Immune–stromal dysregulation and senescence in chronic spontaneous urticaria

Graphical abstract

Abstract

Background

Chronic spontaneous urticaria (CSU) involves mast cells and multiple immune pathways, yet mechanisms sustaining disease chronicity remain incompletely defined.

Objective

We sought to delineate immune, stromal, and senescence-associated transcriptional programs in CSU lesional skin.

Methods

Lesional biopsy samples from 28 adults with CSU and 22 matched healthy controls were analyzed by bulk RNA sequencing, guided topic model–based deconvolution, and p16INK4a immunohistochemistry. Differential expression, gene set enrichment, and immune module scoring were performed.

Results

CSU skin showed 449 differentially expressed genes (309 up, 140 down), including IL4R, LAG3, CXCL8, CASP5, and NLRP12 upregulation, with downregulation of vascular/mesenchymal (CAV1, CD34) and circadian regulators (PER1, NR1D1). Deconvolution revealed expansion of helper, cytotoxic, and regulatory T cells, macrophages, and granulocyte modules, with T_H2/T_H17 polarization. Fibroblasts, keratinocytes, and melanocytes displayed proinflammatory, remodeling-competent states. Senescence-associated transcripts were enriched, and p16INK4a staining confirmed higher senescent cell density in CSU dermis and hypodermis (median, 92.1 vs 36.0 cells/mm²; P = .014).

Conclusions

Adult CSU lesions comprise a chronically inflamed immune–stromal niche, integrating T-cell activation, structural cell reprogramming, and cellular senescence. These data extend current models of CSU and identify senescence as a potential amplifier of inflammation and a testable target for future mechanistic and interventional studies.

Chronic spontaneous urticaria (CSU) is a common debilitating skin disorder characterized by recurrent wheals and/or angioedema for more than 6 weeks in the absence of identifiable triggers.¹ Although classically framed as a mast cell–driven disorder, clinical experience and recent mechanistic studies suggest a broader, persistent inflammatory process.^2,3 While newer agents such as omalizumab, dupilumab, and Bruton tyrosine kinase inhibitors have improved outcomes, identifying drivers of chronicity remains key to improving durable disease control.⁴

Histopathologic and transcriptomic analyses have documented perivascular infiltrates of T cells, eosinophils, basophils, and monocytes in CSU lesions, with increased T_H2 and T_H17 cytokines (IL-4, IL-17, IL-31, IL-33) and evidence of endothelial activation.^2,5,6 These findings implicate cross talk between innate and adaptive immunity, but the contribution of structural cells—fibroblasts, keratinocytes, and melanocytes—to sustaining inflammation remains largely unexplored. By analogy to other chronic inflammatory dermatoses, persistent stromal activation can promote extracellular matrix remodeling, immune recruitment, and barrier disruption, raising the possibility of similar processes in CSU.7, 8, 9

Cellular senescence, a state of stable growth arrest accompanied by a proinflammatory senescence-associated secretory phenotype, has emerged as a potential amplifier of tissue inflammation and fibrosis across autoimmune and inflammatory disorders.¹⁰ Senescent stromal or immune cells can reshape tissue architecture, perpetuate immune activation, and impair resolution, but their presence and functional relevance in CSU are unknown.¹⁰

To address these gaps, we combined bulk RNA sequencing (RNA-Seq) with p16INK4a immunohistochemistry to profile immune, stromal, and senescence-associated programs in adult CSU skin. Guided topic modeling (GTM) was used to deconvolve bulk data into cell type proportions and cell-specific expression profiles. This multicompartment analysis identifies coordinated immune activation, T_H2/T_H17 polarization, stromal remodeling, and enrichment of senescence signatures, reframing CSU as a chronic inflammatory dermatosis with durable tissue alterations.

Methods

Study cohort and sample acquisition

We analyzed lesional skin biopsy samples from 28 adults with CSU and 22 age- and sex-matched healthy controls (HCs) by bulk RNA-Seq, computational immune cell deconvolution, and immunohistochemistry. All specimens were obtained under institutional research ethics board approval (REB 2021-7412) with a waiver of informed consent for the use of anonymized archived tissue.

