Peripheral immune-inducer dendritic cells drive early-life allergic inflammation

Abstract

Atopic diseases associated with allergens, as well as allergic diseases, frequently arise early in life; however, the age-dependent mechanisms governing immune responses to allergens remain poorly understood¹. Here we find that in early life, exposure to common allergens triggers a distinct bifurcated immune response, simultaneously triggering type 17 inflammation in the skin and initiating canonical T helper 2 sensitization in the lymph nodes. This early-life γδ type 17-mediated dermatitis primes the exaggerated allergic lung inflammation upon secondary allergen exposure. Mechanistically, we find dendritic cell (DC)-mediated type 17 activation directly in the skin without requiring migration to lymph nodes; we term this state ‘peripheral immune inducer’ (pii) DC. CD301b⁺ conventional type 2 DCs acquire allergen, adopt the pii-DC state, produce IL-23 and activate local γδ type 17 cells independently of lymph-node engagement. The pii-DC state is enabled by the immature hypothalamic–pituitary–adrenal axis and physiologically low systemic glucocorticoids characteristic of early life^2,3; DC-specific deletion of the glucocorticoid receptor recapitulates the pii-DC phenotype. These findings define a developmental checkpoint, set by neuroendocrine maturation, that enables in situ DC activation and immune induction, thereby shaping age-dependent responses to allergens.

Far from inert, the early-life immune system has a unique composition and function compared with adult immunity⁴. This distinction is further underscored by the high prevalence of allergic immune diseases early in life. In particular, atopic conditions such as eczema, which are tightly linked to allergen exposure, affect nearly one in four children in the Western world and manifest in the skin as early as 8 weeks of life^5,6. Yet, mechanisms of age-dependent immunity and sensitivity to allergens are surprisingly underexplored.

To decode early-life-specific immune mechanisms, we tested how postnatal day 4 (P4) pups respond to a commonly used adult model of MC903-induced atopic inflammation⁷. Unlike adult mice, however, MC903 challenge did not elicit overt skin pathology in pups excluding the use of this model (Extended Data Fig. 1a). The short mouse developmental window of infancy and early childhood (approximately P0–P14) is incompatible with months long sensitization protocols used in adult mice. We thus challenged P4 pups and P60 adult mice with a single intradermal inoculation of house dust mite (HDM) extract, a common allergen associated with atopy⁸. A HDM dose (20 μg g⁻¹) was scaled to body weight at each age. Pups developed a robust inflammatory response with increased transepidermal water loss and marked epidermal pathology (Fig. 1a,b and Extended Data Fig. 1b). Like HDM, challenge with Alternaria alternata (ALT), another common allergen, also elicited robust inflammation in pups, but not in adult mice^9,10 (Extended Data Fig. 1c–e). Furthermore, intradermal challenge with heat-killed Candida albicans, but not with heat-killed Staphylococcus aureus or lipopolysaccharide (LPS), triggered inflammation and pathology in pup skin, indicating that heightened cutaneous immune reactivity in early life is not universal but instead elicited by specific agents (Extended Data Fig. 1f–h).

Fig. 1 |. Heightened immune sensitivity to common allergens in early life.

a, Representative macroscopic images (top) and haematoxylin and eosin (H&E)-stained sections (bottom) of mouse dorsal skin 6 days after intradermal PBS (Ctrl) or HDM inoculation at the indicated ages (N = 5). Scale bars, 0.5 cm (top) and 100 μm (bottom). b, Transepidermal water loss (TEWL) in mice following HDM challenge as in panel a (n = 6). c, Schematic of the lung HDM rechallenge model (top). Pups (P4) were intradermally (i.d.) injected with PBS (1st PBS) or HDM (1st HDM) and re-challenged 7 weeks later with intranasal (i.n.) HDM (2nd HDM) or exposed to intranasal PBS only (Ctrl). Representative lung histology (bottom) is shown (N = 3). Scale bars, 100 μm. d, Quantification of immune cells in bronchoalveolar lavage fluid (BALF) from mice in panel c (N = 3; n = 3 Ctrl, 6 1st PBS and 8 1st HDM). e, Mice treated intradermally with PBS or HDM at P4 or P14 were intranasally challenged with HDM 6 weeks later; immune cells in BALF were quantified and normalized to the 1st PBS + 2nd HDM group from the same cage (N = 2; n = 9 P4–1st PBS, 11 P4–1st HDM, 10 P14–1st PBS and 12 P14–1st HDM). Each dot represents an individual mouse, and data are presented as mean ± s.e.m. (b,d,e). Statistical significance was assessed by two-tailed unpaired Student’s t-test (b) or one-way analysis of variance (ANOVA; d,e). N = independent experiments. Schematics in panels c,e were created using BioRender (https://biorender.com).

Time-course analysis identified 5–6 days post-inoculation as the height of the inflammatory resonse, after which pathology self-resolved (Extended Data Fig. 1i). Furthermore, HDM challenge at mouse developmental milestones — mobility (P14), weaning (P21) and quiescent (P21 and P60) versus active stages of the hair cycle (P4 and P32) — revealed that acute inflammatory responsiveness to HDM was lost by P14 (Extended Data Fig. 1j), underscoring a transient early-life window of immune hyper-reactivity in the skin.

We next tested whether the early-life inflammatory reaction to HDM would impact immunopathology in subsequent encounters. Secondary intradermal HDM rechallenge in adult mice elicited an increase of type 2 lymphocytes in the skin and draining lymph nodes (DLNs) and a potent IgE response in HDM pre-challenged versus naive mice (Extended Data Fig. 1k). Furthermore, after intranasal HDM challenge in adulthood, HDM pre-challenged pups exhibited elevated numbers of T cells and eosinophils in the bronchoalveolar lavage fluid, and an enhanced T helper 2 (T_H2) response in the lung DLNs compared with controls (Fig. 1c,d and Extended Data Fig. 1l). By contrast, intradermal pre-challenge at P14, which does not elicit a skin inflammatory reaction (Extended Data Fig. 1j), resulted in fewer numbers of eosinophils and T cells in the bronchoalveolar lavage fluid of the HDM-rechallenged lung (Fig. 1e), revealing that the inflammatory skin reaction exhibited by pups tunes the magnitude of subsequent systemic responses to HDM. The systemic and heightened long-lasting sensitization induced by early-life HDM challenge offer a unique opportunity to understand and potentially interrupt the allergic and allergen-associated atopic diseases at a key point of origin: the skin.

γδ T17 cells in early-life allergen-induced dermatitis

We next sought to define the specific molecular drivers of the HDM inflammatory response in pup skin; we performed transcriptomics on full-thickness skin from pups and adults at the peak (6 days) of inflammation. HDM-treated pup skin had 774 differentially expressed genes (DEGs; | log₂fold change | ≥ 2, adjusted P ≤ 0.05) and ALT-treated pup skin had 834 DEGs when compared with controls (Fig. 2a and Extended Data Fig. 2a). Conversely, HDM challenge of adult mice resulted in 2 DEGs and ALT challenge in 52 DEGs compared with controls. Consistently, we observed close clustering of control and allergen-treated adult skin samples, whereas HDM-treated and ALT-treated pup groups segregated further away from controls, in line with the overt pathology in early life and lack of inflammation in adults (Extended Data Fig. 2b,c). To examine whether the molecular profile of allergen-challenged pup skin resembled any human inflammatory diseases, we compared DEGs from HDM-challenged and ALT-challenged pups to those from human atopic dermatitis¹¹, psoriasis¹² and nickel contact dermatitis¹³. Both HDM-induced and ALT-induced DEGs shared the greatest overlap with atopic dermatitis (Extended Data Fig. 2d).

Fig. 2 |. Type 17 immunity drives early-life responses to HDM.

a, Volcano plot of DEGs from RNA-seq of full-thickness pup or adult back skin 6 days after PBS (Ctrl) or HDM inoculation. The number of DEGs (| log₂fold change (FC) | ≥ 2, adjusted P ≤ 0.05) are in red. b, Representative KEGG pathways in HDM-challenged pups versus PBS. c, Representative flow cytometry plots and quantification of RORγt⁺ cells within CD45⁺CD90⁺ skin cells 6 days after HDM or Ctrl inoculation (top; N = 2; n = 4 pup HDM and n = 3 for others), and representative flow cytometry plots and quantification of GATA3⁺ T_H2 within CD45⁺βTCR⁺ CD4⁺FOXP3⁻DLNs cells from 6 days after HDM or Ctrl inoculation (bottom; N = 2; n = 5 adult, n = 6 Pup Ctrl and n = 7 Pup HDM). Type 17 is enriched in pup skin, whereas type 2 is enriched in pup and adult DLNs. d–f, Representative images (left) and corresponding TEWL (right) of Rorγt^gfp/gfp mice and WT littermates (d; N = 7; n = 11 WT and 12 Rorγt^gfp/gfp), Tcrd^+/− and Tcrd^−/− littermates (e; N = 4; n = 13 WT and 11 Tcrd^−/−) and IL17rc^EKO mice and WT littermates (f; N = 3; n = 10 WT and 6 IL17rc^EKO), 5 or 6 days after HDM inoculation. Scale bars, 0.5 cm. g, Representative flow cytometry plots and quantification of GATA3⁺ T_H2 cells within CD45⁺βTCR⁺CD4⁺FOXP3⁻ DLNs cells from WT and Il17a/f^−/− pups 5 days after HDM or Ctrl inoculation (N = 2; n = 7 for Il17a/f^−/− HDM and n = 6 for others). h, Summary schematic of this figure. Each dot is an individual animal; data are mean ± s.e.m. (c–g). Statistics used Wald test with Benjamini–Hochberg correction (a), one-tailed Fisher’s exact test (b), two-tailed unpaired Student’s t-tests (d–f), multiple unpaired Student’s t-tests (two-stage step-up Benjamini, Krieger and Yekutieli) comparing HDM versus Ctrl per age group (c) and one-way ANOVA (g). N = independent experiments. Schematics in panels c,h were created using BioRender (https://biorender.com).

