Neonatal-derived IL-17 producing dermal γδ T cells are required to prevent spontaneous atopic dermatitis

Atopic Dermatitis (AD) is a T cell-mediated chronic skin disease and is associated with altered skin barrier integrity. Infants with mutations in genes involved in tissue barrier fitness are predisposed towards inflammatory diseases, but most do not develop or sustain the diseases, suggesting that there exist regulatory immune mechanisms to repair tissues and/or prevent aberrant inflammation. The absence of one single murine dermal cell type, the innate neonatal-derived IL-17 producing γδ T (Tγδ17) cells, from birth resulted in spontaneous, highly penetrant AD with all the major hallmarks of human AD. In Tγδ17 cell-deficient mice, basal keratinocyte transcriptome was altered months in advance of AD induction. Fulminant disease is driven by skin commensal bacteria dysbiosis and highly expanded dermal αβ T clonotypes that produce the type three cytokines, IL-17 and IL-22. These results demonstrate that neonatal Tγδ17 cells are innate skin regulatory T cells. The bifurcation of type 3 cytokine producing skin T cells into the homeostatic, early innate and pathogen-sensing, late adaptive T cell compartments underpin healthy skin and accounts for the dual function of type 3 cytokines in skin maintenance and inflammation.


INTRODUCTION
The incidence of atopic dermatitis (AD, eczema) is on a steep incline in industrialized nations with estimates suggesting as high as a quarter of children affected (1, 2). Clinical and genome wide association studies (GWAS) in humans reveal that dysfunction of key structural components of epidermal barrier, such as filaggrin, and hypersensitive type 2 (IL-4, IL-5, IL-9 and IL-13) and type 3 cytokine responses (IL-17 and IL- 22), are contributing factors to AD onset and progression (3)(4)(5). The contribution of skin-targeting αβ T effector cells to AD pathogenicity is largely understood from the basic focus on damaging cytokine production and inflammatory myeloid cell recruitments. It is widely accepted that aberrant skin barrier integrity and local inflammation orchestrate the activation and recruitment of type 3 cytokine producing αβ Th17/22 cells to the skin, where they are thought to be the arbiters of the major symptoms of the disease, including visible skin damage (6)(7)(8)(9)(10).
Pivotal to the establishment of coordinated skin immunity are αβ and gd T cells, and innate lymphoid cells (ILCs). Dermal ILC2 have been shown to be critical in mobilizing type 2 cytokine responses in AD, but very little is known about the function of innate skin T cells in autoimmunity.
During the neonatal period, skin is populated by several gdTCR + and abTCR + T cell subsets, whose effector functions are thymically programmed to produce IL-17, and to a lesser extent IL-22, upon activation in tissues. IL-17 producing gd T cells (Tgd17) are referred to as innate-like and the gd T cell lineage is subject to the same effector subtype classification (Types 1, 2 and 3 cytokine producers) as adaptive T helper cells and ILCs. Tgd17 cells expressing Vg2TCR (Garman TCRg nomenclature, (11)) are exported from the thymus after birth and rapidly populate the newborn dermis. These cells are part of the neonatal wave of tissue-resident lymphocytes that are not generated efficiently from adult bone marrow hematopoietic stem cells, and are referred to as neonatal innate Tgd17 (nTgd17) cells (12).
Studies to date have established that nTγδ17 cells are the central population of the skin immunocyte subsets and are the most dominant IL-17 producing cells upon acute skin inflammatory perturbations (13)(14)(15)(16)(17). nTγδ17 cells are absolutely required for acute Imiquimod (TLR7-agonist)-induced psoriasis in adult mice. Humans with the loss of function allele of the IL-17R signaling component ACT1 (Traf3ip2) are more susceptible to psoriasis (18), but Act1deficient mice are afflicted with spontaneous skin inflammatory diseases (19,20). Moreover, mice that lack IL-17R on radioresistant epithelial cells develop AD, in genetic background with a type 2 cytokine production bias (21). In the former, skin pathology was attributed to hyper IL-22 production, and in the latter, diminished filaggrin expression and impaired skin barrier was implicated as the cause of AD susceptibility. In both models the apparent disease-protective function of IL-17 in skin homeostasis was not addressed. Thus, while some studies imply that IL-17 can act as a homeostatic factor to prevent aberrant inflammation in the skin, the cellular source and developmental timing of homeostatic IL-17 function is unknown.