CSU cases met international EAACI/GA²LEN/EuroGuiDerm/APAAACI diagnostic criteria (≥6 weeks of spontaneous wheals and/or angioedema without external triggers) and were at least 18 years of age.¹¹ Exclusion criteria included systemic immunomodulatory therapy (eg, omalizumab, corticosteroids, cyclosporine) within 4 weeks of biopsy, a known autoimmune or systemic inflammatory disorder, or concurrent dermatoses involving the biopsy site (see Table E1 in the Online Repository available at www.jaci-global.org).

All samples consisted of formalin-fixed, paraffin-embedded (FFPE) 4 mm punch or excisional biopsies from trunk or proximal extremities. A board-certified dermatopathologist (M.C.), unaware of the patient’s diagnosis, reviewed hematoxylin and eosin–stained slides to confirm CSU histopathology (eg, superficial dermal edema, perivascular lymphocytic infiltrate, absence of vasculitis) and to exclude histologic mimics.

RNA-Seq data from HCs were sourced from our laboratories, comprising both unpublished skin redundancy tissue from histologically normal adults and previously published datasets generated by our group.¹² All HC specimens originated from sun-protected sites, underwent identical RNA extraction and sequencing protocols as CSU samples, and were integrated after batch correction using variance-stabilizing transformation and ComBat.^13,14

RNA extraction

RNA was isolated from four 10 μm FFPE sections per sample with the RNeasy FFPE Kit (catalog 73504, Qiagen, Hilden, Germany). RNA concentration was measured with a NanoDrop ND-1000 (NanoDrop Technologies, Wilmington, Del), and integrity was assessed on an Agilent Bioanalyzer (Agilent Technologies, Santa Clara, Calif). Samples with DV200 ≥ 36.5% were retained.¹⁵ Libraries were prepared with the TruSeq RNA Access Library Prep Kit. Sequencing was performed on an Illumina NovaSeq 6000 (Illumina, San Diego, Calif) platform to generate 2 × 150 bp paired-end reads, targeting ≥30 million reads per sample.

Bioinformatic processing and differential expression analysis

FASTQ files were processed by the GenPipes RNA-Seq pipeline v4.4.5.¹⁶ Adapter trimming and quality filtering were performed by Trimmomatic v0.36.^17,18 Reads were aligned to the GRCh38 human reference genome by STAR v2.7.8a, and gene-level quantification was conducted with HTSeq v2.0.4 using GENCODE v38 annotations.¹⁷ Genes with low counts were filtered before normalization. Differential gene expression (DEG) analysis was performed by DESeq2 v1.34.0.¹⁹ Genes with absolute log₂ fold change (FC) ≥ 1 and false discovery rate (FDR)-adjusted q < 0.05 were considered differentially expressed.²⁰ Visualization was carried out in ‘ggplot2’ v3.5.2 and ‘pheat map’ v1.0.12. packages in R software (www.r-project.org).^21,22

Gene set and pathway enrichment analysis

Gene set enrichment analysis (GSEA) of Gene Ontology (geneontology.org) biological process terms was conducted by ‘clusterProfiler’ v4.14.6 on a preranked gene list ordered by log₂FC.²³ Significant enrichment was defined as FDR-adjusted q < 0.05.²⁰ Enrichment of senescence-associated transcripts was evaluated using a curated gene set from Shi et al, visualized by ComplexHeat map v2.22.0.^24,25

Cross-disease transcriptomic comparison

To assess the disease specificity of CSU-associated transcriptional programs, we compared bulk RNA-Seq differential expression results from CSU skin with published transcriptomic datasets from other inflammatory skin diseases, including lichen planus, atopic dermatitis, psoriasis, neutrophilic dermatoses, Wells syndrome, and cutaneous lupus erythematosus. Because these datasets were generated using a restricted gene panel (NanoString), comparisons were performed by log₂FC values rather than raw expression levels to minimize platform-specific bias. Pairwise Spearman correlations and principal component analysis were conducted using disease-level log₂FC vectors for inflammatory, senescence, and combined gene signatures, as defined in Seremet et al.²⁶

