CD1dhiPD-L1hiCD27+ Regulatory Natural Killer Subset Suppresses Atopic Dermatitis

Effector and regulatory functions of various leukocytes in allergic diseases have been well reported. Although the role of conventional natural killer (NK) cells has been established, information on its regulatory phenotype and function are very limited. Therefore, the objective of this study was to investigate the phenotype and inhibitory functions of transforming growth factor (TGF)-β-producing regulatory NK (NKreg) subset in mice with MC903-induced atopic dermatitis (AD). Interestingly, the population of TGF-β-producing NK cells in peripheral blood monocytes (PBMCs) was decreased in AD patients than in healthy subjects. The number of TGF-β+ NK subsets was decreased in the spleen or cervical lymph node (cLN), but increased in ear tissues of mice with AD induced by MC903 than those of normal mice. We further observed that TGF-β+ NK subsets were largely included in CD1dhiPD-L1hiCD27+ NK cell subset. We also found that numbers of ILC2s and TH2 cells were significantly decreased by adoptive transfer of CD1dhiPD-L1hiCD27+ NK subsets. Notably, the ratio of splenic Treg per TH2 was increased by the adoptive transfer of CD1dhiPD-L1hiCD27+ NK cells in mice. Taken together, our findings demonstrate that the TGF-β-producing CD1dhiPD-L1hiCD27+ NK subset has a previously unrecognized role in suppressing TH2 immunity and ILC2 activation in AD mice, suggesting that the function of TGF-β-producing NK subset is closely associated with the severity of AD in humans.


INTRODUCTION
Atopic dermatitis (AD) is known as a chronic inflammatory skin disease. It is referred to as atopic eczema with typical symptoms such as itchy, red, swollen, and cracked skin lesions. Although AD is widespread and on the rise in developed countries, the exact pathological mechanism of AD is not fully understood yet (1)(2)(3). Currently, it is a typical type 2 helper T (T H 2) cell-mediated hypersensitive immune disorder in which various immune cells are known to participate in the development of skin inflammation (4). In AD, T H 2 cells secrete IL-4, IL-5, and IL-13 known to promote allergic responses (5,6). These T H 2 cytokines can stimulate IgE production from B cells, activate mast cells, and lead to infiltration of eosinophils or other immune cells into inflamed tissues (7). Recent studies have also reported that type 2 innate lymphoid cells (ILC2) have the function of T H 2 cells in peripheral tissues. ILC2 is also well known to initiate and participate in T H 2 cell-mediated responses by secreting T H 2 signature cytokines such as IL-5 and IL-13 in peripheral tissues (8,9).
Natural killer (NK) cells are well known as a type of antimicrobial lymphocytes in innate immunity (10,11). NK cells can release proteolytic enzymes and interferon (IFN)-g to remove virus-infected cells, intracellular bacteria, and tumor cells (12,13). NK cells can be classified into several subsets, depending on the profile of cytokine secretion and their function (14). Typically, NK1 cells (also called conventional NK cells) secrete IFN-g and NK2 cells secrete T H 2 cytokines (15)(16)(17)(18). Subsets of IL-17 secretion (NK17) or IL-22 secretion (NK22) NK cells have also been reported (19)(20)(21). Compared to the past classification of the NK subset, various innate lymphoid cell types have been recently introduced, and in particular innate lymphoid cells (ILCs) have been proposed as representative helper innate immune cells (22). In addition, various T cell-lineage subsets have recently been defined in innate T cells such as NKT or gd T cells that partially share receptors with conventional NK cells (23,24). Therefore, it is necessary to more clearly distinguish the classification of NK. Accumulating evidences have demonstrated that some types of NK cells have a suppressive function like regulatory T (Treg) cells by secreting IL-10 or TGF-b in transplantation, pregnancy, and some infections (25)(26)(27)(28)(29)(30). Although the role of NK cells in allergic diseases including AD is poorly understood, recent studies have reported that the number of circulating NK cells in blood samples of AD patients are generally decreased, but increased in inflammatory skins (31)(32)(33)(34). However, it remains unclear which NK subset secretes suppressive cytokines.
