A macrophage-induced subpopulation of mesenchymal cells expressing Fcer1g contributes to wound-induced fibrosis

Abstract

Fibrosis commonly occurs during adult skin wound healing, characterized by excessive extracellular matrix (ECM), leading to scarring. Mesenchymal cells, the primary ECM-producing population, are heterogeneous with varying fibrotic propensity during healing. While pro-fibrotic embryonically derived mesenchymal lineages have been identified, adult mesenchymal cells responsible for fibrosis are not yet fully characterized. In adult mice with conditional macrophage depletion during the early phase of wound healing, wounds exhibit attenuated fibrosis and a reduction in mesenchymal cell numbers. Here we show that early phase macrophage induces a distinct PDGFRα⁺ mesenchymal population expressing Fcer1g. This cell population expands rapidly after injury, shows high proliferative activity, and is largely absent when macrophages are depleted. Targeted ablation of this cell population does not delay wound closure but results in diminished scarring. Human wound datasets identified a transcriptionally conserved FCER1G-expressing mesenchymal subset, suggesting that this pro-fibrotic mesenchymal state is preserved in human wound healing.

The authors identify a macrophage-induced FCER1G⁺ mesenchymal subset that expands after skin injury and drives fibrosis. Targeted ablation of this cell type reduces scarring, implicating fibronectin and highlighting a conserved target in human wounds.

Introduction

Following full-thickness injury, most adult mammalian skin undergoes a repair process that is accompanied by fibrosis. Wound healing is a highly coordinated, multicellular process progressing through sequential stages: inflammation, proliferation and remodeling¹. The early phase of wound repair (day 0–5) is characterized by inflammation, during which the immune response is activated². Immune cells, including macrophages, are rapidly recruited to the injury site, where they clear debris, combat infection, and initiate signaling cascades for tissue repair. Besides this, increasing evidence supports the correlation between macrophages and fibrotic responses: early fetal wounds, which recruit minimal macrophages, heal with negligible fibrosis³; similarly, wounds in neonatal PU.1-null mice, which lack macrophages and neutrophils, heal without fibrosis⁴; in adult LysMCre;iDTR mice, macrophage depletion during the early phase of wound healing leads to delayed wound closure but significantly reduces fibrosis⁵. These findings suggest that macrophages play an important role in modulating fibrosis.

The proliferative phase of wound healing (day 3-14) is marked by the formation of granulation tissue, characterized by aggregation of mesenchymal cells and deposition of ECM components^6–8. This provisional ECM undergoes remodeling and contributes to scar formation. During the healing process, mesenchymal cells are the main driver of fibrosis as they are the predominant cell type synthesizes and remodels the ECM⁹. Prior lineage tracing experiments revealed the heterogeneity among mesenchymal cells^10,11. Embryonic mesenchymal precursors expressing delta-like non-canonical Notch ligand 1 (Dlk1)¹², Engrailed-1 (En1)¹³ or paired related homeobox 1 (Prrx1)¹⁴ generate pro-fibrotic mesenchymal lineages. Targeted ablation of these embryonic-derived lineages reduces cutaneous fibrosis without delaying wound closure^13,14. These results support the concept that different mesenchymal subpopulations are independently responsible for pro-fibrotic or pro-regenerative roles.

One mechanism by which macrophages modulate fibrosis is through the regulation of mesenchymal cells. During wound healing, macrophages promote migration and proliferation of mesenchymal cells¹⁵. Early-phase macrophage depletion in LysMCre;iDTR reduces mesenchymal cell numbers and their proliferation^5,15. In addition, macrophages activate the expression of ECM components in mesenchymal cells through signaling pathways such as TGF-β¹⁶ and interleukin-4 receptor α (IL-4Rα)¹⁷. However, the pro-fibrotic mesenchymal subpopulation regulated by macrophages remains to be explored.

In this work, we perform single-cell RNA sequencing (scRNA-seq) in Macrophage Fas-Induced Apoptosis (MaFIA) mice, in which macrophages are conditionally depleted during the early phase of wound healing. We identify a macrophage-induced mesenchymal subpopulation that expresses Fcer1g, expands early after injury, and contributes to fibrosis. We further show that targeted ablation of this population attenuates scar formation, identifying it as a key cellular driver of wound fibrosis.

Results

A subpopulation of mesenchymal cells is induced by early phase macrophages during wound healing

To conditionally deplete macrophages, we used the MaFIA mouse, which expresses a mutant human FAS death receptor and green fluorescent protein (GFP) under the control of the colony-stimulating factor-1 receptor (Csf1r) promoter¹⁸. AP20187 is a dimerizer that specifically activates the FAS death receptor to initiate cellular apoptosis. In vivo macrophage depletion during early wound healing (day 0–5) was achieved through daily AP20187 injections for five consecutive days, while control mice received vehicle treatment. On the final day of injection, we generated 5-mm full-thickness wounds on the dorsal skin of adult MaFIA mice using an excisional wound splinting model^19,20. Macrophage depletion efficiency was evaluated three days later to reflect depletion status during the early healing response. GFP expression driven by the Csf1r promoter enabled flow cytometry–based quantification of macrophage depletion by measuring the percentage of GFP⁺ cells (Fig. 1a). At 3 days post injury (dpi), AP20187 treatment depleted at least 80% of GFP⁺ macrophages in the skin (Fig. 1b). To further confirm that GFP⁺ depletion reflects loss of macrophages, we performed flow cytometry for F4/80 and CD11b double-positive cells in the bone marrow, spleen, and skin from vehicle- and AP20187-treated MaFIA mice (Supplementary Fig. 1). These analyses confirmed robust depletion of F4/80⁺CD11b⁺ macrophages in the skin (> 75%) (Supplementary Fig. 1i, j) and moderate depletion in the bone marrow and spleen (~ 50%) (Supplementary Fig. 1a, b, e, f). Immunofluorescence staining further validated the loss of F4/80⁺ macrophages in these tissues (Supplementary Fig. 1c, d, g, h, k, l). By 7 dpi, F4/80⁺ macrophages began to repopulate in the skin¹⁸ (Supplementary Fig. 1m–p).

Fig. 1. A subpopulation of mesenchymal cells induced by early phase macrophages during wound healing.

a Macrophage depletion efficiency in 3 dpi wounds from MaFIA. Representative histogram showing percentage of GFP ⁺ cells in 3 dpi wounds treated with vehicle or AP20187 (n = 4 for vehicle, n = 3 for AP20187). b Quantitative analysis of GFP⁺ cell proportion over total live cells in 3 dpi wounds treated with vehicle or AP20187 (n = 3 per condition, unpaired two-sided t test, P < 0.0001). c–e Excisional wounds from MaFIA mice at 21 dpi treated with vehicle or AP20187 (n = 3 per condition). c: macroscopic images with dashed lines outlining the scar area; d: quantitative analysis of scar area (n = 3 per condition, unpaired two-sided t test, P = 0.0069). e Masson’s trichrome staining of sections (50 μm from the center of the wound) with dashed lines outlining the fibrosis area. f UMAP of global cell landscape in MaFIA wounds at 3 dpi. NK cells: Natural Killer cells. g UMAP of mesenchymal subclusters in MaFIA wounds at 3 dpi. h Split UMAP of the mesenchymal subclusters from vehicle and AP20187-treated MaFIA wounds (n = 3 per condition). Data are presented as mean ± standard deviation (s.d.).

