Atopic dermatitis induces the expansion of thymus‐derived regulatory T cells exhibiting a Th2‐like phenotype in mice

Abstract Atopic dermatitis (AD) is a widespread inflammatory skin disease with an early onset, characterized by pruritus, eczematous lesions and skin dryness. This chronic relapsing disease is believed to be primarily a result of a defective epidermal barrier function associated with genetic susceptibility, immune hyper‐responsiveness of the skin and environmental factors. Although the important role of abnormal immune reactivity in the pathogenesis of AD is widely accepted, the role of regulatory T cells (Tregs) remains elusive. We found that the Treg population is expanded in a mouse model of AD, i.e. mice topically treated with vitamin D3 (VitD). Moreover, mice with AD‐like symptoms exhibit increased inducible T‐cell costimulator (ICOS)‐, cytotoxic T‐lymphocyte antigen‐4 (CTLA‐4)‐ and Glycoprotein‐A repetitions predominant receptor (GARP)‐expressing Tregs in skin‐draining lymph nodes. Importantly, the differentiation of Tregs into thymus‐derived Tregs is favoured in our mouse model of AD. Emigrated skin‐derived dendritic cells are required for Treg induction and Langerhans cells are responsible for the biased expansion of thymus‐derived Tregs. Intriguingly, thymus‐derived Tregs isolated from mice with AD‐like symptoms exhibit a Th2 cytokine profile. Thus, AD might favour the expansion of pathogenic Tregs able to produce Th2 cytokines and to promote the disease instead of alleviating symptoms.


Introduction
Atopic dermatitis (AD) is one of the most common inflammatory skin conditions, predominantly affecting infants and children. It is characterized by pruritus, eczematous lesions and skin dryness. Atopic dermatitis is a complex chronic relapsing inflammatory skin disorder involving immune hyper-responsiveness of the skin, epidermal barrier abnormalities, genetic susceptibility and environmental factors [1]. In its acute phase, AD is characterized by an abnormal production of thymic stromal lymphopoietin (TSLP), an alarmin secreted by keratinocytes and leading to expansion of Th2 cells via activation of Langerhans cells (LCs) [2][3][4]. Moreover, interleukin (IL)-17 has been shown to be present in acute AD [5]. A Th1/Th22 predominant immune response, dermal infiltration with inflammatory dendritic epidermal cells (IDECs), macrophages and eosinophils as well as bacterial superinfection are classical features of the chronic phase of AD [2].
Regulatory T cells (T regs ) play a critical role in the maintenance of peripheral tolerance and in the control of allergic responses. Despite a growing interest in the role of T regs in the pathogenesis of AD, their precise role remains unclear. In both humans and mice, loss-of-function mutations in the FoxP3 gene lead to a multiorgan inflammatory response including skin inflammation resembling AD associated with elevated serum IgE levels, eosinophilia, allergic airway inflammation, food allergies and other autoimmune symptoms [6,7]. These data suggest that lack of functional T regs is sufficient to recapitulate important immunologic features of AD. Accordingly, lower circulating T regs at birth and lower T reg numbers in cord blood predict a higher risk for the development of AD in the first year of life [8]. Furthermore, analysis of thymic tissue from atopic children revealed significantly delayed maturation of thymic T regs as compared to age-matched, non-atopic controls [9]. However, studies in adults seem to dismiss this hypothesis. Some authors reported equal levels of circulating T regs [10][11][12], whereas others found increased circulating T regs , directly correlating with AD disease severity in patients with persisting AD in adulthood as compared to healthy controls [13][14][15][16]. Furthermore, there are con-flicting reports about the presence of T regs in inflammatory infiltrates of AD skin [10,17,18], and controversy exists about the immune suppressive capacity of T regs in AD. T regs from AD patients were shown to exhibit normal suppressive activity at baseline [11,15] or after allergen-stimulation [19], but T regs exhibiting reduced suppressive function have also been identified in patients with AD [13]. Intriguingly, stimulation with staphylococcal superantigen (staphylococcal enterotoxin B) leads to a Th2-dominated cytokine profile in circulating CCR6 À T regs of AD patients [12,14] and there is emerging evidence that T regs can convert to Th2 cells, thereby contributing to AD instead of dampening the immune response [20].
Because the T reg population remains poorly characterized in AD, we here studied the phenotype and the dynamics of thymus-derived versus peripherally derived T regs . Dendritic cells (DCs) are professional antigen-presenting cells and key players in regulating immunity and tolerance, including the instruction of T regs . In light of the association of LCs with AD, we also investigated the role of skin-derived DCs in activating these cells.