In this study, we demonstrated that the population of TGF-b + NK cells was decreased in human PBMC and lymphoid tissues from mice with AD than in healthy control. We further found that TGF-b + NK cells were largely included in CD1d hi PD-L1 hi CD27 + NK cell subset. Of interest, AD severity was relieved after an adoptive transfer of CD1d hi PD-L1 hi CD27 + NK cell subset in mice by inhibiting T H 2 immunity.

Human TGF-b + NK Cell Analysis
Patients were treated at the Department of Allergy, Allergy and Clinical Immunology Center, Cheju Halla General Hospital (Jeju, Korea) between October 2017 and May 2018. Subjects underwent blood tests and skin prick tests as described below and fulfilled the criteria of Hanifin and Rajka (1). The subjects were selected at random regardless of age or sex, and was classified based on the SCORAD index, the amount of IgE, and the number of eosinophils in the blood ( Table 1). This study was approved by the Institutional Review Board of Jeju Halla General Hospital (approval number: CHH-2016-L13-01).

Induction of MC903-Mediated Atopic Dermatitis Model
C57BL/6 (8 to 10 weeks old) female mice were purchased from Orient Bio (Gyeonggi-do, Korea). For MC903 treatment, mice were painted with 2 nmol of MC903 (calcipotriol, Tocris Bioscience, Minneapolis, MN) in 20 mL of ethanol on both ear for 12 consecutive days. At 24 hours after treatment, mice were euthanized and their lymphoid tissues were isolated for flow cytometric analysis. All mice were housed in a pathogen-free facility at Konkuk University (Seoul, Korea). All animal experiments were approved by the Institutional Animal Care and Use Committee (IACUC) of Konkuk University.

Statistical Analysis
Data are expressed as mean ± standard error of the mean (SEM) from three or more independent in vitro or in vivo experiments. All statistical analyses were performed with Student's t-test or one-way analysis of variance (ANOVA) with Tukey's post hoc test. Statistical significance (*p < 0.05 and **p < 0.01) was determined with a GraphPad Prism 7.0 software (GraphPad Inc., San Diego, CA, USA).

Alteration of Population of TGF-b-Producing NK Cells in Mice With Atopic Dermatitis
Accumulating evidences have indicated that NK cells are closely associated with AD progression in humans (37). We further investigated whether TGF-b + NK cells might be involved in AD progression in humans through peripheral blood monocytes (PBMCs) from healthy controls and AD patients ( Table 1). As reported, the population of total NK cells in PBMCs was decreased in AD patients than in healthy controls ( Figures 1A, B).
We also found that the population of TGF-b + NK cells was reduced in PBMCs from AD patients ( Figure 1C). To further characterize the potential role and mechanism of TGF-b + NK cells to control AD symptom, we first checked population changes of TGF-b + NK cells in normal mice ( Figure 1D). We found that the frequency of splenic TGF-b1/latency associated peptide (LAP) + (as TGF-b1) NK cells was higher in NK cells than in T cells or non-NK/T cells ( Figure 1D). Interestingly, the frequency of LAP + NK cells was decreased in spleen and cLN while the population of LAP + NK cells was increased in ear tissues of mice with AD compared to that in normal mice ( Figures 1E, F). We further observed that the expression of TGF-b mRNA from splenic or ear NK cells was changed in mice with AD compared to that in normal mice ( Figure 1G). On the other hand, the population of LAP + CD3 + T cells did not show a significant change in spleen, cLN, and ear according to the development of AD (Supplementary Figure 1). Additionally, we also evaluated the expression of another anti-inflammatory cytokine, IL-10, but this also did not show any significant difference under normal or AD mice in vivo (Supplementary Figure 2A). Taken together, these results confirm that the tissue-specific population of LAP + NK cells has a very closely associated with the disease development of AD mice.

