Research Article

Myofibroblast proliferation and heterogeneity are supported by macrophages during skin repair

See allHide authors and affiliations

Science  23 Nov 2018:
Vol. 362, Issue 6417, eaar2971
DOI: 10.1126/science.aar2971

Myofibroblast diversity with injury and aging

Fibroblasts deposit extracellular matrix (ECM) molecules to regulate tissue strength and function. However, if too much ECM is deposited, fibrosis and scarring results. Shook et al. examined cells during mouse skin wound healing, fibrosis, and aging (see the Perspective by Willenborg and Eming). They identified distinct subpopulations of myofibroblasts, including cells identified as adipocyte precursors (APs). In cellular ablation mouse models, CD301b-expressing macrophages selectively activated proliferation of APs, but not other myofibroblasts. Myofibroblast composition and gene expression changed during aging. Thus, macrophage-fibroblast interactions are important during tissue repair and aging, which may have therapeutic implications for chronic wounds and fibrotic disease.

Science, this issue p. eaar2971; see also p. 891

Structured Abstract


Fibroblasts produce extracellular matrix (ECM) molecules that regulate tissue strength and resilience. Imbalanced ECM maintenance leads to tissue dysfunction. Although multiple populations of fibroblasts support uninjured skin function, the extent of fibroblast diversity and the presence of functionally distinct subsets in adult fibrotic skin are poorly defined. The inability to find a single molecular marker that identifies all activated “myofibroblasts” during wound healing suggests the existence of multiple subsets of ECM-producing myofibroblasts. Furthermore, little is known about the cellular and molecular mechanisms that promote the expansion of individual myofibroblast subsets and support cellular heterogeneity. Although macrophages influence myofibroblast numbers and ECM deposition after injury, subpopulations of wound bed macrophages have only recently been defined, and specific interactions with fibroblasts have not been explored.


Variation in healing and scarring rates in multiple tissues suggests that fibroblast diversity exists. To develop therapies targeting fibrotic responses, functionally distinct subsets of myofibroblasts and the mechanisms that support individual populations must be uncovered. We used a comprehensive, hierarchical fluorescence-activated cell sorting strategy to define myofibroblast subsets in skin wound beds from adult mice. This strategy revealed distinct subsets of wound bed myofibroblasts, including an abundant population that contains the cell surface marker profile of adipocyte precursor cells (APs). We examined myofibroblast subsets in different cutaneous fibrotic contexts and explored mechanisms that selectively promote the expansion of APs.


Genetic lineage tracing and flow cytometry revealed distinct subsets of wound bed myofibroblasts that express smooth muscle actin and collagen. The most abundant populations were CD26-expressing APs and a subset with high cell surface levels of CD29 (CD29High). Although transcriptomic analysis revealed that each myofibroblast subset has a distinct gene expression profile, functional analyses suggest that myofibroblast subsets make both overlapping and distinct contributions to repair. APs were significantly reduced and CD29High cells were more abundant in wound beds from aged mice and skin from mice that underwent bleomycin-induced fibrosis, suggesting that the fibrotic environment influences myofibroblast composition. Injury and repair-related changes in AP transcription implicated macrophage signaling in the modulation of AP gene expression. Genetic ablation and cell transplantations of different myeloid cells revealed that macrophages expressing macrophage galactose N-acetylgalactosamine–specific lectin 2 (Mgl2/CD301b) directly stimulate proliferation in a subset of APs and not in other myofibroblast subsets. By combining in vitro cytokine stimulation with in vivo signaling pathway inhibition, we identified multiple CD301b+ macrophage–secreted factors (platelet-derived growth factor C and insulin-like growth factor 1) that selectively stimulate AP proliferation, thus supporting the heterogeneity of wound bed myofibroblasts.


We identified multiple populations of skin myofibroblasts and observed that the composition of myofibroblasts is dependent upon the fibrotic environment. Distinct interactions allow CD301b+ macrophage–derived signaling to selectively activate the proliferation of APs and not other myofibroblasts. These results have potential for the development of therapies that target multiple cellular populations or signaling pathways under conditions associated with excessive or deficient ECM deposition, such as wound healing and fibrosis.

Regulation of myofibroblast diversity in skin wounds.