Animals
Mice of inbred Balb/c and C57BL/6 strains were purchased from Charles River Laboratories (Sulzfeld, Germany). Mice expressing a diphtheria toxin receptor (DTR) under the control of the Langerin (CD207) gene were bred on a C57BL/6 background as described earlier [21]. All mice were used at 2-4 months of age and animal experiments were carried out according to governmental guidelines.

Analysis of DCs and lymphocytes in skin-draining lymph nodes of mice
Auricular skin-draining lymph nodes (sdLNs) were collected from mice treated with VitD or vehicle on their ears and digested with collagenase D (Roche Diagnostics, Indianapolis, IN, USA) and DNase (Roche Diagnostics) for 25 min. at 37°C. Resulting single cell suspensions were counted in the haemocytometer, stained with mAb and analysed using flow cytometry as previously described [4]. Absolute cell numbers per auricular draining lymph node (LN) were calculated on the basis of flow cytometry analysis and haemacytometer cell counts. Mouse T regs were identified by expression of CD4, CD25 and FoxP3, and distinction between induced and natural T regs was made on the basis of Helios staining. CD11c + CCR7 + cells in sdLNs were considered as emigrated DCs. Expression of CD103 was used to discriminate epidermal LCs (CD11c + CCR7 + CD103 À Langerin + ) from Langerin + dermal DCs (CD11c + CCR7 + CD103 + Langerin + ). CD11c + CCR7 + CD103 +/À Langerin À cells were considered as Langerin À dermal DCs and CD11c + CCR7 À CD103 +/À Langerin À cells as 'other DC'.

Detection of intracellular cytokines
Isolated LN cells were cultured for 4 hrs with 1 lg/ml brefeldin A to block cytokine release; then stained and analysed by flow cytometry.

Flow cytometry and immunohistochemistry
Flow cytometry analysis was performed on a FACScalibur using Cell-Quest software (BD Immunocytometry Systems, San Jose, CA, USA) and results were analysed by FlowJo software (Tree Star, Ashland, OR, USA). Five-and six-colour stainings were carried out with a FACScanto using FACSDiva software (BD Immunocytometry Systems). Epidermal sheets were separated from ear skin with 0.5 M ammonium thiocyanate (Merck, Westchester, PA, USA) as described previously [4], washed and stained with antimouse MHC-class II-FITC mAb for 1 hr at 37°C. Stainings were visualized by an Olympus BX60 epifluorescence microscope using a 409 objective. Fluorochrome-and isotype-matched immunoglobulins of irrelevant specificity served as negative controls.

Statistical analysis
Results are shown as mean AE SD, n represents the number of mice used per group. Data were analysed using a Student's t-test for nor-