We then examined the wound healing outcomes. Consistent with the previous findings in adult LysMCre;iDTR mice⁵, early-phase macrophage depletion led to significantly reduced fibrosis at 21 dpi compared to vehicle-treated controls (Fig. 1c, d and Supplementary Fig. 2a). Masson’s trichrome staining revealed that AP20187-treated wounds exhibited smaller areas of thick, densely packed collagen fibers, indicating attenuated fibrosis (Fig. 1e and Supplementary Fig. 2b). Serial sections demonstrated reduced fibrosis in AP20187-treated wounds spanning from the wound center to the edge, as fibrosis was nearly absent at 150 μm from the center in the depleted group, while still clearly visible at 250 μm in controls (Supplementary Fig. 2c). Moreover, immunofluorescence staining for keratin 10 (K10) and CD31 revealed no significant differences in reepithelialization or angiogenesis between vehicle- and AP20187-treated wounds at 7 dpi and 14 dpi (Supplementary Fig. 3), indicating that the reduction in fibrotic area specifically reflects macrophage-dependent regulation of wound fibrosis. Previous studies have shown that hair follicles in the anagen phase can actively contribute to mouse wound re-epithelialization²¹, making it critical to ensure consistent hair cycle phases across experimental groups. The anagen phase is characterized by active cell proliferation²², as indicated by robust Ki67 expression (Supplementary Fig. 4b), and by activated Wnt/β-catenin signaling²³, evidenced by nuclear β-catenin localization (Supplementary Fig. 5a). Around 7 weeks of age, follicles transition into the telogen phase, marked by the absence of both Ki67 and nuclear β-catenin, and remain in this phase until at least 12 weeks postnatal (Supplementary Figs. 4a, b and 5a). In our study, mice were wounded at approximately 8 weeks of age and analyzed around 11 weeks, during which time hair follicles are expected to be in the telogen phase. Consistent with this, both vehicle- and AP20187-treated groups showed no follicular Ki67 expression throughout the experimental timeline (Supplementary Fig. 4c), indicating limited proliferative activity. Moreover, the absence of nuclear β-catenin in both groups (Supplementary Fig. 5b) further confirmed that all animals remained in the telogen phase during the study period.

To identify mesenchymal subpopulations regulated by macrophages during early wound healing (day 0–5), we performed scRNA-seq on wound tissues from MaFIA mice with or without macrophage depletion at 3 dpi. This time point coincides with substantial macrophage infiltration and active mesenchymal cell proliferation²⁴. UMAP analysis of the scRNA-seq data identified diverse cell populations, which were annotated using published murine skin scRNA-seq data and consensus marker genes from bulk transcriptomic studies^11,25,26 (Fig. 1f, Supplementary Fig. 6 and Supplementary Table 1). The UMAP analysis identified populations including neutrophils (clusters 2, 5 and 15) expressing Mmp8, Retnlg, Efdc21, Ly6g, macrophages (clusters 0 and 1) expressing Pf4, Adgre1, Itgam, and Dendritic/Langerhans cells (clusters 11 and 20) expressing Cd74, Cd80, Cd86. Dendritic/Langerhans cells are a subset of tissue-resident macrophages with antigen-presenting function²⁷, which were hereafter included as a macrophage population.

AP20187 administration significantly reduced macrophage populations (clusters 0, 1, 11, and 20) at 3 dpi (Supplementary Fig. 7a, b). This depletion altered the proportions of several cell types in the wound, including neutrophils, mesenchymal cells and epithelial cells (Supplementary Fig. 7c). This result suggests that macrophages regulate the activation or maintenance of these populations during healing. Among the identified clusters, clusters 4, 9, 14, 24, and 25 were classified as mesenchymal cells, because they expressed canonical mesenchymal markers, including Col1a1, Col1a2, Pdgfrα, Col5a1, Loxl1, Lum, Fbln1 and Cd34^12,28,29. To further examine the specific mesenchymal subpopulation regulated by macrophages, we re-anchored, re-integrated and re-clustered the mesenchymal cells separately. This approach identified 16 mesenchymal subpopulations (Fig. 1g). Among them, subcluster 4 had high expression of Dpp4 (CD26), a marker associated with adipocyte precursor cells (APs)¹⁵ (Supplementary Fig. 8a). Subcluster 6 had high expression of Acta2 (smooth muscle actin), a hallmark of myofibroblasts³⁰ (Supplementary Fig. 8b). The proportions of most mesenchymal subpopulations were not altered by macrophage depletion (Supplementary Fig. 8c). However, subcluster 9 was almost completely absent in wounds from macrophage-depleted mice (Fig. 1h and Supplementary Fig. 8c), suggesting a macrophage-dependent regulation of this mesenchymal subpopulation.

The subpopulation of mesenchymal cells induced by macrophages expresses Fcer1g

To identify unique gene markers that are both highly and broadly expressed in mesenchymal subcluster 9, we used the FindMarkers function in Seurat^31,32. Comparisons were restricted to subcluster 9 versus other mesenchymal subclusters. Genes were evaluated based on expression level (log2 fold change) and prevalence (percentage of cells expressing the gene). Specifically, we calculated a uniqueness score (U-score) using the formula: U-score = log2 fold change × (percent expression in subcluster 7 − percent expression in all other subclusters), with criteria of log2 fold change ≥ 0 and adjusted p-value < 0.05. Using this approach, we identified several genes uniquely enriched in subcluster 9, including Fcer1g, Cd53, Lyz2, and Lcp1, with Fcer1g exhibiting the highest U-score (Fig. 2a, b and Supplementary Data File 1).

Fig. 2. The subpopulation of mesenchymal cells induced by macrophages expresses Fcer1g.

a Differentially expressed (DE) genes in subcluster 9 identified using DESeq2 (Wald test, two-sided; Benjamini-Hochberg adjustment for multiple comparisons; DE: adjusted P-value < 0.05; Non-DE: adjusted P-value ≥ 0.05). b Split UMAP showing Fcer1g expression in mesenchymal subclusters from 3 dpi wounds treated with vehicle or AP20187. c Representative immunofluorescence image showing channel-splitted and merged GFP, FCER1G, and PDGFRα staining in vehicle-treated MaFIA wounds at 3 dpi (n = 3). White arrow: GFP⁺ cells; pink arrow: FCER1G⁺ PDGFRα⁺ cells. d,Representative immunofluorescent images of skin sections stained for FCER1G and PDGFRα during unwounded and wounded stages (n = 2 for unwounded, 15 dpi, and 21 dpi; n = 3 for 3 dpi and 7 dpi). White arrow: FCER1G⁺ PDGFRα⁺ cells. e Stacked bar graph showing the proportional distribution of FCER1G⁺ PDGFRα⁺ cells across PDGFRα⁺ cells and total cells. f, g Comparison of vehicle and AP20187-treated MaFIA wounds at 3 dpi. f Representative immunofluorescence images of FCER1G and PDGFRα staining.White arrow: FCER1G⁺ PDGFRα⁺ cells. g Quantitative analysis of FCER1G⁺ PDGFRα⁺ cells as a percentage of total PDGFRα⁺ cells (n = 3 per condition, unpaired two-sided t test, P = 0.0114). h Representative histogram showing Fcer1g-PE fluorescence intensity in PDGFRα⁺ cells from 3 dpi MaFIA wounds treated with vehicle and AP20187 (n = 3 per condition). i–k NIH/3t3 MDF cultured with or without bone-marrow-derived macrophages isolated from MaFIA mice. i Representative microscopic images showing the MDF-macrophage co-culture. Blue: CellTrace^TM labeled MDF; Green: GFP⁺ macrophage isolated from MaFIA mice. j Representative FACS plot showing the Live/CTV⁺ FCER1G⁺ PDGFRα⁺ cells under MDF-macrophage co-culture. k Quantification of FCER1G⁺ PDGFRα⁺ cells out of total MDF (n = 3 per condition, unpaired two-sided t test, P = 0.0008). Data are presented as mean ± s.d.

Fcer1g encodes the Fc fragment of the IgE receptor (FCER1G). scRNA-seq analysis revealed that among all mesenchymal subpopulations, Fcer1g-expressing mesenchymal cells uniquely expressed hematopoietic markers, particularly the myeloid-related marker Lyz2 (Fig. 2a). Previous studies reported that some wound-associated mesenchymal cells express myeloid-related genes¹¹. Further analysis of published time-course scRNA-seq data on wound healing in mice confirmed that Fcer1g-expressing mesenchymal cells form a distinct cluster during early wound healing, co-expressing Tyrobp, Lyz2, and Lcp1 (Supplementary Fig. 9a)²⁶.