Next, we examined expression of common immune pathways between HDM-challenged and ALT-challenged pups in our transcriptomics data. This pathway analysis revealed an enrichment of NF-κB signalling, C-type lectin receptor signalling, NOD-like receptor signalling and, strikingly, IL-17 signalling in allergen-treated pups (Fig. 2b and Extended Data Fig. 2e). Consistently, clinical immunophenotyping of human early-onset atopic dermatitis lesions and adult atopic dermatitis lesions has revealed a striking enrichment of type 17 immune pathways in early-onset atopic dermatitis^14,15. Type17 immunity-associated gene signatures are also enriched in infants that are at high risk to develop atopic dermatitis¹⁶. Indeed, HDM challenge led to an upregulation of type 17 cytokines, Il17a/f and Il2, as well as type 1 and 2 cytokines, Ifng and Il4, but not Il5 and Il13, in pup skin (Extended Data Fig. 2f). ALT-challenged and heat-killed C. albicans-challenged pups also upregulated type 17 cytokines (Extended Data Fig. 2g,h).

Flow cytometric analysis at day 6 of HDM challenge revealed an enrichment of RORγt⁺ type 17 cells in pup, but not in adult skin, and no increase in GATA3⁺ type 2 cells in either the pup or adult skin relative to control (Fig. 2c and Extended Data Fig. 2i). By contrast, and consistent with previous literature^17,18, both pup and adult skin DLNs showed higher frequencies of T_H2 cells following HDM challenge, but no difference in T_H17 cells (Fig. 2c and Extended Data Fig. 2i). This unexpected finding revealed a compartmentalized bifurcation of immunity: early-life skin pathology is propelled by type 17 programs, whereas age-independent T_H2 sensitization proceeds in the DLNs.

Further profiling uncovered that dermal Vγ6⁺γδ type 17 cells constituted 70% of HDM-elicited RORγt⁺ type 17 cells (Extended Data Fig. 3a–d). These innate-like T cells robustly produced IL-17A without ex vivo TCR stimulation (Extended Data Fig. 3e,f). Accordingly, RORγt-deficient (RORγt^gfp/gfp), Rag1-deficient and Tcrd-deficient pups, but not Tcrb-deficient pups, had diminished responses to HDM compared with their heterozygote or wild-type (WT) littermates (Fig. 2d,e and Extended Data Fig. 3g–j). Consistent with our mouse data, and in addition to type 17 cytokines, γδTCR gene sequences were enriched in skin transcriptomics data from early-onset atopic dermatitis lesions compared with both non-lesional skin and adult atopic dermatitis lesions^19,20 (Extended Data Fig. 4a).

Following HDM challenge in pups, we observed a marked expansion of IL-17⁺ γδ T cells (γδ T17 cells) that intercalated into the epithelium (yellow arrows), suggesting that direct γδ T17 cell–epithelial crosstalk drives skin pathology (Extended Data Fig. 4b). Therefore, we generated epidermal-specific IL-17RC-deficient pups (Krt14^Cre;Il17rc^fl/fl; herein called Il17rc^EKO). The loss of epidermal IL-17RC phenocopied the RORγt^gfp/gfp, Rag1-deficient and Tcrd-deficient pups and abrogated the response to HDM (Fig. 2f and Extended Data Fig. 4c).

In worm infections and hapten-induced inflammation, IL-17 signalling paradoxically promotes the T_H2 response^21,22. Thus, we tested the link between the acute γδ type 17 reaction in pups and T_H2 induction in the lymph node. IL17a/f^−/− pups had two times fewer T_H2 cells in skin DLNs than WT pups after HDM inoculation (Fig. 2g). Collectively, these data reveal an early-life window of a type 17 dominant immune reactivity to HDM that drive acute skin inflammation and augments long-term T_H2 sensitization in DLNs (Fig. 2h).

Adult skin also has γδ T17 cells, so we wondered why they did not respond to HDM like pup γδ T17 cells. We outlined two possibilities: (1) early-life γδ T17 cells are more intrinsically sensitive to activating cytokines, and (2) regulatory T (T_reg) cells in the adult skin may limit γδ T17 response (Extended Data Fig. 4d). To test the former, we purified and cultured dermal γδ T cells from pup and adult skin. Stimulation with classical type 17 activating cytokines IL-1β and IL-23 elicited comparable levels of IL-17A from pup and adult γδ T cells (Extended Data Fig. 4e). To test the latter, we made use of Foxp3-DTR mice, which specifically and rapidly deplete T_reg cells upon diphtheria toxin administration and only have lethal reactions 1 week after depletion²³. Acute loss of T_reg cells from adult skin did not facilitate a reaction to HDM (Extended Data Fig. 4f). Thus, the divergent early-life and adult responses to HDM could not be traced to γδ T cell intrinsic sensitivity or T_reg action and directed us to examine the upstream innate immune conductors of inflammation.

pii-DCs drive HDM inflammation in early life

Single-cell RNA sequencing (scRNA-seq) of skin antigen-presenting cells revealed an enrichment of CD206⁺ macrophages in pup skin relative to adult (Extended Data Fig. 5a). Yet, clodronate liposome-mediated macrophage depletion did not dampen pup responses to HDM (Extended Data Fig. 5b,c). Loss of mast cells, in Kit^W-sh mice, also did not curb inflammatory responses to HDM (Extended Data Fig. 5d). We next used CD11c-DTR mice to deplete DCs before inflammation (Fig. 3a and Extended Data Fig. 5e,f).

Fig. 3 |. pii-DCs trigger HDM skin inflammation in early life.

a, Representative images of diphtheria toxin (DT)-treated WT and CD11c-DTR mice (left; N = 2), and the corresponding TEWL measurement (right; n = 10 WT and 8 CD11c-DTR) 5 days after HDM inoculation at P4. Scale bars, 0.5 cm. b,c, Experimental schematic (top), representative flow cytometry plots (bottom left) and quantitative pie charts (bottom right) showing A647–HDM-labelled DC subsets in pup (P4) and adult (P60) skin (b) and DLNs (c). Skin samples (b) were analysed 12 h after challenge, and DLNs (c) were analysed 24 h after challenge (N = 2). CD301b⁺ cDC2 dominantly take up HDM and migrate in pup and adult. d–g, Schematic of the experiment (left), representative images (middle) and the corresponding TEWL measurement (right) of HDM-treated Ccr7^fl/fl and Ccr7^DCKO littermates (d; N = 3; n = 12 Ccr7^fl/fl and 7 Ccr7^DCKO); K14-VEGFR3Ig mice and WT littermates (e; N = 2; n = 7 WT and 6 K14-VEGFR3Ig); splenectomized Ltb^+/− and Ltb^−/− littermates (f; N = 2; n = 8 Ltb^+/− and 6 Ltb^−/−); and DMSO-treated or FTY720-treated littermates (g; N = 3; n = 6), 5 days after intradermal HDM inoculation at P4. Lymph nodes are indispensible for HDM-induced neonatal skin inflammation. Scale bars, 0.5 cm. h, Schematic summary of data shown in this figure. Peripheral DCs trigger early life type 17 responses in pup skin. Each dot represents an individual mouse; data are mean ± s.e.m., and statistical significance was assessed using two-tailed unpaired Student’s t-tests (a,d–g). N = independent experiments. Schematics in panels b,d–h were created using BioRender (https://biorender.com).

The skin houses specialized DC subsets including epidermal Langerhans cells, CD103⁺ type 1 conventional dendritic cells (cDC1s) and CD301b⁺ type 2 conventional dendritic cells (cDC2s)²⁴. We bioorthogonally labelled HDM with Alexa Flour 647 (AF647–HDM) and tracked subset-specific antigen uptake. CD301b⁺ cDC2s constituted more than 75% of AF647–HDM-labelled cells in both the skin and DLNs of pups and adult mice (Fig. 3b,c and Extended Data Fig. 6a).

The observed partition of skin type 17 and DLN type 2 responses lead us to hypothesize that the inflammatory reaction to HDM may be induced directly in the skin. We therefore examined the requirement for DLN reactions using three complementary genetic strategies: (1) DC-specific Ccr7 deletion generated by crossing CD11c–Cre with Ccr7^fl/fl mice (CD11c–Cre;Ccr7^fl/fl; herein called Ccr7^DCKO); (2) K14-VEGFR3Ig mice that developmentally lack skin-draining lymphatics and thus do not have DC migration or antigen drainage to DLNs following HDM (Extended Data Fig. 6b); and (3) splenectomized lymphotoxin-β-deficient mice (Ltb^−/−) that lack skin DLNs. Ccr7^DCKO, K14-VEGFR3Ig and splenectomized Ltb^−/− pups all mounted equivocal responses to HDM as their littermate controls (Fig. 3d–f).

FTY720 treatment to block lymph node egress of T cells into the bloodstream and then skin also did not halt HDM-induced inflammation and confirmed that skin-dwelling γδ type 17 cells were sufficient for this response²⁵ (Fig. 3g and Extended Data Fig. 6c). Consistently, skin γδ type 17 cells expressed early activation markers, CD25 and CD69, after HDM challenge (Extended Data Fig. 6d). Allergens stimulate DC migration through the activation of sensory neurons in adult skin¹⁷. Underscoring the dispensability of DC egress for the skin HDM reaction, a loss of TRPV1⁺ sensory neurons or heat inactivation of HDM did not affect responses of pups to HDM (Extended Data Fig. 6e,f). Together, these experiments identify a previously unappreciated form of DC activation and immune induction in the skin, occurring in parallel to the migratory DC response in early life, a state we termed pii-DC (Fig. 3h).