After injury, multiple subsets of fibroblasts become activated myofibroblasts that contribute to tissue repair and scar formation. Wound bed macrophages expressing CD301b selectively activate proliferation in APs and not other myofibroblasts. With age, impaired healing is associated with a reduction in CD301b+ macrophages (Mϕ) and APs. These findings identify distinct cellular and molecular interactions that support myofibroblast heterogeneity.


During tissue repair, myofibroblasts produce extracellular matrix (ECM) molecules for tissue resilience and strength. Altered ECM deposition can lead to tissue dysfunction and disease. Identification of distinct myofibroblast subsets is necessary to develop treatments for these disorders. We analyzed profibrotic cells during mouse skin wound healing, fibrosis, and aging and identified distinct subpopulations of myofibroblasts, including adipocyte precursors (APs). Multiple mouse models and transplantation assays demonstrate that proliferation of APs but not other myofibroblasts is activated by CD301b-expressing macrophages through insulin-like growth factor 1 and platelet-derived growth factor C. With age, wound bed APs and differential gene expression between myofibroblast subsets are reduced. Our findings identify multiple fibrotic cell populations and suggest that the environment dictates functional myofibroblast heterogeneity, which is driven by fibroblast-immune interactions after wounding.

Tissues sustain resilience and strength through the maintenance of extracellular matrix (ECM) molecules by mesenchymal cells. Under disease states, profibrotic conditions lead to excessive and disordered ECM deposition that impairs tissue function (1, 2). Additionally, dysregulated ECM is associated with aged skin and age-related defective wound healing (38). Variability in rates of wound healing, scarring, and fibrosis may result from functionally distinct mesenchymal cells (9, 10). Thus, identifying distinct mesenchymal cell populations that contribute to fibrosis and the mechanisms that drive cellular diversity has substantial implications for disease treatment (2, 1113).

Prior experiments in mice have demonstrated that embryonic mesenchymal precursors expressing Engrailed (En1) or Delta-like homolog 1 (Dlk1/Pref1) generate skin fibroblast and adipocyte lineages (1416). During skin repair after injury, myofibroblasts expressing alpha smooth muscle actin (SMA), Pdgfra, Sca1, Itga8, CD34, and Dpp4 (CD26) migrate, proliferate, and deposit ECM (1719). Myofibroblasts do not form lipid-filled adipocytes within regenerated tissue after a standard small skin injury (14, 15, 20, 21) but can form adipocytes that regenerate hair follicles in large wounds (20). How environmental conditions alter functional cellular diversity and the contribution of mesenchymal subsets to tissue fibrosis are not well understood.

In this study, we uncover unappreciated heterogeneity within wound bed myofibroblasts that is dependent on the tissue environment. In particular, we show that the predominant population of myofibroblasts is adipocyte precursor cells (APs) derived from En1 lineage–traced fibroblasts (14, 15, 21, 22), which contribute to tissue repair and ECM production and modulation. We show that, in wound beds of aged mice, APs are markedly reduced and wound bed myofibroblast subpopulations become more homogeneous in their gene expression profiles and localization. Our data indicate that CD301b+ macrophage–derived platelet-derived growth factor C (PDGFC) and insulin-like growth factor (IGF) signaling contributes to myofibroblast heterogeneity by selectively promoting the proliferation of wound bed APs and not other myofibroblast subsets. These findings define major subsets of wound bed myofibroblasts and identify immune and molecular interactions that promote functional cellular heterogeneity under distinct fibrotic conditions.

Mesenchymal cell heterogeneity under fibrotic conditions

Myofibroblasts within wound beds express PDGF receptor α (PDGFRα), CD34, and SCA1 and derive from embryonic precursors that express En1 or Dlk1/Pref1 (1416) (14, 15, 23, 24) (Fig. 1A). Because PDGFRα, CD34, and SCA1 define APs (SCA1+;CD34+;CD29+) (25, 26), we sought to determine whether APs were derived from En1- and Dlk1-expressing precursors. We confirmed that En1Cre;Rosa26-LSL-tdTomato– and Dlk1CreER;mT/mG–traced cells expressed CD26 and SCA1 (14, 15) (Fig. 1B). Ninety-six percent of tdTomato+;SCA1+;CD26+ cells in En1Cre;tdTomato and 97% of GFP+;SCA1+;CD26+ cells in Dlk1CreER;mTmG nonwounded skin express AP markers CD29 and CD34 (Fig. 1B and fig. S1).

Fig. 1 Dermal mesenchymal cell heterogeneity changes after injury.