To exclude the possibility that the reduction of Fcer1g-expressing mesenchymal cells was due to Csf1r-driven apoptosis, we examined the expression of Csf1r-driven Fas and GFP in this population. Since FCER1G is widely expressed in non-mesenchymal cells, such as T cells and macrophages^33,34, and immunohistochemistry confirmed its expression in multiple immune cell types within the wound bed (Supplementary Fig. 9b), we used platelet-derived growth factor receptor alpha (PDGFRα) as an additional marker because PDGFRα is broadly expressed in dermal mesenchymal cells during wound healing^35,36. Immunostaining of FCER1G⁺ PDGFRα⁺ mesenchymal cells in MaFIA wounds did not show GFP expression (Fig. 2c), indicating no Csf1r-driven mutant FAS receptor expression. Therefore, their reduction was not a direct consequence of AP20187-induced apoptosis but rather a result of interactions with macrophages.

We next used immunostaining to examine the spatial distribution of Fcer1g-expressing mesenchymal cells in unwounded and wounded skin, with wound bed and periwound strictly defined (Supplementary Fig. 10a). In unwounded skin, FCER1G⁺ PDGFRα⁺ mesenchymal cells accounted for less than 1% of the total cell population and less than 20% of total PDGFRα⁺ mesenchymal cells (Fig. 2d, e and Supplementary Fig. 10b, c). Following injury, their numbers increased significantly, first appearing in the periwound dermis at 3 dpi, peaking at 7 dpi, and gradually declining thereafter, plateauing at a level around two times of the unwounded skin (Fig. 2d and Supplementary Fig. 10b, c). By 7 dpi, this population was enriched in the wound bed with approximately 80% of PDGFRα⁺ cells expressing FCER1G (Fig. 2e and Supplementary Fig. 10b, c). From 7 dpi onward, FCER1G⁺ PDGFRα⁺ cells remained predominantly localized in the wound bed, with fewer cells in the periwound area (Fig. 2d and Supplementary Fig. 10c).

To assess the spatial relationship between FCER1G⁺ PDGFRα⁺ mesenchymal cells and vascular structures, we performed triple immunofluorescence staining for CD31, PDGFRα, and FCER1G. FCER1G⁺ PDGFRα⁺ cells were frequently located in proximity to CD31⁺ endothelial structures but did not colocalize with them, indicating that these mesenchymal cells are perivascularly associated rather than endothelial in identity (Supplementary Fig. 10d).

Having established the spatiotemporal dynamics of Fcer1g-expressing mesenchymal cells during wound healing, we next examined how their presence was affected by macrophage depletion. In AP20187-treated MaFIA mice, FCER1G⁺ PDGFRα⁺ mesenchymal cells were significantly reduced in the periwound area at 3 dpi compared to vehicle-treated controls (Fig. 2f–h and Supplementary Fig. 11a). However, at 7 dpi and 14 dpi, when macrophage numbers rebounded (Supplementary Fig. 1m–p), FCER1G⁺ PDGFRα⁺ mesenchymal cells reappeared in the wounds (Supplementary Fig. 11b–d).

To determine whether macrophages directly induce the emergence of FCER1G⁺ PDGFRα⁺ mesenchymal cells, we performed an in vitro co-culture assay using NIH/3T3 mouse dermal fibroblasts (MDFs) and bone marrow–derived macrophages isolated from MaFIA mice. MDFs were labeled with CellTrace™ Violet (CTV) and co-cultured with macrophages at a 1:1 ratio for 24 h. Flow cytometric analysis of CTV⁺ MDFs revealed the appearance of a distinct FCER1G⁺ PDGFRα⁺ subpopulation in the co-culture but not in MDFs cultured alone (Fig. 2i–k). Quantification confirmed a significant increase in FCER1G⁺ PDGFRα⁺ mesenchymal cells upon macrophage co-culture, indicating that macrophages directly promote the induction of this mesenchymal subset.

Targeted ablation of Fcer1g-expressing mesenchymal cells leads to less fibrosis

To examine the role of Fcer1g-expressing mesenchymal cells in wound healing, we used STOCK Igs7^{tm2(tetO-tdTomato,-DTA*G128D)Rdiez}/AljJ (Dragon-DTA) mouse. The Dragon-DTA mouse allows doxycycline (Dox)-dependent and recombinase-activated expression of diphtheria toxin A (DTA), leading to cell death³⁷. We crossed Dragon-DTA with C57BL/6-Pdgfra^{em1(rtTA)Xsun}/J (Pdgfra-rtTA) and C57BL/6J-Tg(Fcer1g-cre)NIDA168Htz/Mmucd (Fcer1g-Cre) to generate Pdgfra-rtTA; Fcer1g-Cre; Dragon-DTA mouse (Fig. 3a). In this mouse, only cells co-expressing Pdgfrα and Fcer1g undergo DTA-mediated ablation upon Dox administration (Supplementary Fig. 12a). To achieve targeted ablation, Dox was administered through supplemented food and intraperitoneal injections with optimized dosage and regime (Supplementary Fig. 12b, c) to Pdgfra-rtTA; Fcer1g-Cre; Dragon-DTA (hereafter referred to as “depleted” group) and their control littermates. At 7 dpi, at least 90% of FCER1G⁺ PDGFRα⁺ mesenchymal cells were ablated in the wound bed of the depleted group, while FCER1G^- PDGFRα⁺ cells were unaffected (Fig. 3b, c). Immunofluorescence staining for smooth muscle actin (αSMA) at 7 dpi revealed no significant difference between depleted and control wounds (Supplementary Fig. 12d, e), confirming that SMA⁺ mesenchymal cells were unaffected by the depletion of FCER1G⁺ PDGFRα⁺ mesenchymal cells. Macrophage numbers remained unchanged following FCER1G⁺ PDGFRα⁺ mesenchymal cell depletion (Supplementary Fig. 12f, g), suggesting a unidirectional regulation in which macrophages regulate this cell population without reciprocal feedback.

Fig. 3. Targeted ablation of Fcer1g-expressing mesenchymal cells leads to less fibrosis.

a Schematic of the Pdgfra-rtTA; Fcer1g-Cre; Dragon-DTA mouse, which allows DTA-mediated ablation of Fcer1g-and Pdgfrα-expressing cells upon Dox administration. b, c Depletion efficiency of Fcer1g-expressing cells in Pdgfra-rtTA; Fcer1g-Cre; Dragon-DTA mice. b Representative immunofluorescent images of 7 dpi wounds in control and depleted groups. White arrow: FCER1G⁺ PDGFRα⁺ cells. c Quantitative analysis of FCER1G⁺ PDGFRα⁺ mesenchymal cells per high-power field (HPF) at 7 dpi (n = 3 per group, unpaired two-sided t test, P = 0.0129). d, Representative macroscopic images of control and depleted wounds over time (control: n = 11, depleted: n = 8 for all time points). e Wound area expressed as a percentage of the wound area immediately post-injury (control: n = 11, depleted: n = 8 for all time points). f, g Macroscopic analysis of scar area in control and depleted groups at (f) 21 dpi (control: n = 12, depleted: n = 11, unpaired two-sided t test, P < 0.0001) and (g) 8 wpi (n = 11 per group, unpaired two-sided t test, P = 0.0011). h H&E staining of wounds at different time points (7 dpi: control n = 4, depleted n = 5; 11 dpi: n = 6 per group; 21 dpi: n = 11 per group; 8 wpi: n = 11 per group; 50 μm from the center of the wound). i, Scar index (scar area/average thickness, µm) of 21 dpi wounds in control and depleted groups (5 slides per n, n = 10 per group, unpaired two-sided t test, P = 0.0190). j Total collagen quantified from 21 dpi wounds in control and depleted groups (control: n = 7, depleted: n = 5; unpaired two-sided t test, P = 0.0308). k t-SNE plot visualizing ECM organization in unwounded skin and wounds from control and depleted groups at 8 wpi; clusters for each condition are highlighted with shaded regions (unwounded: n = 19, control: n = 18, depleted: n = 18). Data are presented as mean ± s.d.