Identification of Surface Phenotype for TGF-b + NK Cells
To find unique phenotypical surface makers of TGF-b + NK cells, expression levels of potential NK cell surface makers in LAP + and LAP − splenic NK cells were compared. As shown in Figure 2, expression levels of CD1d, CD2, CD18, CD27, CD49b, PD-L1, NKG2D, and MHCII were increased in LAP + NK cells than in LAP − NK cells ( Figure 2). It has been reported that the CD27 + NK cell subset is a unique subset for the production of effector cytokines and that CD11b + NK cell subset has a strong cytotoxicity (38,39). In our results, LAP + NK cells showed high CD27 expression and low CD11b expression compared to LAP − NK cells. Therefore, LAP + NK cells might be capable of producing cytokines rather than causing cytotoxicity in NK subsets. In addition, we found that expression levels of CD1d and PD-L1 in LAP + NK cells were higher than those in LAP − NK cells ( Figure 2C).

TGF-b + NK Cells Are Largely Included in
CD1d hi PD-L1 hi CD27 + NK Cell Subset The above results prompted us to investigate whether the development of TGF-b + NK cells might be associated with expression levels of CD1d, CD27, and PD-L1 on NK cells. CD27 was highly expressed on LAP + NK cells than on LAP − NK cells (Figures 3A, B). The population of CD1d high and PD-L1 high NK cell subset was observed in 28.7 ± 1.5% of LAP + CD27 + NK cells ( Figure 3C). Additionally, we analyzed the population of LAP + NK cells in other NK cell subsets such as CD1d hi PD-L1 h i CD27 + NK, CD1d l o PD-L1 l o CD27 + NK, CD1d h i PD-L1 hi CD27 -NK, and CD1d lo PD-L1 lo CD27 -NK subsets. Among them, the CD1d hi PD-L1 hi CD27 + NK subset had the highest frequency (32.6 ± 2.0%) of LAP + NK cells in healthy control mice ( Figures 3D, E). These results suggest that the expression of CD27, CD1d, and PD-L1 is closely associated with TGF-b production in NK cells.

Populations of TGF-b + NK Cells Are Correlated With the CD1d hi PD-L1 hi CD27 + NK Subset in Peripheral Tissues
In the above phenotypical analysis, we found that CD1d hi PD-L1 hi CD27 + NK subset had the highest expression than other NK subsets ( Figure 3). Next, we checked how unique CD1d hi PD-L1 hi CD27 + NK subsets were for TGF-b production compared to other NK subsets. In the experiment using CD1d hi PD-L1 hi CD27 + and CD1d lo PD-L1 lo CD27 − NK subsets ( Figure 4A), we further found that the expression of TGF-b was much higher in CD1d hi PD-L1 hi CD27 + NK subsets than in CD1d lo PD-L1 lo CD27 − NK subsets ( Figure 4B).
Next, we tested whether the expression of TGF-b in NK cells was associated with population change of CD1d hi PD-L1 hi CD27 + NK subset in AD mouse model. Consistent with the above results ( Figures 1E, F), we further discovered that the frequency and number of CD1d hi PD-L1 hi CD27 + NK subsets were also decreased in the spleen and cLN but increased in ears of mice with AD ( Figures 4C-F). As in the proportion of total TGF-b + NK cells in Figure 1F, the proportion of CD1d hi PD-L1 hi CD27 + NK subsets was also decreased in lymphoid tissues and increased in the target skin lesions. TGF-b production in CD1d hi PD-L1 hi CD27 + NK subsets was also synergistically changed. These results indicate that population changes of TGF-b + NK cells are closely associated with those of CD1d hi PD-L1 hi CD27 + NK subsets in mice with AD. Like a Supplementary Figure 2A inflammatory cytokine, in the suggested CD1d hi PD-L1 hi CD27 + NK subsets, but showed no AD-dependent changes in each lymphoid tissue (Supplementary Figure 2B). Altogether, these results suggest that the CD1d hi PD-L1 hi CD27 + NK subset contains a large portion of TGF-b-producing NK subset.
