Early-life inflammation primes a T helper 2 cell-fibroblast niche in skin

Abstract

Inflammation early in life can prime the local immune milieu of peripheral tissues, which can cause lasting changes in immunological tone that confer disease protection or susceptibility¹. The cellular and molecular mechanisms that prompt changes in immune tone in many nonlymphoid tissues remain largely unknown. Here we find that time-limited neonatal inflammation induced by a transient reduction in neonatal regulatory T cells causes a dysregulation of subcutaneous tissue in mouse skin. This is accompanied by the selective accumulation of type 2 helper T (T_H2) cells within a distinct microanatomical niche. T_H2 cells are maintained into adulthood through interactions with a fibroblast population in skin fascia that we refer to as T_H2-interacting fascial fibroblasts (TIFFs), which expand in response to T_H2 cytokines to form subcutaneous fibrous bands. Activation of the T_H2-TIFF niche due to neonatal inflammation primes the skin for altered reparative responses to wounding. Furthermore, we identify fibroblasts in healthy human skin that express the TIFF transcriptional signature and detect these cells at high levels in eosinophilic fasciitis, an orphan disease characterized by inflammation and fibrosis of the skin fascia. Taken together, these data define a previously unidentified T_H2 cell niche in skin and functionally characterize a disease-associated fibroblast population. The results also suggest a mechanism of immunological priming whereby inflammation early in life creates networks between adaptive immune cells and stromal cells to establish an immunological set-point in tissues that is maintained throughout life.

Extended Data Fig. 1 |. Characteristics of cutaneous inflammation induced by neonatal Treg reduction.

(a) Weights of PBS- and DT-treated FoxP3^DTR mice from the time of treatment to adulthood. n = 3–7 mice per data point, 51 total. (b) Quantification of Tregs in skin and skin-draining lymph nodes (sdLN) following neoTreg depletion. n = 3–9 animals per data point (60 total), 1 experiment per time point. (c–f) Representative flow cytometry and quantifications of immune cell populations in skin and skin-draining lymph nodes during the inflammatory phase of neonatal Treg depletion at P25 and adult Treg depletion at P67 (both 10 days after DT). Gating: CD8 T cells (CD3⁺ CD8⁺); CD4 Teffs (CD3⁺ CD4⁺ FoxP3⁻); Ly6C monocytes (Ly6G⁻ Siglec F⁻ CD64⁺ CD11c⁻ CD11b⁺ Ly6C⁺). n = 18 animals. (g) Representative histology of wildtype neonatal mice treated with PBS or DT and sacrificed 10 days post-injection. (h) Histology of selected organs in ΔneoTreg and control mice at P25. gWAT – perigonadal white adipose tissue. Data are displayed as mean +/− SD from one independent experiment, representative of 2–3 repeats. *p < 0.05, **p < 0.01, ***p < 0.001 (all two-sided); repeated measures two-way ANOVA (a); two-way ANOVA with Šídák multiple comparison test (b-c); one-way ANOVA with Tukey’s multiple comparisons test (c–f).

Extended Data Fig. 2 |. Resolution of inflammation and return to homeostasis following neonatal Treg reduction.

(a) Representative skin histology of control and neonatal Treg-depleted FoxP3^DTR mice at P25, P35, and P50 (10, 20, and 35 days post-DT treatment). Fibrous bands are outlined with dotted lines and regenerating adipocytes are marked by arrows. (b) Abundance of selected inflammatory immune cell populations in skin 0–90 days after neoTreg depletion. n = 3–9 animals per data point (60 total), 1 experiment per time point. (c) T helper and CD8⁺ T cell cytokine production 0–35 days after neoTreg depletion, quantified by intracellular cytokine staining. n = 3–9 animals per data point (51 total), 1 experiment per time point. (d–f) Sample gating and quantification of dermal γδ T cells, ILC2s, and Tc2 cells 0–90 days after neoTreg depletion in Il5^Red5/+; FoxP3^DTR mice. n = 3–9 animals per data point (51 total, d-e); 3–6 animals per data point, (19 total, f), 1 experiment per time point. Data are displayed as mean +/− SD from one independent experiment, representative of 2–3 repeats. *p < 0.05, **p < 0.01, ***p < 0.001 (all two-sided); Student’s t test at selected time points during adulthood (b-e); two-way ANOVA with Šídák multiple comparison test (f). All results were reproduced over 2–3 independent experiments.

Extended Data Fig. 3 |. CD49d marks subcutis-resident Th2 cells that are associated with subdermal eosinophilia and age-specific fibrous band formation.