AD-like inflammation is associated with increased numbers of T regs
To study T regs in AD, we topically treated mice with VitD to trigger high TSLP expression in the epidermis as observed in AD lesions [4,22]. The inflammatory phenotype in these mice is similar to that observed in other TSLP-overexpressing mice and is characterized by an AD-like cutaneous inflammation containing Th2 CD4 + T cells expressing cutaneous homing receptors and by elevated serum IgE levels [4,[22][23][24][25]. Figure 1 depicts the kinetic of T reg (CD4 + CD25 + FoxP3 + ) and non-regulatory (CD4 + CD25 +/À FoxP3 À ) CD4 + cell expansion in sdLNs of mice upon treatments. Results show that VitD treatment significantly enhanced numbers of T regs at all time-points when compared to vehicle treatment (Fig. 1A). In contrast, numbers of other CD4 + lymphocytes increased on days 3 and 5 but not on day 10 in sdLNs of VitD-treated mice when compared to controls (Fig. 1B). Maximal cell numbers were reached for both subsets on day 5 ( Fig. 1A and B). Thus, although the expansion of T regs was continuously promoted by VitD, the expansion of other CD4 + lymphocytes regressed between day 5 and day 10 ( Fig. 1A and B). The variations in cell percentages ( Fig. 1C and D) with decreased percentages of T regs on day 3 and decreased percentages of other CD4 + lymphocytes on day 10 additionally suggest that the expansion of non-regulatory CD4 + lymphocytes precedes the expansion of T regs . Because AD-like symptoms in mice treated with VitD enhance over time, our data show that expansion of T regs parallels symptom development in this AD model.

T regs display an activated phenotype in AD-like inflammation
To better characterize the CD4 + CD25 + FoxP3 + T reg population in the VitD AD model, we measured the expression of various surface markers which are involved in T reg function in mice with overt AD symptoms, i.e. on day 10 of treatment. Percentages of CD4 + CD25 + FoxP3 + T regs expressing ICOS, CTLA-4 and GARP at their cell surface were increased in sdLNs of mice with AD when compared to controls ( Fig. 2A and B). Hence, peripheral activated T regs are observed in AD-like inflammation.
LCs are the first DC subset to emigrate to sdLNs to potentially expand T regs in AD Earlier work has demonstrated that epidermal TSLP is overexpressed after topical treatment with the VitD analogue MC903 for 4 days [4]. As opposed to dermal DCs, LCs acquire an activated phenotype in the skin after MC903 treatment and increased migration to sdLNs [4]. These findings suggested that the activation of LCs may be associated with a biased Th2 response prior to the development of clinical signs of AD [4]. On the other hand, activation and migration of LCs might also contribute to increased T reg expansion in early AD. We therefore analysed total numbers of CD4 + CD25 + FoxP3 + T regs , LCs, Langerin + and Langerin À dermal DCs in sdLNs on days 0, 3 and 5 of topical VitD treatment. Numbers of CD4 + CD25 + FoxP3 + T regs started to significantly increase in sdLNs as early as on day 3 (Fig. S1A), similar to emigrated LCs (Fig. S1B). Numbers of Langerin + and Langerin À dermal DCs and other DCs were not altered on day 3, but started to increase on day 5 ( Fig. S1C-E), according to the literature [4,26,27]. Taken together, our results establish that, in the VitD AD model, LCs are the first skin DC subset to reach the sdLNs, which coincides with the beginning of T reg expansion. It was previously reported that VitD directly induces T regs in vitro [28], challenging the requirement of skin-derived DCs in the development of T regs after topical application of VitD. To address this issue, we removed the application sites (ears) 4 hrs after a single topical application of vehicle or VitD (3 nmol/ear) to prevent any skin DC migration [29]. As depicted in Figure S1, removal of the application sites prevented the increase of both T regs (Fig. S1F) and DCs (Fig. S1G) in sdLNs of VitD-treated as compared to vehicle-treated mice. We verified that one-time application of 3 nmol VitD per ear elicits the same effects on the numbers of various DC subsets and T regs in sdLNs than a daily treatment with 1 nmol VitD for 3 days (data not shown). Therefore, expansion of CD4 + CD25 + FoxP3 + T regs is not because of a direct effect of VitD in this mouse model of early AD, but instead requires the migration of skin-derived DCs to sdLNs.
To further assess the tolerogenic function of skin DCs, we topically applied VitD to mice deficient for Langerin + DCs, including LCs (Fig. S2). Numbers of CD4 + CD25 + FoxP3 + T regs in sdLNs were increased on days 5 and 10 in VitD-treated mice depleted of Langerin + DCs as compared to vehicle-treated controls (Fig. S3A and B). Therefore, the overall induction of T regs might be initially dependent on LCs reaching firstly the sdLNs, whereas later on, other skin-derived DC subsets contribute to this expansion.