Increases in Tγδ17 cells in patients with aberrant skin inflammation have been observed in AD (13,22,23), but accurate assessments of their contribution to human disease has lagged, in part due to challenges of isolating these cells from human tissues (24). Possible dual homeostatic and inflammatory roles for IL-17 and IL-22, or cells that can produce them, have also limited the use of cytokine and T cell deficient mice to unveil their context-dependent contribution to skin disease pathogenesis. We show here that mice specifically lacking IL-17-producing nTgd17 cells succumb to a highly penetrant spontaneous AD that captures most characteristic disease features of human AD. Fulminant disease in the mice is associated with hyperactive ILC2 and requires both skin commensal bacteria (CB) and expansion of clonal αβ T cells. The initial trigger for the disease is linked to aberrant keratinocyte differentiation at young ages. Thus, nTgd17 skin sentinels are essential to maintain skin homeostasis, in part by promoting normal keratinocyte barrier formation in perinatal period.

Spontaneous AD in Sox13 -/mice specifically lacking Vg2TCR+ dermal nTgd17 cells
To study the role of V2 Tγδ17 cells in skin immunity, we have generated mice deficient in Sox13, an HMG box transcription factor (TF) essential for their development (15,25). In the immune system Sox13 expression is restricted to early hematopoietic stem/progenitors and gd T cells. Mice lacking Sox13 have a highly selective defect in V2 Tγδ17 cell development with all other hematopoietic cell types normally preserved (15,16). One exception is innate iNKT17 cells that are partially affected (26), but these cells are rare in the skin. Incompatible with the proinflammatory nature of V2 Tγδ17 cells, >90% of Sox13 -/mice (>250 mice cumulatively tracked over several years) of both sex develop visible dermatitis in the muzzle, ears, eyes and elsewhere around three to four months of age, displaying the hallmarks of human AD (2,(27)(28)(29).
Pathophysiology included epidermal thickening (acanthosis, Fig. 1A, Top), marked accumulation of immunocytes in skin epithelial lesions leading to eosinophilia, neutrophilia, and increases monocytes (Mo) and Mo-derived dendritic cells (DCs) in the skin (Fig. 1B-F). Crucially, agedependent increases in IgE titer, evident by 3 months of age of the mice (Fig. 1H), before visible signs of disease, captured one of the major symptoms of human AD. In addition, expanded ILC2 (GATA3 hi ) associated with human AD (29)(30)(31), and their capacity to produce the type 2 cytokines IL-5 and/or IL-13, was recapitulated in Sox13 -/mice (Fig. 1G, Supp Fig. 1A-C). Conversely, in young Rora -/mice lacking in ILC2 (32) there is an increase in nTgd17 cells with enhanced capacity to produce type 3 cytokines (Supp Fig. 1D), suggesting a possible counter-regulation between nTgd17 cells and ILC2.

Expanded ab T cells are required for AD
Significant expansion of αβ T cells in the skin of Sox13 -/mice was evident starting ~3 months of age, prior to any visible skin inflammation. Both CD4 + and CD8 + T cells increased in numbers up to 10-fold by 6 months of age, depending on skin sites ( Fig. 2A). Notably, CD4 neg CD8 neg (double negative, DN) T cells accounted for 10-20% of TCRb + cells in the skin of both LMC and Sox13 -/mice, with a significant expansion observed in Sox13 -/skin ( Fig. 2A). Utilizing the MR1/5-OP-RU tetramer, we identified that the DN subset in both healthy and AD skin consisted primarily of MAITs (Fig. 2B). CD4 + or CD8 + MAITs were rare in the skin of WT mice, with only marginal increase in CD8 + MAITs in Sox13 -/skin (Supp Fig. 2A). In the skin draining LNs (dLNs), only subtle increase in the frequency of MAITs was observed in Sox13 -/mice, with the majority being the CCR6 + CD4 + subset in all mice (Supp Fig. 2B). iNKT cells were rare in the skin and no significant alterations were observed in Sox13 -/mice (Supp. Fig. 2C).
The majority of ab T cell subsets in AD were associated with enhanced capacity to produce both IL-17 and IL-22, whereas in control mice very few CD4 + or CD8 + ab T cells were capable of IL-17 production, and even more constrained IL-22 secretion was evident (Fig. 2C). DN MAIT cells were primed for IL-17 in both LMC and Sox13 -/mice. In contrast to the enhanced type 3 cytokine production, the frequency of Th2 cells was not altered significantly in Sox13 -/skin, although numerically they were also enhanced. Similarly, although the frequency of skin FOXP3 + regulatory T cells (Tregs) was decreased in the ear, but not muzzle, skin of Sox13 -/mice, their numbers were comparable to controls (Supp Fig. 2D), indicating preferential expansion of effector populations. Matching the T cell expansion in skin there was an ~8 fold expansion in cellularity in dLNs (Fig. 2D). The trend to this increase was evident before visible skin lesions, at ~3 months of age, and was associated with greatly increased numbers of spontaneous germinal centers (GCs), To determine whether the expansion of skin T cells was correlated to more efficient display of skin antigens in dLN, melanocyte-specific antigen presentation in Sox13 -/mice was assessed. Naïve PMEL17 CD8 + TCR transgenic T cells specific for a melanocyte antigen (34) were labeled with CFSE and transferred into Sox13 -/and WT hosts, and their proliferation was analyzed by CellTrace Violet dilution (Supp Fig. 2H, I). We observed a 3-fold increase in PMEL17 T cells proliferation in skin dLNs of Sox13 -/mice compared to controls. In contrast, no differences in proliferation were observed at distal sites, including the spleen. Finally, to demonstrate that ab T cells are required for AD in Sox13 -/mice, skin pathology in Sox13 -/-Tcrb -/was monitored. The absence of ab T cells prevented AD development with no visible evidence of skin inflammation and skin histology was grossly normal, including lack of epidermal hyperplasia (Fig. 2G).
