh feces from 10 mice were pooled, mixed with sterile normal saline, and immediately homogenized. The homogenate was centrifuged at 100 × g for 5 minutes at 4°C and the supernatant was used for transplantation. The mice received 100 µL of the supernatant orally from 2 weeks before primary sensitization until the study endpoint.26

The Lactobacillus rhamnosus (Lcr35) strain used in this study was obtained from Lyocentre^® Laboratory (Aurillac, France) and prepared according to the manufacturer's directions. Lcr35 cells were suspended in saline and administered orally from 2 weeks before primary sensitization until the study endpoint.1126

Dorsum lesions were scored for erythema, scaling, and excoriation after each sensitization event using a 0-3 scoring system, where 0 = no lesion, 1 = mild lesion, 2 = moderate lesion, and 3 = severe lesion.11 The same 2 investigators randomly performed all scoring evaluations throughout the study.

To determine whether epidermal permeability barrier function is altered as a result of OVA-induced AD, we measured transepidermal water loss (TEWL) at baseline at the beginning of the experiment, and then after each sensitization event, using a vapometer (SWL-3; Delfin Technologies Ltd, Kuopio, Finland).

The dorsal skin of the experimental mice was removed on the final day of the schedule, fixed in 10% phosphate-buffered formalin, and embedded in paraffin. Serial paraffin sections (4.5 mm thick) were stained with hematoxylin and eosin for the evaluation of edema.

Serum samples were obtained from blood taken during exsanguination of the mice after completing the sensitization, and stored at −80°C until use. Total IgE levels in sera were detected using the Mouse IgE ELISA kit (eBioscience, San Diego, CA, USA) in duplicate. The optical density was measured at 450 nm.

To measure interleukin (IL) 4 expression in mouse skin, RNA was extracted from the dorsal skin of the experimental mice using the RNeasy kit (Qiagen, Valencia, CA, USA). Real-time PCR was performed using the TaqMan method on an ABI 7900 system (Applied Biosystems, Piscataway, NJ, USA). Each signal was normalized to that of GAPDH in the same sample.

Intraepithelial lymphocytes (IELs) were isolated from pooled mouse intestines as previously described.27 Briefly, the intestines were cut lengthwise into short segments and shaken in Roswell Park Memorial Institute (RPMI)-1640 containing 1 mM dithiothreitol, 2 mM ethylenediaminetetraacetic acid, and 2% (v/v) fetal calf serum (FCS) for 15 minutes at 37°C to remove the epithelial layer. The tissue remaining from the epithelial stripping was minced and digested in RPMI-1640 containing 1.5 mg/mL collagenase II (Gibco), 50 μg/mL DNase I (Sigma-Aldrich), and 1% FCS for 40-45 minutes at 37°C. The digested tissue was then washed and filtered at least twice to obtain a single-cell suspension. To harvest IELs from the epithelial layer, the cells were further spun through a 40:70 Percoll gradient, and IELs were isolated from the interphase layer. For intracellular cytokine staining, immediately after isolation, the cells were incubated for 4 hours with 50 ng/mL PMA (Sigma-Aldrich), 750 ng/mL ionomycin (Sigma-Aldrich), and 10 μg/mL GolgiPlug (BD Biosciences, Mountain View, CA, USA) in a tissue culture incubator at 37°C.

Next, the cells were fixed and permeabilized using the Intracellular Fixation & Permeabilization Buffer set from eBioscience and stained with the following antibodies: APC-anti-FOXP3 (FJK-16s), PE-anti-IL-17A (eBio17B7), and FITC-anti-CD4 (RM4-5) (all from eBioscience). For 3 innate lymphoid cell (ILC3) staining, isolated IELs were incubated with the following antibodies: APC-eFluor780-anti-CD19 (eBio1D3), PerCP-eFluor710-anti-CD3 (17A2), PE-anti-ROR gamma (t) (B2D), APC-anti-CD45 (30-F11), and FITC-anti-CD335 (NKp46) (29A1.4) (all from eBioscience). The stained cells were then analyzed by flow cytometry using a fluorescence-activated cell sorting (FACS) Calibur with CellQuest software (BD Biosciences).