(A) Molecular markers of wound bed myofibroblasts identified by using genetic lineage tracing (14, 15). (B) FACS analysis and quantification of CD34 and CD29 subsets of SCA1+;CD26High lineage-traced cells in nonwounded skin (n = 4 samples). (C) FACS plots and quantification of cellular subsets in nonwounded skin (n = 8 samples) and 5-day (n = 19 samples), 7-day (n = 4 samples), and 14-day (n = 7 samples) wound beds. (D and E) Real-time quantitative polymerase chain reaction analysis of SMA (Acta2) and Col I (Col I) in mesenchymal subsets isolated from nonwounded skin (D) or 5-day wound beds (E). (F) Representative flow cytometry histograms and quantification of SMA, CD90, and Col I in mesenchymal subsets (n = 3 samples). Error bars indicate the mean ± SEM. NW, nonwounded; WB, wound bed; pc, panniculus carnosus; dwat, dermal white adipose tissue.

Flow cytometry analysis of immune and endothelial lineage–negative cells (Lin) isolated from uninjured dermis and 5-day wound beds revealed four populations of cells: CD29+;CD34+, CD34+, CD29High (cells with high surface levels of CD29), and CD29Low cells (Fig. 1C). We define CD29+;CD34+ cells as APs because 90% retain SCA1+ expression after injury and have adipogenic potential in vitro and in vivo (2628) (fig. S2A). Although AP and CD29Low cells were the most abundant populations in nonwounded skin, wound beds contained increased proportions of CD29High cells (Fig. 1C).

To determine which populations were myofibroblasts, we analyzed the expression of SMA and collagen I (Col I). Within wound beds, each cell population up-regulated SMA and Col I mRNA expression compared with that in cell populations from uninjured skin (Fig. 1, D and E); however, flow cytometry revealed that only APs and CD29High cells were enriched for SMA, Col I, and the fibroblast marker CD90 in wound beds (Fig. 1F).

To further analyze the fibrotic nature of these cell populations, we examined the expression of profibrotic proteins SCA1, CD9, CD26, and PDGFRα (1416, 22, 29) (figs. S1B and S2, A to D). A greater percentage of APs and CD29High and CD29Low cells expressed CD9 after injury (fig. S2D), and CD9+ APs have decreased in vitro adipogenic potential compared with CD9 APs (fig. S2, E and F). These data suggest that fibrotic cells are heterogeneous and distinct between uninjured and injured skin. Further, at least two major populations of fibrotic mesenchymal cells exist in skin wounds: APs and CD29High cells.

We next examined CD29 and CD34 populations in bleomycin-induced fibrosis. CD29High cells increased more robustly and fewer APs were observed after bleomycin treatment than in the context of wound healing (fig. S2, G and H). Colocalization with other profibrotic markers was not markedly changed (fig. S2, B to D). Thus, the profibrotic cellular composition in bleomycin-treated skin is distinct from that in wound healing, suggesting that distinct strategies are required to treat tissue fibrosis under different pathological conditions.

Because SMA and CD9 expression increased in multiple populations of mesenchymal cells within skin wounds, we sought to design a comprehensive hierarchical marker panel to delineate mesenchymal heterogeneity in skin wounds by using six fibrotic markers (Fig. 2, A and B, and fig. S1). Lin cells were subdivided on the basis of PDGFRα and SCA1 (14, 15). Whereas PDGF signaling is central to fibrosis (2, 21, 3032), we included PDGFRα cells in our analysis because PDGFRα profibrotic cells may also contribute to repair (14, 33). We further subdivided populations on the basis of CD29 and CD34 expression and then by the presence of CD26 (high or low) and CD9 (CD9+ or CD9). This analysis revealed that 54% of Lin cells in nonwounded skin contained surface markers that prospectively identify APs: PDGFRα+;SCA1+;CD29+;CD34+ (Fig. 2A and figs. S2A and S3A). Within the AP pool, 66% were CD26High;CD9 cells (Fig. 2, A to C, and fig. S4, A and B). The only other nonimmune or nonvascular populations that contribute more than 3% of total cells in nonwounded skin were PDGFRα;SCA1;CD29High (CD29High) cells (~7%) and PDGFRα;SCA1;CD29Low (CD29Low) cells (~15%) (Fig. 2, A to C, and figs. S2A and S3A).

Fig. 2 Skin wounds contain multiple myofibroblast subsets.

