A basophil-fibroblast pro-inflammatory axis fuels type 2 skin inflammation

Abstract

Chronic inflammatory skin diseases arise from dysregulated interactions between tissue-resident and infiltrating cells, the complexity of which hinders disease understanding and treatment. To address this, here, we present a single-cell spatiotemporal atlas of murine type 2 skin inflammation using MERFISH and scRNA-seq. Analyzing ∼430,000 cells during MC903- and oxazolone-induced dermatitis, we identify 39 cell types, including pro-inflammatory fibroblasts that resemble those in human atopic dermatitis. Spatial neighborhood analyses reveal basophils as potent activators of pro-inflammatory fibroblasts, with basophil-derived oncostatin-M (OSM) and IL-4 synergizing fibroblast-mediated feedforward basophil and immune recruitment. While fibroblast-specific deletion of the IL-4Rα receptor disrupts inflammation in vivo, the addition of pharmacologic gp130 inhibition, a core component of the OSM receptor, results in synergistic reduction of inflammation. Our study establishes a basophil-fibroblast circuitry as a critical regulator of type 2 skin inflammation, redefining basophil biology and positioning fibroblasts as dynamic immune regulators and therapeutic targets in inflammatory skin disease.

Keywords: CP: Immunology; MERFISH; atopic dermatitis; basophil; fibroblast heterogeneity; inflammatory skin disease; pro-inflammatory fibroblast; single-cell RNA sequencing; skin inflammation; spatial transcriptomics; type 2 skin inflammation.

Figure 1.. Single-cell spatiotemporal analysis of type 2 skin inflammation

(A) MERFISH and scRNA-seq profiling of mouse type 2 skin inflammation with MC903 and oxazolone (OXA). (B) Split UMAP of MERFISH (left) and scRNA-seq (right), labeled by cell types annotated from clustering. Abbreviations: DC, dendritic cell; FIB, fibroblast; DP, dermal papilla; HF, hair follicle; GL, germinative layer; IRS, inner root sheath; ORS, outer root sheath; KC, keratinocyte; IFE, interfollicular epidermis; Diff, differentiating; LC, Langerhans cell; LEC, lymphatic endothelial cell; MAC, macrophage; MEL, melanocyte; SG, sebaceous gland; SM, smooth muscle; VEC, vascular endothelial cell. (C) Spatial plot of single cells in tissues labeled by cell types. Red box = zoom-in field-of-view below. Dotted box = zoom-in field-of-view for (D). Black scale bars, 1 mm; white scale bars, 500 μm. (D) Spatial single-cell gene expression of markers for each cell type as indicated. (E) Left, dot plot of top markers for each cell type. Right, bar plots for cell type proportion across experimental conditions in MERFISH data. (F) Bar plots of cell type densities in each tissue group across time. See also Figure S1, Tables S1 and S2.

Figure 2.. MERFISH spatially localizes lymphocyte subsets and a pro-inflammatory fibroblast subpopulation

(A) Split UMAP of lymphocyte subpopulations identified in MERFISH (left) and scRNA-seq (right). (B) Dot plot of marker gene expression in lymphocyte subsets. (C) Lymphocyte subset density (cells/mm²) from MERFISH across normal and MC903- and OXA-treated skin. (D) Cytokine expression in scRNA-seq data across immune cell populations in MC903- and OXA-treated skin. (E) Split UMAP of fibroblast subpopulations identified in MERFISH (left) and scRNA-seq (right). (F) Dot plot of marker gene expression in fibroblast subpopulations. (G) Spatial distribution of fibroblast subpopulations in normal and MC903-treated skin. Non-epidermal tissue layers are labeled. (H) Proportion of fibroblast subpopulations across normal and MC903- and OXA-treated skin identified by MERFISH. (I) Subset of mesenchymal cell populations (fibroblasts and pericytes) from integrated human AD scRNA-seq datasets. NL, non-lesional. (J) Annotated clusters of fibroblast and pericyte subpopulations from integrated human AD scRNA-seq datasets. DP/DS, dermal papilla/dermal sheath; ProInf, pro-inflammatory. (K) Violin plots of pro-inflammatory gene expression in human AD pro-inflammatory fibroblasts compared with non-lesional (NL) and healthy fibroblast counterparts. See also Figure S2 and Table S2.

Figure 3.. Type 2 skin inflammation features the emergence of multicellular inflammation-associated neighborhoods of spatially proximal cells