Thymus-derived T regs are enhanced in AD
Helios, an Ikaros transcription factor preferentially expressed in human and mouse CD4 + FoxP3 + T regs , was shown to allow discrimination between thymus-derived and peripherally induced T regs [30,31]. In VitD-induced AD-like inflammation, the percentages of thymus-derived (Helios + ) T regs were enhanced, whereas peripheral (Helios À ) T regs were reduced in sdLNs on day 10 of topical VitD treatment (Fig. 3A). Accordingly, thymus-derived T regs displayed higher absolute numbers (Fig. 3B). Kinetic analysis revealed a predominant expansion of thymus-derived T regs in sdLNs of mice as early as day 5 of topical VitD treatment (Fig. 4). Thus, our results indicate an early imbalance of the T reg compartment towards predominating thymusderived T regs in AD. Intriguingly, percentages of thymus-derived T regs failed to increase after depletion of Langerin + DCs in mouse skin at day 5 of VitD-treatment (Fig. 5). To identify the molecules providing the tolerogenic function to DCs, we screened for expression of various costimulatory and co-inhibitory molecules, without identifying significant changes in the expression of PD-L1, ICOS-L, GITR-L and IDO by VitD-exposed DCs (data not shown). Moreover, we detected only trace amounts of IL-10 production by skin-derived DCs in our experiments (data not shown). Thus, thymus-derived T regs are increased in the VitD AD mouse model and Langerin + DCs are required for the early expansion of thymus-derived T regs in AD, via a still elusive mechanism.

Thymus-derived T regs exhibit a Th2-like phenotype in AD
We first measured the percentages of overall T regs producing IL-10 and IL-13 in the VitD model of AD. T regs isolated from AD mice (day 10 of treatment) produced larger amounts of both IL-10 and IL-13 than T regs isolated from healthy controls, regardless of the presence of Langerin + DCs ( Fig. S3C and D). In contrast, the production of IL-13 but not of IL-10 by T regs was significantly increased earlier during the development of AD, i.e. 5 days after the start of VitD treatment ( Fig. 6A and B). Depletion of Langerin + DCs did not alter the produc-  (Fig. 6B), but increased the secretion of IL-10 by T regs at early time-points (Fig. 6A). Moreover, the production of IL-13 was more strongly induced in T regs than in CD4 + effector T cells in VitDinduced early and overt AD (Fig. S4). To further dissect the production of cytokines within the T reg compartment, we measured percentages of thymus-and peripherally derived T regs producing IL-10 and IL-13. Thymus-derived T regs were identified as the main source of IL-10 and IL-13 ( Fig. 6C-E). Notably, we observed similar numbers of thymus-derived T regs and effector T cells producing IL-13 in the VitD AD model (Fig. 6F). Moreover, both numbers of thymus-derived T regs and effector T cells producing IL-13 were higher than numbers of peripherally derived T regs . These findings emphasize the potential role of thymus-derived Th2-polarized T regs in driving the pathogenic events leading to or sustaining AD and suggest that peripherally derived T regs are small contributors to the overall Th2 cytokine production in VitD AD model ( Fig. 6D and F). Furthermore, depletion of Langerin + DCs did not alter the production of cytokines by thymusderived T regs in the VitD model of AD (Fig. 6C and D). In conclusion, we identified activated, IL-10-producing thymus-derived T regs , concomitantly exhibiting a Th2-like phenotype in the VitD model of AD.