(A and B) FACS plots detailing the gating strategy to define mesenchymal subpopulations. (C to E) Quantification of the relative abundances of prevalent profibrotic subsets (n = 6 samples) (C) and colocalization with SMA (D) and Col I (E) in nonwounded skin and 5-day wound bed mesenchymal subsets (n = 3 samples). (F) Pipeline for processing immunostained tissue sections to infer the locations of APs (CD29+;CD26High) and CD29High cells in day 5 wound beds. Yellow lines delineate wound edges. Scale bar, 250 μm. Error bars indicate the mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. AU, arbitrary units; LUT, look-up table.

In 5-day wound beds, the relative abundance of APs decreased as PDGFRα+;CD29High cells and CD29Low cells became more abundant (Fig. 2, A to C, and figs. S3 and S4, A and B). Because CD34+ cells were mostly negative for SMA (Fig. 1F) and not abundant during repair, we excluded them from further analysis. CD29High and CD29Low cells remained predominantly CD26Low;CD9+ and CD26Low;CD9, respectively (Fig. 2, A to C, and fig. S4, A and B). Forty-six percent of APs were CD26High;CD9+ cells in 5-day wound beds, compared with 24% in nonwounded skin (Fig. 2, A to C, and fig. S4, A and B). The shift from CD9 to CD9+ APs persisted 14 days after injury (fig. S3E). Immunofluorescent staining of fluorescence-activated cell sorting (FACS)–isolated cells and wound beds showed greater intensity of CD26 staining in APs than in other cells, confirming our flow cytometry results (fig. S4, C and D). SMA and Col I were expressed by APs and CD29High cells in wound beds regardless of the cells’ CD26 or CD9 expression; however, few CD29Low cells expressed SMA, Col I, CD90, or ER-TR7 (Figs. 1, D to F, and 2, D and E, and fig. S4, C to H).

To explore the spatial organization of myofibroblast subsets, we regionally dissected wound beds for flow cytometry and examined the signal intensity of CD26 and CD29 in nonimmune (CD45) cells in tissue sections. Although APs are found throughout the wound bed, CD29High cells were biased toward the most superficial region of the outer wound bed edge (Fig. 2F and fig. S5). CD29High cells were more abundant in the upper dermis of nonwounded skin, which contributes to the superficial regenerating dermis (15). Human skin has a composition of mesenchymal cells similar to that of mouse skin. However, greater percentages of these populations were CD26High and CD9+ cells (fig. S6), indicating that human skin may be more biased toward fibrotic responses.

To determine whether repair-related myofibroblast heterogeneity arises from conversion between cellular subsets, we performed genetic lineage tracing by using inducible Cre-lox mouse lines that label profibrotic cells: Dlk1CreER;mT/mG (postnatal labeling) and PdgfraCreER;mT/mG (adult labeling) (fig. S7A) (15). In uninjured skin, several mesenchymal populations were labeled in Dlk1CreER mice, whereas PdgfraCreER mice predominantly labeled APs (~94%) (fig. S7, B and C). PdgfraCreER lineage–traced cells in 5-day wound beds contributed to APs, SCA1;CD29Low cells, and a rare population of SCA1+;CD29High cells (fig. S7, C to F). SCA1+;CD29High cells are CD9+ during repair and similar in size to APs (fig. S7, E and F), suggesting that they are profibrotic. Two weeks after injury, PdgfraCreER lineage–traced cells comprised ~80% APs, ~10% SCA1+;CD29High cells, and 7% SCA1;CD29Low cells (fig. S7D). These data suggest that APs contribute to multiple myofibroblast subpopulations; however, they do not contribute to the expansion of SCA1;CD29High cells. Yet, because non-AP, CD29+ cells are labeled in PdgfraCreER mice, we cannot rule out the possibility that the proliferation of CD29+ cells (fig. S12) also contributes to myofibroblast heterogeneity after injury.

Myofibroblast subsets have distinct gene expression profiles

The comparison of transcriptional profiles of CD29Low and CD29High cells and APs that were either CD9 positive or negative by RNA sequencing (RNA-seq) (n = 2 profiles for each population) confirmed significant diversity among wound bed myofibroblasts (Fig. 3, A and B). Although transcriptomic analysis revealed that CD9+ and CD9 APs were similar, each mesenchymal subset expressed distinct mRNAs (figs. S8 and S9 and table S1). For instance, CD29High cells had elevated expression of Pdgfrβ, CD146, NG2, and other perivascular cell or pericyte markers (34) compared with APs and CD29Low cells and had elevated Acta2 expression in nonwounded skin (Fig. 1D).