(A) UMAP of MERFISH cells labeled by cell type (top) and neighborhood (bottom) consisting of five normal neighborhoods (NNs) and six inflammation-associated neighborhoods (IANs). (B) Left, heatmap of cell type proportion distribution across neighborhoods. Right, heatmap of neighborhood proportion distribution across tissue samples. (C) H&E of MC903-treated neck skin at indicated day time points (top) and MERFISH of serial sectioned adjacent tissues of cells labeled by neighborhood. (D) Left, proportion of epidermal-associated neighborhoods NN1, IAN1, IAN2, and IAN5 across time in MC903 and OXA tissues. Right, spatial highlights of NN1, IAN1, and IAN2 neighborhood cells. DEJ = dermo-epidermal junction. (E) Keratinocyte and fibroblast cell proportion shifts during epidermal NN1 to IAN1 and IAN2 transition. Inner donut charts indicate NN1 cellular proportions from UMAP in (A), while outer donut chart indicates cellular proportion changes of KCs in IAN1 (left) and both KCs and fibroblasts in IAN2 (right). (F) Spatial position of NN1 keratinocytes that are replaced in the epidermis and DEJ by stressed KCs in IAN1 and IAN2 and of fibroblasts that are replaced by pro-inflammatory fibroblasts in IAN2. (G) Left, proportion of dermis-associated neighborhoods NN3, IAN2, IAN3, and IAN4 across time in MC903 and OXA tissues. Right, spatial highlights of NN3, IAN2, and IAN3 neighborhood cells. DEJ = dermo-epidermal junction. (H) Fibroblast and immune cell proportion shifts during dermal NN3 to IAN2 and IAN3 transition. Inner donut charts indicate NN3 cellular proportions from UMAP in (A), while outer donut chart indicates cellular proportion changes of fibroblasts and immune cells in IAN2 (left) and IAN3 (right). (I) Spatial position of NN3 neighborhood fibroblasts that are replaced by pro-inflammatory fibroblasts in IAN2 and IAN3 and of immune cells that are newly recruited into IAN2 and IAN3. See also Figure S3.

Figure 4.. Cell-cell interaction analysis within neighborhoods identifies basophil-fibroblast IL-4 and OSM signaling

(A) Global signaling of all pathways between cell types in normal and inflammation-associated neighborhoods by scRNA-seq CellChat. For (A–C), arrows indicate the direction of sending cell type to receiving cell type. Thickness of arrows is proportional to aggregate predicted activity among all ligand-receptor pathways. (B) Immune signaling to fibroblast populations in MC903 IAN2 by scRNA-seq CellChat. (C) Fibroblast signaling to immune populations in MC903 IAN3 by scRNA-seq CellChat. (D) Top basophil signaling pathways to pro-inflammatory fibroblasts in MC903 by scRNA-seq CellChat. For (D–E), the thickness of arrows is proportional to the relative contribution of each ligand-receptor pathway to predicted signaling. (E) Top CXCL and CCL chemokine R-L signaling from pro-inflammatory fibroblasts to immune populations by scRNA-seq CellChat. (F) Communication probability of receptor-ligands from top predicted immune populations to PIFB (left) and PIFB to immune populations (right) by scRNA-seq CellChat. (G) Pro-inflammatory fibroblast (red) and basophil (yellow) co-localization (left) along with detected transcripts (right) in d12 MC903 skin by MERFISH. Scale bars, 10 μm. In (G–I), the dotted white line marks the epidermis and hair follicle boundary from the dermis. (H) Representative immunofluorescence of d6 MC903 skin depicting basophil (MCPT8) and pro-inflammatory fibroblast (PDGFRA/TNC) co-localization. Scale bars, 50 μm. (I) Immunofluorescence of pSTAT3 (red), PDGFRA (green), and nuclei (DAPI;, blue) across time, quantified at right. Low-magnification scale bars, 50 μm; high-magnification scale bars, 10 μm. (J) Experimental design for in vitro mouse primary fibroblast (FB) stimulation by recombinant cytokines IL-4 (100 ng/mL) and OSM (20 ng/mL) for 12 h (K) Western blot of cell lysates from mouse in vitro stimulation with relative intensity quantification. Data are mean ± S.E.M. *p < 0.05, **p < 0.01, ***p < 0.001, one-way ANOVA with Dunn’s multiple comparisons test. (L) qPCR of mouse primary fibroblast cytokine stimulation. Data are mean ± S.E.M. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, one-way ANOVA with Dunn’s multiple comparisons test. (M) Experimental design for in vitro human primary FB stimulation by recombinant cytokines IL-13 (100 ng/mL) and OSM (20 ng/mL) for 12 h (N) Western blot of cell lysates from human in vitro stimulation with relative intensity quantification. Data are mean ± S.E.M. *p < 0.05, **p < 0.01, ***p < 0.001, one-way ANOVA with Dunn’s multiple comparisons test. (O) qPCR of human primary FB cytokine stimulation. Data are mean ± S.E.M. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, one-way ANOVA with Dunn’s multiple comparisons test. See also Figure S4 and Table S3.