To ascertain corresponding changes in the expression of secreted inflammatory mediators RNA was isolated from the muzzle skin at 6 months of age and select cytokine and chemokine gene expression was assessed by quantitative RT-PCR (Supp Fig. 1E). A coordinate induction of the cytokines IL-1b, IL-6 and IL-23, which promote type 3 cytokine producing lymphocytes, was prominent. A simultaneous increase in the danger associated molecular pattern molecule IL-33 was observed, which has been associated with skin inflammation (30,35) Collectively, these results indicated that prior to the onset of visible diseases, B and T cells expand, with evidence for IgE hyperproduction. With the progression of disease, the skin displays a prominent type 3 effector inducing cytokine milieu with attendant expansion of Th17 cells and IL-17 + MAITs. Thus, fulminant AD in Sox13 -/mice is predicated on Th17 and Th17-like cells of ab T cell lineage.

Altered basal keratinocyte differentiation program in Sox13 -/mice
To map the sequence of early cellular and molecular alterations in Sox13 -/mice that can account for the eventual inflammatory immune landscape, we first assessed the impact of the loss of nTgd17 cells on differentiating keratinocytes. For this we undertook a whole transcriptome analysis of basal CD49f + (Itga6) keratinocytes of Sox13 -/mice at 3 and 7 weeks (wks), well before the onset of aberrant skin inflammation starting in ~3 months old (mo) mice. This population was chosen because they contain keratinocyte stem cells and progenitors (36,37) and the two timepoints coincide with the hair follicle catagen cycle, characterized by active keratinocyte differentiation followed by the relatively quiescent telogen cycle, respectively (38). In all 261 genes were differentially expressed (>2 fold changes, p<0.05) between 3 wk WT vs Sox13 -/basal keratinocytes (Fig. 3A). Gene Ontology (GO) enrichment analysis revealed pronounced cell apoptosis signatures and stress responses in Sox13 -/basal keratinocytes (Fig. 3B). At 7wk the difference was muted with 50 genes differentially expressed ( Fig. 3A) with no significant clustering of these genes into specific biological processes, likely reflecting the resting state of basal keratinocytes in the telogen phase. Expression of only 3 genes, Igfbp3, Mir-17hg (Mir-17-92) and 4930480K23Rik (non-coding RNA), was altered at both ages. Igfbp3 and Mir-17hg (Mir-17-92) have been shown to be associated with skin inflammations (39,40) and their expression was initially decreased in Sox13 -/basal keratinocytes, but this pattern was flipped at 7wk. Sox13 -/mice prior to 2 months do not show any significant alterations in skin immune subsets or visible damage, and consistent with this Sox13 -/basal keratinocytes showed no significant alterations in the expression of inflammatory mediators of immunocytes at 3 and 7 wks. Genes encoding for the structural components of the skin barrier including gap junction proteins, extracellular matrix (except collagens at 3wk) and keratins, were also not altered in expression. However, expression of several genes critical for normal differentiation of basal keratinocytes was altered at 3wk, including diminished expression of the IL-17 target Blimp1 (Prdm1) (41,42), Sox9 (43), Runx1, Irf3/6, S100a11, and increased expression of Myc (44), Dlx3,Trp73 and Maf. In addition, genes in the TGFb, Lymphotoxin and the JAK-STAT signaling pathways had lower levels of expression in Sox13 -/basal keratinocytes. Genes controlling barrier fitness, such as Trex2, Epcam, Adam17, Itga2, Cdh3, Tgm4, Il31ra, Il1rn and Jup, were decreased in expression, whereas Def, Lrrc31 and Tsc22d3 (GILZ) were increased in Sox13 -/keratinocytes (Fig. 3C). Together, these results indicate that Tgd17 cells are critical for establishing normal developmental program of basal keratinocytes during the catagen cycle, and in their absence the data suggests that altered keratinocyte differentiation signatures are linked to increased propensity to apoptosis.