We then examined the wound healing process and outcomes in control and depleted groups. At 3 dpi, immunofluorescence analysis demonstrated comparable levels of F4/80⁺ macrophages (Supplementary Fig. 13a, b), CD31⁺ endothelial cells (Supplementary Fig. 13c, d), and K10⁺ keratinocytes between groups (Supplementary Fig. 13g, h), indicating normal inflammatory, angiogenic, and reepithelialization responses. In contrast, PDGFRα staining was reduced in depleted wounds at this early stage (Supplementary Fig. 13e, f), consistent with loss of the Fcer1g-expressing mesenchymal subset. Wound size measurements indicated that wound closure rates were comparable between the control and depletion groups from 0 to 15 dpi (Fig. 3d, e). However, depleted wounds exhibited significantly smaller scar sizes at 21 dpi (Fig. 3d, f), persisting through 8 weeks post-injury (wpi) (Fig. 3d, g and Supplementary Fig. 14a). Hematoxylin and eosin (H&E) staining showed similar healing kinetics between groups, but markedly reduced fibrotic tissue in the depleted groups at 21 dpi (Fig. 3h). Moreover, increased hair follicle formation was observed in the depleted wounds at 8 wpi, suggesting a more regenerative healing outcome. To further quantify the fibrotic outcome, we performed histomorphometric analysis of Masson’s trichrome-stained sections to calculate the scar index (scar area normalized to average dermal thickness) (Supplementary Fig. 14b, c). In 21 dpi wounds, Fcer1g-expressing mesenchymal cell depletion reduced the scar index by over 20% (Fig. 3i), indicating attenuated fibrosis and improved tissue remodeling. Furthermore, total collagen quantification of 21 dpi wounds revealed a 30% reduction in collagen content in the depleted group (Fig. 3j).

To examine the ECM organization, we applied an image-processing algorithm^38–40 to profile picrosirius red-stained histological sections from unwounded skin and 8 wpi wounds in both the control and depleted group. Picrosirius red-stained images were deconvoluted into ECM parameters, including color composition (red for mature fibers, green for immature fibers) and fiber properties (length, width, persistence, alignment, etc.). These parameters were quantified independently and analyzed using t-SNE to assess ECM network properties. This analysis revealed that ECM organization in control wounds deviated significantly from that of unwounded skin, with minimal overlap in principal component (PCA) analysis (Fig. 3k). In contrast, ECM organization in the depleted group showed substantial overlap with unwounded skin, indicating reduced fibrosis and a more regenerative matrix structure.

We confirmed that both the control and FCER1G⁺ PDGFRα⁺ mesenchymal cell–depleted groups were in the telogen phase of the hair cycle, ruling out potential confounding effects from differences in hair cycle stage. This was supported by the absence of Ki67 immunostaining in hair follicles (Supplementary Fig. 15a, b) and the lack of nuclear β-catenin staining throughout the experimental timeline (Supplementary Figs. 5a and 16).

Fcer1g-expressing mesenchymal cells are highly proliferative

Previous results revealed a significant increase in Fcer1g-expressing mesenchymal cells following injury (Fig. 2d, e). To determine whether this expansion was driven by proliferation, we administered EdU intraperitoneally to wounded C57BL/6 J mice and assessed EdU incorporation in dermal mesenchymal cells 18 hours later. At 3 dpi, Fcer1g-expressing mesenchymal cells exhibited significantly higher EdU incorporation compared to other mesenchymal populations (Fig. 4a–c). These findings indicate that this cell population is highly proliferative during the early phase of wound healing, likely driving its expansion following injury. To examine whether migration also contributed to this expansion, we performed Boyden Chamber assays to evaluate the migratory ability of Fcer1g-expressing mesenchymal cells. The results showed that Fcer1g-expressing mesenchymal cells had significantly lower migratory ability than other mesenchymal cells (Fig. 4d, e). Collectively, these findings indicate that Fcer1g-expressing mesenchymal cells are highly proliferative after injury, and the expansion of Fcer1g-expressing mesenchymal cells is primarily driven by proliferation rather than migration.

Fig. 4. Fcer1g-expressing mesenchymal cells are highly proliferative.

a–c EdU incorporation in FCER1G⁺ PDGFRα⁺ and FCER1G^- PDGFRα⁺ cells from 3 dpi wounds. a Representative histogram of EdU-488 fluorescence intensity. b Quantitative data of EdU⁺ cells as a percentage of total live cells (n = 3 per condition, paired two-sided t test, P = 0.0017). c: Quantitative data of median fluorescence intensity of EdU (n = 3 per condition, paired two-sided t test, P = 0.0005). d, e Boyden chamber assay measuring the migration ability of FCER1G⁺ PDGFRα⁺ and FCER1G^- PDGFRα⁺ cells from 3 dpi wounds. d Representative images of migrated cells stained with crystal violet. e Quantitative analysis of the number of migrated cells (n = 6 per condition, paired two-sided t test, P = 0.0148). Data are presented as mean ± s.d.

EDA fibronectin-deficient wounds have significantly fewer Fcer1g-expressing mesenchymal cells

We next investigated how macrophages induce and regulate Fcer1g-expressing mesenchymal cells. Fcer1g-expressing mesenchymal cells were first mapped onto the comprehensive UMAP, where they clustered as a distinct subpopulation, designated as cluster 28 (Supplementary Fig. 17a). To assess interaction between macrophages (clusters 0, 1, 11, and 20) and Fcer1g-expressing mesenchymal cells (cluster 28), we performed CellChat analysis⁴¹. This analysis identified FN1-SDC4 as the top ligand-receptor pair with the highest communication probability between macrophages and Fcer1g-expressing mesenchymal cells at 3dpi (Supplementary Fig. 17b). Fibronectin (FN1) is known to modulate fibrosis by regulating cell adhesion, migration, differentiation, and intracellular pathways such as TGF-β and β-catenin signaling^42–45. In our scRNA-seq dataset, Fn1 was predominantly expressed in macrophages (clusters 0, 1, and 11) and mesenchymal cells (clusters 4 and 9) (Supplementary Fig. 17c). Macrophage depletion significantly reduced Fn1 expression in 3 dpi wounds (Supplementary Fig. 17d), consistent with previous findings that macrophages are a primary source of fibronectin during early wound healing⁴⁶. These results support a mechanism by which macrophages regulate Fcer1g-expressing mesenchymal cells via fibronectin.

The EDA splice variant of fibronectin is identified as a wound-specific isoform with altered integrin-binding properties in postnatal mice⁴⁷. Loss of the EDA domain reduces fibrosis, tensile strength, and mesenchymal cell recruitment in wounds^45,48. To study whether fibronectin can induce Fcer1g-expressing mesenchymal cells in vivo, we examined 8 dpi wounds in genetically modified B6.129-Fn1^tm2Feb mice⁴⁹, which lack the EDA splice variant of fibronectin (Eda Fn^-/-), and their wild-type littermates (Eda Fn^+/+). While the number of FCER1G⁻ PDGFRα⁺ cells was comparable between the two groups (Supplementary Fig. 17e), FCER1G⁺ PDGFRα⁺ mesenchymal cells were significantly reduced in Eda Fn^⁻/⁻ wounds (Supplementary Fig. 17f-h).