Fig. 3 Myofibroblast subsets can distinctively regulate repair.

(A) Transcriptomic principal components analysis of myofibroblast subsets. (B) Table of genes with statistically significant differences in expression between cellular subsets. (C) Wound healing–related genes enriched in APs (CD9 and CD9+ AP populations) or CD29High cells. (D and E) Quantification of hydroxyproline content (n = 7 samples) (D) and lysyl oxidase (LOX) activity (n = 4 samples; *P = 0.0416) (E) in cells from day 5 wounds. (F) Migration distance of APs (CD26High) (asterisks) and CD29High cells (arrowheads) from cultured wound beds (n ≥ 250 cells from 3 wound beds). Scale bar, 10 μm. Error bars indicate the mean ± SEM.

Although transcriptomes were distinct between cellular subsets, Ingenuity Pathway Analysis (IPA) predicted common active biofunctions and similar upstream activators of gene expression profiles (figs. S10 and S11 and table S2), suggesting some functional redundancy among myofibroblasts. However, many genes differentially expressed between APs and CD29High cells have been implicated in wound healing (Fig. 3C and figs. S8 and S9). Additionally, each myofibroblast population was enriched for different ECM components and modifiers (Fig. 3C and figs. S8 and S9). Both CD9+ and CD9 APs were enriched for many cytokine genes (Ccl2, Cxcl1, Cxcl10, and Cxcl12) and ECM components (Col5a2, Fbln1, Fbn1, Has1, and Loxl1) that promote rapid ECM deposition (3537). The enrichment of genes involved in repair and fibrosis changed among myofibroblast populations between day 5 and day 14 of repair (fig. S8B), indicating that myofibroblast subsets can distinctively influence both the proliferative and maturation phases of tissue repair.

To determine whether myofibroblast subpopulations were functionally distinct, we examined collagen production, collagen cross-linking, and the migration of APs and CD29High cells. We did not observe differences in collagen production or cellular migration; however, we detected an increased ability of APs to cross-link collagen compared with CD29High cells, consistent with elevated Lox expression in APs (Fig. 3, C to F, and fig. S8A).

Because myofibroblast numbers increase after injury (Fig. 1C and fig. S3), we examined in vivo proliferation within the different mesenchymal subsets (APs and CD29High and CD29Low cells) during tissue repair. Proliferation increased in APs and CD29High and CD29Low cells after injury, and CD26Low cells were more proliferative than CD26High cells within each cellular subset (fig. S12, A to D). Taken together, these data demonstrate that the dermis contains tremendous heterogeneity within profibrotic cells, which have distinct functions during tissue repair.

Myofibroblast composition and gene expression are altered during aging

Age-related defects in repair are associated with reduced myofibroblasts and dysfunctional ECM deposition (36) (fig. S13, A and B). To determine whether mesenchymal populations were altered with age, we analyzed 5-day wound beds in young and aged mice. The relative abundance of APs decreased and CD29High cells increased in wound beds from aged mice (Fig. 4, A to D), with reduced percentages of CD9+ cells in all mesenchymal populations (Fig. 4C), suggesting that fibrotic cells are lost or unstimulated with age.

Fig. 4 Myofibroblast composition and gene expression is altered during aging.

(A) FACS plots for 5-day wounds from young and aged mice. (B) Quantification of the relative abundance of prevalent profibrotic subsets in 5-day wounds (n = 4 samples). (C) Pie charts depicting CD9 and CD26 colocalization. (D) Pipeline for processing immunostained wound bed sections to infer AP and CD29High cell locations in day 5 wound beds from aged mice. Yellow lines delineate wound edges. Scale bar, 250 μm. (E) Genes with altered differential expression with age. Black and red text indicates enrichment in young and aged mice, respectively. Error bars indicate the mean ± SEM. **P < 0.01, ***P < 0.001, ****P < 0.0001.