Skin commensal bacteria dysbiosis in Sox13 -/mice is responsible for AD
Analysis of differentiated keratinocytes at 2 months or later does not allow for clear distinction between impaired barrier function arising from keratinocyte-intrinsic defects or from inflammatory immunocyte-mediated degradation. In patients with AD, expansions of Staphylococcus and Corynebacteria species are often observed in skin lesions (5,9,45,46) and mouse models of AD with barrier defects replicate the AD-associated microbiome dysbiosis. Thus, one prediction of the altered keratinocyte differentiation and barrier function well before the onset of chronic inflammation in young Sox13 -/mice is that the homeostasis of skin commensal bacteria (CB) with the barrier will be disrupted, with the resultant dysbiosis driving the immune responses. We tested this possibility by first establishing skin microbiota of Sox13 -/mice at 3 and 6 mo by 16S rRNA sequencing, followed by assessment of antibiotic treatment (Abx) on AD onset and progression. As in human AD patients, AD in Sox13 -/mice was associated with dysbiosis of Staphylococcus and Corynebacteria, but with distinct kinetics (Fig. 4A). Most Sox13 -/mice showed an early bloom of Corynebacteria (C. mastitis), with the expansion maintained in some mice, but for the majority returning to the LMC frequencies at 6 mo. Expansion of Staphylococcus was pronounced at the frank phase of disease but was not obvious at 3 mo. These results largely recapitulate skin CB dysbiosis in two mouse models of AD (9,21).
To determine whether skin CB is necessary for AD initiation and/or progression in Sox13 -/mice we treated the mice from birth or starting at 3 mo with a combination of antibiotics (cefazolin and enrofloxacin in drinking water) previously used for a similar purpose (9). Skin commensal sequencing of Abx mice confirmed that Staphylococcus and Corynebacterium species were significantly reduced (Supp Fig. 3A). Regardless of regiments, the Abx Sox13 -/mice were protected from AD. All pathophysiological features of AD were absent, with resolution of acanthosis ( Fig. 4B), decreased serum IgE concentrations (Fig. 4C), and suppression of myeloid expansion (Fig. 4D). While CD4 + cells remained at an elevated frequency, IL-17 and IL-22 production was significantly reduced (Fig. 4E). Further, all disease-associated phenotypes of the dLN were corrected by Abx treatment, leading to reduction of total cell number, and the normalization of Tfh, GC B cell, and plasma cell frequencies ( Fig. 4F-I). We also tested whether the disease initiation is restricted to a narrow developmental window spanning neonatal-juvenile stages. For this, Sox13 -/mice were treated with the antibiotic cocktail from birth and then the treatment was terminated at 3 wks of age. AD development was not prevented in mice treated only acutely at birth, suggesting that continuous skin commensal-immunocyte crosstalk contributes to the disease postnatally and delayed/altered commensal interactions during neonatal stage do not permanently remodel skin pathophysiology.

Tgd17 cells respond to skin CB by IL-1 and IL-23 secreted by APCs
Commensal dysbiosis is known to result from impaired barrier functions. That nTgd17 cells themselves normally respond to Corynebacteria/Staphylococcus and the absence of nTgd17 cells also directly contributes to the aberrant microbiome expansion was assessed next. A recent report of nTgd17 cell activation in SPF mice topically colonized with C. accolens (47) strongly supported this possibility. There are two Tgd17 subsets in mice. Along with Vg2 + nTgd17 cells, the dermis contains the canonical Vg4TCR + fetal derived Tgd17 (fTgd17) cells, which are not dependent on Sox13 for populating the skin (15). Thus, an obvious question is why dermal nTgd17 cells are functionally non-redundant in suppressing AD initiation. Whereas fTgd17 cell persistence is dependent on CB (48) and parallels dermal Th17 and Tc17 cells (14), nTgd17 cells were not, as assessed in germ free (GF) mice (Supp Fig 3B). Abx WT mice also showed the loss of skin fTgd17 cells (Vg2 neg Vg3 neg quadrant, Supp Fig 3C) and the loss of tonic Il17a transcription by residual fTgd17 cells (Vg2 neg ) in Abx WT mice. In contrast, constitutive Il17a transcription in nTgd17 cells was not suppressed by Abx (Supp Fig 3C, D). These results indicate unique homeostatic activation requirements for dermal nTgd17 cells.
To determine how nTgd17 cells normally react to skin CB, they were isolated from dLNs and stimulated with a diverse set of Staphyloccus and Corynebacteria species in transwell cultures with antigen presenting cells (APCs). While Corynebacteria consistently stimulated copious IL-17 but not IFNg, production from Vg2 + nTgd17 cells, so did Staphyloccus species, albeit with a consistent diminution of IL-17 amounts per cell (Fig. 5A). For comparison, fTgd17 cells showed indistinguishable pattern of CB reactivity. This Tgd17 activation was not T-APC contact dependent, as similar levels of IL-17 production was elicited when CB-activated APCs were separated from nTgd17 cells in transwells, indicating sufficiency of trans-acting factor(s) (Fig. 5B).