CD1d hi PD-L1 hi CD27 + NK Subsets Suppress Symptoms of AD via

Suppression of ILC2s in Mice
There are three types of innate lymphoid cells (ILCs) such as type 1, type 2, and type 3 ILCs (40,41). Among them, type 2 ILC is well recognized to be able to induce allergic inflammation by secreting IL-4, IL-5, and IL-13 as by T H 2 cells in various allergic responses (42)(43)(44). In particular, ILC2 has been accepted as a major effector cell type for the secretion of IL-5 and IL-13 in MC903-induced AD mice (8,9). Hence, we checked whether CD1d hi PD-L1 hi CD27 + NK subset could suppress the population of ILC2 in the AD mouse model. To test this, we adoptively transferred CD1d hi PD-L1 hi CD27 + NK subsets or CD1d lo PD-L1 lo CD27 − NK subsets as the control into MC903-induced AD mice. Adoptive transfer of CD1d hi PD-L1 hi CD27 + NK subsets but not CD1d lo PD-L1 lo CD27 − NK subsets largely suppressed symptoms of AD in mice ( Figure 5A). The thickness for representing ear swelling was reduced largely by the transfer of CD1d hi PD-L1 hi CD27 + NK subsets, but not by CD1d lo PD-L1 lo CD27 − NK subsets ( Figure 5B). These results suggest that TGF-b-producing CD1d hi PD-L1 hi CD27 + NK subsets play a pivotal role in the inhibition of MC903-induced AD symptoms. Besides, adoptive transfer of CD1d hi PD-L1 hi CD27 + NK subsets but not CD1d lo PD-L1 lo CD27 − NK subsets significantly inhibited numbers of IL-13 + ILC2s in spleen, cLN, and ear tissues of AD mice (Figures 5C-E). In more detail, these results showed that although the distribution of ILC2 in each tissue was suppressed by administration of CD1d hi PD-L1 hi CD27 + NK subsets ( Figure 5C), IL-13 expression in ILC2 was not restricted ( Figure 5D). These results suggested that the adoptive transfer effect of the CD1d hi PD-L1 hi CD27 + NK subset resulted from the suppression of the increase in the number of IL-13 + ILC2 in peripheral tissues of mice with AD ( Figure 5E).

CD1d hi PD-L1 hi CD27 + NK Subsets Suppress the Development of Atopic Dermatitis Through Inhibition of T H 2 and Maintenance of Treg in T Cell Immunity
To determine how CD1d hi PD-L1 hi CD27 + NK subsets affected the development of AD, we observed changes in the population of IL-4 + T H 2 and Foxp3 + Tregs cell in CD1d hi PD-L1 hi CD27 + NK subsets transferred AD mice. We found that the number of T H 2 cells was significantly reduced in CD1d hi PD-L1 hi CD27 + NK subsets transferred AD mice compared to that in CD1d lo PD-L1 lo CD27 − NK subsets transferred AD mice ( Figures 6A, B). It is generally accepted that TGF-b can induce Treg cells activity in murine and human (45). Thus, we further tested whether the effect of adoptive transferred CD1d hi PD-L1 hi CD27 + NK subsets influenced the change of Treg cell population. Results showed that adoptive transfer of CD1d hi PD-L1 hi CD27 + NK subsets did not cause any change in the number of Treg cells in spleen or cLN ( Figure 6B). However, reduced number of Treg cells was observed in ear tissues after adoptive transfer of CD1d hi PD-L1 hi CD27 + NK subsets ( Figures 6C, D) probably caused by a decrease in total CD4 + T H cells in skin lesions. It seems to be a phenomenon that the infiltration of immune cells into the target tissue is reduced as much as the reduction of AD aggravation. We also found that the population and number of spleen-derived CD4 + T H cells were not different ( Figure 6E). However, the ratio of Treg cells per effector T H 2 cells was much higher after the adoptive transfer of CD1d hi PD-L1 hi CD27 + NK subsets compared to that of CD1d lo PD-L1 lo CD27 − NK subsets ( Figure 6F). Furthermore, we applied a TGF-b neutralizing antibody to AD mice to check whether the inhibitory effect of the CD1d hi PD-L1 hi CD27 + NK subsets was indeed TGF-bdependent in vivo. When the TGF-b neutralizing antibody was treated, the inhibitory effect of the adoptively transferred NK subsets was restored in the ear swelling of the MC903-induced AD mouse model ( Figure 6G). Altogether, these results suggest that CD1d hi PD-L1 hi CD27 + NK subsets can suppress T H 2 cells development in MC903-induced AD mouse model and affect the balance of inflammatory or regulatory T cells. And it was confirmed that it was TGF-b-dependent to limit the exacerbation of AD mouse model.

