All experimental procedures using human cord blood derivatives, including hUCB-MDSCs, were conducted following the guidelines approved by the Korea National Institute for bioethics policy (IRB no. P01-202010-31-008). hUCB-MDSCs were isolated and cultured as previously described ( Figure 1A ) (23, 24).

Five-week-old female NC/Nga mice (Japan SLC, Hamamatsu, Japan) were allowed to acclimatize for 1 week before the experiments. The mice were maintained under specific pathogen-free conditions in the animal care facility of the Catholic University of Korea. The Animal Care and Use Committee of the Research Institute at the Catholic University of Korea approved the experiments (IACUC no.: CUMC-2021-0114-04). Df body ointment was prepared by Biostir, Inc. (Kobe, Japan). One gram of the ointment contained 136.4 mg protein, 234 μg Df 1, and 7 μg Df 2.

Splenocytes from NC/Nga mice were seeded at a density of 1 × 10⁶ cells/well into 24-well plates (Falcon, Corning, NY, USA) and stimulated with CD3/CD28 DYNABEADS (Gibco, Thermo Fisher Scientific, Waltham, MA, USA) to analyze T cell proliferation. hUCB-MDSCs were cocultured with spleen cells at ratios of 0.5:1, 0.25:1, and 0.1:1. Thereafter, cell proliferation was assessed using the Cell Trace CFSE Cell Proliferation Kit (Invitrogen, Thermo Fisher Scientific) per the manufacturer’s instructions.

To analyze T-cell differentiation, purified CD4⁺ T cells from NC/Nga mice were seeded at a density of 3.12 × 10⁵ cells/well into 24-well plates (Falcon) and differentiated using the Th2 differentiation kit (STEMCELL Technologies, Vancouver, Canada) or Th17 differentiation kit (R&D systems, Minneapolis, MN, USA). hUCB-MDSCs were cocultured with CD4⁺ T cells at ratios of 1:1, 0.5:1, and 0.25:1. Th2 and Th17 cell cultures were incubated for 6 and 5 days, respectively, according to the manufacturer’s instructions.

Splenocytes or lymph node (LN) cells were harvested and stained with eFluor780-fixable viability dye (eBioscience), Pacific Blue-anti-CD3e (BioLegend, San Diego, CA, USA), PE-cyanine (Cy)7-anti-CD8 (BioLegend), and PerCP-Cy5.5-anti-CD4 (BioLegend) antibodies. After fixation and permeabilization, the cells were stained with PE-anti-IL-13, PE-anti-IL-22, allophycocyanin-anti-IL-5, PE-Cy7-anti-IL-4, Alexa Fluor 488-anti-IL-17A, and BV785-anti-IFN-γ antibodies (BioLegend). The data were analyzed using FlowJo (Tree Star, Ashland, OR, USA).

AD was induced using Df body ointment (25). The back hair of NC/Nga mice was shaved, and the mice were subjected to skin hair removal treatment (Niclean, Ildong, Korea). The skin barrier was disrupted by treating the shaved dorsal skin and surfaces of both ears with 4% sodium dodecyl sulfate 3 h before the application of Df body ointment (100 mg/mouse). These procedures were repeated two times a week for 4 weeks. After 14 days of the initial Df sensitization, the mice were randomly divided into six groups (n = 5 mice per group): normal (untreated), Df-alone (negative control), Df+Dexa (positive control), Df+hUCB-MDSCs (1 × 10⁴ cells per mouse), Df+hUCB-MDSCs (1 × 10⁵ cells per mouse), and Df+hUCB-MDSCs (1 × 10⁶ cells per mouse). hUCB-MDSCs were intravenously administered to these mice two times a week for 2 weeks. Dexa (2 mg/kg) in water was orally administered to these mice five doses per week for 3 weeks by gavage (oral-zoned needle). Additional information is provided in the Supplementary Materials and Methods .

Erythema/hemorrhage, scarring/dryness, edema, and excoriation/erosion were scored as 0 (none), 1 (mild), 2 (moderate), or 3 (severe) per the observed symptoms. The total skin score was defined as the sum of individual scores (26).

Mouse serum was collected 1 week after the final administration of hUCB-MDSCs, and total IgE concentration was measured using an enzyme-linked immunosorbent assay (ELISA) kit (Yamasa, Tokyo, Japan), per the manufacturer’s instructions. IL-4, IL-5, IL-13, IL-17, TSLP, and TARC levels were measured using ELISA kits (R&D Systems).

The dorsal skin was resected, fixed in 10% formalin solution, and embedded in paraffin. The embedded specimens were then serially sectioned (5 μm) with a microtome (HM 325; Thermo Fisher Scientific) and stained with hematoxylin–eosin to observe the histopathological features or with Masson’s trichrome stain to examine the variable deposition of collagen fibers (blue) and skin fibrosis in the lesioned skin. Mast cells and eosinophils were stained with toluidine blue and Congo red, respectively. Additional information is provided in the Supplementary Materials and Methods .