Fcer1g-expressing mesenchymal cells are conserved in human wound healing

To test whether FCER1G⁺ PDGFRα⁺ mesenchymal cells are conserved in human wound healing, we reanalyzed a previously published human skin wound healing scRNA-seq dataset (GSE241132) collected at days 0, 1, 7, and 30 post-injury⁵⁰. Within this dataset, PDGFRA and PDGFRB marked mesenchymal populations (Supplementary Fig. 18a, b), which were further subclustered into 13 distinct subclusters (Fig. 5a). Using gene signatures derived from mouse Fcer1g-expressing mesenchymal cells, as well as FCER1G expression alone, we identified a human mesenchymal subcluster (cluster 11) that uniquely expressed both the mouse Fcer1g signature (Fig. 5b, c) and FCER1G itself (Supplementary Fig. 18c, d). The abundance of human FCER1G-expressing mesenchymal cells (cluster 11) peaked at day 7 post-injury (Fig. 5d, e), consistent with the kinetics observed in the mouse model (Supplementary Fig. 10b). FCER1G also exhibited one of the highest U-scores in human mesenchymal subcluster 11 (Supplementary Fig. 18e and Supplementary Data file 2). Moreover, the top differentially expressed genes (DEGs) of FCER1G⁺ PDGFRα⁺ mesenchymal cells were highly conserved between human and mouse, suggesting this represents an evolutionarily conserved mesenchymal population that responds to skin injury (Fig. 5f and Supplementary Data file 3).

Fig. 5. FCER1G-expressing mesenchymal cells are conserved in human wound healing.

a UMAP visualization of PDGFRA⁺PDGFRB⁺ mesenchymal subclusters from the human skin wound healing scRNA-seq dataset (GSE241132). b, c Expression of a mouse-derived Fcer1g mesenchymal gene signature (top 25 genes with highest uniqueness scores) in human mesenchymal cells, calculated using Seurat’s AddModuleScore(). b Feature plot. c Violin plot. d, e Temporal dynamics of human mesenchymal subcluster 11 across days post injury (0, 1, 7, and 30 dpi). d Feature plots showing subcluster 11 distribution. e Quantification of subcluster 11 as a percentage of total mesenchymal cells, (n = 3 per condition, One-way ANOVA with Dunnett’s post-hoc test for multiple comparison against 7 dpi. D7 vs. D0 Padj = 0.098; D7 vs. D1 Padj = 0.023; D7 vs. D30 Padj = 0.233). f Cross-species correlation of differentially expressed genes (DEGs) in FCER1G-expressing mesenchymal cells between mouse and human, based on Seurat FindMarkers() output. Spearman’s correlation coefficient and p-value are shown. g Normalized ssGSEA enrichment scores of mice Fcer1g signature genes (top 100 by log fold change) across wound states in human bulk RNA-seq dataset GSE178411, including normal skin (n = 24), early wound (n = 22), late wound (n = 29), chronic wound (n = 3), and hypertrophic scar (n = 28) (One-way ANOVA with Dunnett’s post-hoc test for multiple comparisons against normal skin. Padj < 0.0001 for all comparisons). Data are presented as mean ± s.d.

To evaluate whether FCER1G⁺ PDGFRα⁺ mesenchymal cells are associated with human scar formation, we further reanalyzed another publicly available bulk RNA-seq dataset (GSE178411), which includes 108 human samples across uninjured skin (n = 26), early wounds (n = 22), late wounds (n = 29), chronic wounds (n = 3), and hypertrophic scars (n = 28). The top 100 DEGs derived from both mouse and human FCER1G⁺ PDGFRα⁺ mesenchymal cells effectively distinguished wound and scar samples from uninjured skin (Fig. 5g and Supplementary Fig. 18f). ssGSEA analysis revealed that FCER1G-associated signatures peaked during the early (acute) wound phase and gradually declined as wounds progressed toward hypertrophic scarring (Supplementary Fig. 19a), consistent with the temporal dynamics observed in mice. Furthermore, FCER1G expression alone recapitulated this pattern (Supplementary Fig. 19b), supporting its role as a key marker of a pro-fibrotic mesenchymal subset involved in hypertrophic scar formation during wound healing.

Discussion

Here, we identified a macrophage-induced adult mesenchymal subpopulation expressing Fcer1g that drives fibrosis during wound healing. This cell population is induced following full-thickness injury, localizes to the wound bed, and exhibits high proliferative activity. Targeted ablation of this cell population reduced fibrosis without impairing wound closure. Moreover, less-fibrotic wounds from EDA fibronectin-deficient mice contained significantly fewer Fcer1g-expressing mesenchymal cells, suggesting that macrophages may induce Fcer1g-expressing mesenchymal cells through fibronectin, which then drives fibrosis in this context. Finally, this pro-fibrotic mesenchymal subset is conserved between mouse and human, highlighting its potential relevance for translational therapeutic targeting.

Fibroblasts comprise heterogeneous subpopulations with distinct functional roles in wound repair. For example, αSMA⁺ myofibroblasts are essential for contractile repair, as their targeted depletion completely blocks wound closure⁵¹. In contrast, pro-fibrotic Prrx1⁺ fibroblasts mainly drive scar formation in the ventral dermis, and their ablation reduces fibrosis without affecting closure¹⁴. Engrailed-1 lineage fibroblasts contribute to both repair and fibrosis; their depletion delays re-epithelialization while decreasing collagen deposition¹³. In this context, our findings show that depletion of FCER1G⁺ PDGFRα⁺ mesenchymal cells in dorsal wounds did not alter inflammatory, angiogenic, or epithelial responses during the early wound-healing phase and resulted in similar closure kinetics throughout the healing process, indicating that these cells are not required for the efficacy of wound healing. However, their loss markedly reduced scar size and collagen deposition at later stages, suggesting FCER1G⁺ PDGFRα⁺ cells as a pro-fibrotic mesenchymal subset primarily responsible for fibrosis.

Fcer1g-expressing mesenchymal cells uniquely express several fibrosis-associated genes, including Lcp1, Il1b, and Ctss. Lcp1 encodes L-plastin, which enhances macrophage-mediated inflammation and promotes lung fibrosis via Il1b induction⁵². Consistent with this mechanism, Il1b is markedly upregulated in Fcer1g-expressing mesenchymal cells. Il1b encodes Interleukin-1β (IL-1β), a cytokine that drives fibrosis in vivo^53,54. Overexpression of IL-1β in rat lungs led to an acute inflammatory response with severe progressive tissue fibrosis. Ctss, encoding cathepsin S, contributes to ECM remodeling and is upregulated in fibrotic tissues across the lung, liver, kidney, and intestine^55–58. Pharmacological or genetic inhibition of cathepsin S significantly reduces fibrosis in vivo.

Although Fcer1g-expressing mesenchymal cells express hematopoietic markers such as Ptprc and Lyz2, they are distinct from previously described hematopoietic-derived mesenchymal populations. Those cells predominantly express mesenchymal markers at 7 dpi, a stage marked by re-epithelialization and strong αSMA expression^11,59–62. In LysMCre-Rosa^mT/mG mice, mesenchymal markers like Fsp1 and Col1a1 are significantly higher in GFP⁺ cells at 7 dpi compared to 3 dpi. In contrast, Fcer1g-expressing mesenchymal cells emerge at 3 dpi with strong mesenchymal marker expression, suggesting they are either not hematopoietic-derived or represent a distinct, earlier-arising subset. CD26⁺ APs is another mesenchymal subpopulation regulated by CD301b⁺ macrophages at 5 dpi¹⁵. However, in our scRNA-seq results, APs did not decrease following macrophage depletion at 3dpi. This could be due to the difference in the examined time point. APs may be regulated by macrophages at a later stage of healing, such as 5 dpi or beyond. In contrast, Fcer1g-expressing mesenchymal cells are regulated by macrophages at an earlier stage, indicating this cell population is involved in the earlier phase of fibrosis initiation. Together, these findings indicate that Fcer1g-expressing mesenchymal cells define a distinct population detectable as early as 3dpi.