DISCUSSION
NK cells are well recognized as innate immune cells that participate in various immune responses to viral infections or cancers by exhibiting anti-viral or anti-tumor effects (46)(47)(48).
Although NK cells are lymphocytes, like other types of innate lymphoid cells (ILCs), they are generally involved in both innate immunity and adaptive immunity. NK cells are fundamental immune cells that play important roles in the initiation of immune responses and body homeostasis (49). They usually secrete signature cytokine such as IFN-g. They are involved in the formation of T H 1 immunity in the body. They are known to play a crucial role in eliminating infected or tumor cells through secretion of intrinsic digestive enzymes (50,51). Previous studies have shown that the frequency of NK cells in the body is generally decreased in various cancer diseases. This trend is considered to be due to the immune escape mechanism of cancer cells from host immune surveillance (52,53). Recent studies have reported that NK cells are decreased in PBMCs from AD patients (34). This trend predicts that maintaining the balance of the number of NK cells and the immune circumstance will be important for the pathogenesis of allergy diseases such as AD. Our results also showed that the frequency of NK cells (CD56 + CD3 − ) in the blood of AD patients was reduced ( Figures 1A, B). Previous studies have found that NK cells are decreased in the blood of atopic dermatitis patients but increased in AD lesions, suggesting that NK cells can migrate to control peripheral T H 2 responses (34). However, the definite cause and mechanism of the alteration of NK cells in AD remain unclear. Therefore, we focused on another perspective to understand NK cells in atopic dermatitis. In addition to the well-known classical function, NK cells are known to have a subset of regulatory functions like other immune cells (14,54).
In the past, researchers have paid attention to the etiological aspect caused by changes in the distribution and activity of a subset of specific immune cells due to breakdown of the balance of immune state in the body. In the 2000s, it was found that some sub-phenotypes of immune cells, including regulatory T cells (Tregs), were regulatory subsets that could restore the body's immune status to normal through immunomodulation or induction of immune tolerance (55). Previous studies have shown that T cells are Tregs, B cells are regulatory B cells (Bregs), monocyte/neutrophils are myeloid-derived suppressor cells (MDSCs), macrophages are M2 or alternative macrophages, dendritic cells (DCs) are tolerogenic DCs, and ILCs are regulatory ILCs. Thus, a regulatory phenotype has been reported for each immune cell (56)(57)(58)(59)(60)(61). These regulatory immune cells can differentiate to have their own characteristic functions by secreting anti-inflammatory cytokines such as IL-10 and TGF-b to control pro-inflammatory cells (62). It is also known that NK cells have a regulatory type (called NKreg or NK3) (14). In addition to dividing effector functions into NK1 or NK2 cells according to inflammatory types, it has been proposed that secreting IL-10 or TGF-b can control excessive inflammatory responses (27,28). It has been reported that NK cells can secrete anti-inflammatory cytokines such as IL-10 and TGF-b through secretion profiling, although there are only a small population of NK cells in the whole body. Especially, IL-10-producing NK cells are known to control the activation of T cells while Prf1 −/− mice-derived IL-10 + NK cells regulate CD8 + T cells and contribute to immunological maintenance in mouse cytomegalovirus (MCMV) infection (25,26). In human studies, a regulatory NK subset that can secrete IL-10 and TGF-b has been reported (26,27). It has been confirmed that IL-10 and TGF-bproducing NK cells exist in peripheral blood mononuclear cells and decidua in pregnant women. In particular, TGF-bproducing NK cells are significantly increased in decidua. It has been suggested that these regulatory NK cells contribute to pregnancy tolerance (26,27). The present study showed that TGF-b-producing NK cells were present in human PBMCs. It also revealed that not only NK cells, but also TGF-b + NK cells were decreased in allergic patients ( Figure 1C). We employed MC903, a representative atopic dermatitis mouse model, to find changes in TGF-b + NK cells in atopic dermatitis mice. Although there was a difference in the total number of lymphocytes, interestingly, TGF-b expression was higher in NK cells than in other lymphocytes of mouse spleen. Thus, we could predict that NK cells are an important source of TGF-b ( Figure 1D). After MC903 treatment, TGF-b + NK cells were significantly reduced in mouse spleen and cervical LN (cLN), a draining lymph node (LN), whereas these cells were increased in ear tissues, a peripheral target site of the disease (Figures 1E-G). As mentioned above, these patterns are similar to the migration pathway of classical IFN-g + NK cells in atopic dermatitis (34).
Previous studies have predicted that NK cells also have a regulatory type that can secrete anti-inflammatory cytokines such as IL-10 and TGF-b and that they are expected to be involved in immune tolerance or regulation (14). However, elucidation of the characteristic phenotype of NKreg is still insufficient. Through this study, we analyzed the expression of major receptors in mouse TGF-b + NK cells ( Figure 2) and found high TGF-b expression (32.6 ± 2.0%) in CD1d hi PD-L1 hi CD27 + NK subsets (Figure 3). However, it is necessary to evaluate whether our proposed TGF-beta-producing NK subsets overlap with T cells, NKT cells, gdT cells or helper ILC1s. Therefore, as a result of the confirmation, it can be seen that the TGF-betaproducing NK subsets is CD3 − CD49a − CD49b + Eomes + T-bet + conventional NK cells (Supplementary Figure 3).
We proposed that these highly TGF-b-expressing NK cells were splenic CD1d hi PD-L1 hi CD27 + subsets. Several human NK cell studies have suggested an antigen-presenting role of NK cells. CD1d is a non-polymorphic, MHC class I-like molecule (63,64). It is well known that CD1d usually presents antigens such as glycolipids, including a-galactosylceramide (a-GC), to CD1d-restricted NKT cells (65). CD1d is mostly expressed in innate immune cells such as DC, macrophages, B cells, and ILCs. In our study, it was confirmed that CD1d expression of TGF-b + NK cells was high than TGF-b − NK cells (66,67). PD-L1 is well known to be expressed in NK cells as a representative immunosuppression marker (68). The murine CD27 expressing NK cell is exhibits potent cytokine production and high migratory capacity (39). Their correlation with the distribution pattern of TGF-b-producing NK subsets in AD was also determined. As a result, with AD development, CD1d hi PD-L1 hi CD27 + NK subsets were decreased in lymphoid organs such as spleen and cLN, but markedly increased in the ear ( Figure 4). Based on these results, we can explain that the CD1d hi PD-L1 hi CD27 + NK1.1 + CD3 − phenotype in mice is due to TGF-b-producing regulatory NK subsets. In addition, the distribution of CD1d hi PD-L1 hi CD27 + NK subsets was decreased in spleen and draining LN from AD mouse model, suggesting that this distribution might be increased in skin lesions where type 2 inflammation appears.