Given that IL-1 and IL-23 from activated APCs was linked to Tgd17 effector cytokine production in peripheral tissues (49), both cytokines were quenched by Ab in the same culture to test whether they are the trans activating factors in the skin. Transwell cultures in which nTgd17 (or fTgd17) cells were cultured in a separate compartment from CB-APC and then blocked with Abs against the cytokines showed significantly reduced IL-17 production ( Fig. 5C and data not shown).
Collectively, these results indicate that nTgd17 and fTgd17 cells respond comparably to skin CB that are altered in AD, and that this reactivity can occur independently of direct contact with CB-APCs. Thus, biases in CB recognition by nTgd17 and fTgd17 per se are unlikely to explain the necessity of nTgd17 cells for skin homeostasis. To date, type 3 cytokine producing T cells with established functions in the skin have been shown to require CB for persistence. However, nTgd17 cells can be maintained and function in the skin independent of CB, a distinguishing characteristic that likely underpins the non-redundancy of nTgd17 cells in controlling aberrant skin inflammation.
The expanded ab T cells in Sox13 -/mice are required for AD progression. If the expansion is antigen driven a prediction would be that there would be restricted TCR repertoire in skin infiltrating ab T cells of Sox13 -/mice. To test this, we first assessed TCRVb chain repertoire of CD4 + T cells by flow cytometry. While the TCRVb usage of dLN T cells of WT and Sox13 -/mice was indistinguishable, skin CD4 + T cells in Sox13 -/mice were dominated by the usage of Vb4 TCR, starting at 3 months of age and reaching a plateau at ~5-6 months (Supp Fig. 4A, B). As skin inflammation progressed to overt disease (~5 mo), the frequency of Vβ4 + CD4 + T cells increased ~3-fold and in 5-6 mo Sox13 -/mice the total number of skin CD4 + T cells was more than 10-fold greater in Sox13 -/mice than WT mice, depending on the skin site, with up to 50% of these cells expressing Vβ4 TCR ( Fig. 6 A-C). In comparison, TCR Vβ skewing was not consistently observed for any other Vβs or for any TCRs associated with FOXP3 + Tregs or CD8 + T cells (Supp Fig. 4B, C). The increased cellularity in diseased Sox13 -/skin, combined with the strong Vβ4-bias and increased proliferation of skin Vb4 + CD4 + T cells in Sox13 -/mice (Supp Fig. 4D), suggested that these CD4 + T cells were undergoing expansion in the skin.
Cytokine production from skin-infiltrating CD4 + T cells was assessed to correlate effector function with the TCR Vβ repertoire. In WT mice, Th17 cells (IL-17 + and IL-17/22 + ) were found in both Vβ4 + and Vβ4dermal CD4 + T cell populations, but there was a biased representation of these effectors within Vb4 + T cells (Fig. 6D, E). ~10% of WT skin CD4 + T cells were geared for IL-13 and/or IL-4 production, but there were negligible numbers of skin Th1 and Th22 cells (data not shown). In contrast, Sox13 -/-AD skin lesions were enriched in Th17 subset and a larger population of dual IL-17 + /22 + Th17 cells, which were strongly biased to Vβ4 + T cells (Fig. 6D, E). Moreover, another AD-associated Th subset was the IL-22-only Th22 cells (10,28), which predominantly expressed Vβ4 (Fig. 6E). Frequencies and TCR Vβ repertoire of skin Th2 cells (~10% of CD4 + T cells) and cytokine producing skin CD8 + T cells (from WT and Sox13 -/mice) were unchanged in Sox13 -/mice at 3 and 6 mo (data not shown). To demonstrate that the expanded CD4 + T cells critically contribute to AD, Sox13 -/mice were treated with CD4 T cell depleting Abs starting at 3 mos of age for 3 mos. Skin inflammation significantly improved, including substantially reduced epidermal hyperplasia (Fig. 6F, G) and amelioration of eosinophil and neutrophil infiltration (data not shown). Collectively, these results indicate that CD4 + ab T cells are the major driver of AD in Sox13 -/mice and Vβ4 + CD4 + T cell expansion with enhanced IL-22 production is the primary distinguishing feature of ab T cells in AD, dovetailing with findings in human severe AD (10).