Cell type and immune module deconvolution

Cell composition was estimated by GTM implemented in SpaTM (custom Rcpp implementation of GTM-decon), trained on a publicly available single-cell RNA-Seq reference of adult human skin containing 20 unique cell types.27, 28, 29 The reference dataset comprised healthy adult skin and did not include CSU or other inflammatory dermatoses. The number of highly variable genes used to train the model was tuned by leave-one-out cross-validation. The trained model was then applied to bulk RNA-Seq data from CSU and HC samples to estimate relative cell type abundances and deconvolute cell-specific gene signatures. For each cell type population with ≥1% (1 significant figure) median proportion, DESeq2 was applied to GTM-inferred cell type–specific profiles, adjusting for sample batch and cell type proportion. Log₂FC shrinkage was applied, genes with Benjamini-Hochberg–adjusted q < 0.05 were considered differentially expressed and were ranked by log₂FC to identify the top DEGs. DEGs were visualized using volcano plots with −log₁₀(P) values capped at 10 to improve visualization. Immune module scores (T_H1, T_H2, T_H17, M0, M1, M2 macrophages, eosinophils, basophils, neutrophils) were calculated from well-established transcriptional markers.^26,30,31 Because basophils, eosinophils, and neutrophils were not annotated as distinct cell types in the reference dataset, basophil-associated activity was inferred by using transcriptional marker modules rather than estimated cell proportions. The untrained model parameters were initialized as a gene-by-topic matrix where each topic represents a module. A binary indicator is used to assign genes to their corresponding module before normalizing each module by its sum. A background topic, comprising non–differentially expressed genes (DEGs; |log₂FC| < 1 and P > .05), was included to regularize module activity estimates. This configuration was used to infer relative immune program activity between CSU and HC samples. Downstream assessment was achieved by computing the Spearman correlation of inferred cell type proportions and the normalized expression of known cell type markers. All computational analyses were conducted by investigators with formal training and expertise in computational biology and biostatistics (Y.L., A.O.).

Immunohistochemistry and image analysis

FFPE tissue sections (4 μm) were stained for p16INK4a with the CINtec p16 Histology Kit (Roche Diagnostics, Mannheim, Germany). Detection was performed with OmniMap Anti-Mouse HRP and visualized with the ChromoMap DAB (3,3′-diaminobenzidine) Kit (Roche Diagnostics). Slides were counterstained with hematoxylin. Additional immunohistochemistry for CD4 and CD8 (Roche Diagnostics) was performed to assess T-cell infiltration.

Whole-slide images were acquired and analyzed in QuPath v0.6.0.³² For p16INK4a, dermal and hypodermal regions were manually annotated, excluding the epidermis. Nuclei were identified by a positive cell detection algorithm applied to optical density–transformed images (pixel size, 0.5 μm), with nuclear area thresholds of 10 to 400 μm² and a DAB detection threshold of 0.1. Shape-based splitting was enabled, and membrane staining was excluded. p16INK4a-positive nuclei were quantified as cells/mm².

Statistical analysis

All analyses were conducted by R v4 software. Continuous variables were compared by Mann-Whitney U test and categorical variables by Fisher exact test.

Code availability

All scripts and workflows are available on request. Analysis scripts involving the GTM pipeline can be found at www.github.com/aosakwe/CSU_GTM/.

Results

Demographics

We analyzed lesional skin biopsy samples from 28 adults with CSU and 22 age- and sex-matched HCs (Fig 1). The CSU group had a mean age of 53.8 ± 20.3 years (mean ± SD; range, 18-91 years), and 64.3% were female (n = 18). The HC group had a similar sex distribution (63.6% female, n = 14) and a slightly older mean age of 58.9 ± 18.8 years (mean ± SD; range, 21-76 years).

Fig 1.