This study not only suggested a phenotype of TGF-b + NK cells, but also confi rmed that the proposed TGFb + CD1d hi PD-L1 hi CD27 + NK subsets could regulate type 2 i n fl a m m a t i o n l i k e A D . T o d e m o n s t r a t e t h e immunomodulatory effect of TGF-b-producing CD1d hi PD-L1 hi CD27 + NK subset in vivo, we separated CD1d hi PD-L1 hi CD27 + NK subsets and CD1d lo PD-L1 lo CD27 − NK subsets and adaptively transferred before MC903 induction of AD using a mouse model. Compared to vehicle mice (MC903+PBS i.v.) and CD1d lo PD-L1 lo CD27 − NK subsets transferred mice, significant inhibitory effects were observed in CD1d hi PD-L1 hi CD27 + NK subsets ( Figures 5A, B).
Type 2 innate lymphoid cells (ILC2s) have recently been proposed as important effector cells in allergic responses. They are involved in peripheral allergic conditions through secretion of IL-4, IL-5, and IL-13 (42)(43)(44). We tested how CD1d hi PD-L1 hi CD27 + NK subsets could affect the activity of ILC2 in the AD mouse model and found a decrease of systemic ILC2 (Figures 5C-E).
AD is well known as a typical chronic allergic disease. The classical immune system is initiated as allergen, resulting in a hypersensitivity reaction mediated by innate immune cells followed by a T H 2-cell mediated chronic inflammatory response (4). It is difficult to adequately overcome atopic dermatitis by controlling only the initial hypersensitivity reaction by innate immune cells. Thus, we tested how TGF-bproducing CD1d hi PD-L1 hi CD27 + NK subsets could affect the activity of T H 2 effector cells in AD through in vivo adaptive transfer. Results confirmed that the addition of TGF-bproducing NK subsets induced a decrease in T H 2 cells without controlling the number of total T cells in lymphoid organs, thereby blocking the activity of T cells polarizing with T H 2 ( Figure 6).
It is well known that TGF-b can induce T cells into Foxp3 + regulatory T cells (iTregs) among several immunomodulatory functions (36,69). In addition, Treg cells are the most representative immunomodulatory cells known to regulate various allergic disorders such as atopic dermatitis through several studies (70,71). Therefore, we considered whether the control of T H 2-mediated inflammatory responses such as ILC2s and T H 2 cells by CD1d hi PD-L1 hi CD27 + NK subsets could be correlated with Foxp3 + Treg cells. In our results, CD1d hi PD-L1 hi CD27 + NK subsets did not directly control the number of Foxp3 + Treg cells in lymphoid tissues compared to its inhibitory effect on T H 2-mediated inflammatory cells. On the other hand, the number of Foxp3 + Treg cells was decreased in in the ear from the CD1d hi PD-L1 hi CD27 + NK-treated group (Figures 6C, D). It was found that Foxp3 + Treg cells were maintained by inhibiting T H 2 cells activity without changing the distribution or the number of CD4 + T cells in the spleen. Therefore, administration of CD1d hi PD-L1 hi CD27 + NK subsets in an ADinduced state can increase the ratio of Foxp3 + Treg cells compared to PBS control ( Figures 6E, F). The decrease of Foxp3 + Treg cells in CD1d hi PD-L1 hi CD27 + NK subsets administered ear appeared to result in reduced infiltration of CD4 + cells in ears. It can be seen that the disease improvement effect of CD1d hi PD-L1 hi CD27 + NK subsets administration lies in the inhibition of infiltration of peripheral CD4 + T cells rather than the induction of an increase of peripheral Treg cells. According to this, CD1d hi PD-L1 hi CD27 + NK subsets mainly controls T cell-mediated inflammatory responses in lymphoid tissues. Thus, treatment with CD1d hi PD-L1 hi CD27 + NK subsets decreased T H 2 but increased the ratio of Treg cells in the spleen and draining lymph node.
In conclusion, we found that CD1d hi PD-L1 hi CD27 + was a unique TGF-b-producing NK subset and that treating such regulatory subset in a mouse AD disease model inhibited T H 2mediated effector cells and helped improve disease exacerbation. Although more diverse mechanism studies and mutual evaluation in human studies are needed, results of this study suggest that NK cell-derived regulatory subset can be used in various ways as a novel immune disease treatment strategy.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding authors.