We found that less-fibrotic wounds from EDA fibronectin-deficient mice contained significantly fewer Fcer1g-expressing mesenchymal cells. This suggests that the presence of this pro-fibrotic population depends on fibronectin, particularly the EDA isoform. Fibronectin is known to promote cell adhesion, migration, proliferation, and survival^63–65. Previous studies showed that the EDA-containing isoform modulates mesenchymal cell behavior through integrin signaling and pathways like TGF-β and β-catenin signaling^44,45. These functions may explain why Fcer1g-expressing mesenchymal cells are reduced when fibronectin is absent. Without fibronectin, this population may have impaired migration, reduced survival, or limited expansion in the wound. However, the current EDA-fibronectin knockout model deletes fibronectin globally rather than specifically in macrophages, and thus the reduction of Fcer1g-expressing mesenchymal cells likely reflects a combined effect from loss of both macrophage- and non-macrophage-derived fibronectin. Importantly, our findings highlight a context-specific reduction in Fcer1g-expressing mesenchymal cells. This links fibronectin to the emergence of this distinct, macrophage-induced population that drives fibrosis. Together, these results suggest a previously unrecognized mechanism by which fibronectin contributes to fibrosis during wound healing.

The findings from our mouse model were further supported by analyses of two independent human skin wound healing cohorts, encompassing both bulk transcriptomic and single-cell RNA-sequencing datasets. These results indicate that FCER1G⁺ PDGFRα⁺ mesenchymal cells are functionally conserved in human wound healing. The observation that depletion of FCER1G⁺ PDGFRα⁺ mesenchymal cells in mice markedly reduces scar formation highlights the potential clinical relevance of targeting this population to improve wound repair and minimize fibrosis in patients with acute skin injury.

In this study, we identify an adult mesenchymal population activated by macrophages that contributes to fibrosis. The gene expression profile of this population requires further investigation to better understand the mechanisms by which it drives fibrosis. This knowledge may enable the development of targeted strategies to selectively attenuate fibrosis without impairing tissue repair.

Methods

Mice

We obtained institutional ethical approval through Duke University Institutional Animal Care and Use Committee (IACUC) of Duke University for all animal experiments under the protocol number A1512307. Young adult mice (8 weeks of age) in both sexes were used in this study. C57BL/6 J, MaFIA (C57BL/6-Tg (Csf1r-EGFP-NGFR/FKBP1A/TNFRSF6)2Bck/J, Strain#005070)¹⁸, Pdgfra-rtTA (C57BL/6-Pdgfra^{em1(rtTA)Xsun}/J, Strain# 034459)⁶⁶ and Dragon-DTA (STOCK Igs7^{tm2(tetO-tdTomato,-DTA*G128D)Rdiez}/AljJ, Strain# 034778)³⁷ were purchased from The Jackson Laboratory. Fcer1g-Cre (C57BL/6J-Tg(Fcer1g-cre)NIDA168Htz/Mmucd, RRID: MMRRC_043798-UCD)⁶⁷ was obtained from the Mutant Mouse Resource and Research Center at the University of California at Davis. Both male and female mice were included in all experiments; however, the study was not powered for sex-specific analysis, and sex differences were not a predefined endpoint. Mice were housed under conditions at 22 °C, with a humidity between 30% to 70%, and a light cycle of 12h-12h on-off set. Primers used for genotyping of genetically engineered mouse models were included in Supplementary Table 2.

Macrophage depletion

To conditionally deplete macrophages, we used the MaFIA mouse. AP20187 (Takara, Cat#532522) was dissolved in 100% ethanol to a concentration of 62.5 mg/mL and stored at − 20 °C. This stock solution was further diluted to a 2.5 mg/mL in an injection solution of 10% PEG-400, and 2% Tween-20 in DI water. Retro-orbital injection of AP20187 at a dose of 10 mg per kg body weight or vehicle were used to ablate macrophages. Five days of treatment effectively depletes 90% of macrophage cells in the bone marrow⁶⁸.

Depletion of Fcer1g-expressing mesenchymal cells

To deplete Fcer1g-expressing mesenchymal cells in the skin, we generated Pdgfra-rtTA; Fcer1g-Cre; Dragon-DTA mouse. As Pdgfra-rtTA mice were firstly reported to target lung mesenchymal cells⁶⁶, we evaluated their efficiency in targeting dermal mesenchymal cells. Dragon-DTA includes a tdTomato (tdTM) reporter, allowing visualization of cells targeted by the rtTA system. In Pdgfra-rtTA; Dragon-DTA mice, Dox administration induced tdTM⁺ cells in the wounded skin (Supplementary Fig. 12c). The highest targeting efficiency was achieved by combining Dox-supplemented food (625 mg/kg diet) and intraperitoneal injections (100 mg/kg body weight). At this dosage, over 80% of tdTM⁺ cells exhibited PDGFRα immunoreactivity, confirming efficient targeting of dermal PDGFRα⁺ mesenchymal cells in Pdgfra-rtTA mice. We then followed the optimized dosage and regimen (Supplementary Fig. 12b) to administrate Dox to Pdgfra-rtTA; Fcer1g-Cre; Dragon-DTA and their control littermates.

Excisional wound splinting model

The excisional wound splinting model was conducted according to an established protocol⁶⁹. Mice were anesthetized with isoflurane, hairs were clipped, the skin site was disinfected and two full-thickness excisional wounds on the dorsal skin was created with a sterile 5 mm biopsy punch. Sterilized silicone splints (14 mm/7 mm) were centered over the wounds and anchored with nylon sutures. The wounds were then be covered with a transparent occlusive dressing (Tegaderm) to maintain moisture, and the mouse was wrapped in Vetwrap before being placed in a single cage.

Cell dissociation

Mice were euthanized by CO₂ exposure followed by cervical dislocation. The dorsal skin was disinfected, shaved, depilated, and excised. The excised skin was cut into small pieces (approximately 0.2 cm × 0.2 cm) and transferred into a 15 mL conical tube. Each skin piece was incubated in 5 mL of dissociation buffer at 37 °C for 60 min. The dissociation buffer contained 5 mL RPMI 1640 (no calcium or magnesium) with 232U DNAse I, 0.25 mg/mL Liberase stock (2.5 mg/mL), 23.2 mM HEPES, and 2.32 mM sodium pyruvate. The digestion was stopped by adding 10 µL of 0.5 M EDTA and 400 µL FBS, and the samples were transferred to GentleMacs tubes (Miltenyi Biotec, Cat#130-093-237) for dissociation (program: m_impTumor_01.01).

The digested samples were filtered through a 70 µm filter into a 50 mL conical tube, rinsed with 10 mL of RPMI media containing FBS, and filtered again through a fresh 70 µm filter. The samples were centrifuged at 500 × g for 10 min at 4 °C, and the supernatant was carefully aspirated. The cell pellets were resuspended in flow cytometry buffer, which was prepared using 500 mL PBS, 2 mL of 0.5 M EDTA, 25 mL FBS, and 5 mL non-essential amino acids (100x stock).

scRNA sequencing of mouse model

Three MaFIA mice per condition were studied. Cell dissociation was prepared as above, and then filtered through 30 μm cell strainer. Propidium iodide (PI) was added to each sample before loading to a DiVa cell sorter (BD Biosciences) for flow cytometry analysis or cell sorting. 18 K live cells were sorted and loaded on the 10x Genomics Chromium using the Chromium Single Cell 3’ Reagent V3 Kit, and the sequencing libraries were constructed following the user guide.

Cell Ranger version V3.0.2 (10x Genomics) was used to process raw sequencing data before subsequent analyses. These RNA sequencing reads were then aligned against the refdata-cellranger-mm10-3.0.0 transcriptome to quantify the expression of transcripts in each cell to create feature-barcode matrices. There were around 12 K cells recovered per sample. The median genes per cells were around 2200. The analyses of processed scRNA-seq data were carried out in R version 4.4.0 using the Seurat v4 for downstream analysis^31,32.