Axillary LNs from normal or Df-induced AD-NC/Nga mice were isolated, and a single-cell suspension was prepared. Cell proliferation was assessed using the Cell Trace CFSE Cell Proliferation Kit (Invitrogen), per the manufacturer’s instructions. LN cells (1 × 10⁶ cells) were stimulated with 1 mg/mL Df in a 24-well flat-bottom microplate at 37°C for 4 days. hUCB-MDSCs (5 × 10⁵ cells) were co-cultured either directly on LN cells or in a Transwell of pore size 0.4 μm (Costar, Corning Inc., Corning, NY, USA). ARG1 (Nor-NOHA 0.5 mM; Selleckchem, Houston, TX, USA) or iNOS inhibitor (1400W 0.1 mM; Medchem Express, Monmouth, NJ, USA) and hUCB-MDSCs was simultaneously treated. After incubation, the cells were harvested for flow cytometry.

Data were analyzed for statistical significance using Prism 6.0 (GraphPad Software, San Diego, CA, USA). Unless otherwise indicated, a two-sided unpaired Student’s t-test was used to compare two groups, and a one-way analysis of variance with a Dunnett post-hoc test was used for multiple group comparisons. The results are presented as mean ± standard deviation. Results with p < 0.05 were considered significant.

hUCB-MDSCs cultured with hGM-CSF/hSCF for 6 weeks were immune-positive for CD11b, CD33, and CD14, but negative for CD3, CD19, CD34, and CD56 ( Figure 1A ). The levels of ARG1, IDO, and iNOS, the main factors involved in MDSC-mediated immune suppression, were increased in these cells ( Figure 1B ).

The carboxyfluorescein succinimidyl ester (CFSE) dilution assay confirmed that hUCB-MDSCs effectively inhibited human CD4⁺ and CD8⁺ T cell proliferation ( Figure 1C ). hUCB-MDSCs downregulated the differentiation of human CD4⁺ T cells into Th2 and Th17 cells ( Figure 1D ). Furthermore, they markedly inhibited mouse CD4⁺ and CD8⁺ T cell proliferation stimulated with anti-CD3/CD28 beads at a 1:0.5 (mouse T-cell:hUCB-MDSC) ratio ( Figure 2A ). The addition of hUCB-MDSCs at ratios of 1:1 and/or 1:0.5 to Th2- and Th17-polarizing conditions significantly decreased the IL-4, IL-13, and IL-17 levels compared with those of the untreated group ( Figure 2B ). These results demonstrate the in vitro immunosuppressive properties of hUCB-MDSCs against human and mouse T cells.

To investigate the distribution of hUCB-MDSCs, PKH26-labeled hUCB-MDSCs were injected into both atopic and normal control mice and subsequently identified in various organs. Notably, although hUCB-MDSCs were observed in the lungs, spleen, and LNs of both groups, infiltration of hUCB-MDSCs in skin tissue was found only in atopic dermatitis mice ( Supplementary Figure S1 ).

The schematic diagram of the experiment is illustrated in Figure 3A . Dexamethasone (Dexa), an anti-inflammatory and immunosuppressive drug used for treating AD, served as a positive control. The symptom severity scores of the Df+hUCB-MDSC (1 × 10⁵ and 1 × 10⁶ cells) and Df+Dexa groups were considerably improved compared with those of the Df-alone group ( Figure 3B ), whereas no difference was observed compared with those of the Df+hUCB-MDSC (1 × 10⁴ cells) group. When hUCB-MDSCs were injected once or twice, the effect of hUCB-MDSCs (1 × 10⁵ and 1 × 10⁶ cells) was not observed ( Supplementary Figure S2 ). Df-induced skin inflammation in mice presented as epidermal hyperplasia, hyperkeratosis, and lymphocyte infiltration into the epidermis and dermis, which are typical histopathological characteristics of human AD (2, 3). The Df+hUCB-MDSC (1 × 10⁵ and 1×10⁶ cells) and Df+Dexa groups exhibited amelioration of histopathological characteristics ( Figure 4A ). Epidermal thickness considerably decreased in the Df+hUCB-MDSC and Df+Dexa groups compared with that in the Df-alone group ( Figure 4A ). Treatment with Df+hUCB-MDSCs (1 × 10⁵ and 1 × 10⁶ cells) led to a significant decrease in the number of infiltrating mast cells compared with that in the Df+Dexa and Df-alone groups ( Figure 4B ). Furthermore, the extent of decrease in eosinophil counts following Df+hUCB-MDSC treatment was similar to that observed in the Df+Dexa group ( Figure 4C ). Remarkably, even in a Df-stimulated AD animal model under relapse conditions, hUCB-MDSCs exhibited sustained therapeutic efficacy, in contrast to Dexa which rapidly lost its effectiveness, as confirmed by clinical skin scores ( Supplementary Figure S3 ).