(a-c) Expression of CD69, CD103 (α_E integrin), and CD49d (α₄ integrin) on skin lymphocytes in ΔneoTreg mice aged to adulthood. Bulk Teff defined as IL5^Red5− FoxP3⁻ CD4⁺ T cells. n = 4 animals. (d–f) ΔneoTreg mice were aged to adulthood, subcutis was separated from the dermis/epidermis, and the two skin fractions were analyzed separately by flow cytometry. Th2 localization over time (d), lymphocyte localization at P50 (e), and CD49d expression in dermal vs. subdermal lymphocytes (f) is shown. n = 4–5 animals per data point, 14 total (d); 4 animals (e); 5 animals (f). (g–i) Quantification of CD49d⁺ Th2 cells (g) and eosinophils (h) in ΔneoTreg and ΔadTreg mice during the inflammatory phase of Treg reduction at 10 days post-DT. Myeloid cell localization in ΔneoTreg mice (i) was quantified by flow cytometry of dissected skin layers. n = 19 animals (g-h); 5 animals (i). (j–l) ΔneoTreg mice were treated every other day with FTY720 or vehicle from P8 to P25. Histology (j), Th2 numbers (k), and eosinophil numbers (l) are shown at P25. n = 7 animals. (m–n) CD49d expression on IL5-Red5⁺ Th2 cells and frequency of CD49d⁺ Th2 cells in ΔneoTreg and ΔadTreg mice during (D10 / P25) and after (D35 / P50) inflammation. n = = 23 animals. Data are displayed as mean +/− SD from one independent experiment, representative of 2–3 repeats. *p < 0.05, **p < 0.01, ***p < 0.001 (all two-sided); repeated-measures ANOVA with Dunnett’s multiple comparison test (a-c, e); 2-way repeated-measures ANOVA with Sidak multiple comparison test (f); Welch’s ANOVA with Dunnett’s multiple comparison test (g-h); Student’s t-test (k-n). All results were reproduced over 2–3 independent experiments.

Extended Data Fig. 4 |. Single-cell transcriptomic characterization of skin stroma in control and ΔneoTreg mice.

(a) Expression of marker genes for skin stromal clusters in control mice. (b) Expression of fibroblast markers and ECM genes in control mouse skin stroma. (c) Expression of immune-related genes in mouse skin stroma, split by control (PBS) and ΔneoTreg (DT) sample. (d) Differential gene expression analysis of Il13ra1⁺ FBs (TIFFs) in ΔneoTreg vs. control mice. (e) Expression of MHC class II-related transcripts in skin stromal clusters (ctrl and ΔneoTreg samples combined). (f) Expression of cell surface markers used to design the Il13ra1⁺ FB flow cytometry gating strategy in main figure 2d. (g) Representative flow cytometry of fibroblast markers PDPN and PDGFRα within subsets of Lin⁻ skin stromal cells from P25 control mice.

Extended Data Fig. 5 |. Il13ra1⁺ FBs / TIFFs are transcriptomically similar to Pi16⁺ fibroblasts found across mouse organs.

Published data were downloaded from a mouse cross-tissue fibroblast atlas, containing twenty-eight 10X scRNAseq datasets across 16 murine tissues that were aligned, filtered, and analyzed with a standardized methodology to minimize batch effects.. (a) Steady-state atlas of all 16 tissues was re-plotted, demonstrating similar clustering to published meta-analysis. (b) Expression of TIFF markers from control skin in cross-tissue clusters defined by the atlas. (c) Expression of the top 50 TIFF markers (ranked by log-fold change) in cross-tissue atlas clusters. (d) Cross-tissue atlas cluster representation in selected tissues. SAT – subcutaneous adipose tissue; VAT – visceral adipose tissue. (e) Enrichment of all skin TIFF markers with logFC > 0.25 (n = 313) among cross-tissue fibroblast atlas clusters, calculated using the Seurat AddModuleScore function. (f) Geneset enrichment analysis of the skin TIFF gene set among cross-tissue fibroblast atlas clusters.

Extended Data Fig. 6 |. Anatomic characterization of Il13ra1⁺ fibroblasts (TIFFs).