Discussion
In this study we discovered higher numbers of overall T regs with a specific expansion of thymus-derived T regs in mice with AD. Furthermore, our results indicate that AD is associated with the expansion of thymus-derived T regs exhibiting a Th2 phenotype and that LCs seem to be responsible for this biased T reg differentiation. However, other cells or factors from the microenvironment in sdLNs might also contribute to shaping the unusual cytokine profile of thymus-derived T regs . Irrespectively, our data strongly suggest that Th2-like T regs actively contribute to the development of AD [32].
Several groups reported increased T regs in the peripheral blood [13][14][15][16] and skin lesions [10] of AD patients, whereas others did not [10][11][12]18]. In support of the former, we here report increased numbers of CD4 + CD25 + FoxP3 + T regs in sdLNs of mice at different dis-ease stages of AD development (Fig. 1). The T-cell-specific costimulatory molecule ICOS is up-regulated after cell activation and binding to its ligand (ICOS-L). This step is essential for T reg survival, proliferation and memory rather than for their activation [33]. Indeed, reduced numbers of CD4 + FoxP3 + T regs have been observed in ICOS knockout mice in the steady-state and upon immunization [34]. CTLA-4 is a CD28 homologue that is up-regulated after activation of effector T cells and T regs . It is associated with T reg suppressive function, although this remains controversial [35]. GARP (or LRRC32), a T reg -specific activation marker, is part of the receptor for latencyassociated peptide/latent transforming growth factor-b complex [36].
T regs can be divided into two subcategories, namely thymusderived and peripherally induced T regs . While Helios is expressed in all CD4 + CD8 À FoxP3 + mouse thymocytes [30], neuropilin discriminates thymic T regs from peripheral T regs only in the steady-state [37]. Therefore, Helios is currently the most discriminative marker for thymusderived T regs [30]. We found that the ratio of thymus-derived (Helios + ) T regs over peripherally derived (Helios À ) T regs was increased in sdLNs during AD-like inflammation (Fig. 3), with enhanced expansion of thymus-derived (Helios + ) T regs starting early in the development of the disease (Fig. 4). Thymus-derived T regs are involved in self-tolerance and were shown to be activated by microbes [32]. Moreover, they are important for the control of Th1 immune responses [38]. Thus, thymic T regs might be less efficient at counteracting Th2-related diseases such as AD when compared to peripherally induced T regs .
The T reg population is heterogeneous. Indeed, T regs can acquire alternative effector or hybrid fates, associated with promotion rather than inhibition of inflammation under certain conditions [32]. Accordingly, increased production of IL-5 and IL-13 has been described in skin-homing T regs of AD patients [12,14]. High expression of GATA3, as observed in T regs located at barrier sites such as the skin and gut [39], might enable T regs to produce Th2 cytokines. In our VitD AD model, we found increased percentages of IL-13-producing T regs in sdLNs ( Fig. 6B and D, Figs S3D and S4), similar to Th2 T regs in the skin of AD patients [12,14]. This might potentially confer a prorather than an anti-inflammatory phenotype to T regs in AD.
When we analysed the cytokine production by T regs more thoroughly, we found that thymus-derived T regs were the main source of IL-10 and IL-13 in the VitD AD model (Fig. 6C-E). Therefore, we demonstrate here for the first time that expansion of thymus-derived T regs exhibiting a Th2-like phenotype is promoted in AD. In fact, the numbers of thymic T regs secreting IL-13 were similar to the numbers of effector T cells producing IL-13 (Fig. 6F) and the percentages of thymus-derived T regs secreting IL-13 were significantly increased compared to effector T cells (Fig. 6D and E). This strongly suggests a pathogenic role of thymus-derived Th2-polarized T regs in AD. Indeed, these T regs might exert poor immunosuppressive properties despite their capacity to produce IL-10 and consequently contribute to the development of AD-like inflammation. It would be of particular interest to test this hypothesis by assessing the overall immunosuppressive capacity of these Th2 T regs . Unfortunately, due to the nuclear Fig. 4 Thymus-derived regulatory T cells expand in murine AD-like inflammation. Percentages (A and C) and numbers (B and D) of thymus-derived (Helios + ) and peripherally derived (Helios À ) regulatory T cells (CD4 + (CD25 + ) FoxP3 + ) in sdLNs of ETOH versus VIT D-treated mice on day 0, day 3, day 5 and day 10 of treatment. Data are representative of two to three independent experiments and were analysed with a Student's t-test or a Mann-Whitney U-test, n = 3-12. n.s. not significant. localization of Helios, the lack of Helios-EGFP mice, and the missing specificity of neuropilin as a surface marker during inflammation, it is not possible to purify Th2 thymus-derived T regs for further in vitro immunosuppressive assays.