Early-life cDC2s adopt pii state to activate γδ T17 cells

We next used genetic strategies for DC subset-specific ablation to determine which subset (or subsets) mediate the HDM response. Consistent with their dominant uptake of labelled HDM, diphtheria toxin-mediated depletion of CD301b⁺ cDC2s in Mgl2-DTR mice completely protected pups from HDM-induced inflammation (Fig. 4a and Extended Data Fig. 7a). Despite low levels of AF647–HDM uptake in Langerhans cells, loss of Langerhans cells in huLangerin-DTA mice also protected pups from HDM-induced inflammation (Fig. 4b). By contrast, loss of cDC1s in Batf3^−/− mice did not alter responses to HDM (Fig. 4c). Thus, both cDC2s and Langerhans cells simultaneously adopt the pii-DC program and elicit inflammation in response to HDM in early life.

Fig. 4 |. pii-cDC2-derived IL-23 activates γδ T17 cells directly in pup skin after HDM exposure.

a–c, Representative images (left) and corresponding TEWL (right) of HDM-treated and diphtheria toxin-treated WT and Mgl2-DTR (a; N = 4; n = 7 WT and 17 Mgl2-DTR), huLangerin-DTA and WT (b; N = 2; n = 4 WT and 5 huLangerin-DTA) and Batf3^−/− and Batf3^+/− (c; N = 4; n = 11 Batf3^+/− and 10 Batf3^−/−) littermates, 5 days after HDM inoculation at P4. Scale bars, 0.5 cm. d, scRNA-seq of pup skin 9 h after intradermal PBS (Ctrl) or HDM inoculation. UMAP of Ptprc⁺ immune cells, and HDM-induced gene pathways in Langerhans cells (pii-LCs) and cDCs (pii-cDCs; log₂FC ≥ 1 and adjusted P ≤ 0.05). e,f, Representative images (left) and corresponding TEWL (right) of HDM-inoculated pups treated with anti-IL-12p40 (e; N = 2; n = 3 isotype and 4 anti-IL-12p40) or anti-IL-23p35 (f; N = 2; n = 4 isotype and 3 anti-IL-23p35) or isotype antibodies, 5 days after HDM. Scale bars, 0.5 cm. g, Histogram and AF647 mean fluorescence intensity (MFI) of AF647–HDM⁺CD301b⁺ cDC2 from pup and adult skin 12 h after AF647–HDM inoculation (N = 2; n = 4 pup and 3 adult). NC, negative control without HDM injection. h, Lymph node migratory DCs (MHCII^high) as a proportion of CD45⁺ CD90⁻CD19⁻ cells 24 h after HDM (N = 2; n = 4 pup and 3 adult). i,j, Heatmaps of migratory genes (i) and pii-cDC2 activation-associated genes (j) from sort-purified CD301b⁺ cDC2, 9 h after challenge. Key activation genes are marked in red. Each column is one adult or four pooled pups (n ≥ 4). Each dot is an individual mouse; data are mean ± s.e.m. (a–c,e–g). Statistics used two-tailed unpaired Student’s t-test (a–c,e–g), one-tailed Fisher’s exact test with Benjamini–Hochberg false discovery rate correction (d) and multiple unpaired Student’s t-tests (two-stage step-up Benjamini, Krieger and Yekutieli; h). N = independent experiments.

To determine how cDC2s and/or Langerhans cells directly triggered γδ T17 cells in the pup skin, we performed scRNA-seq of all skin cells, 9 h after control PBS and HDM inoculation. Uniform manifold approximation and projection (UMAP) projections segregated into 20 distinct cellular clusters (Extended Data Fig. 7b). Within Ptprc⁺ immune cells, both cDCs and Langerhans cells had notable shifts in UMAP location between the control and HDM groups (Fig. 4d). Pathway analysis of DEGs revealed that pii-Langerhans cells were enriched for type 1 interferon (IFN) and viral signalling pathways, whereas pii-cDCs were enriched for chemokine and cytokine signalling including IL-12 and IL-23 pathways (Fig. 4d).

Inhibition of the type 1 IFN signalling with an anti-IFNAR antibody²⁶ exacerbated skin pathology following HDM (Extended Data Fig. 7c), suggesting a protective role for this pathway in early-life inflammation. Conversely, Il12b^−/− mice were protected from HDM pathology (Extended Data Fig. 7d). Anti-IL-12p40 (encoded by Il12b) and anti-IL-23p19 (encoded by Il23a) blocking antibodies similarly ameliorated HDM pathology (Fig. 4e,f). Furthermore, feature plots confirmed that cDCs and not Langerhans cells express Il12b and, to a lesser extent, Il23a (Extended Data Fig. 7e), indicating that the role of Langerhans cells in driving HDM pathology is independent of direct γδ type 17 cell activation. These findings support a model in which HDM uptake triggers cDC2s to adopt the pii-DC state, produce IL-23, activate γδ T17 cells locally and fuel skin pathology in early life.

Given the γδ type 17 cell-driven inflammation confined to early life, but a shared T_H2 response in both pups and adults induced by HDM, we hypothesized that the pii-DC state is specific to pups, whereas migratory cDC2s would be present in both pups and adults^17,27. Early-life and adult skin CD301b⁺ cDC2s had comparable uptake of HDM (Fig. 4g), confirming that any age-related functional differences were not due to difference in antigen uptake. Consistent with their similar T_H2 upregulation (Fig. 2c), we observed comparable DC migration induced by HDM in both pup and adult skin DLNs (Fig. 4h).

To assess the pii state, we evaluated the transcriptional response of purified cDC2s from pups and adult skin 9 h after HDM or PBS inoculation. The shared DEGs between the adult and early-life groups were enriched for genes associated with the migratory state (Fig. 4i), in line with both groups migrating to lymph nodes^28–30. However, only cDC2s from pup skin had higher expression of Il12b and other key activation genes including antigen-presenting genes (H2-M2 and Cd1d) and chemokines (Ccl12, Ccl17 and Ccl22) in skin, demarcating the pii state (Fig. 4j). By contrast, adult cDC2s failed to express the key activation genes including Il12b required to induce γδ T17 cells in the skin following HDM (Extended Data Fig. 8a,b). The same pattern of pii-cDC2 activation was observed with orthogonal heat-killed C. albicans challenge in pup, but not in adult, skin (Extended Data Fig. 8c,d).

Our results thus far illustrate that the pii-DC state and local initiate type 17 inflammation as a hallmark of early-life sensitivity to HDM. We next asked what confers this capacity in early life. scRNA-seq revealed no rapid, pronounced changes in stromal or epithelial compartments accompanying the emergence of pii-DCs after HDM (Extended Data Fig. 7b). Moreover, the concurrent adoption of the pii-DC state by epidermal Langerhans cells and dermal cDC2s, despite their distinct location in the skin, argued against a microenvironment-restricted cue and instead points to systemic drivers of pii-DC induction in early life.

Hypothalamic–pituitary–adrenal axis enables pii-DC state

To define the underlying mechanism of the pii-DC state, we first performed a time course to precisely determine the age at which the pii-DC phenotype is lost. Signature genes from the pii-cDC2 program (Il12b, Ccl12, Ccl17 and Apol7c), started to weaken at P7 and gradually declined until P14, when the response was comparable with baseline (Fig. 5a). The window between birth and P14 is marked by rapid organismal growth and development, accompanied by dynamic changes in neuroendocrine and growth factors that distinguish this stage from adulthood³¹. Accordingly, we detected significantly lower levels of plasma corticosterone, thyroid hormone triiodothyronine (T3), vitamin D and a higher level of growth hormone (GH) in pups than adults (Fig. 5b). To test whether these factors modulate the pii-cDC2 state, corticosterone, T3 and vitamin D (calcitriol) were administered to pups and GH to adults 30 min before HDM inoculation (Fig. 5c). Of the factors tested, low-dose corticosterone administration in pups effectively suppressed pii-cDC2 activation (Fig. 5c).

Fig. 5 |. Developing HPA and low systemic corticosterone levels in early life enable the pii-DC state.

a, Quantitative PCR (qPCR) of pii-cDC2 activation genes 12 h after PBS or HDM (n = 5 P4, 7 P7, 3 P10 and 5 P14). pii-DC activation wanes by P14. b, Plasma hormones (N = 2; n = 7 corticosterone or T3, n = 4 vitamin D or GH). c, Experimental schematic (left) and qPCR (right) of pii-DC genes in hormones or vehicle groups 12 h post-HDM (N = 2; n = 6 corticosterone, 7 calcitriol, 5 T3 and 3 human GH (hGH)). i.p., intraperitoneal. d, Plasma corticosterone levels across ages (N = 2; n = 14 P4, 10 P7, 3 P10 and P14, and 7 P16 and P60). e, GR expression in pup (P4) and adult skin cDC2s, MFI on plots (N = 2; n = 4). f, UCell score of GR-target genes³⁴ in pup and adult cDC2s from Extended Data Fig. 5a. g, Experimental schematic (top) and heatmaps (bottom) of pii-cDC2 activation-associated genes in purified cDC2 from the indicated groups. Key activation genes are highlighted in red. Each column is one adult or a pool of four or more pups. h, qPCR of Il12b and Il17a in WT and GR^DCKO littermates treated with DMSO or corticosterone 30 min before HDM at P4. Samples were collected 12 h or 24 h after HDM for Il12b (N = 8; n = 23 DMSO + HDM, 10 WT + corticosterone + HDM and 13 GR^DCKO + corticosterone + HDM), and 24 h after HDM for Il17a (N = 6; n = 12 GR^DCKO + corticosterone + HDM and n = 17 for others). Gene expression was normalized to the DMSO + HDM group. i, Schematic of the study findings. In panels b,d,h, each dot is an individual mouse; data are mean ± s.e.m. (a–d,h). Statistics used two-tailed unpaired Student’s t-tests for HDM versus Ctrl per timepoint (a), hormone versus vehicle (c) and panel b; and one-way ANOVA (d,h). N = independent experiments. Schematics in panels c,g–i were created using BioRender (https://biorender.com).