(a) IF microscopy of mouse back, ear, and tail skin to identify fascia (CD26), adipocytes (PLIN1), and skeletal muscle (MyoIV). (b) Variable layering of fascia, adipocytes, and skeletal muscle at two different back skin locations at in P22 control mice. (c) Control skin was dissected from P22 mice and the subcutis was manually separated from the dermis and epidermis. Il13ra1⁺ FBs (TIFFs) were quantified in each fraction. (d) Fgf18 expression by scRNAseq in control Lin⁻ skin stromal cells. (e) Confocal microscopy of adult skin from Fgf18^CreRET2; Rosa26^tdTomato mice injected with tamoxifen for five days prior to harvest. Top row: Z-projection with tdTomato signal thresholded for visualization. Bottom row: inset of fascia with original tdTomato fluorescence. All results were reproduced over 2–3 independent experiments.

Extended Data Fig. 7 |. Further characterization of Th2-TIFF interactions in the subdermal niche.

(a) IF microscopy of skin from wildtype mice injected for five days with IL-13 or IL-33 starting at P21 with quantification of fascial proliferation by Ki67. n = 9 animals. (b) Skin histology from ΔneoTreg mice crossed to IL4/13- or IL33-deficient strains at P25 (10 days post-DT). (c) Expression of IL4RA on TIFFs from wildtype neonate (P25) and adult (P50) mice. n = 10 animals. (d–f) Adult mice were injected for 7 days with type 2 cytokines and the indicated cell populations in skin were quantified by flow cytometry. n = 13 animals (d); 7 animals (e-f). (g–h) TIFFs and dermal fibroblasts were sorted from P21 mouse skin and co-cultured with sorted IL-5^Red5+ skin Th2 cells from ΔneoTreg mice for four days. n = 12 samples (h). (i–j) IL-18R1 and ST2 expression in skin lymphocyte subsets with quantification of IL-18R1 expression (see main fig. 3g for ST2). n = 4 animals. (k–l) ΔneoTreg mice were aged to adulthood, the subcutis was separated from the dermis/epidermis, and expression of alarmin receptors was quantified across lymphocyte subsets. n = 5 animals. Data are displayed as mean +/− SD from one independent experiment, representative of 2–3 repeats. *p < 0.05, **p < 0.01, ***p < 0.001 (all two-sided); Welch’s ANOVA with Dunnett Multiple Comparisons Test (a, d, h, j); Student’s t-test (c, e-f); two-way ANOVA with Šídák multiple comparison test (l).

Extended Data Fig. 8 |. Th2-TIFF interactions and niche priming across multiple models of early life subcutaneous inflammation.

(a-d) Mice were immunized subcutaneously (s.c.) with ovalbumin (OVA) and papain. CD49d⁺ Th2 cells (b, c) and TIFFs (d) were quantified at indicated timepoints. n = 25 animals I; 10 animals (d). (e–f) Pdgfra^Cre; Il4ra^f/f mice and Cre⁻ controls were injected with OVA-papain at P8 and P15. IL4RA expression (e) and TIFF frequency (f) were quantified at P25. n = 13 mice. (g–j) Mice were infected s.c. with Nippostrongylus brasiliensis at P8 and boosted with s.c. N. brasiliensis allergen at P15. CD49d⁺ Th2 cells (g, i) and TIFFs (j) were quantified at indicated timepoints. n = 13 mice (h); 12 mice (j). Data are displayed as mean +/− SD from one independent experiment; each experiment was reproduced 2–3 times. *p < 0.05, **p < 0.01, ***p < 0.001 (all two-sided); Student’s t-test (c-d, f, i-j); two-way ANOVA with Šídák multiple comparison test (e).

Extended Data Fig. 9 |. Neonatal Treg reduction primes skin for Th2-driven tissue reparative responses during adulthood.