Impaired skin barrier function and skin fibrosis are important targets for AD treatment (3, 27). The proportion of dermal collagen matrix considerably increased in the Df-alone group compared with that in the normal group. However, compared with the Df-alone group, the Df+hUCB-MDSC (1 × 10⁵ and 1 × 10⁶ cells) and Df+Dexa groups showed a markedly reduced proportion of dermal collagen matrix ( Figure 5A ). In both Dexa and hUCB-MDSC groups, the epidermal protein levels were higher than those in the Df-alone group ( Figure 5B ). Notably, the levels of epidermal proteins such as FLG, IVL, LOR, and CK10 were significantly higher in the Df+hUCB-MDSC (1 × 10⁶ cells) group than in the positive control group treated with Dexa ( Figure 5B ). Overall, hUCB-MDSCs display potential as an alternative treatment strategy to avoid the adverse effects of Dexa-associated skin atrophy.

The total serum IgE level was considerably lower in the Df+hUCB-MDSC (1 × 10⁵ and 1 × 10⁶ cells) and Df+Dexa groups than in the Df-alone group; however, no notable difference was observed between the Df+hUCB-MDSC (1 × 10⁴ cells) and Df-alone groups ( Figure 6 ). The Th2 (IL-4, IL-5, and IL-13) and Th17 cytokine (IL-17) levels markedly decreased in the Df+hUCB-MDSC (1 × 10⁵ and/or 1 × 10⁶ cells) and Df+Dexa groups compared with those in the Df-alone group. Moreover, the serum thymic stromal lymphopoietin (TSLP) and thymus- and activation-regulated chemokine (TARC) levels considerably decreased in the Df+hUCB-MDSC (1 × 10⁵ and/or 1 × 10⁶ cells) and Df+Dexa groups. Conversely, there was no considerable difference between the Df+hUCB-MDSC (1 × 10⁴ cells) and Df-alone groups in terms of TSLP and TARC levels.

The spleen weight and size of the Df-alone group were notably higher than those of the normal control group. Compared with the Df-alone group, the Df+hUCB-MDSC (1 × 10⁵ and 1 × 10⁶ cells) and Df+Dexa groups showed considerably reduced spleen weight and size ( Figure 7A ). Analysis of effector CD4⁺ T cell subsets in the mouse spleen revealed that the number of IL-4-, IL-5-, IL-13-, and IL-17-producing CD4⁺ T cells in the splenocytes of Df-induced NC/Nga mice was notably lower in the Df+hUCB-MDSC (1 × 10⁵ cells and/or 1 × 10⁶ cells) group than in the Df+hUCB-MDSC (1 × 10⁴ cells) and Df-alone groups ( Figure 7B ). Additionally, IL-5 production was decreased in the Df+hUCB-MDSC (1 × 10⁴ cells) group, whereas IFN-γ production was increased in the Df+hUCB-MDSC (1 × 10⁵ cells) group.

The proportion changes of effector CD4⁺ T cell subsets caused by hUCB-MDSCs in the LNs are illustrated in Supplementary Figure S4 . The number of IL-4-, IL-5-, IL-13-, and IL-17-producing CD4⁺ T cells in the LNs of Df-induced NC/Nga mice was decreased in the Df+hUCB-MDSC (1 × 10⁵ cells and 1 × 10⁶ cells) and/or Df+hUCB-MDSC (1 × 10⁴ cells) groups, whereas IFN-γ production was increased in the Df+hUCB-MDSC (1 × 10⁵ cells and 1 × 10⁶ cells) groups.

Restimulation of lymph node cells collected from mice with Df-induced AD increased the proliferation of CD4⁺ and CD8⁺ T cells, and hUCB-MDSCs downregulated the proliferation ( Figure 8A ). However, the Transwell assay revealed that hUCB-MDSCs no longer exhibited immunomodulatory functions in the context of blocked cell-cell interactions, suggesting that secreted immunosuppressive substances had little effect on AD mice ( Figure 8A ). We validated our results using iNOS and ARG1 inhibitors (Nor-NOHA, 1400W), which also did not affect the immunosuppressive function of hUCB-MDSCs ( Figure 8A ). The same results were observed in a subset of CD4⁺ T cells ( Figure 8B ). These findings suggest that the suppressive function of hUCB-MDSCs in AD is primarily dependent on cell–cell interactions rather than on secreted immunosuppressive substances.