Glucocorticoid production is regulated by the hypothalamic–pituitary–adrenal (HPA) axis and has a major role in both organismal homeostasis and responses to stress³². Concurrent with postnatal maturation of the brain and neuronal circuits, the HPA axis is immature at birth and undergoes postnatal maturation to establish homeostatic levels of circulating glucocorticoids^3,33. Indeed, pups had 10 times lower levels of plasma corticosterone throughout the circadian cycle (Extended Data Fig. 9a). Baseline corticosterone levels progressively increased starting at P7 and reached adult levels by P14–P16 (Fig. 5d). A stress-hyporesponsive period, or the inability to respond to mild stress, is a defining characteristic of the postnatal HPA immaturity³³. To gauge this, we used a new cage stress test, in which pups or adults were singly housed in a new cage for 30 min before evaluating corticosterone levels³³. Adult mice responded to this acute stress by rapidly increasing plasma corticosterone, whereas pups failed to respond entirely (Extended Data Fig. 9b). Accordingly, the adult hypothalamus had lower levels of glucocorticoid receptor (GR; encoded by NR3C1) expression than pups (Extended Data Fig. 9c). The adult hippocampus, known to modulate HPA activity, has structurally defined regions with strong expression of GR, whereas the pup hippocampus had nearly undetectable levels (Extended Data Fig. 9c), highlighting the requirement for postnatal brain development in functional maturation of the HPA.

Glucocorticoid signalling via the GR induces a well-defined gene signature³⁴. Despite expressing comparable levels of GR surface protein, cDC2s from pup skin had lower GR gene expression signature than cDC2s from adult skin at steady state (Fig. 5e,f), These results favoured a model of limited GR ligand availability in early life and raised the tantalizing possibility that transient corticosterone would be sufficient to restrain the pii-cDC2 program and HDM-induced skin pathology. Corticosterone administration led to an acute uptick in plasma corticosterone at 3 h and a rapid return to baseline 6 h after administration (Extended Data Fig. 9d). One low-dose corticosterone injection 30 min before HDM inoculation was sufficient to dampen the pii-cDC2 program and rescue the pups from type 17 inflammatory pathology 5 days later (Extended Data Fig. 9e).

Delving deeper into the mechanism, we asked whether corticosterone restrained the pii-cDC2 state by modulating expression of key activation genes. Towards this end, we sort purified and transcriptionally profiled HDM-challenged pups with and without corticosterone preconditioning and adult CD301b⁺ cDC2s from skin. Corticosterone pre-treatment depressed cytokine, chemokine and antigen-presenting genes, including Il12b, which is responsible for triggering γδ type 17 cells following HDM challenge (Fig. 5g).

To test whether GR signalling directly on DCs is responsible for modulating their activation post-corticosterone, DC-specific GR deletion generated by crossing CD11c-Cre with Nr3c1^fl/fl mice (CD11c–Cre;Nr3c1^fl/fl, herein called GR^DCKO; Extended Data Fig. 9f). Loss of GR on DCs phenocopied the early-life response to HDM and entirely restored Il12b expression in pii-cDC2 and type 17 cytokine production (Fig. 5h and Extended Data Fig. 9g). Without excluding a potential role of GR in T cells, our data reveal that GR signalling in DCs is necessary for corticosterone-mediated regulation of pii-cDC2 activation and type 17 inflammation induced by HDM exposure in early-life skin.

Together, these findings reveal that heightened innate immune sensitivity to allergens in early life is rooted in a previously unappreciated state of DC activation and immune induction in peripheral tissue, without lymph node involvement. This pii-DC state results from a suppressed early-life HPA axis and low basal levels glucocorticoid levels (Fig. 5i), underscoring the essential role of developmental context in understanding mechanisms of inflammation in early life.

Discussion

Sensitization to environmental allergens is a major risk factor for atopic dermatitis, allergic rhinitis, asthma and food allergy. Here we have showed that during the postnatal developmental window, the innate immune system of the skin is tuned for heightened reactivity to allergens. This finding provides a biological rationale for the early onset of allergic and allergen-associated atopic disease, particularly in the skin, and its progression across tissues during childhood.

Our data are consistent with evidence that immune responses in early life are amplified towards cytokines and environmental stimuli^35–37. Like adults, pups generated a T_H2 response in the DLNs after allergen exposure. However, during early life, acute cutaneous inflammation further amplifies this lymph node T_H2 response and imprints a durable, systemic susceptibility to subsequent allergen challenge. Of note, HDM exposure at P14, when mice no longer mount a type 17 inflammatory response, provokes weaker pulmonary immunity than priming at P4. Future work will be required to define how the age at first allergen encounter calibrates the magnitude and type of secondary responses.

Mechanistically, our results extend the classical dendritic cell paradigm. In adult systems, T cell activation is initiated in DLNs by migratory or lymph node resident DCs that capture antigen³⁸. By contrast, pii-DC represents an additional mode of local DC activation that occurs directly in peripheral tissues without requiring lymph node migration. This state may have evolved to ensure rapid immune defence before lymph node organization and adaptive immunity are fully developed^39–42. Of note, pii-DCs engage innate-like γδ T cells, which are enriched in early life. Thus, suggesting that postnatal antigen-presenting cell network is poised to interface with skin unconventional T cells, which are highly cytokine-responsive and poised to mount rapid, broad-spectrum responses.

The loss of the pii-DC state coincides with developmental increases in systemic glucocorticoid levels. Our findings indicate that low endogenous corticosteroid levels in early life permit pii-DC activation and local production of cytokines to trigger inflammation. Selective loss of the GR in DCs recapitulated this state, establishing a causal link between homeostatic glucocorticoid signalling and pii-DC suppression. Indeed, individuals with atopic dermatitis or long COVID show reduced cortisol responses to physiological stress, suggesting that the levels of endogenous corticosteroids act as key regulators of immune tone across distinct inflammatory contexts^43,44.

These findings add to an emerging view that developmental neuroendocrine cues are integral to immune function^32,45. Recent work has shown that maternal immune stress and heightened GR signalling in utero can program fetal mast cell dysfunction, leading to exaggerated responses to mechanical stress in the skin in adult animals⁴⁵. Our study extends this paradigm to the postnatal period, revealing that the developing HPA axis and reduced GR activity in early life underpin a unique window of allergen sensitivity. Together, these findings suggest that precise calibration of GR signalling, neither excessive nor insufficient, is essential for optimal immune function. Beyond glucocorticoids, immune cells express receptors for multiple growth factors and hormones, raising the possibility that additional endocrine factors modulate this formative period of immune reactivity. Understanding how these hormonal networks interact with the maturing immune system may further explain the early-life predominance of allergen-associated atopic and allergic diseases and inform strategies to prevent their long-term consequences.

Methods

Mice

The following mouse strains (strain number) were purchased from The Jackson Laboratory: C57BL/6J (000664), B6.129P2(Cg)-Rorc^tm2Litt/J (Rorγt^gfp/gfp; 007572), B6N.Cg-Tg(KRT14-cre)1Amc/J (018964), B6.Cg- Il17rc^tm1.1Koll/J (031002), B6.129P2-Tcrd^tm1Mom/J (Tcrd^−/−; 002120), B6.129P2- Tcrb^tm1Mom/J (Tcrb^−/−; 002118), B6.FVB-1700016L21RikTg(Itgax-HBEGF/EGFP)57Lan/J (Cd11c-DTR; 004509), B6.129(Cg)-Foxp3^{tm3(Hbegf/GFP)Ayr}/J (Foxp3-DTR; 016958), B6.Cg-Tg(Itgax-cre)1–1Reiz/J (Cd11–Cre; 008068), STOCK Ccr7^tm1.1Iaai/MbogJ (036464), B6.129S7-Rag1^tm1Mom/J (Rag1^−/−; 002216), C57BL/6-Gt(ROSA)26Sor^{tm1(HBEGF)Awai}/J (iDTR; 007900), B6.129- Trpv1^tm1(cre)Bbm/J (Trpv1^Cre/+; 017769), B6(FVB)-Mgl2tm1.1(HBEGF/EGFP)Aiwsk/J (Mgl2-DTR; 023822), B6.Cg-Nr3c1^tm1.1Jda/J (GR^flox; 021021), B6;129-Ltb^tm1Flv/J (Ltb^−/−; 003530), B6.Cg-Kit^W-sh/HNihrJaeBsmJ (Kit^w-sh; 030764), B6.Cg-Il17a/Il17f^tm1.1Impr Thy1^a/J (Il17a/f^−/−; 034140), C.129S1-Il12b^tm1Jm/J (Il12b^−/−; 002694), B6.FVB-Tg(CD207-Dta)312Dhka/J (huLangerin-DTA; 017949) and Tg (Rorc-EGFP)1Ebe (Rorγt-EGFP) mice were a gift from G. Eberl (Institut Pasteur). K14-VEGFR3Ig mice⁴⁶ were from A. Lund (NYU). CD169-DTR (RikenBRC: 04395) mice were from K. M. Khanna (NYU). Mice were bred and maintained under specific pathogen-free conditions at the AAALAC-accredited facilities at NYU Langone Health Center, Icahn School of Medicine at Mount Sinai and Weill Cornell Medicine. Mice were housed in accordance with the procedures outlined in the guide for the care and use of laboratory animals. Mice were housed in a dark–light cycle of 12 h Zeitgeber time (ZT); 0 = 6:00, ZT12 = 18:00) with ad libitum access to food and water under 22–24 °C temperature and 30–70% humidity. Mice of similar age and the same sex were randomly assigned to experimental groups prior to treatment. Investigators were not blinded during experiments. Sample sizes were determined on the basis of the biological variability observed in preliminary assays. Breeding mice were 6–20 weeks of age, and mice used for experiments ranged from postnatal day 0 to postnatal day 60, as indicated. All animal procedures were conducted in accordance with protocols approved by the NYU Langone Health, Icahn School of Medicine at Mount Sinai and Weill Cornell Medicine Institutional Animal Care and Use Committee, and all efforts were made to minimize animal suffering. Age-matched and sex-matched littermate controls were used for all in vivo experiments.