To eliminate empty droplets, doublets, and dead cells, as well as cells with high mitochondrial DNA expression, we applied a filtering criterion of 200 < nFeature < 2500, percent.mt < 5 and doublets were removed by DoubletFinder⁷⁰. During preprocessing, we also tested different thresholds for the number of features, setting them at 1800 and 7000 based on density plots (Supplementary Fig. 6a), which corresponded to the first enriched peak and the entire cell population, respectively. However, these thresholds did not provide a clear and distinct separation between clusters in the downstream analysis when compared to the 2500 threshold.

We then normalized, anchored and integrated the filtered cells and performed the non-linear dimensional reduction by UMAP. We selected mesenchymal cell clusters by canonical mesenchymal markers, including Col1a1, Col1a2, Pdgfrα, Col5a1, Loxl1, Lum, Fbln1 and Cd34. We then re-anchored and re-integrated to define the mesenchymal subpopulations. Differentially expressed genes with top 100 log2FC from the FindMarkers function were used as the signature genes in ssGSEA.

CellChat was used to analyze and visualize the cell-cell communication network⁴¹. In general, the unique Fcer1g-expressing mesenchymal cell population was first mapped to the comprehensive UMAP as cluster 28 (Supplementary Fig. 17a), which was saved and loaded as a new scRNA-seq matrix for further analysis. Clusters 0, 1, 11, 20 and 28 from the vehicle group were extracted to analyze the interaction between macrophages and Fcer1g-expressing mesenchymal cells.

scRNA sequencing re-analysis of publicly available cohort

Data were obtained from GSE241132. Analyses were performed in Seurat (v5.3.0) following the user guide. Briefly, low-quality cells were filtered out, and expression values were normalized. Datasets were then integrated using an anchor-based approach. Dimensionality reduction was performed with Uniform Manifold Approximation and Projection (UMAP), and cells were clustered using a shared-nearest-neighbor (SNN) modularity-optimization algorithm. Clusters were annotated based on marker-gene expression. After extracting the fibroblast cluster from the full dataset, dimensionality reduction and clustering were repeated as above. Cell-cycle scores (S and G2M) were computed from cell-cycle–related gene expression. Differential expression analysis was performed with the FindMarkers function, comparing the FCER1G-high cluster with resting clusters. Differentially expressed genes with top 100 log2FC from the FindMarkers function were used as the signature genes in ssGSEA.

Bulk RNA sequencing re-analysis of publicly available cohort

A bulk RNA-seq transcriptomic dataset for human acute skin wound were downloaded from GSE178411. The downloaded FASTQ files were aligned to GRCh38 reference genome using STAR⁷¹, and the raw gene counts were computed as default –quantMode. Following normalization and differentially expressed gene calling was performed using the DESeq2 R package⁷² with default parameters. VST normalization of DESeq2 counts were performed prior to assessing single gene expression levels. ssGSEA with the human or mouse FCER1G⁺ signature was performed using the default parameters, with an alpha value equal to 0.25. NES enrichment score was calculated based on GSEA with 10,000 permutations, which is shown by the running sum statistics plot.

Histological analysis

Wound tissues were harvested using a 7 mm biopsy punch at designated time points after complete wound closure, ensuring capture of the entire wound bed. Each wound sample was bisected through the center, fixed immediately in ice-cold 4% paraformaldehyde (PFA) for 36 h under gentle agitation at 4 °C, processed and embedded such that cross-sections were oriented with the wound center facing downward. Serial sections were then collected in 5 μm and deparaffined before subject to staining. Tissues from B6.129-Fn1^tm2Feb mice (Eda Fn^-/-) and their wild-type littermates (Eda Fn^+/+) were collected and prepared following an established procedure⁴⁵. Serial sections through the circular wound were examined to identify the center, and the staining was performed on the wound center to ensure fair comparison.

For Masson’s trichrome staining, we followed the EMS technical Data sheet (Cat#26367-series). In brief, slides were mordanted in Bouin’s Fixative for 1 h at 56 °C. After washing, they were stained with Weigert’s Iron Hematoxylin for 5 min and washed in distilled water for 10 min. Sections were then stained in Biebrich Scarlet-Acid Fuchsin for 15 minutes, followed by treatment with Phosphomolybdic Acid-Phosphotungstic Acid for 15 min. Aniline Blue was applied for 10 to 20 min, and differentiation was performed using 1% Acetic Acid for 3 to 5 min. For picrosirius red staining, we followed the Abcam protocol (Cat#ab150681). In brief, slides were mordanted in picrosirius red solution for 1 h and rinsed quickly with 2 changes of 0.5% acetic acid. Finally, tissues were dehydrated in alcohol, cleared in xylene, and mounted.

For immunofluorescence staining, slides were steamed with 1% citrate-based antigen retrieval (Vector Laboratories, H-3300) for 10 mins and blocked with 5% donkey serum and 0.3% Triton-X-100 in 1xPBS overnight at 4 °C prior to addition of the following primary antibodies: Rat-F4/80 (Biorad, Cat#MCA497GA, clone A3-1, 1:500), Rabbit-Fcer1g (Invitrogen, Cat#PA5-115222, 1:200), Goat-Pdgfra (R&D, Cat#AF1062; 1:100), Rabbit-Keratin 10 (BioLegend, Cat# 905403, 1:100), Rat-CD31 (BD Biosciences, Cat#557355, clone MEC 13.3, 1:100), Rabbit-Ki67 (Novus, Cat#NB500, 1:1000) and Rabbit-alpha smooth muscle actin (Proteintech Cat#80008-1-RR, Clone No.5H7 1:500). Slides were steamed with 1% Citrate-based antigen retrieval (Vector Laboratories, H-3300) for 10 mins, incubated with 4%PFA for 8 mins to increase permeabilization, and blocked with 3% bovine serum albumin, 0.5% Donkey serum in 1xTBST for 1 h at room temperature prior to Chicken-GFP (Abcam, cat#ab13970, 1:100) primary antibody. Rat-IgG (Invitrogen, Cat#02-9602), Rabbit-IgG (R&D, Cat#AB105-C), Goat-IgG (R&D, Cat#AB108-C) and Chicken-IgY (R&D, Cat#AB101-C) were used at corresponding concentrations as negative controls. Slides were then incubated for 1 h with Goat anti-rat Alexa Fluor 647 (Invitrogen, Cat#A-21247, 1:500), Goat anti-rat Alexa Fluor 594 (BioLegend, Cat#A-405422, 1:500), Donkey anti-goat Alexa Fluor 647 (Abcam, Cat#ab150131, 1:500), Goat anti-rabbit Alexa Fluor 594 (Invitrogen, Cat#A-11037, 1:500), Goat anti-chicken (Invitrogen, Cat#A-11039, 1:500), Goat anti-rabbit Alexa Fluor 647 (Invitrogen, Cat#A-32733, 1:500). Finally, slides were mounted in ProLong™ Glass Antifade Mountant with NucBlue™ Stain (Invitrogen, Cat#P36981).

For immunohistochemistry, slides were steamed with 1% citrate-based antigen retrieval (Vector Laboratories, H-3300) for 15 mins, blocked with 3% H₂O₂ in methanol for 10 mins, and subsequently blocked in 1× PBST containing 1.5% normal goat serum for 60 mins prior to following primary antibodies: Rabbit-Fcer1g (Invitrogen, Cat#PA5-115222, 1:200) and Rabbit-beta Catenin (Invitrogen, Cat#71-2700, 1:400). Rabbit-IgG (R&D, Cat#AB105-C) were used at corresponding concentrations as negative controls. Slides were then incubated overnight at 4 °C in a humidified chamber. The next day, slides were equilibrated to room temperature, washed with PBST, and incubated with biotinylated secondary antibody for 30 min, followed by Vectastain ABC reagent for an additional 30 min. After PBST and water washes, signals were developed using Vector ImmPACT DAB, with color development monitored microscopically. Reactions were stopped in water, and sections were counterstained with hematoxylin, rinsed, and briefly incubated in PBS. Slides were then dehydrated through graded ethanol, cleared, air-dried, and mounted with Cytoseal.