To test the possibility of clonal TCRVb4+ T cell expansion, we used high throughput sequencing to identify TCR Vβ clonotypes expressed on conventional ab T cells of WT and Sox13 -/mice at 5 months of age. We interrogated cells expressing Vβ4 TCRs, as well as ones expressing Vβ2, 6 and 8. Collectively, these T cells represent 50-70% of CD44 + CD4 + T cell repertoires. Analyses of skin from healthy mice revealed a single, dominant clonotype (CDR3β: CASSQDSSAETLYF) expressed on ~70% of all CD44 + Vβ4 + CD4 + T cells (Fig. 6H). Notably, CD4 + T cells expressing this TCR Vβ clonotype along with a related Vb4 sequence (CDR3β: CASSPDSSAETLYF) were strongly expanded in diseased Sox13 -/mice, making up >75% of Vβ4 + conventional CD4+ T cells. These clonotypes, which we denote as the common Vβ4 (comVβ4), were less frequent in activated/memory T cells in dLNs (2-3% of CD4+ T cells), and detectable only at minute frequencies in naïve T cells (< 0.1%). In comparison, TCRβ chains expressed on skin-resident CD8 + T cells in WT and Sox13 -/mice were oligoclonal (data not shown).
To begin to identify TCRαβ clonotypes, 15 CD4 + skin T cell lines were established from Sox13 -/-AD skin and converted to hybridomas. Although this approach was inefficient, four Vβ4 + T hybridomas expressed the comVb4 chain, all of which were paired with a conserved Vα4.9 chain (comVa4, CALSDNTGNYKYVF). TCRα deep sequencing of total skin CD4 + T cells confirmed that >90% of the Va4+ cells expressed the comVa4 chain in the skin of diseased Sox13 -/mice (Fig.   6I), while these clonotypes were rare in dLNs. Together, these studies reveal that ~25% of all skin CD4 + T cells in AD mice express two related TCR clonotypes composed of Vb4Va4.9 TCR, indicative of antigen-specific clonal expansion.

DISCUSSION
When there are systemic structural breakdowns of the skin barrier, dysregulated immunity leads to uncontrolled inflammation. Most mouse models of AD to date involve either a systemic breakdown of the skin barrier (e.g filaggrin/matted (50,51) or Adam17-deficient (9)), or rely on heavy manipulations of the skin (e.g. tape stripping followed by antigen challenge (52) or topical applications of inflammatory cytokines, such as IL-23 (53)) and they may not reflect natural progression of AD. In particular, physiological events that contribute to skin barrier damage in postnatal animals have not been modeled for experimentation. Moreover, identification of pathogenic CD4 T cell clones and events that trigger adaptive T cells that culminate in AD have not yet been systematically investigated. Here, removal of one innate dermal T cell sentinel subset that normally populates the neonatal skin is sufficient to cause spontaneous, highly penetrant AD, with all the major hallmarks of the human disease. Early changes in basal keratinocyte transcriptome, well before the onset of fulminant disease, are consistent with an altered barrier formation that is likely to have linkage CB dysbiosis and the damaging immune responses that ensue. Our model thus serves to close fundamental gaps in understanding of AD and identify dermal innate nTgd17 cells as skin regulatory T cells.
AD in Sox13 -/mice is driven by Th17 cells, and transfer studies using CD44 hi T cells from dLNs of diseased Sox13 -/mice did result in AD-like symptoms in Sox13 -/-, but not in LMC host (data not shown). However, the hosts need to be primed for the disease transfer, by sublethal irradiation and skin scarring, and the kinetics of disease induction and severity were variable. Predictable kinetics of AD transfer using dermal T cells from diseased Sox13 -/mice would be ideal, and we are currently attempting to establish more physiological priming conditions for these studies. mice (58). However, gd T cell subset-specific function in these disease models were unknown and Tgd17 cells have been considered principally as an inflammatory/pathogenic cell type, required for psoriasis and EAE (49), ocular responses to C. mastitidis to protect against fungal infection (59), and intestinal responses to Listeria (60). In the frontline mucosal tissues there are several innate and conventional T cells that can produce IL-17, including Tgd17, MAITs, iNKT17, ILC3, Tc17 and Th17 cells. While the homeostatic role of IL-17 in the skin was documented (19-21) the critical cell source of IL-17 was not known. We show that Vg2TCR + nTgd17 cells is that source necessary to prevent the skin CB dysbiosis dependent inflammation cascade.
Given that there exists multiple innate type 3 cytokine producing T cells, it remains unclear why nTgd17 cells are indispensable in skin homeostasis. The alternate fetal-derived PLZF + Vg4TCR + fTgd17 cells in adipose tissues are required for normal thermogenic responses (61) and they are also present in most mucosal tissues (62). While Sox13 -/mice generate reduced numbers of fTgd17 cells from the postnatal thymus, their numbers in peripheral tissues normalize over time (15). The emerging model of AD progression is then that tonic IL-17/22 produced by nTgd17 cell recognitions of CB and other skin-specific cues promote normal development of keratinocytes in postnatal mice. In the absence of this lymphoid-epithelial crosstalk, skin CB dysbiosis develops in conjunction with altered skin barrier, driving APC activation and setting in motion aggressive activation, infiltration and expansion of type 3 cytokine producing T cells that are primarily focused on dealing with altered CB, but also result in collateral skin degradation. In parallel, damaged skin releases DAMPs, such as IL-33 that activates ILC2, which in turn promote Th2 responses (30). Cytokines and chemokines copiously produced by activated skin lymphocytes perpetuate eosinophilia and neutrophilia that chronically worsen skin damage. In this setting, skin Tregs do not significantly impact the disease progression, as their sustained depletion does not impact AD amelioration caused by conventional CD4 T cell ablation (Fig. 4).