Allergen and microbial challenge

HDM extracts from Dermatophagoides pteronyssinus (XPB82D3A2.5) and Dermatophagoides farinae (XPB81D3A2.5), and ALT extract (XPM1D3A2.5) were obtained from Stallergenes Greer. Extracts were reconstituted in PBS and administered via intradermal injection into the dorsal skin at a dose of 20 μg protein per gram of body weight. A fixed allergen concentration was used across animals, regardless of weight. The maximum volume per intradermal injection site was limited to 40 μl; animals requiring larger volumes received injections at multiple dorsal skin sites. Heat inactivation of HDM was performed by incubation at 95 °C for 60 min¹⁸. Lipopolysaccharide from Escherichia coli O55:B5 (L2880, Millipore Sigma) was reconstituted in PBS and injected intradermally at 400 ng g⁻¹ body weight. S. aureus (USA300 LAC strain) was cultured in tryptic soy broth, washed, resuspended in PBS and heat inactivated at 65 °C for 30 min. A dose of 10⁷ colony-forming units in 30 μl PBS was used for intradermal injection in P4 pups; for adult mice, the dose was scaled proportionally to body weight. Heat-killed C. albicans (tlrl-hkca, InvivoGen) was prepared in PBS at 10⁷ colony-forming units per 30 μl and similarly injected intradermally in pups, with scaling applied for adult mice.

In vivo treatments

MC903 treatment.

MC903 (Tocris 2700) were dissolved in ethanol and diluted into 30 μM working solution. MC903 working solution (7.5 μl) per gram of body weight or ethanol control were topically applied onto shaved adult or pup mice back skin daily for 9 days.

Foxp3-DTR mice.

Adult mice were injected intraperitoneally with diphtheria toxin (D0564, Millipore Sigma) at 40 ng g⁻¹ body weight on days −2 and −1 before HDM administration (day 0), and again on days 1 and 3 post-treatment.

Liposome clodronate.

P3 pups received intradermal injections of 50 μl liposome clodronate or control liposomes (LIPOSOMA). Mice were treated with HDM 24 h later.

CD11c-DTR mice.

P3 pups were injected intradermally with diphtheria toxin (0.5 ng g⁻¹) and treated with HDM 24 h later.

Fingolimod (FTY720).

FTY720 (10006292, Cayman Chemical) was dissolved in DMSO and diluted into PBS containing 2% β-hydroxypropyl-cyclodextrin to prepare the working solution. Pups were injected intraperitoneally with 2 mg kg⁻¹ FTY720 at the time of HDM treatment (day 0), and again on days 1 and 3.

Trpv1^Cre-iDTR mice.

Starting at P1, mice were injected intraperitoneally with diphtheria toxin (40 ng g⁻¹) once daily for 3 consecutive days. HDM was administered on P4, followed by an additional diphtheria toxin injection 24 h later.

Hormone treatments.

Calcitriol (HY-10002, MedChemExpress) and T3 (T2877, Millipore Sigma) were prepared in DMSO, diluted in PBS with 2% β-hydroxypropyl-cyclodextrin, and injected intraperitoneally at 5 μg g⁻¹ (calcitriol) or 500 ng g⁻¹ (T3) into P4 pups, 30 min before HDM or PBS treatment. Human GH (100-40-500UG, PeproTech) was administered intraperitoneally to P60 adults at 4 μg g⁻¹, 30 min before HDM treatment. Skin was harvested 12 h later for quantitative PCR.

Corticosterone treatment.

Corticosterone (HY-B1618, MedChemExpress) or vehicle (DMSO/PBS with 2% β-hydroxypropyl-cyclodextrin) was injected intraperitoneally into P4 pups at 15 μg g⁻¹. Thirty minutes later, pups were treated with HDM or PBS. Skin was collected at 12 h for quantitative PCR and at 5 days for histological analysis. Blood was collected at different time points post-treatment to assess plasma corticosterone levels.

Antibody treatments.

Anti-IFNAR1 (MAR1–5A3; BE0241, Bio X Cell) or isotype control (BE0083) was injected intradermally at 24 μg g⁻¹ into pup skin 30 min before HDM on day 0, and again on days 1 and 3. Tissues were collected on day 5 for pathology. Anti-IL-23p19 (BE0313), anti-IL-12p40 (BE0051) or isotype controls (BE0088 and BE0089) were injected intradermally at 60 μg g⁻¹ 30 min before HDM on day 0. Additional antibody doses were given on days 1, 2 and 3. Skin was harvested on day 5 for analysis.

TEWL measurement

Mice were euthanized 5–6 days after allergen or control treatment. For animals older than P14, dorsal hair was removed using electric clippers before dissection. Skin was harvested and positioned epidermal side up for TEWL measurement. A Tewameter TM 300 probe (Courage + Khazaka Electronic) was used to assess water loss. Readings were recorded continuously for 5 min without disturbance. TEWL values were calculated as the average stable reading (g m⁻² h⁻¹) over a 30-s window beginning 3 min after probe placement, when measurements had stabilized.

Bulk RNA-seq and analysis

Full-thickness skin biopsies from pup and adult mice were flash-frozen in liquid nitrogen. Total RNA was extracted using the RNeasy Mini Kit (Qiagen), following the manufacturer’s instructions. RNA-seq libraries were prepared using the NEXTFLEX Rapid Directional RNA-Seq Kit 2.0 and sequenced on the Illumina NovaSeq 6000 platform with 100-bp paired-end reads. For DC transcriptomic profiling, CD301b⁺ cDC2 cells were fluorescence-activated cell sorting (FACS)-sorted from mouse skin, and total RNA was extracted using the RNeasy Plus Micro Kit (Qiagen). Libraries were generated using the Revelo RNA-Seq High Sensitivity Library Preparation Kit and sequenced on the Illumina NovaSeq X Plus system (100-bp paired-end).

Sequencing output was demultiplexed and converted to FASTQ format using Illumina bcl2fastq software. Adapter trimming was performed with Trimmomatic (v0.36) in paired-end mode with the following parameters: minimum read length of 35 bp, trailing quality threshold of 5 and sliding window trimming set to 4:15. Trimmed reads were aligned to the Mus musculus reference genome (mm10/GRCm38) using the STAR aligner⁴⁷. Gene-level counts were obtained, and differential expression was tested using negative binomial generalized linear models implemented in the DESeq2 R package⁴⁸.

For DC datasets, Cook’s distance was used to identify outlier genes; a filtering threshold corresponding to the 95th percentile of the distribution was applied. Correlation heatmaps were generated using deepTools⁴⁹, using the computeMatrix and plotHeatmap functions with bigWig input files. Functional enrichment analysis of DEGs (adjusted P ≤ 0.05) was performed using Enrichr⁵⁰. All computational analyses were conducted using the Minerva supercomputing cluster at the Icahn School of Medicine at Mount Sinai⁵¹.

Skin immunofluorescence and histology imaging

Immunofluorescence staining was performed using protocols adapted from previously described methods⁵². Tissues were fixed in 4% paraformaldehyde (PFA) for 1 h at 4 °C, washed three times with PBS and incubated in 30% sucrose overnight at 4 °C. Samples were then washed in PBS, embedded in optimal cutting temperature compound, frozen, sectioned at 10–14 μm thickness and mounted on Superfrost Plus microscope slides. Sections were blocked and incubated with primary antibodies followed by fluorophore-conjugated secondary antibodies. Nuclei were counterstained with DAPI.

Images were acquired using either a Zeiss Imager.Z2 upright fluorescence microscope equipped with a Plan-Apochromat ×20/0.80 NA objective (420650–9902) or a Leica Stellaris 8 Falcon laser scanning confocal microscope using a HC PL APO ×20/0.75 CS2 air objective. Tiled and stitched sagittal images were generated using Zen software (Carl Zeiss) or Leica LAS X software. Images were analysed with Fiji (ImageJ v1.54f).

For histological analysis, tissue was fixed in 4% PFA for 24 h at 4 °C, transferred to 70% ethanol, and processed for paraffin embedding, sectioning and haematoxylin and eosin staining. Brightfield images were acquired using standard pathology workflows.