Figure 5.. Basophil depletion in vivo leads to a reduction in pro-inflammatory fibroblasts

(A) Experimental outline of IgG control vs. anti-FcεRI (MAR-1) in MC903 dermatitis. i.p., intraperitoneal. (B) Basophil quantification in IgG control vs. MAR-1 at d6 MC903. Data are mean ± S.E.M. ***p < 0.001, Mann-Whitney test. Scale bars, 50μm. (C) Representative clinical gross images and H&E staining from d6 MC903. (D) Skin thickness quantification of IgG control vs. MAR-1-treated animals at d6 MC903. Data are mean ± S.E.M. ****p < 0.0001, two-ANOVA. (E) CD45⁺ proportion in IgG control vs. MAR-1-treated mice at d6 MC903. Data are mean ± S.E.M. ****p < 0.0001, Mann-Whitney test. (F) Principal component analysis (PCA) of bulk RNA-seq of tissues. Each point is an individual animal. (G) Gene Ontology (GO) terms enriched in downregulated genes in MAR-1 vs. IgG control-treated animals. (H) Volcano plot of differentially expressed genes in MAR-1 vs. IgG control-treated animals. (I) Representative IF staining of pSTAT3 (red) and PDGFRA (green) in IgG control vs. MAR-1-treated tissues at d6 MC903, with quantification at right. Low-magnification scale bars, 50 μm; high-magnification scale bars, 10 μm. **p = 0.0022, Mann-Whitney test. In (I–K), the dotted white line marks the epidermis and hair follicle boundary from the dermis. (J) Same as (I), but for IL4RA (green) and PDGFRA (red). Scale bars, 50 μm. *p = 0.024, Mann-Whitney test. (K) Same as (I) and (J), but for TNC (red) staining. Scale bars, 50 μm. **p = 0.0023, Mann-Whitney test. (L) qPCR in IgG control vs. MAR-1-treated animals at d6 MC903. Data are mean ± S.E.M. *p < 0.05, Mann-Whitney test. See also Figure S5 and Table S4.

Figure 6.. Fibroblast-specific deletion of Il4ra decreases basophil and other immune recruitment

(A) Experimental outline of Il4ra^fl/fl control and Il4ra^ΔPdgfra mice in MC903 dermatitis. (B) Representative IF staining of IL4RA (green) and PDGFRA (red) Il4ra^fl/fl vs. Il4ra^ΔPdgfra mice. Scale bars, 50 μm. (C) Representative gross clinical images and H&E from Il4ra^fl/fl vs. Il4ra^ΔPdgfra mice at d7 MC903. Scale bars, 50 μm. (D) Skin thickness quantification of Il4ra^fl/fl control vs. Il4ra^ΔPdgfra mice at d7 MC903. Data are mean ± S.E.M. ****p < 0.0001, two-way ANOVA. (E) CD45⁺ proportion in Il4ra^fl/fl vs. Il4ra^ΔPdgfra mice at d7 MC903. Data are mean ± S.E.M. ***p < 0.001, **p < 0.01, Mann-Whitney test. (F) Principal component analysis (PCA) of bulk RNA-seq of tissues. Each point is an individual animal. (G) Gene Ontology (GO) terms enriched in downregulated genes in Il4ra^fl/fl vs. Il4ra^ΔPdgfra mice at d7 MC903. (H) Volcano plot of differentially expressed genes in Il4ra^fl/fl vs. Il4ra^ΔPdgfra mice d7 MC903. (I) Immune cell quantification of Il4ra^fl/fl vs. Il4ra^ΔPdgfra mice at d7 MC903. **p < 0.01, ***p < 0.001, Mann-Whitney test. (J) Representative IHC staining of MCPT8 in Il4ra^fl/fl vs. Il4ra^ΔPdgfra mice at d7 MC903, with quantification at right. Scale bars, 50 μm. ***p = 0.0008, Mann-Whitney test. See also Figure S6, Figure S7, and Table S4.

Figure 7.. Inhibition of gp130, or IL6ST, with fibroblast-specific Il4ra deletion leads to significant improvement in inflammation

(A) Experimental outline of Il4ra^fl/fl vs. Il4ra^ΔPdgfra mice treated with vehicle or SC144 during MC903. (B) Representative IF staining of pSTAT3 (red) and PDGFRA (green) in Il4ra^fl/fl and Il4ra^ΔPdgfra with vehicle or SC144 in d7 MC903, quantified at right. *p < 0.05, **p < 0.01, one-way ANOVA with Dunn’s multiple comparisons test. (C) Representative gross clinical images and H&E from Il4ra^fl/fl vs. Il4ra^ΔPdgfra mice with vehicle or SC144 at d7 MC903. (D) Skin thickness measurements of Il4ra^fl/fl vs. Il4ra^ΔPdgfra mice with vehicle or SC144 at d7 MC903. Data are mean ± S.E.M. *p < 0.05, ***p < 0.001, one-way ANOVA with Tukey’s multiple comparisons test. (E) Skin thickness quantification of Il4ra^fl/fl vs. Il4ra^ΔPdgfra mice with vehicle or SC144 at d7 MC903. Data are mean ± S.E.M. *p < 0.05, ***p < 0.001, one-way ANOVA with Tukey’s multiple comparisons test. (F) CD45⁺ proportion in Il4ra^fl/fl vs. Il4ra^ΔPdgfra mice at d7 MC903. Data are mean ± S.E.M. *p < 0.05, **p < 0.01, one-way ANOVA with Tukey’s multiple comparisons test. (G) Proposed model of basophil-induced OSM and IL-4/13 synergistic stimulation of pro-inflammatory fibroblasts that recruit immune cells to promote inflammation.