Six-week-old female NC/Nga mice were purchased from Central Lab Animal Incorporation (Seoul, Korea) and housed under specific pathogen-free conditions with a stable temperature (22±3℃) and humidity (55±15%). After a week of stabilization, the hair on the back was removed using electric clippers and hair removal cream. One day after hair removal, 150 µL of 4% sodium dodecyl sulfate (SDS) was topically applied to disrupt the skin barrier. Two hours later, Biostir AD cream [Dermatophagoides farinae (D. farinae) body extracts, Biostir, Kobe, Japan] was applied to the dorsal surface. The removal of hair, application of SDS, and treatment using Biostir AD cream were repeated twice a week for 8 weeks. NC/Nga mice without dust mite application were used as a control group.

The severity of AD-like skin lesions was measured using the SCORAD index for mice. This index ranges from 0 to 12. In brief, the SCORAD index includes scores based on the presence of erythema or hemorrhage, scarring or dryness, excoriation or erosion, and edema. Each symptom was graded on a scale of 0 to 3 (0, none; 1, mild; 2, moderate; and 3, severe). The score was the sum of individual item scores. The severity of dermatitis was assessed once weekly by two independent researchers.

Total immunoglobulin E (IgE) levels in serum were measured with an enzyme-linked immunosorbent assay (ELISA) MAX™ Deluxe Set (BioLegend, San Diego, CA, USA) in accordance with the manufacturer's instructions. In brief, wells of a 96-well plate were coated with an IgE-specific monoclonal antibody and then incubated overnight at 4℃. Standards and serum samples were added to the plate, which was then incubated at room temperature for 2 h. Captured IgE molecules were detected using biotinylated anti-mouse IgE detection antibody. Avidin-horseradish peroxidase was subsequently added, followed by TMB substrate solution. Absorbance (as optical density) of each well was measured at 450 nm with a microplate reader.

To identify differentially expressed proteins in peripheral induced Treg cells in AD and in naturally derived Treg cells in normal controls, CD4⁺CD25⁺ Treg cells were isolated from thymus tissue of normal mice and the spleens of AD mice. The spleen and thymus were removed from each mouse and disrupted using a syringe plunger to prepare a cell suspension. Blood cells were lysed using blood cell lysis buffer (Sigma-Aldrich, St. Louis, MO, USA) containing 8.3 g/L ammonium chloride and 0.01 M Tris-HCl buffer (pH 7.5±0.2). Murine CD4⁺CD25⁺ T cells were isolated from splenocytes and thymocytes using a CD4⁺CD25⁺ Regulatory T cell Isolation kit (Miltenyi Biotec, Bergisch Gladbach, Germany) according to the manufacturer's protocol. The cells obtained were about 80% pure, as determined via FACS analysis.

Isolated Treg cells were centrifuged at 850×g for 2 min, and the supernatant was then removed. Membrane proteins were extracted using a Mem-PER^TM Eukaryotic Membrane Protein Extraction Reagent Kit (Thermo Scientific, Waltham, MA, USA). In brief, Reagent A was added to the pellet in order to lyse the cells. Two parts Reagent C along with one part Reagent B was then added, and after centrifugation and incubation, the hydrophilic top layer was discarded.

After membrane protein extraction, 100 µg of protein was equally divided into two halves and labeled with two different TMT reagents (Thermo Scientific) following the manufacturer's standard protocol. The membrane protein samples from the spleens of six-week-old AD mice were labeled with TMT-126 and TMT-130, whereas those from the thymuses of normal mice were labeled with TMT-127 and TMT-131. The four samples labeled with different TMT reagents were mixed, dried, and then re-solubilized with water containing 0.5% formic acid for one-dimensional liquid chromatography/tandem mass spectrometry (1DLC/MS/MS) analysis.