Brain tissue preparation and confocal imaging

Mice were deeply anaesthetized with an overdose of pentobarbital (Euthasol; 390 mg ml⁻¹ pentobarbital and 50 mg ml⁻¹ phenytoin; 2 μl g⁻¹ body weight) and transcardially perfused with PBS containing 10 mg l⁻¹ heparin (H3393, Sigma-Aldrich), followed by 4% PFA. Brains were immediately dissected and post-fixed in 4% PFA overnight at 4 °C. Fixed tissue was rinsed in PBS and sectioned sagittally at 100 μm using a vibrating microtome (Leica VT1000 S). Free-floating sections were blocked and incubated with primary and corresponding secondary antibodies, then mounted with Fluoromount-G containing DAPI (00-4959-52, Thermo Fisher Scientific).

Imaging was performed on a Zeiss LSM 800 confocal microscope using either a ×10 Plan-Apochromat 0.45 NA air objective (512 × 512 resolution) for montage acquisition or a ×40 Plan-Neofluar 1.3 NA oil objective (1,024 × 1,024 resolution) for high-resolution imaging. Z-stacks were collected at 12.6-μm intervals for ×10 images and 7.2-μm intervals for ×40 images. All samples compared within an experiment were acquired using identical imaging parameters and magnification. Maximum intensity Z-projections were generated using Fiji (ImageJ v1.54 f).

Tissue digestion and flow cytometry

Back skin was excised and digested in Liberase TL (05401020001, Millipore Sigma) as previously described⁵². In brief, approximately 1.5 × 2 cm of skin was minced and incubated in 2 ml PBS containing 55 μM β-mercaptoethanol, 10 mM HEPES and 250 μg ml⁻¹ Liberase TL at 37 °C with shaking for 2.5 h. For IL-17A detection, 1,000× Golgi Plug (51–2301KZ, BD Biosciences) was added directly to the digestion buffer. Lymph nodes were digested in the same buffer for 40 min at 37 °C. Digestions were quenched with 12.5 mM EDTA and passed through 70-μm cell strainers, then washed with PBS containing 4% FBS and 10 mM HEPES (FACS buffer).

Single-cell suspensions were pre-incubated with anti-CD16/32 to block Fc receptors, followed by staining with fluorescent-conjugated and/or oligo-tagged surface antibodies in 100 μl FACS buffer per 10⁷ cells. Cells were then resuspended in DAPI-containing FACS buffer before sorting or analysis. For intracellular staining, cells were first labelled with fixable Live/Dead dye (L34961, Thermo Fisher) before surface marker staining. Transcription factor and intracellular cytokine staining were performed using the Foxp3/Transcription Factor Staining Buffer Set (00-5523-00, Thermo Fisher) and IC Fixation Buffer (00-8222-49, Thermo Fisher), following the manufacturer’s instructions.

For in vitro γδ T cell cultures, skin from P5 pups and adult females was pre-treated with dispase (5 U ml⁻¹; 7913, StemCell) for 30–45 min at 37 °C with shaking to separate the epidermis before Liberase TL digestion. Flow cytometry data were acquired on BD FACS Symphony A5 analyzers (BD Biosciences) and analysed using FlowJo software. Cell sorting was performed using BD FACSAria and Symphony S6 cell sorters. A full list of antibodies is provided in Supplementary Table 1.

scRNA-seq

Sample preparation and subclustering.

CD45⁺ cells were isolated from pooled P5 pups (n = 3) and adult P60 mice (n = 2) from the CD169-DTR background without diphtheria toxin treatment. Following FACS, cells from each age group were labelled with distinct hashtag oligonucleotides, pooled at a 1: 1 ratio, and processed using the 10x Genomics 3′ scRNA-seq platform. After quality control filtering (see below), antigen-presenting cell populations, including macrophages and dendritic cells, were computationally subclustered for downstream analysis.

Library preparation and pooling strategy.

P4 WT pups were injected intradermally with HDM or PBS. Skin tissue was harvested 9 h post-injection, and single-cell suspensions were prepared. Cells from two pups (PBS group) or four pups (HDM group) were pooled per condition to obtain sufficient cell numbers. CD301b⁺ cDC2 cells were enriched from each treatment group to ensure adequate representation in the final mixture. For each condition, cells were combined in a fixed ratio of CD301b⁺ cDC2:other CD45⁺ immune cells:epidermal cells:lineage-negative cells at 1:2:2:2. Each treatment group was processed and sequenced independently.

Computational analysis.

Raw FASTQ files were processed with Cell Ranger (v6.1.2) to demultiplex libraries, align cDNA inserts to the mm10 (GRCm38) reference genome, remove PCR duplicates and invalid barcodes, and generate gene–barcode matrices from reads bearing unique molecular identifiers. Downstream analyses were carried out in R (v4.3.2) with Seurat (v4). Cells expressing fewer than 200 or more than 7,000 genes, or with more than 20% mitochondrial transcripts, were excluded. Data were log normalized, the 2,000 most variable features were selected (FindVariableFeatures, default parameters) and expression values were scaled. Principal component analysis (RunPCA) was performed on the variable gene set; the first 15 principal components were used for graph-based clustering (FindNeighbours, FindClusters; resolution = 0.4) and UMAP visualization (RunUMAP). Cluster identities were assigned by reference to marker genes identified with FindAllMarkers (Wilcoxon rank-sum test, Benjamini–Hochberg adjusted P ≤ 0.05). Differential expression within the Langerhans cell and DC clusters between HDM-exposed and control mice was assessed with FindMarkers. UMAP and feature plots were output from ShinyCell (v2.1.0).

HDM protein labelling

Proteins in HDM extract were fluorescently labelled using the Alexa Fluor 647 Protein Labeling Kit (A20173, Thermo Fisher Scientific), following the manufacturer’s instructions. Following conjugation, unbound dye was removed and the buffer exchanged to PBS using Zeba Spin Desalting Columns (89889, Thermo Fisher Scientific). Final protein concentration was determined using the DC Protein Assay Kit (5000112, Bio-Rad) according to the manufacturer’s protocol.

Ex vivo γδ T cell stimulation

U-bottom 96-well plates were coated with anti-CD3ε antibody (1 μg ml⁻¹; 100340, BioLegend) at 37 °C overnight. Single-cell suspensions were prepared from pooled dermal tissue of pup and adult mice as described above. Immune cells were enriched using CD45 MicroBeads (130-052-301, Miltenyi Biotec) according to the manufacturer’s protocol, followed by FACS to isolate dermal γδ T cells. A total of 2,500–5,000 γδ T cells were cultured per well in 200 μl of complete IMDM (IMDM supplemented with 10% FBS, 1 mM non-essential amino acids, 1 mM sodium pyruvate, 55 μM β-mercaptoethanol, 10 mM HEPES and 1 mM GlutaMAX). All wells received 1 μg ml⁻¹ anti-CD28 (102116, BioLegend); control wells received PBS, and cytokine stimulation wells received 20 ng ml⁻¹ IL-1β and 20 ng ml⁻¹ IL-23. After 72 h, Golgi Plug (51–2301KZ, BD Biosciences) was added for the final 5 h of culture. Cells were harvested and processed for intracellular IL-17A staining by flow cytometry.

Quantitative PCR

Full-thickness skin biopsies were flash-frozen in liquid nitrogen and stored at −80 °C until RNA extraction. Total RNA was isolated using the RNeasy Fibrous Tissue Mini Kit (Qiagen) for adult skin and the RNeasy Mini Kit (Qiagen) for pup skin, following the manufacturer’s protocols. Equal amounts of RNA were reverse transcribed using the Maxima First Strand cDNA Synthesis Kit (K1641, Thermo Scientific). Quantitative PCR was performed using SYBR Green chemistry on a QuantStudio 3 or 6 Real-Time PCR System (Thermo Fisher Scientific). Gene expression was normalized to Actb as the reference gene. Control conditions for each comparison are indicated in the corresponding figure legends. A complete list of primers is provided in Supplementary Table 2.

Plasma hormone and immunoglobulin quantification

All samples were collected between ZT8 and ZT10 (14:00–16:00). Trunk blood from pups was obtained immediately following decapitation, whereas adult blood was collected via retro-orbital bleed within 2 min of restraint. Samples were collected into EDTA-coated tubes and stored at −80 °C until analysis. Plasma corticosterone and vitamin D levels were measured in individual mice using the Corticosterone Enzyme Immunoassay ELISA Kit (K014-H1, Arbor Assays) and the Vitamin D ELISA Kit (501050, Cayman Chemical), respectively, following the manufacturer’s protocols. Pooled plasma from pups and individual adult plasma samples were assayed for T3 and GH using the T3 ELISA Kit (K056-H1, Arbor Assays) and the Mouse Growth Hormone ELISA Kit (NBP3–08143, Novus Biologicals), respectively. Plasma IgE levels were measured using the ELISA MAX Deluxe Set Mouse IgE (BioLegend), according to the manufacturer’s instructions. Data were acquired on FlexStation 3 Microplate Reader and analyzed with SoftMax Pro 7.0 software.

Pup splenectomy

P1 pups were anaesthetized by wrapping in sterile gauze and placing on ice for 10 min. The abdominal area overlying the spleen was sterilized with ethanol, and a 1–3-mm incision was made using a sterile disposable scalpel. Gentle pressure was applied to the right flank to externalize the spleen through the incision. The spleen, along with the associated fat pad, was excised in a single cauterized cut using a high-temperature fine-tip sterile cautery tool (Aaron Bovie Disposable Sterile Cautery, High Temp, Fine Tip). The incision was sealed by manually approximating the skin edges and applying cyanoacrylate-based tissue adhesive (Super Glue Gel, 3 M). Pups were placed on a 50 °C warming pad for 10 min to recover and then returned to their home cage. HDM was administered at P4.