Schematic summarizing study design and analytical workflow. (A) Lesional skin biopsy samples from 28 adults with CSU and skin samples from 22 age- and sex-matched HCs underwent bulk RNA-Seq and p16INK4a immunohistochemistry. (B) GTM trained on reference single cell dataset was used to predict and compare cell type composition of CSU and HC samples based on their bulk expression profiles. (C) GTM also inferred cell type–specific expression profiles, enabling differential gene expression analysis by cell type. (D) For cell types absent from reference data, immune signatures were inferred by initializing GTM with literature-derived markers. (E) Gene set enrichment and p16 staining were applied to evaluate immunosenescence at CSU onset.

Global transcriptional alterations in CSU

Bulk RNA-Seq identified 449 DEGs between CSU and HC skin at FDR q < 0.05, including 309 upregulated and 140 downregulated genes (Fig 2, A, and see Table E2 in the Online Repository available at www.jaci-global.org). Unsupervised clustering of the top 50 DEGs demonstrated clear separation of CSU and HC samples, with consistent intragroup patterns (Fig 2, B).

Fig 2.

Global transcriptional alterations in CSU skin. (A) Volcano plot of DEGs in lesional skin from CSU (n = 28) vs HCs (n = 22). Significantly upregulated and downregulated genes (|log₂FC| ≥ ±1, FDR < 0.05) are shown in red and blue, respectively; nonsignificant genes are shown in gray. (B) Unsupervised hierarchical clustering of top 50 DEGs by absolute log₂FC, with expression values shown as z scores. Samples are annotated by disease status (red = CSU, green = HC), sex (red = female, blue = male) and age.

Among the most upregulated genes were neuronal and neuroimmune mediators, including SNAP25 (synaptosomal vesicle fusion), NEFM and NEFL (axonal structural proteins), and SYNE4 (nuclear–cytoskeletal linker). Immune activation and checkpoint-associated genes were strongly elevated, such as IL4R, CCR3, CD80, CD40, LAG3, and HAVCR2, alongside inflammasome components CASP5 and NLRP12. Myeloid/lymphoid modulators (CD300C), inflammatory signaling molecules (MYD88), cytokine regulators (CISH, PIM1, OSMR), and noncanonical NF-κB pathway components (MAP3K14, BCL3) were also enriched. Chemokines including CXCL8 and CXCL3, histamine–related mediators (HDC), adhesion molecules (ICAM1), matrix remodeling factors (MMP1), and coagulation cascade (F2) were upregulated.

Downregulated genes included matrix inhibitor TIMP2, vascular/mesenchymal markers CAV1, MGP, GNG11, and CD34, and circadian regulators PER1, DBP, NR1D1, and NR1D2. These patterns suggest concurrent immune activation, stromal dysregulation, and impaired circadian control in CSU lesions.

Pathway enrichment analysis

GSEA of Gene Ontology biological processes, performed by a permutation test, revealed significant upregulation (FDR q < 0.05) of immune-related pathways, including chemokine-mediated signaling, leukocyte-mediated cytotoxicity, and regulation of adaptive immune responses (see Fig E1, A, in the Online Repository available at www.jaci-global.org). Downregulated genes were enriched for structural and proliferative processes, such as nucleosome organization, mesenchymal cell proliferation, and elastic fiber assembly (Fig E1, B).

Cell composition changes in CSU

GTM-based cell type deconvolution showed significant expansion of cytotoxic T cells (P = .00037), helper T cells (P = 3.6 × 10⁻⁷), regulatory T cells (P = 8.2 × 10⁻⁹), and macrophages (P = 9.7 × 10⁻⁸) in CSU skin compared to HCs (Fig 3, A-C and G-H). Structural cell types were depleted, with keratinocytes (P = .0012) and fibroblasts (P = .0016) reduced in relative abundance (Fig 3, E and F). Mast cell composition was unchanged (Fig 3, D).

Fig 3.

Estimated cell type composition in CSU skin. Box plots showing guided topic modeling–derived estimates of immune and structural cell type proportions in lesional skin from CSU (n = 28) and HCs (n = 22). Each dot represents individual sample. P values from Mann-Whitney U tests are shown above each comparison.