Dendritic cells are antigen-presenting cells regulating immunity and tolerance, respectively, by priming effector T cells and expanding T regs [40]. In the VitD AD model, T reg expansion in sdLNs required the presence of skin-derived DCs (Fig. S1F and G). The time course of T reg induction revealed that, unexpectedly, LCs are the first cutaneous DC subset to reach sdLNs in our experimental setup (Fig. S1A-E). VitD-induced production of TSLP by keratinocytes might primarily trigger LCs, whereas dermal DCs migrate to sdLNs more quickly after Fig. 6 Langerin + dendritic cells are dispensable for cytokine-producing phenotype of thymus-derived regulatory T cells in murine AD-like inflammation. Production of IL-10 (A) and IL-13 (B) by overall CD4 + FoxP3 + T regs and production of IL-10 (C) and IL-13 (D) by thymus-and peripherally derived (CD4 + FoxP3 + ) T regs in sdLNs from ETOH versus VIT D-treated Langerin-DTR mice, injected with PBS (+ Langerin + DC) or diphtheria toxin (À Langerin + DC), at day 5 of treatment. Percentages of effector T cells (E) and numbers of effector T cells, thymus-and peripherally derived T regs (F) producing IL-13 in sdLNs of ETOH or VIT D-treated mice, at day 5 of treatment. Data are representative of one to three independent experiments and were analysed with a Student's t-test, n = 6-8. n.s. not significant. skin immunization with DNFB [4,41]. Previous results established that LCs promote T reg proliferation upon RANK signalling [42]. However, depletion of LCs and Langerin + dermal DCs did not affect the size of the overall CD4 + CD25 + FoxP3 + T reg population in sdLNs of VitD-induced AD mice (Fig. S3A and B) but specifically abolished the induction of thymus-derived T regs (Fig. 5). Moreover, depletion of LCs and Langerin + dermal DCs did not affect the production of IL-13 by T regs , regardless of Helios expression ( Fig. 6B and D, Fig. S3D). Thymus-derived T regs are involved in antimicrobial responses [32] and LCs are a privileged DC subset sensing microbe-derived antigens in AD [43]. Thus, the expansion of thymus-derived T regs might be attributed to LCs, while their cytokine production might rather be determined by other cells or factors within the microenvironment of the sdLNs. Langerhans cell-derived IL-10 can promote T reg expansion [44,45], but in our experiments, IL-10 was not detectable in skinderived DCs following VitD treatment (data not shown). Furthermore, expression of PD-L1, ICOS L, GITR L and IDO by skin-derived DCs might only have a supporting role in T reg expansion in the VitD AD mouse model (data not shown). Thus, the question how LCs or other DCs promote the expansion of T regs or otherwise impact on their phenotype remains unanswered.
In summary, our work represents the first study demonstrating a preferential expansion of activated CD4 + CD25 + FoxP3 + thymusderived T regs exhibiting a Th2 phenotype in a mouse model of AD. Furthermore, differentiation of thymus-derived T regs seems to depend on LCs, while their cytokine profile might rather be determined by DC phenotype and their microenvironment. Hence, T regs in AD might contribute to the disease rather than playing their role of immunosuppressive cells and thus might represent potential new therapeutic targets. laboratory. We would like to thank Nikolaus Romani and Patrizia Stoitzner for critically reading of the manuscript. We appreciate the support of the European SKINBAD COST Action BM0903.

Conflicts of interest
Christoph H. Tripp was employed by the COMET Center ONCOTYROL, which is funded by the Austrian Federal Ministries BMVIT/BMWFJ (via FFG) and the Standortagentur Tirol. We further appreciate the participation of the TILAK hospital holding company, who serves as a partner in the Oncotyrol research program.

Supporting information
Additional Supporting Information may be found in the online version of this article:

Figure S4
Percentages of IL-13-producing effector and total T regs in sdLNs of ETOH or VIT D-treated mice at day 5 (A) and day 10 (B) of treatment.