HDM proteolytic activity assay

Heat inactivation of HDM extract was performed by incubating the extract at 95 °C for 60 min. Both native and heat-inactivated HDM samples were diluted to 200 μg ml⁻¹ in reaction buffer (100 mM sodium phosphate, pH 6.0, 10 mM EDTA and 1 mM dithiothreitol) to a final volume of 50 μl. Proteolytic activity was assessed using the fluorogenic substrate Boc-Gln-Ala-Arg-AMC (BML-P237, Enzo; 200 μM in reaction buffer). Substrate solution (50 μl) was added to each HDM preparation, and AMC fluorescence was monitored kinetically using a PerkinElmer EnVision Multimode Plate Reader (2015 model). Protease activity was quantified as the rate of fluorescence increase over the initial 10 min of the reaction. Reaction buffer lacking dithiothreitol served as a negative control.

Lung tissue sample collection

Mice were euthanized by CO₂ inhalation, and bronchoalveolar lavage was performed by intratracheal insertion of a 21-gauge catheter followed by instillation and retrieval of 1 ml sterile PBS. Bronchoalveolar lavage fluid was centrifuged at 500g for 5 min at 4 °C, and the cell pellet was counted and processed for flow cytometry. Mice were then perfused with 10 ml PBS via the right ventricle, and whole lungs were harvested and fixed in 10% neutral-buffered formalin for 24 h. Tissues were processed for paraffin embedding and sectioned for haematoxylin and eosin staining. Mediastinal lymph nodes were collected and digested using Liberase TL (Roche) as described for skin-draining lymph nodes, and single-cell suspensions were prepared for flow cytometric analysis.

Statistical analysis and reproducibility

Data are presented as mean ± s.e.m. Group sizes were determined based on results from preliminary experiments. Mice were randomly assigned to experimental groups; no blinding was performed. Details of statistical tests used for each experiment are provided in the corresponding figure legends. In all figure legends, n refers to biologically independent animals per group, and N indicates the number of independent experimental replicates. For bulk RNA-seq analyses, differential gene expression was assessed using the DESeq2 R package with a cut-off adjusted P ≤ 0.05. All other statistical analyses were performed using GraphPad Prism, DESeq2 or R. No data were excluded from the analyses.

Extended Data

Extended Data Fig. 1 |. Pup skin exhibits hypersensitivity to allergens and fungal stimuli.

a, Representative H&E-staining and TEWL of ethanol (Ctrl)- or MC903-treated mice (N = 2; n = 6 Pup, 3 Adult Ctrl, 4 Adult MC903). b, Epidermal thickness 6 days after HDM/PBS (Ctrl) (n = 3). c, Representative images and H&E-staining of skin 6 days after intradermal ALT/PBS (Ctrl) inoculation at indicated ages (N = 3). d-e, TEWL (d; n = 9 Pup Ctrl, 6 Pup ALT, 3 Adult) and epidermal thickness (e; n = 3) following ALT as (c). f–h, Representative images and H&E-staining of skin 5 days after intradermal HK Candida albicans (C. albicans; f), HK Staphylococcus aureus (S. aureus; g), LPS (h) or PBS (Ctrl) (N = 3) inoculation at P4. Red dashed line (f) demarcates the inflammation. i, Representative images of HDM-/PBS-inoculated pups (N ≥ 3). D, days (D0, start inoculation). j, Representative images and TEWL fold change (HDM versus PBS) across ages, 5/6 days after inoculation (N = 3; n = 6 (P1, 4, 14, 60), 5 (P21, 32)). k, Experimental schematic. Plasma IgE (N = 2; n = 4), skin CD45⁺CD90⁺GATA3⁺ cells and DLNs Th2 frequencies (N = 2; n = 3) were quantified. Ctrl, naïve animals. l, Representative flow plots of mediastinal LNs αβTCR⁺CD4⁺FOXP3⁻ T cells and quantification of GATA3⁺ Th2 proportions (N = 3; n = 3 Ctrl, 6 1^st PBS, 8 1^st HDM). In a, b, d, e, j, k and l, each dot represents an individual mouse, data are mean ± s.e.m. Statistics used multiple unpaired Student’s t-tests (Two-stage step-up Benjamini, Krieger and Yekutieli (BKY)) (a), two-tailed unpaired Student’s t-tests (b, d, e, j, k: GATA3⁺ skin and LNs Th2), one-way ANOVA (k: IgE, l). Scale bars: 100 μm (H&E), 0.5 cm (dorsal skin images). N = independent experiments. Schematics in panels i,k were created using BioRender (https://biorender.com).

Extended Data Fig. 2 |. Transcriptomic profile of skin response to allergens.

a, Volcano plots of RNA-sequencing DEGs from skin 6 days after ALT/PBS (Ctrl) inoculation in pup (P4) or adult. DEGs (| log2FC | ≥ 2, adj P ≤ 0.05) number in red. b, Pairwise sample correlation heatmap. c, Principal component analysis of transcriptomes from a. d, Proportion of DEGs enriched in human diseases (AD: Ewald, D. A. et al.¹¹, Psoriasis: Suarez-Fariñas, M. et al.¹², Contact dermatitis: Dhingra, N. et al.¹³) overlapping with DEGs enriched in ALT-/HDM- inoculated pup skin. e, Venn diagram of genes upregulated (Log2FC ≥ 2, adjusted P ≤ 0.05) in ALT-/HDM- pup skin, and representative shared KEGG pathways. f-h, Cytokines genes qPCR of HDM- (f; N = 3; Il17a: n = 4 Ctrl, 6 HDM; Il17f: n = 6 Ctrl, 6 HDM; Il22: n = 6 Ctrl, 7 HDM, Il4: n = 5 Ctrl, 6 HDM, Il5: n = 8 Ctrl, 9 HDM Il13: n = 5 Ctrl 6 HDM, Ifng: n = 4). ALT- (g; N = 2; n = 5), HKC. albicans- (h; N = 3; n = 8) or PBS (Ctrl)- treated pup skin, 5/6 days after inoculation. The y axis shows ΔΔC values normalized to Ctrl. i, Quantification of GATA3⁺ in CD45⁺CD90⁺ skin cells (N = 2; n = 4 Pup HDM, n = 3 others) and RORγt⁺ Th17 cells in CD45⁺βTCR⁺CD4⁺FOXP3⁻ DLNs cells (N = 2; n = 5 for adult, n = 6 Pup Ctrl, n = 7 Pup HDM), 6 days after HDM/PBS inoculation at P4. In (f–i), each dot represents an individual mouse; data are mean ± s.e.m. Statistics used Wald test with Benjamini–Hochberg correction (a), one-tailed Fisher’s exact test (e), two-tailed unpaired Student’s t-tests (f–h) and multiple unpaired Student’s t-tests (Two-stage step-up BKY) (i). N = independent experiments.

Extended Data Fig. 3 |. Dermal γδ T17 cells are enriched in HDM-inflammation.

a, Gating strategy for RORγt⁺ cell subsets from pup skin 5 days after PBS (Ctrl)- or HDM-treatment. b, RORγt⁺ cell number in day 5 Ctrl- or HDM-challenged pup and adult skin (N = 3; n = 8 Pup, 7 Adult). c, Pie charts of RORγt⁺ lymphocyte subsets proportion from (a) (N = 2). d, γδTCR⁺ RORγt⁺ subset cell numbers (N = 2, n = 5) in day 5 Ctrl- or HDM-challenged pup skin. TCRVγ1,2 and TCRVγ4 double negative cells are labelled TCRVγ6 cells. e-f, Representative flow plot of IL-17A⁺ in CD45⁺CD90⁺ cells (e), and number of IL-17A⁺ cells (n = 5) (f) from pup skins 5 days after Ctrl or HDM challenge (N = 3). g, Representative H&E-stained sections and quantification of epidermal thickness of Rorγt^gfp/gfp and WT littermates 6 days after HDM inoculation at P4. Scale bars, 100 μm, (N = 7, n = 3). h, Representative images of Rag1^+/− and Rag1^−/− littermates and corresponding TEWL measurement 6 days after HDM inoculation at P4. Scale bars, 0.5 cm, (N = 4; n = 28 Rag^+/−, 13 Rag^−/−). i, Representative H&E-stained sections and quantification of epidermal thickness of Tcrd^−/− and WT littermates 5 days after HDM inoculation at P4. Scale bars, 100 μm, (N = 4, n = 5). j, Representative images of WT and Tcrb^−/− and corresponding TEWL measurement 5 days after HDM inoculation at P4. Scale bars, 0.5 cm, (N = 2; n = 4 WT, 5 Tcrb^−/−). Each dot in (b, d, f-j) represents an individual biological animal, data are mean ± s.e.m. Statistics used two-tailed unpaired Student’s t-test in (f-j) and multiple unpaired Student’s t-test (Two-stage step-up BKY) in (b, d). N = independent experiments.

Extended Data Fig. 4 |. Dermal γδ T17 cells drive allergen-induced early life inflammation.

a, TCR transcripts in lesional skin from adult and infant AD patients, normalized to age-matched healthy controls (Renert-Yuval, Y. et al.¹⁹; Del Duca, E. et al.²⁰). b, Representative immunofluorescence images of γδTCR⁺ RORγt⁺ cells in skin from RORγt–GFP transgenic pups 5 days after PBS (Ctrl) or HDM inoculation at P4. Yellow, anti-GFP (RORγt); magenta, anti-TCRδ; blue, DAPI. White dashed line indicates epidermal basement membrane, and white arrows indicate TCRδ⁺ RORγt⁺ cells within the epithelium. Scale bars, 100 μm (main images) and 50 μm (A1 insets). Asterisk indicates non-specific magenta staining (N = 2). c, Representative H&E-stained sections and quantification of epidermal thickness in IL17rc^EKO and WT littermates 6 days after HDM inoculation at P4 (N = 3; n = 7 WT, 6 IL17rc^EKO). Scale bars, 100 μm. d, Representative flow cytometry plots and quantification of FOXP3⁺ cells from resting pup (P4) and adult (P60) skin (n = 3). e, Experimental schematic and representative flow cytometry plots of PBS- or IL-1β + IL-23–treated, purified TCRδ⁺ cells from pup and adult skin (N = 3). f, Experimental schematic (top). Mice were challenged with HDM at day 0; DT was used for Treg depletion. Depletion efficiency of skin Tregs (bottom left) was assessed at day 0 (n = 5). Representative images (bottom middle; N = 3) and corresponding TEWL measurements (bottom right; n = 6) 5 days after HDM inoculation. Scale bars, 0.5 cm. In c, d and f, each dot represents an individual biological animal, and data are presented as mean ± s.e.m. Statistics performed using two-tailed unpaired Student’s t-tests. N = independent experiments. Schematics in panels e,f were created using BioRender (https://biorender.com).