The resultant peptides were analyzed using 1DLC-MS/MS. Peptides were identified using MS/MS with a nano-LC-MS system consisting of a Nano Acquity U7PLC system (Waters, Milford, MA, USA) and an LTQ Orbitrap elite mass spectrometer (Thermo Scientific) equipped with a nanoelectrospray source. An autosampler was used to load 5-µL aliquots of the peptide solutions onto a C18 trap-column (i.d. 300 µm, length 5 mm, and particle size 5 µm; Waters). The peptides were desalted and concentrated on the column at a flow rate of 5 µL/min. Then, the trapped peptides were back-flushed and separated on a 200-mm homemade microcapillary column, consisting of C18 matrix (Aqua; particle size 3 µm) packed into 100-µm silica tubing with an orifice i.d. of about 6 µm. The mobile phases, A and B, were composed of 0 and 100% acetonitrile, respectively, and each contained 0.1% formic acid. The LC gradient began with 5% B for 5 min and was increased to 15% B over 5 min, to 50% B over 100 min, and then to 95% B over 5 min, at which point it remained at 95% B for 5 min and then decreased to 5% B for another 5 min. The column was re-equilibrated to 5% B for 15 min before the next run. The voltage applied to produce the electrospray was 2.2 kV. During the chromatographic separation, the LTQ Orbitrap Elite was operated in a data-dependent mode under direct control of Xcalibur software (Thermo Scientific). The MS data were acquired using the following parameters: 10 data-dependent collision-induced dissociation (CID) MS/MS scans per every full scan in label-free mode; 10 data-dependent higher energy collision-induced dissociation (HCD) MS/MS scans per every full scan in TMT; CID scans acquired in LTQ with two-microscan averaging; full scans and HCD scans acquired in Orbitrap at the resolutions of 30000 and 15000, respectively, with two-microscan averaging; 35% normalized collision energy in CID and in HCD; ±1.5-Da isolation window; and dynamic exclusion enabled with a ±1.5-Da exclusion window. All 1DLC-MS/MS analyses for TMT-labeling quantification were performed in duplicate for each sample.

A probability-based (and error-tolerant) protein database search of MS/MS spectra against the latest IPI rat protein database (IPI rat v3.70) was performed using a local MASCOT server (2.3, Matrix Science, London, UK) to identify and quantify the analyzed proteins. The rate of decoy hits in the combined forward and reverse database was less than 1% of the forward hits at both the peptide and the protein levels in each of these experiments. The following search criteria were used: 20 ppm precursor ion mass tolerance; 0.5-Da product ion mass tolerance; two missed cleavages; trypsin as the enzyme; TMT modification at the N-terminus and lysine residues as well as carbamidomethylation at the cysteine residues as static modifications; oxidation at methionine; phosphorylation at serine, threonine, and tyrosine as variable modifications; an ion score threshold of 20; and TMT-6 plex for quantification. Quantification was based on the averaged signal-to-noise ratio of TMT reporter product ions of more than two unique peptides. In TMT experiments, reporter ions for peptide identification were extracted from small windows (±20 ppm) around their expected m/z in the HCD scan. As a single sample was individually labeled with two TMT reagents, peptides with similar ratios in the comparison of the intensity of reporter ions within 30% were selected for protein quantitation. The abundance ratio of a protein was estimated using the ratio between the total intensities of 12 proteins in different reporter ion channels. Given the distributions of protein log₂ ratios, proteins showing ≤-0.4 or ≥0.4 were considered to be differentially expressed.

Human blood samples were obtained from three non-AD healthy controls and from four AD patients who were diagnosed according to the criteria of Hanifin and Rajka.15 The Institutional Review Board approved this study (IRB no: 4-2013-0624), and all subjects provided written informed consent to participate in the study. PBMCs from the subjects were isolated by centrifugation on a Lymphoprep gradient (density 1.077 g/mL) and centrifuged at 800×g for 15 min at 4℃. Cells from the interphase were then washed three times with phosphate-buffered saline (PBS) containing 5 mM EDTA. Isolated PBMCs were used for flow cytometric analysis.

Cells were washed with PBS and stained with a fixable viability dye. After washing, cells were labeled at 4℃ for 30 min with anti-CD3, -CD4, -CD8, and -CD25 antibodies conjugated with fluorescent dye (eBioscience, San Diego, CA, USA), anti-CD47 (eBioscience), and PerCP-Cy5.5 anti-rat secondary antibody. For intracellular labeling, cells were fixed and permeabilized with cytofix/cytoperm buffer (eBioscience) and labeled with anti-human FOXP3 antibody (eBioscience) conjugated with FITC. Labeled cells were quantified using a BD FACSVerse flow cytometer, and the data were analyzed using FlowJo Software (BD Bioscience, San Jose, CA, USA).