Human inflammatory skin diseases also involve Tgd17 cells (22). While type 3 cytokine producing

Cell isolation and stimulation and antibodies
Ears and muzzle skin were first treated with Nair for 2 min, and then Nair was gently wiped away with a PBS-moistened cotton-tip applicator, and tissue was subsequently rinsed extensively with PBS prior to digestion. For this study, muzzle tissue is demarcated by the boundaries of the vibrissiae. Ears were split into dorsal and ventral halves, and muzzle tissue was removed of subcutaneous tissue. Skin was finely minced and then digested with 1 U/mL Liberase TL (Roche) + 0.5 mg/mL Hyaluronidase (Sigma-Aldrich) + 0.05 mg/mL DNAse (Roche) dissolved in HBSS (with Ca 2+ /Mg 2+ , Corning) + 10 mM HEPES (Gibco) + 5% FBS (Sigma-Aldrich) for 90 min at 37°C with gentle shaking. After digestion, EDTA (Sigma-Aldrich) was added at 5-10 mM, and then tissue was strained through a 100 µm cell strainer. Cell were washed in FACS buffer (DPBS, Ca 2+ /Mg 2+ -free + 0.5% BSA [Fisher Scientific] + 2 mM EDTA) and then plated for antibody staining. Mandibular and parotid dLN were mechanically homogenized between etched glass slides (Fisher Scientific) and strained through 70 µm mesh prior to plating for antibody staining. . MR1 and CD1d tetramers were provided by the NIH Tetramer Core Facility at Emory University. All samples were labeled with a fixable viability dye (ThermoFisher) prior to analysis. The combinatorial TCR Vb staining strategy has been described previously (68), and all Vb epitopes were found to be resistant to the enzymes used for digestion when tested on dLN cells (data not shown). CellTrace Violet was purchased from ThermoFisher, and cell were labeled as recommended by the manufacturer. For intracellular cytokine staining, cells were fixed/permeabilized with Cytofix/Cytoperm buffer (BD Biosciences) and then stained in permeabilization buffer. For intranuclear transcription factor staining, cells were fixed/permeabilized and then stained using the FoxP3/transcription factor Staining Buffer Set (eBioscience).

Histology and immunofluorescence microscopy
For H & E staining, muzzle tissue was first fixed in 10 % neutral-buffered formalin for 24 h, and then paraffin embedded, sectioned, and stained by the UMMS DERC Morphology Core. Epidermal thickness was calculated using ImageJ, taking the average of 3 measurements per image to record as 1 data point. For immunofluorescence microscopy, dLN were fixed in 4% paraformaldehyde (diluted from 16% ampules, Electron Microscopy Sciences) in PBS for 6-8 h at 4°C, washed three times in PBS, equilibrated in 30% sucrose in PBS overnight, and then frozen in OCT compound (Sakura Tissue-Tek). Cryosections were cut to 7 um thickness, blocked in PBS + 0.3% Triton X-100 + 5% normal mouse serum for 1 h at RT, then endogenous biotin was blocked using the Avidin/Biotin Blocking System (BioLegend) as recommended. Primary antibody labeling was performed in blocking buffer overnight at 4°C in a humidified chamber using the following antibodies: anti-CD4 Alexa Fluor 647 (BioLegend), goat anti-IgD purified (Cedarlane Labs), anti-GL7 Alexa Fluor 488 (BioLegend), and anti-CD11c Brilliant Violet 421 (BioLegend). Slides were washed 3x in PBS, and then labeled with donkey anti-goat Cy3 (Jackson ImmunoResearch) in blocking buffer for 1 h at RT. Slides were rinsed 3x in PBS and mounted using Fluoromount-G (Southern Biotech). Images were acquired on a Zeiss Axio Observer with LED excitation using ZEN software (Zeiss) and displayed using best-fit parameters.