(a–b) Control and ΔneoTreg mice were aged to adulthood and then treated with two shots DT (identical to neonatal dosing regimen). Representative histology is shown with fibrous bands outlined. (c) Th2 and ILC2 cell numbers in adult ctrl (Il5^Red5/+; FoxP3^DTR + PBS), ΔneoTreg (Il5^Red5/+; FoxP3^DTR + DT), and ΔneoTreg/ΔTh2 (Il5^Red5/+; Rosa26^DTA/+; FoxP3^DTR) mice. n = 16 animals. (d) Wound bed ILC2 numbers (gated as CD45⁺ CD3⁻ CD4⁻ CD8⁻ Thy1⁺ IL5^Red5+). n = 7–10 animals per data point in 2 pooled experiments per time point (108 total). (e–n) Control, ΔneoTreg, and ΔneoTreg/ΔTh2 mice were aged to adulthood and subjected to full-thickness cutaneous wounding. (f) Th2 numbers in skin 0 − 10 days post wounding (dpw). n = 4–11 animals per data point in 2 pooled experiments per time point (89 total). (g) Th2 frequency in paired skin biopsies taken at dpw0 and dpw7. n = 22 animals in one experiment. (h–i) Wound area was quantified daily, fit to a one-phase exponential decay model, and tested for equivalence of the rate constant (h). Rate constants of curves fit to each biological replicate are shown in (i). (j–k) Alternatively activated macrophage (AAM) and eosinophil frequency in dpw10 wounded skin. n = 42 animals in 2 pooled experiments. (l) TIFF abundance in wound beds at 0 −10 dpw. n = 4–16 animals in 2 pooled experiments (97 total). (m–n) Flow cytometric quantification (m) and IF microscopy (n) of CD26^hi TIFFs in wounds at dpw10. n = 39 animals in 2 pooled experiments. (o–u) ΔneoTreg mice were aged to adulthood, wounded, and treated with FTY720 every other day. Wound closure (p-q) and flow cytometric quantifications of Th2 cells, alternatively activated macrophages (AAMs), eosinophils, and TIFFs are shown (r-u) are shown. n = 11 animals. (v-ab) Neonatal mice were immunized with OVA/papain, aged to adulthood, and wounded. Wound closure (w-x) and flow cytometric quantifications of Th2 cells, alternatively activated macrophages (AAMs), eosinophils, and TIFFs are shown (y-ab). n = 22 animals. Data are displayed as mean +/− SD (c-d, f, i-m, q-u, x-ab) or SEM (h, p, w). Results were reproduced over 2 independent experiments (a-d; o-ab) or 4 independent experiments pooled into two separate analyses (e-n). *p < 0.05, **p < 0.01, ***p < 0.001 (all two-sided); ANOVA with Dunnett Multiple Comparisons Test (c, i-k, m, x-ab); least-squares quadratic regression with extra sum-of-squares F test (f, l); mixed-effects analysis with Šídák multiple comparison test (g); nonlinear one-phase exponential decay regression (h, p, w); Student’s t-test (q-u).

Extended Data Fig. 10 |. Characterization of human skin stroma.

(a) Gating strategy used to FACS-purify human Lin⁻ stromal cells (CD45⁻ CD31⁻ Ecad⁻, CD235a⁻). (b) Expression of top 50 murine TIFF (mTIFF) orthologs in human skin stroma, ranked by fold-change. (c) Expression of mouse TIFFs and Thbs4+ FB markers in healthy human stromal clusters. (d) Human and ortholog-converted mouse scRNAseq data were integrated and co-clustered. Cluster identities from single-species analyses (main fig. 4a–b) are shown projected onto the cross-species UMAP. (e) Sample histology of healthy and eosinophilic fasciitis lesional skin with IHC staining for GATA3 and CD4.

Fig. 1 |. Transient neoT_reg cell reduction causes temporary derangement of stromal architecture and lasting T_H2 cell accumulation in the skin subcutis.

a, Schematic of the anatomical layers of murine skin. dWAT, dermal white adipose tissue; PC, panniculus carnosus. b, Schematic of the experiment. FoxP3^DTR mice were treated with either PBS (Ctrl) or DT at indicated postnatal days. c–e, Representative skin histology images of mice euthanized at 10 days (c, d) or 35 days (e), with fibrous bands outlined. Scale bars, 100 μm. f, g, Flow cytometry analysis (f) and quantification (g) of T_H2 cells in skin from adult (P50) Il5^Red5/+;FoxP3^DTR mice following neoT_reg cell reduction. n = 3–7 animals per data point (47 total), one experiment per time point. h, Quantification of skin T helper subsets by intracellular cytokine staining at P50. n = 17 animals. i, Lymphocyte integrin expression in ΔneoT_reg mice at P50. j, k, Representative images (j) and quantification (k) of T_H2 and ILC2 cell localization in skin of ΔneoT_reg Il5^Red5/+; Rosa26^tdTomato/+;FoxP3^DTR mice. Cells were segmented by tdTomato and CD3 expression. Dotted lines denote the boundary between the dermis and subcutis. Scale bar, 100 μm. n = 3 animals. Data are displayed as the mean ± s.d. from one independent experiment, representative of two to three repeats (b–j). NS, not significant. *P < 0.05, ***P < 0.001 (all two-sided); two-way analysis of variance (ANOVA) with Šidák’s multiple comparisons test (g, h) or Student’s t-test (k).