For the measurement of SCFAs in mouse feces, 100 mg of fecal matter was first homogenized in 4 volumes of deionized water. Next, the supernatants were collected and an internal standard solution (100 μL of 10 mg/mL propionic acid-2,2-d2, Sigma-Aldrich Co. Ltd, Yongin, Korea), 20 μL of 0.6 N HCl, and 2 mL diethyl ether were added. The upper layer was then collected and the step was repeated as needed. After that, the solutions collected were completely dried under a vacuum and a 100-μL aliquot of 58 mg/mL pentafluorobenzyl bromide in acetonitrile and 20 μL of N,N-diisopropylethylamine were added to the derived matter. The mixtures were vortexed well and incubated at room temperature for 20 minutes. Finally, the sample solutions were vacuum-dried and reconstituted with 100 μL acetonitrile.

The SCFA levels were analyzed using a gas chromatography-mass spectrometry (GC-MS) system (7890A/5975A, Agilent) and HP-5 MS 30 m × 250 μm × 0.25 μm column (Agilent 19091S-433). Helium (99.999%) was used as the carrier gas. The initial temperature was 50°C and was raised to 100°C in 10°C/min increments after a 3-minute hold time. Next, the temperature was then raised to 130°C at 2.5°C/min and then to 230°C at 30°C/min which was maintained for 9 minutes. A 6-minute solvent delay and an SIM mode were applied. The extracted ion chromatogram corresponding to specific SCFAs was used for quantitation.

Significant differences were calculated (GraphPad Prism 5.0, San Diego, CA, USA) using 1-way analysis of variance, followed by Tukey's multiple comparison test. A P value < 0.05 was considered significant.

A mouse model of AD was developed using repeated OVA sensitization and injury as previously described.24 An antibiotic cocktail was administered orally for 2 weeks to induce dysbiosis in the intestinal microbiota of the mice prior to OVA sensitization. Either fecal matter from healthy mice or probiotics (L. rhamnosus) was administered throughout the study to rescue the gut imbalance (Fig. 1).

Antibiotic treatment elicited significant swelling and erythematous skin lesions, as well as increased clinical scores, in the AD model mice (Fig. 1A and B). Significant TEWL was also observed in the antibiotic-treated AD mice (Fig. 1C). The OVA-sensitized groups treated with fecal matter from healthy mice and probiotics showed a significant improvement in clinical signs (Fig. 1A and B) and relevant suppressive pattern on TEWL (Fig. 1C).

Antibiotic-treated AD mice showed a significantly increased total IgE level compared only to the OVA-sensitized animals and this effect was suppressed by treatment with both fecal matter from healthy mice and probiotics (Fig. 2A). The levels of the Th2 cytokine IL4 in the skin were assayed by real-time PCR and were found to be significantly increased in the antibiotic-treated AD mice compared to the control (Fig. 2B).

To investigate whether antibiotics affect the intestinal metabolites processed by microbiota, SCFA levels in the guts of the AD mice were measured by GC-MS. Antibiotic treatment of OVA-sensitized animals led to a significant suppression of the SCFA levels, which re-covered with administration of fecal matter from healthy mice or probiotic treatment (Fig. 3). The effect of each treatment on the production of SCFAs was different. Notably, propionate (C3) and valerate (C5) were found to be significantly associated with the fecal and probiotic treatments in the AD mouse model (Fig. 3).

To investigate whether antibiotics affect the gut mucosal immune cells, we measured the levels of ILC3s and CD4⁺IL17⁺, and CD4⁺FOXP3⁺ regulatory T (Treg) cells in the small intestine of the AD mice by FACS. Antibiotics evoked a significant increase in the number of CD4⁺IL17⁺ cells (Fig. 4A and C) in the AD mice, but not administration of fecal matter from healthy mice or probiotic treatment (Fig. 4B and D). In addition, the CD4⁺FOXP3⁺ cell levels were significantly increased by treatment with fecal matter from healthy mice. The ILC3 number was also significantly increased by antibiotics and suppressed by fecal matter from healthy mice or probiotic treatment (Supplementary Fig. S2). These results suggested that the altered ILC3 and CD4⁺IL17⁺ cell profiles in the antibiotic-treated AD mice were representative of gut dysbiosis in these animals and that CD4⁺FOXP3⁺ cell levels can be used as a marker for the restoration of gut microbial homeostasis.