After membrane protein extraction as described above, equal amounts of cellular proteins were mixed with 5x sample buffer and heated at 100℃ for 5 min. Proteins were then resolved on an 8% SDS-polyacrylamide gel. After electrophoresis, the proteins were transferred onto an ECL nitrocellulose membrane (GE Healthcare, Buckinghamshire, UK) using Tris buffer [0.025 M Tris-HCI (pH 6.8), 0.192 M glycine, and 20% MeOH]. The membrane was blocked for 1 h at room temperature with 5% skim milk in TBS-Tween 20, incubated overnight at 4℃ with anti-CD47 antibody (BD Bioscience) and anti-GAPDH antibody (Santa Cruz Biotechnology, Santa Cruz, CA, USA), and then incubated with horseradish peroxidase-conjugated antirat secondary antibody (Santa Cruz Biotechnology) for 1 h at room temperature. Finally, the membrane was developed using enhanced chemiluminescence Western blotting detection reagents (Santa Cruz Biotechnology) and quantified via densitometry.

A one-way ANOVA was used to assess the SCORAD results (Fig. 1B) and CD47 expression in NC/Nga mice (Fig. 3B and D). Bonferroni correction was used for post-hoc analysis. A two-way ANOVA was used to assess the results for serum total IgE (Fig. 1C). Pearson correlation was used for the analysis of CD47 expression in AD patients (Fig. 4C). Deviations were considered statistically significant when p<0.05.

Mild erythema was observed following a 2-week application of D. farinae ointment, and significant scarring and crusts were observed after 3 weeks of topical application of D. farinae extracts (Fig. 1A). The SCORAD scores exhibited a rapid and significant increase, with the highest score (7.8±0.73) noted after 6 weeks of D. farinae treatment. After the 6 weeks of D. farinae ointment treatment, a slight improvement in the skin was observed, and the SCORAD score decreased to 5.5±1.19 at 8 weeks of D. farinae topical application (p<0.001) (Fig. 1B).

In accordance with the clinical findings, repeated topical application of D. farinae caused a significant increase in serum IgE levels in NC/Nga mice compared to normal mice. The serum IgE level was 53±8.8 ng/mL after 2 weeks of application, and it increased at the 4- and 6-week time points before reaching a plateau (525.4±12.4 ng/mL after 4 weeks and 679.2±20.1 ng/mL after 6 weeks; p<0.001) (Fig. 1C).

To identify Treg cells specifically involved in AD pathogenesis, we isolated CD4⁺CD25⁺ T cells from AD spleens and normal thymuses with 80% purity. More than 90% of these cells expressed Foxp3, suggesting that these cells were Treg cells (Fig. 1D). Thus, we analyzed the cell surface proteins from these purified cells using TMT-labeling proteomic analysis.

Quantification of 510 proteins was achieved via LC-MS/MS analysis and based on the peak area of precursor ions of identified peptides. Among the 510 quantified proteins, 63 were membrane proteins (Table 1), and 16 were plasma membrane proteins (Table 2). Considering the distributions of the protein log2 ratios, proteins that exhibited values of ≤-0.4 or ≥0.4 were considered to be differentially expressed. These criteria allowed for the identification of six upregulated proteins, including H-2 class II histocompatibility antigen, h-2 class I histocompatibility antigen K-W28 alpha chain-like isoform 1, D-P alpha chain-like isoform 4, protein tyrosine phosphatase receptor type C-associated protein, isoform 3 of receptor-type tyrosine-protein phosphatase C, and isoform 1 of the leukocyte surface antigen CD47 (Table 2) in Treg cells in AD. Among these, CD47 has been reported to be associated with immune regulation, particularly T cell costimulation16 and phagocytosis.1718 Therefore, we further evaluated CD47 expression in Treg cells.

CD47 was expressed in all CD4⁺ T cells and CD8⁺ T cells (Fig. 2A). In addition, 99% of the CD4⁺CD25⁺Foxp3⁺ Treg cells also expressed CD47 (Fig. 2B), and there was no difference between AD mice and control mice. However, the expression level of CD47 was significantly higher in Treg cells from the AD mice than in those from the control mice (Fig. 3A). Moreover, its expression was much higher in Treg cells from the AD spleens than in those from the AD thymuses (Fig. 3B). Increased expression of CD47 in the Treg cells of AD mice relative to controls, especially in the spleen, were also confirmed on Western blot analysis (Fig. 3C and D).

To provide support for the mice data results, the expression of CD47 was also analyzed in human PBMC. All CD4⁺CD25⁺Foxp3⁺ Treg cells expressed CD47 in both AD and healthy controls (Fig. 4A). However, the expression level of CD47 was higher in Treg cells in AD than in healthy controls (Fig. 4B). We also analyzed the correlation of the expression level of CD47 with AD severity. The expression of CD47 in Treg cells showed positive correlation with AD severity measured via EASI score (Pearson correlation r=0.84) (Fig. 4C). Higher levels of CD47 were shown to correlate significantly more severe AD.