Fig. 2 |. neoT_reg cell reduction causes outgrowth of an Il13ra1⁺ fibroblast population in skin fascia.

a–d, scRNA-seq of Lin⁻ stromal cells (EpCAM⁻CD31⁻CD45⁻) sorted from skin of control and ΔneoT_reg mice at P22. Il13ra1⁺ fibroblast (FB) clusters are circled. n = 2 animals pooled per group. prolif, proliferative. a, Unsupervised clustering. b, Relative cluster abundance in ΔneoT_reg skin. Arrows indicate Il13ra1⁺ FB clusters. c, Gene expression by cluster in control sample. d, Gating strategy for Il13ra1⁺ fibroblasts in P25 control skin. The top plot shows total live cells. e, f, Expression of selected Il13ra1⁺ FB markers identified by scRNA-seq (e) was measured by quantitative PCR of sorted cells (f) to validate the gating strategy in d (bulk SCA1⁺: CD34⁺ SCA1⁺CD26^+/−CD9⁺; other stroma: CD34⁻SCA1⁻; f). n = 4 animals. g, Localization of Il13ra1⁺ fibroblasts stained for CD26 in back skin of P25 mice. Scale bar, 50 μm. h, Flow cytometry quantification of Il13ra1⁺ fibroblasts at P22. n = 12 animals. Data are displayed as the mean ± s.d. from one independent experiment, representative of one (a–c, e) or two-to-three repeats (d, f–h). *P < 0.05, **P < 0.01 (all two-sided); repeated measures ANOVA with Geisser–Greenhouse correction and Dunnett’s multiple comparisons test (f) or Student’s t-test (h).

Fig. 3 |. Reciprocal interactions between skin T_H2 cells and Il13ra1⁺ fibroblasts (TIFFs) drive fascial expansion and T_H2 cell maintenance.

a, b, Representative images (a) and quantification (b) of control mice treated with subcutaneous IL-33 or a complex of IL-4–anti-IL-4 antibody (IL-4c) with IL-13 from P15 to P21. Dotted lines denote fascial expansion. n = 16 animals. Scale bar, 100 μm. c, d, TIFF quantification by flow cytometry after neoT_reg depletion in Il4/Il13^−/−;FoxP3^DTR mice (c) and Il33^−/−;FoxP3^DTR mice (d). n = 9 (c) or 8 (d). e, f, TIFFs and dermal fibroblasts were sorted from P21 mouse skin and co-cultured with sorted IL-5^Red5+ skin T_H2 cells from P28 ΔneoT_reg mice for 4 days. n = 12 (e) or 18 (f) samples. g, Expression of ST2 in skin lymphocytes from ΔneoT_reg mice aged to P50. h, Expression of IL33 in skin stromal populations from Il33^mch/+; FoxP3^DTR mice aged to P50. n = 4 animals. i, Control Il33^mch/+ skin at P50. Inset boxes, magnified views of epidermis/dermis (1) and fascia (2). Scale bar, 100μm. j, Skin T_H2 cell frequency in ΔneoT_reg mice crossed with the IL-33-deficient Il33^mch/mch background. n = 8 animals. k, Skin T_H2 cell frequency in ΔneoT_reg mice that were aged to adulthood and treated with subcutaneous PBS or IL-33. n = 8 animals. Data are displayed as mean ± s.d. from one independent experiment, representative of two to three repeats (a–k). *P < 0.05, **P < 0.01, ***P < 0.001 (all two-sided); Welch ANOVA with Dunnett’s multiple comparisons test (b, e–h) or Student’s t-test (c, d, j, k).

Fig. 4 |. TIFF-like cells are present in healthy human skin and in a fibroinflammatory disease of the fascia.

a, Unsupervised clustering of scRNA-seq from FACS-purified stromal cells in healthy human skin (n = 3 donors) and in back skin of control mice (genes converted to human orthologues). VSMC, vascular smooth muscle cell. b, Expression of mTIFF markers in human skin stroma. c, Module score for the mTIFF signature in human stromal cells. d, Enrichment of the mTIFF signature in each human stromal cluster. Dotted line denotes an odds ratio of 1 (the null hypothesis). e, Integration of stromal cells sequenced from EF skin (n = 1 donor) with healthy controls. f, Frequency of major stromal cell types in healthy skin and EF skin. n = 4 donors. g, h, Images (g) and quantification (h) of GATA3⁺CD4⁺ T-cell localization in EF skin and controls. Areas of expanded fascia are outlined with dotted lines. Scale bar, 800 μm. n = 6 samples. Data are displayed as mean ± s.d. from one independent experiment (a–h). *P < 0.05, ***P < 0.001 (all two-sided); Fisher’s exact test with Benjamini–Hochberg correction (d).