TCR CDR3 deep sequencing
The strategy for deep sequencing of TCR Vb4 CDR3 regions has been described previously (69). Cells from pooled muzzle and ear skin of 6 mo LMC and Sox13 -/mice with AD were sorted via FACS as Live CD45 + TCRb + CD4 + CD25 -GITR lo to exclude Treg cells. RNA was extracted using Trizol (ThermoFisher), and cDNA generated using oligo dT priming and OminScript reverse transcriptase (Qiagen) per the manufacturers' recommendations. PCR was performed using a Vb4or Va4-specific forward primer containing adapter and barcode sequences combined with a Cb or Ca reverse primer. Multiple forward primers were used for Va4 to ensure coverage of the entire Va4 family. Sequencing was performed on an Illumina MiSeq at the Deep Sequencing Core Lab. For analysis, low quality (Q score <25) reads were removed and then sequences were parsed based on the sample barcode using fastq-multx. TCR V and J nucleotide sequences were converted to amino acid sequences using TCRKlass, using the conserved Cys residue of TCR Vb to identify CDR3 position 1.

Microbiome sequencing, antibiotics, and in vitro bacterial/gd cell cultures
To sequence the muzzle microbiome of LMC and Sox13 -/mice, sterile cotton-tip applicators were swabbed across both sides of the muzzle and then placed into sterile Eppendorf tubes and placed onto dry ice. Muzzle swabs were sent to Molecular Research LP (MR DNA, Shallowater, TX) for DNA extraction and sequencing on an Illumina MiSeq. Extracted DNA was used to amplify the 16S V4 region, and then amplicons were purified for library generation. For analysis, low quality and short sequences (<150bp) were removed. Operational taxonomic units were identified and classified using BLASTn and a curated database derived from NCBI, RDPII, and GreenGenes. Count files were then converted to percentages by dividing the number of counts for a given phylum/species by the sum of all counts. For antibiotic treatment, Sox13-/-breeders were placed on drinking water containing 0.5 mg/mL enrofloxacin and 0.5 mg/mL cefazolin (hereafter Abx). Weaned mice were then placed on Abx water and analyzed at 6 mo. To assess gd cell responses to skin commensals, LN gd T cells were isolated from WT 129 mice by negative selection (without the use of anti-TCRd Abs). CD11c + cells were isolated from spleens using CD11c microbeads (Miltenyi Biotec). Corynebacteria were grown on brain heart infusion agar (BHI) with 1% Tween-80, then grown in BHI broth with 1% Tween-80 overnight. Staphylococcus was grown on trypticase soy agar, then grown in BHI broth overnight. C. accolens was purchased from ATCC. C. bovis and C. mastitidis were kindly provided by K. Nagao (National Institute of Arthritis and Musculoskeletal and Skin Diseases, (9)). S. lentus was isolated from the muzzle skin of a Sox13 -/mouse with AD by streaking onto mannitol salt agar, followed by re-streaking of an isolated, mannitol-fermenting colony. Species identification was determined by sequencing analysis of 16S V1-V3 followed by BLAST. The day of the experiment, bacterial cultures were subcultured 1:100 for 2-4 hours to permit recovery into exponential growth phase. Culture density was determined by OD600, and then bacteria were resuspended in PBS and heat-killed at 56°C for 1 hour. DC, gd T cells, and bacteria were cultured at 1:1:10 ratio for 16-18 h, and then GolgiStop and GolgiPlug were added for an additional 4 h prior to FACS analysis. In some cases, anti-IL-23 (MMp19B2, BioLegend) and anti-IL-1R (JAMA-147, Bio X Cell) or isotype control antibodies were added for the entire culture duration. To assess contact dependency, DC and bacteria were placed in the top chamber of a 0.4 µm TransWell apparatus (Corning) and gd T cells in the bottom well.

Gene expression analysis
For RT-qPCR analysis of whole skin, skin was excised and stored in RNALater (ThermoFisher) overnight at 4°C. The next day, the sample was homogenized in Trizol using an Omni Tissue homogenizer, and then RNA isolated. RNA was converted to cDNA using oligo dT priming and AffinityScript reverse transcriptase (Agilent). qPCR was performed using iQ SYBR green Supermix and a CFX96 thermal cycler (Bio-rad), followed by thermal melt curve analysis to confirm specific amplification. Primers used in this study were synthesized by Integrated DNA Technologies and are reported in Supplemental Table 1. For RNA sequencing analysis, epidermal keratinocytes were purified by first separating dorsal and ventral halves of dissected ears and floating dermis down on 5 U/mL dispase (Sigma-Aldrich) with 0.05 mg/mL DNAse I for 50 min at 37 °C. Epidermis was then peeled away, and the dermis discarded. The Epidermis was further minced and then digested for an additional 30 min with 2 mg/mL Collagenase IV (Worthington) with 0.05 mg/mL DNAse I. Epidermal single cell suspensions were then labeled with anti-CD49f to identify basal keratinocytes, anti-CD45 to exclude leukocytes, and 7-AAD to exclude dead cells. Keratinocytes were double-sorted for purity, with the second sort into cell lysis buffer for RNA extraction at 10 4 cell equivalents. Samples were generated in triplicates. RNAseq analyses were performed by the Immunological Genome Project, using the standard operating protocol