Brightfield images were acquired with a Leica DM2000 LED microscope, while fluorescent images were acquired with a Leica DM6000 SP5 upright confocal microscope and a Leica Stellaries 8 confocal microscope. Images were processed in Fiji (ImageJ) using identical settings across experimental groups; no nonlinear adjustments were applied.

Measuring cell proliferation in vivo

To assess cell proliferation, EdU (Invitrogen: A10044) was injected intraperitoneally at 25 mg/kg body weight into mice 18 hours before euthanasia. Cells were dissociated following the procedures as described earlier in the Cell Dissociation section. We followed the protocol from the Click-iT^TM EdU Flow Cytometry Assay Kit (Invitrogen, Cat# MP 35002) for cell permeabilization and EdU detection. Finally, cells were analyzed by flow cytometry.

Cell culture

For the Boyden chamber assay to assess cell migration, mesenchymal cells were sorted from 3dpi wound and were seeded at 4000 cells per 8 µm 24-well insert (Falcon, Cat#353097) with serum-free medium, while the bottom wells were filled with medium with FBS. After 2 days, the inserts were collected, washed with PBS, and fixed with 4% PFA for 2 min. Cells were permeabilized with 100% methanol for 20 minutes and stained with 0.1% crystal violet (20% methanol in PBS). Non-migrated cells were scraped off, and the membrane was imaged.

To differentiate macrophages from bone marrow stromal cells, mice were euthanized by CO2 exposure followed by cervical dislocation. The femur and tibia were dissected, and the ends of the bones were removed. The bones were placed cut-side down in a 0.6 mL tube with a hole at the bottom, inside a 1.5 mL Eppendorf tube. Bone marrow was extracted by centrifugation at 10,000 x g, resuspended in 5 mL of plating medium (10% FBS, 1% Pen/Strep in alpha-MEM), and centrifuged at 500xg for 5 minutes. After discarding the medium, 1 mL of red blood cell lysis buffer was added for 1 min at room temperature. The cells were washed with 5 mL of plating medium, centrifuged again, and resuspended in 10 mL of fresh medium. The cell suspension was filtered through a 70 µm sterile cell strainer, plated in a 10 cm dish with 40 ng/mL M-CSF (R&D, Cat# 416-ML), and cultured for 5 days. Macrophage identity was confirmed by F4/80 and CD11b staining.

NIH/3T3 murine dermal fibroblasts (MDFs) were maintained in DMEM supplemented with 10% FBS and 1% penicillin–streptomycin at 37 °C in a humidified incubator with 5% CO₂. Cells were passaged every 2–3 days at 70–80% confluence. For passaging, cultures were washed with PBS and incubated with 0.05% trypsin–EDTA for 5 min at 37 °C. Trypsinization was quenched by adding 5 mL complete medium, and cells were collected by centrifugation at 500 × g for 5 min before resuspension and replating.

For co-culture experiments, macrophages were dissociated with 0.25% trypsin-EDTA on day 5, counted, and replated with mesenchymal cells. NIH/3t3 MDF were counted and stained with CTV dye (1 µL of 5 mM CTV for 1 million cells in 1 mL) from the CellTrace™ Violet Cell Proliferation Kit (Invitrogen, Cat# C34557) for 20 minutes at 37 °C in the dark. After washing, MDF were co-cultured with macrophages at a 1:1 ratio (150 K cells each) or plated alone (300 K cells) in a 6-well plate for 24 h. Afterward, cells were dissociated and analyzed by flow cytometry.

Flow cytometry

Skin cells were dissociated as described earlier in the Cell Dissociation section and incubated with the antibodies for 30 min at room temperature, protected from light. Antibodies used in this study include PDGFRA-APC (Invitrogen, Cat# 17-1401-81, 1:50), FCER1G-PE (Miltenyi Biotec, Cat# 130-118-760, 1:200), F4/80-AF647 (BD Biosciences, Cat# 565853, 1:100) and CD11b-PE/Cy7 (BioLegend, Cat# 101216, Clone#M1/70, 1:100). After incubation, cells were washed twice with 1x PBS and filtered through a 30 μm filter. Flow cytometry was then performed using a FACSCanto II flow cytometer (BD Biosciences), and data were analyzed with FlowJo v.10 (Tree Star). For cell sorting, a DiVa cell sorter (BD Biosciences) was used. A representative gating strategy is provided in Supplementary Fig. 20.

Total collagen quantification

We followed the protocol from the Total collagen assay kit (Abcam, Cat#ab222942) to quantify the total collagen in the wounded skins. In brief, minced tissues were hydrolyzed with 10 N NaOH at 120 °C for 3 h to yield hydroxyproline, which was further oxidized and developed. The absorbance was detected by SpectraMax i3x plate reader (Molecular Devices) at OD 560 nm and compared to standards to calculate the final concentration.

Morphometric analysis

The macroscopic wound area was quantified by processing of photographs taken at various time points and was calculated as the percentage of the wound area immediately post-surgery. The scar index was used to assess the extent of scar formation by calculating the ratio of the contoured area of the scar (from a sagittal view) to the average thickness of normal skin.

Quantitative analysis of ECM organization

Skin sections that were 60 µm from the center of the wound were stained with picrosirius red as described above. Wounded skin and surrounding unwounded skin area were imaged for a minimum of 18 images per experiment condition and subject for quantitative analysis of ECM network properties^38,39. In general, picrosirius Red-stained images was first deconvoluted to isolate different color contributions corresponding to mature and immature connective tissue fibers, excluding cellular elements. The deconvoluted images are then processed with an adaptive Wiener filter to reduce noise, followed by binarization and morphological filtering to extract fiber-shaped objects. The fiber network is skeletonized, and various properties like fiber length, width, and alignment are measured using the regionprops function. Finally, dimensionality reduction of the quantified data is performed using t-SNE to visualize group differences.

Statistical analysis

Data are presented as mean ± standard deviation (s.d.), with individual values shown for each sample. Statistical analysis of scRNA-seq cell distribution percentages was conducted using a two-way repeated measures ANOVA in JMP Pro 16 with the Full Factorial Repeated Measures ANOVA Add-In. Other statistical analyses were performed using GraphPad Prism 10, applying unpaired or paired two-tails t tests as appropriate, following confirmation of normal distribution.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

X.M. and E.W. contributed equally to this project. X.M. and B.A.A. conceived the project. X.M., E.W., V.P., C.Y., and K.I. designed the experiments. X.M., P.N., E.W., Y.L., K.I., M.N., and Z.H. performed the animal studies. X.M. and E.W. performed cell culture. E.W. performed flow cytometry experiments. E.W., X.M., and E.S. carried out the bioinformatics analyses. X.M., Z.S., X.L., and Z.L. performed histological experiments. X.M., Z.S., and X.L. conducted quantification analyses. X.M., E.W., and B.A.A. prepared the manuscript. B.A.A. and X.-F.W. supervised the study.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

The scRNA-seq data derived from 3 dpi MaFIA wounds treated from vehicle and AP20187 generated in this study have been deposited in the GEO database under accession code GSE280334. scRNA-seq for human skin wound healing cohort used in this study are available in the GEO database under accession code GSE241132. Source data are provided in this paper.

Code availability

No custom code was generated for the single-cell RNA-seq analysis. All analyses were performed using publicly available functions implemented in the Seurat R package (https://satijalab.org/seurat/). Quantitative analysis of extracellular matrix (ECM) organization was performed using an established MATLAB pipeline, with the full script available at https://github.com/shamikmascharak/Mascharak-et-al-ENF. Any additional minor processing scripts are available from the corresponding author upon reasonable request.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-69449-2.