As orthogonal qualitative validation of deconvolution-based T-cell estimates, immunohistochemistry for CD4⁺ and CD8⁺ revealed increased accumulation of helper and cytotoxic T cells, respectively, in CSU lesions compared to HC skin. Quantitative assessment further demonstrated weak-to-moderate (CD4) and moderate (CD8), but statistically significant, Spearman correlations between inferred T-cell proportions and normalized expression of corresponding marker genes, with consistent separation between CSU and HC samples (see Figs E3 and E4 in the Online Repository available at www.jaci-global.org), supporting the directionality of the deconvolution-based T-cell findings.

Immune module activity

Immune module scoring revealed strong enrichment of T_H2 (P = .0082) and T_H17 (P = 2 × 10⁻⁵) programs in CSU, whereas T_H1 activity was unchanged (P = .15) (Fig 4, A-C). Both M1 (P = .00077) and M2 (P = 3.2 × 10⁻⁸) macrophage modules were elevated, while M0 signatures were unchanged (P = .85) (Fig 4, D-F). Transcriptional activity of basophil (P = .0082), eosinophil (P = 3.2 × 10⁻⁸), and neutrophil (P = 2.7 × 10⁻⁷) modules were also significantly enriched (Fig 4, G-I).

Fig 4.

Immune module activity in CSU skin. Box plots showing GTM-derived activity scores for T_H1 (A), T_H2 (B), T_H17 (C), M0 macrophage (D), M1 macrophage (E), M2 macrophage (F), basophil (G), eosinophil (H), and neutrophil (I) transcriptional modules in lesional skin from CSU (n = 28) and HCs (n = 22). Signature scores were obtained from bulk RNA-Seq data and reflect relative enrichment of each immune program across samples. P values from Mann-Whitney U tests are shown above each comparison.

Cell-type–specific transcriptional programs in CSU skin

A complete list of the cell type–specific DEGs derived from GTM deconvolution is provided in Table E3 in the Online Repository available at www.jaci-global.org. Genes identified as differentially expressed in both bulk tissue and GTM-inferred cell type–specific analyses are explicitly labeled in Fig 5; some bulk DEGs were not included in cell type–specific analyses as a result of feature selection during model training.

Fig 5.

Cell type–specific differential gene expression in CSU skin. Volcano plots showing differential gene expression results for GTM-derived cell type–specific expression profiles from lesional skin in CSU (n = 28) vs HCs (n = 22). Results are shown for (A) fibroblasts, (B) keratinocytes, (C) melanocytes, (D) helper T cells, (E) macrophages, (F) regulatory T cells, and (G) mast cells. Red and blue points indicate significantly upregulated and downregulated genes, respectively (absolute |log₂FC| ≥ 1, FDR < 0.05); gray points indicate nonsignificant genes.

Fibroblasts

Fibroblasts exhibited a proinflammatory, matrix-remodeling-associated transcriptional state, with upregulation of MMP1, MMP10, ADAMTS4, COL14A1, CH25H, and ICAM1. Innate immune sensing genes (STING1, GBP1, ISG15) and chemokines (CCL2, CCL26) were elevated, along with signaling intermediates (PLCG2, RHOB, PIM1) and chromatin regulators (H1-4, SESN1, RGS16) (Fig 5, A).

Keratinocytes

Keratinocytes showed increased expression of epithelial cytokines (IL20, TSLP, CCL26), inflammatory/stress genes (TNF, S100A7A, SNHG15), and innate immune mediators (HAS3, GADD45A, STEAP). Structural/differentiation markers (KRT6A, KRT16, MT4), signaling regulators (RUNX1, RHCG, PIM1), and chromatin modifiers (H1-4, TXNIP, SESN1) were also upregulated (Fig 5, B). Several transcripts linked to cytoskeletal remodeling (MYLK, TAGLN), lipid metabolism (CH25H, FABP3, CRYAB), and oxidative stress (RHOB, TXNIP) were also differentially expressed, consistent with a reactive epidermal transcriptional state.