Extended Data Fig. 5 |. Skin macrophages and mast cells are dispensable for HDM-induced inflammation.

a, From left to right: experimental schematic; UMAP plots of pup and adult antigen-presenting cells, and pie charts showing proportions of antigen-presenting cell subsets. The schematic was created using BioRender (https://biorender.com). b, Frequency of CD206⁺ macrophages 24 h after treatment with control liposomes (LipoCtrl) or clodronate liposomes (Clodronate) (N = 2; n = 3). c, Representative images of LipoCtrl- and Clodronate-treated pups and corresponding TEWL measurements (N = 3, n = 8) 5 days after HDM inoculation at P4. Scale bars, 0.5 cm. d, Representative images and corresponding TEWL measurements of HDM-treated Kit^W-sh mast-cell-deficient mice and WT littermates 5 days after inoculation at P4 (N = 3; n = 9 WT, 8 Kit^W-sh). Scale bars, 0.5 cm. e, Frequency of skin CD11c⁺MHCII⁺ dendritic cells 24 h after DT injection in WT and CD11c-DTR pups (N = 2; n = 5 WT, 4 CD11c-DTR). f, Representative H&E-stained sections and quantification of epidermal thickness from DT-injected WT and CD11c-DTR littermates 5 days after HDM inoculation at P4 (N = 2, n = 4). Scale bars, 100 μm. In b–f, each dot represents an individual biological animal; data are mean ± s.e.m. Statistics used two-tailed unpaired Student’s t-tests. N = independent experiments.

Extended Data Fig. 6 |. Dendritic cells trigger HDM- inflammation directly in pup skin.

a, Experimental schematic and representative flow cytometry plots of AF647–HDM-labelled DCs. DCs were isolated from the skin and DLNs of PBS (Ctrl)- or AF647–HDM-injected pups, harvested 12 h (skin) and 24 h (DLNs) post-injection. DN, double negative for CD301b and CD103 (N = 2). The schematic was created using BioRender (https://biorender.com). b, Quantification of MHCII^high migratory DC numbers in pooled DLNs 24 h after PBS treatment of WT mice (Ctrl) and HDM treatment of WT and K14–VEGFR3Ig mice (n = 4). c, Frequency of blood TCR⁺ cells within CD45⁺ cells from DMSO- and FTY720-treated mice (n = 4). d, Representative histograms (top) and quantification of MFI (bottom) of T cell activation markers on dermal γδTCR⁺ cells from Ctrl- and HDM-treated pup skin 5 days after inoculation at P4 (n = 3). e, qPCR analysis (left) of Trpv1 expression in spinal cord from iDTR and Trpv1^Cre/+ iDTR pups after three consecutive DT injections (n = 4); representative images (middle) of DT-injected iDTR and Trpv1^Cre/+ iDTR pups (N = 5) and corresponding TEWL measurements (right; n ≥ 9) 5 days after HDM treatment at P4. Scale bars, 0.5 cm. f, Protease activity (left) of HDM and heat-inactivated HDM (HI-HDM) measured with the Boc–Gln–Ala–Arg–AMC fluorogenic substrate; negative control reaction without DTT (w/o DTT). Representative images (middle) of HDM- and HI-HDM-treated mice (N = 2) and corresponding TEWL measurements (right; n = 6) 5 days after inoculation at P4. Scale bars, 0.5 cm. In b–f, each dot represents an individual biological animal; data are mean ± s.e.m. Statistics used two-tailed unpaired Student’s t-tests (c–e and f, TEWL) and one-way ANOVA (b and f, protease activity). N = independent experiments.

Extended Data Fig. 7 |. pii-cDC2 direct skin inflammation locally.

a, Representative H&E-stained sections and quantification of epidermal thickness in DT treated WT and Mgl2-DTR littermates 5 days after HDM inoculation at P4 (N = 4, n = 5). Scale bar, 100 μm. b, Workflow for scRNA-seq of pup skin cells with enrichment of CD301b⁺ dendritic cells, and UMAP plot of skin cells from PBS (Ctrl)- and HDM-inoculated pups. The schematics was created using BioRender (https://biorender.com). c, Representative images of pups treated with isotype control (Isotype) or α-IFNAR antibody and corresponding TEWL measurements (right; n = 4) 5 days after HDM treatment at P4. Scale bars, 0.5 cm. d, Representative images of Il12b^−/− mice and WT littermates and corresponding TEWL measurements (N = 5, n = 17) 5 days after HDM treatment at P4. Scale bars, 0.5 cm. e, scRNA-seq feature plots of Il12b and Il23a expression in immune-cell subclusters shown in (b). In a, c and d, each dot represents an individual biological animal; data are mean ± s.e.m. Statistics used two-tailed unpaired Student’s t-tests. N = independent experiments.

Extended Data Fig. 8 |. pii-cDC2 activation in pup, but not adult skin after HDM and heat killed C albicans challenge.

a, scRNA-seq feature plots of Il12b, Ccl22, Ccl17 and Apol7c (hereafter referred to as pii-cDC2 genes) 9 h after PBS (Ctrl) or HDM challenge, showing preferential expression in the cDC cluster (outlined in black). b, qPCR analysis of pii-cDC2 genes in pup (P4) and adult skin 12 or 24 h after PBS (Ctrl) or HDM inoculation (n = 5); y axis shows expression values normalized to the Ctrl group of the corresponding age. c-d, qPCR analysis of pii-cDC2 genes in pup and adult skin collected 16 h after intradermal inoculation with HKS.aureus (c) or HKC.albicans (d) and PBS (Ctrl) (n = 6); y axis shows expression values normalized to the Ctrl group of the corresponding age. In b–d, each dot represents an individual biological animal; data are mean ± s.e.m., and statistics used multiple unpaired Student’s t-tests (Two-stage step-up BKY).

Extended Data Fig. 9 |. Immature HPA axis in early life permits pii-cDC2 response to allergens.

a, Plasma Corticosterone at indicated times of the day (N = 2; n = 5 for 4 pm P4, n = 4 for others). b, Plasma Corticosterone of mice in resting or subjected to 30 min novel-cage stress (N = 2; n = 3 P4, 4 P60). c, Representative immunofluorescence images of brain sections (GR; white) and DAPI (blue) (N = 2). Subregions of hippocampus (Yellow). Scale bars, 500 μm (hippocampus) and 50 μm (hypothalamus). d, Experimental schematic and plasma corticosterone levels after DMSO (Ctrl)- or CORT- injection (N = 2; n = 8 for Ctrl, 5 for 3 h and 24 h, 4 for 6 h and 12 h, 7 for 9 h). e, Experimental schematic and qPCR of Il17a at 24 h (n = 3); representative images and corresponding TEWL at 5 days (N = 3; n = 9) of DMSO- and CORT-treated mice after HDM inoculation at P4. Scale bars, 0.5 cm. The schematics were created using BioRender (https://biorender.com). f, GR expression in CD90⁺CD45⁺ lymphocytes and CD301b⁺ cDC2 cells from WT and GR^DCKO skin (n = 4). g, qPCR of pii-cDC2 activation genes in pups skin injected with DMSO or CORT 30 min prior HDM and collected 12/24 h after HDM inoculation at P4; gene expression was normalized to DMSO + HDM group (N = 8; n = 23 DMSO + HDM, 10 (Ccl22) and 9 (Apol7c) for WT + CORT + HDM, 13 (Ccl22) and 12 (Apol7c) for GR^DCKO + CORT + HDM). In (b, e–g), each dot represents an individual biological animal; in (a, b, d, e–g) data are mean ± s.e.m. Statistics used multiple unpaired Student’s t-tests (Two-stage step-up BYK) (a, P60 versus P4) and one-way ANOVA (d, g). two-tailed unpaired Student’s t-tests (e), multiple unpaired Student’s t-tests (Two-stage step-up BKY) (b, f). N = independent experiments.

Supplementary Material

Supplementary information The online version contains supplementary material available at https://doi.org/10.1038/s41586-026-10162-x.

Data availability

All data to support the conclusions in this article can be found in the main text, extended data, supplementary information and source data. All genomic data generated in this study are publicly available at the Gene Expression Omnibus under accession numbers GSE273131, GSE299934 and GSE273161. The M. musculus reference genome (mm10/GRCm38) was used for sequencing alignment. We reanalysed the following publicly available dataset: GSE180542, GSE30999, public-accessible supplementary tables from Ewald et al.¹¹, Dhingra et al.¹³, Renert-Yuval et al.¹⁹ and Del Duca et al.²⁰. Source data are provided with this paper.

Code availability

The computational code used in this work can be found on GitHub (https://github.com/Naiklab/Pup-allergen-hypersensitivity-analysis-scripts).