Melanocytes

Melanocytes, which are typically immune quiescent, displayed stress/neuroimmune activation markers ICAM1, SNCA, CD24, and LPL, as well as TXNIP, SESN1, KLF6, and inflammatory regulators TSC22D3 and PLAAT3 (Fig 5, C). Modest increases in KRT6A, CTSL, GADD45A, and interferon-responsive transcripts (TNFAIP2, STING1) were also detected, indicating engagement of stress and innate immune–related pathways.

Helper T cells

Helper T cells expressed T_H2/T_H17-polarized signatures with IL23A, TNF, IL9R, CSF2, BATF, GATA3, and RGS16, alongside STING1, ISG15, GBP1, RUNX1, and PIM1 reflecting sustained activation and metabolic engagement (Fig 5, D).

Macrophages

Macrophages showed a hybrid activation profile with proinflammatory chemokines (CCL3L1, CXCL1, CCL8) and cytokines (IL24, TNF) as well as immune regulators (IDO1, SLAMF9), and lipid metabolism genes (CH25H, OLR1) (Fig 5, E).

Regulatory T cells

Regulatory T cells expressed canonical suppressive markers (FOXP3, TIGIT), immune checkpoint receptors (LAG3, TNFRSF4 [OX40], TNFRSF9), inflammatory mediators (IL23A, TNF, IL9R, CH25H), stress/metabolism genes (TXNIP, SESN1, MTHFD2, STING1), and interferon-stimulated transcripts (TRIM22, GBP5), consistent with a chronically stimulated, partially exhausted phenotype (Fig 5, F).

Mast cells

Although mast cells were not enriched in relative proportion, GTM-inferred mast cell transcriptional profiles showed an upregulation of activation- and signaling-associated transcripts in CSU. Upregulated transcripts included cytokines and mediators (TNF, TNFSF9, PTGS2), signaling molecules (PLCG2, PIK3R5, SPHK1), and surface receptors (IL1R1, CD83), indicating proinflammatory activity. Negative feedback regulators such as SOCS3, RGS1, and CBLB were also induced, consistent with active FcεRI-associated signaling pathways. Stress and metabolic response genes (GADD45A, SESN1, MTHFD2, NR4A3) were prominent, suggesting chronic stimulation (Fig 5, G).

Senescence-associated transcriptomic signatures

GSEA showed significant enrichment of senescence-associated genes in CSU (Normalized enrichment score = 1.72, P = 1.66 × 10⁻⁴) (Fig 6, A). Senescence scores, calculated as the mean expression of a curated set of 68 senescence-associated genes (see Table E4 in the Online Repository available at www.jaci-global.org), were significantly higher in CSU (mean, 8.44 ± 0.54) than in HC (mean, 7.79 ± 0.51; P = .00009) (Fig 6, B). Leading-edge genes included matrix metalloproteinases (MMP1, MMP3, MMP10, MMP12, MMP13), chemokines (CXCL1, CXCL8, CCL2), cytokines (TNF, IL-1α, IL-6, IL-15), cell cycle regulators (CDKN2A, CDKN2D, CCNE1, E2F3), adhesion molecules (ICAM1, VCAM1), and epigenetic modulators (CBX2, CBX8) (Fig 6, C).

Fig 6.

Senescence-associated transcriptional and histologic changes in CSU skin. (A) GSEA showing enrichment of senescence-associated transcripts in CSU (n = 28) compared to HCs (n = 22). (B) Box plots of ComBat-corrected senescence scores, calculated as mean expression of curated senescence genes; each dot represents 1 sample. Statistical comparison was performed by Mann-Whitney U test. (C) Heat map of differentially expressed senescence-associated genes (log₂FC, CSU vs HC). (D) Representative immunohistochemistry (IHC) images showing increased p16INK4a⁺ cell density in CSU lesional dermis and hypodermis. Scale bar = 100 μm. (E) Quantification of p16INK4a⁺ nuclei (cells/mm²) by QuPath.

Histologic validation of senescence

Immunohistochemistry for p16INK4a restricted to dermal and hypodermal compartments revealed increased nuclear staining in CSU lesional skin compared to HC (Fig 6, D, and see Fig E5 in the Online Repository available at www.jaci-global.org). Quantitative image analysis showed a median (interquartile range) of 92.1 (335.7) p16INK4a-positive cells/mm² in CSU versus 36 (19.2) in HC (P = .0144) (Fig 6, E).

Activation of immune and senescence pathways across skin disorders

We compared the log₂FCs of bulk DEGs in CSU to those reported for other inflammatory skin disorders (lichen planus, atopic dermatitis, psoriasis, neutrophilic diseases, Wells syndrome, cutaneous lupus erythematosus) (see Fig E2 in the Online Repository available at www.jaci-global.org).²⁶ A correlation and principal component analysis–based analysis showed that CSU most resembled Wells disease and neutrophilic diseases, with the other conditions showing lower correlations and a greater separation in the computed principal components.

Discussion

This study provides an integrated transcriptomic and histopathologic profile of lesional skin in adult CSU, revealing a chronically altered tissue ecosystem characterized by persistent immune activation, circadian disruption, T_H2/T_H17 polarization, stromal remodeling, and enrichment of cellular senescence markers. These findings support the concept that sustained immune–stromal dysregulation and senescence may contribute to disease persistence.

The transcriptomic signature encompassed not only mast cell–associated or mast cell–interacting mediators (IL4R, HDC, ICAM1) but also immune checkpoint molecules (LAG3, HAVCR2), chemokines (CXCL8, CXCL3), and inflammasome components (CASP5, NLRP12), implicating both adaptive and innate immune axes in disease persistence.³³ Downregulation of vascular/mesenchymal markers (CAV1, MGP, CD34) and circadian regulators (PER1, DBP, NR1D1, NR1D2) suggests impaired vascular homeostasis and temporal regulation. Given that mast cell activation and FcεRI expression are under circadian control, disruption of core clock components can abolish time-of-day variation in mediator release and lower activation thresholds during the biological night.^34,35 Such misalignment could plausibly contribute to the nocturnal or early-morning symptom flares reported in CSU and provide rationale for aligning therapeutic timing with circadian biology to enhance symptom control—an approach warranting targeted investigation.^35,36

Cell composition estimates revealed enrichment of helper, cytotoxic, and regulatory T cells, macrophages, and granulocyte-associated modules, accompanied by T_H2 and T_H17 transcriptional programs aligned with previous studies.³⁷ In both helper and regulatory T cells, coexpression of activation markers (BATF, GATA3, FOXP3), immune checkpoints (LAG3, TIGIT, TNFRSF4 [OX40], TNFRSF9), and stress/metabolic regulators (TXNIP, SESN1) suggests chronic stimulation with features of functional adaptation and partial exhaustion. This phenotype may sustain cytokine production while limiting complete effector clearance, creating a self-reinforcing loop in which T cells perpetuate mast cell–dependent and –independent inflammation.³⁸ Although mast cell abundance was unchanged, GTM-inferred mast cell transcriptional profiles showed increased expression of mediators and signaling nodes classically engaged downstream of FcεRI (TNF, PTGS2/COX-2, PLCG2, SPHK1), together with inducible brakes on activation (CBLB, broader negative regulators in FcεRI pathways). These findings are consistent with previous studies showing that mast cell activation in CSU is primarily driven by enhanced signaling and mediator release rather than by mast cell proliferation.³⁹

Macrophages exhibited a hybrid activation profile, combining proinflammatory chemokines (CCL3L1, CXCL1) with regulatory mediators (IDO1, SLAMF9), consistent with ongoing engagement in an inflamed stromal niche. Fibroblasts and keratinocytes were transcriptionally reprogrammed toward proinflammatory, matrix-remodeling states, marked by MMP1, MMP10, ICAM1, ISG15, and CCL2/CCL26, suggesting active roles in immune recruitment and extracellular matrix turnover. Even melanocytes, classically immune quiescent, displayed stress and immune-interaction signatures (ICAM1, SNCA, TXNIP), indicating potential participation in neuroimmune signaling.⁴⁰ Collectively, these findings point to structural and pigment cell compartments as active amplifiers of inflammation, rather than passive bystanders, contributing to the persistence of CSU lesions.

Senescence-associated gene sets were significantly enriched in CSU, encompassing extracellular matrix modifiers, inflammatory cytokines, and cell cycle regulators. Quantitative p16INK4a staining confirmed an increased density of senescent cells in lesional dermis and hypodermis. Because p16INK4a can be induced by nonsenescent cellular stress states, immunohistochemical findings were interpreted in conjunction with a broad senescence-associated transcriptional program rather than as a stand-alone marker of senescence. In other chronic inflammatory dermatoses, senescent cells propagate inflammation through the senescence-associated secretory phenotype.¹⁰ While our cross-sectional design cannot distinguish whether senescence is a cause or consequence of inflammation in CSU, its presence alongside sustained immune activation suggests a potential role in maintaining the inflammatory niche. We hypothesize that chronic exposure to T_H2/T_H17 cytokines, oxidative and metabolic stress, or repeated activation of stromal and immune cells may drive the senescence observed in CSU. Once established, senescent stromal/immune cells can reinforce chronicity by releasing proinflammatory mediators (senescence-associated secretory phenotype), impairing clearance of effector lymphocytes (eg, IL-15⁺ effector memory cells) and disrupting tissue homeostasis, thus positioning senescence as a candidate contributor to chronicity and/or treatment resistance.²

Together, these findings support a model in which CSU lesions constitute a multicompartment inflammatory ecosystem: Mast cells trigger effector responses, but chronicity is sustained by reciprocal interactions with T cells, macrophages, and reprogrammed stromal cells, with senescent cells serving as persistent amplifiers. This framework suggests that therapies directed solely at mast cells may be insufficient in refractory disease, and that approaches targeting immune checkpoints, restoring stromal balance, or clearing senescent cells could complement existing treatments and move the field closer to cure.

Limitations

Our study has limitations. Bulk RNA-Seq lacks spatial resolution, and deconvolution depends on the accuracy of reference datasets. As a result, the cellular source of broadly expressed mediators such as ICAM1 and HDC cannot be definitively assigned and may reflect activation across multiple immune or stromal compartments rather than changes in mast cell abundance per se. In addition, the deconvolution framework was trained on single-cell data from healthy skin, which may incompletely capture disease-specific transcriptional states present in CSU. The cohort comprised adult patients from a single center, limiting generalizability to other CSU subtypes or pediatric populations. Functional validation of exhaustion and senescence phenotypes was beyond the present exploratory scope.

Future studies integrating spatial transcriptomics, multiplexed imaging, and ex vivo functional assays will be essential to delineate immune–stromal interactions and test the causal roles of senescent or checkpoint-expressing cells in CSU persistence. Mechanistic insights from such work could inform clinical trials of checkpoint modulators, stromal-targeted agents, or senolytic compounds, with the ultimate goal of improving outcomes of CSU patients.

Conclusion

Our analysis suggests that CSU is a persistent, multicompartment inflammatory dermatosis in which mast cell–initiated responses are reinforced by T cells, macrophages, stromal remodeling, and senescent cell amplification. The associated T_H2/T_H17 polarization, circadian dysregulation, and senescence challenge the view of CSU as a purely episodic mast cell disorder. These hypothesis-generating findings warrant functional studies to test whether targeting immune checkpoints, restoring stromal balance, or modulating senescence could complement mast cell–directed therapy in refractory disease.

Key messages.

• •CSU lesional skin exhibits persistent T-cell activation, stromal reprogramming, and senescence-associated changes in addition to well-established mast cell activity.
• •These tissue programs may help generate hypotheses regarding disease chronicity and treatment resistance and suggest testable adjunct strategies, such as modulation of T-cell/checkpoint pathways, correction of stromal dysfunction, and senolytic approaches.
• •These findings require mechanistic and interventional validation.

Disclosure statement

Supported by the Canadian Dermatology Foundation and a Fonds de recherche du Québec–Santé (FRQS) Junior 1 Clinician Scientist Award to E.N.

Disclosure of potential conflict of interest: The authors declare that they have no relevant conflicts of interest.