Analysis of transcriptional changes in myofibroblasts during aging by RNA-seq (n = 2 profiles) revealed fewer differentially expressed genes between myofibroblast subsets (fig. S14, A to C) because of age-related down-regulation of many genes within individual populations (fig. S14D). Comparing the transcriptomes of myofibroblast populations in young versus aged mice revealed age-related changes in gene expression of extracellular molecules (Fig. 4E) and increased expression of multiple metalloproteases in myofibroblasts, consistent with the ability of aged fibroblasts to break down ECM faster than young fibroblasts and impair healing (4, 6).

APs become fibrotic after injury

To identify molecular mechanisms regulating AP myofibroblasts during repair, we isolated APs from uninjured skin and 5-day wound beds and performed RNA-seq (n = 2 profiles). Injury and repair up-regulated Acta2 (SMA) and several secreted factors implicated in tissue repair (fig. S15, A to C). Several adipogenic genes and in vitro adipogenic potential were reduced in wound bed APs (fig. S15, D and E). Thus, APs displayed dramatic alterations within the wound environment that limit their adipogenic potential, promote myofibroblast gene expression, and may explain the myofibroblast origin of adipocytes in large wounds (20).

Macrophage signaling selectively activates proliferation of wound bed APs

Because delayed healing in aged mice is associated with decreased APs and APs rapidly increase from days 3 to 7 after injury, when new dermal tissue is generated, we investigated potential signaling pathways that could affect AP numbers during repair. IPA predicted that injury-related changes in AP gene expression could result from monocyte-macrophage–derived ligands (fig. S15F). Macrophage ablation reduces wound bed myofibroblast numbers, impairs myofibroblast function, and impairs wound healing (3841); however, the underlying mechanisms are ill defined. To examine the contribution of monocytes and macrophages to myofibroblast heterogeneity, we ablated macrophages by using LysMCre;iDTR mice (38, 4143) (Fig. 5A). Ablating monocytes and macrophages reduced all AP subsets (fig. S16) and diminished AP proliferation in wounds without significantly changing CD29High or CD29Low populations (Fig. 5B and fig. S16A). Additionally, pharmacological reduction of macrophages decreased the percentage of dividing APs from 25 to 9% in controls (fig. S17), indicating that the myeloid lineage in 5-day wound beds selectively activates the proliferation of APs and not other myofibroblasts.

Fig. 5 CD301b+ macrophages selectively stimulate AP proliferation during wound healing.

(A) Quantification of wound bed macrophage depletion (n ≥ 3 samples; P = 0.004). (B) Quantification of myofibroblast proliferation in wound beds (n = 4 samples). (C) Quantification of cell proliferation in wound beds of CD301b+ macrophage–depleted mice (Mgl2DTR) (n = 3 samples). (D) Quantification of EdU-incorporating APs in mice receiving DT on days 2, 4, and 6 after injury (left) (n = 3 samples, P = 0.0469) and DT on days 2 and 3 or 3, 4, and 6 after injury (right) (n = 3 samples, P = 0.0116). Mice were given two injections of EdU per day from days 3 through 7 after injury. (E) FACS plots of immune cell populations isolated for transplants. (F and G) Quantification of EdU-incorporating cells after injection of select immune cell subsets in vivo (n = 5 samples, P = 0.0146) (F) or Transwell coculture (n = 6 samples, P = 0.001) (G). Error bars indicate the mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001. IV DT, intravenous DT; Mo/Mϕ, monocyte-macrophage.

CD301b+ macrophages activate AP proliferation during wound healing

During the mid-phase of wound healing (days 3 to 7), the myeloid lineage is composed predominantly of monocyte and macrophage subsets (41, 44, 45) (fig. S18A). We have previously shown that wound bed macrophages expressing macrophage galactose-type C-type lectin 2 (Mgl2/CD301b) contribute to repair by promoting proliferation and fibroblast repopulation and that CD301b+ macrophages are ablated in Mgl2DTR mice (41, 46, 47). Whereas ablating all macrophages in LysMCre;iDTR mice decreased the proliferation of all subsets of APs, ablating CD301b-expressing macrophages reduced the proliferation of CD26Low APs in 5-day wounds, with no change in CD29High or CD29Low cell proliferation (Fig. 5C). This suggests that diversity in wound bed macrophages (48) allows the proliferation of different AP subsets to be differentially regulated, thus promoting myofibroblast heterogeneity. Reduced AP proliferation in diphtheria toxin (DT)–treated Mgl2DTR mice resulted in a ~50% reduction in 5-ethynyl-2′-deoxyuridine–positive (EdU+) APs in 7-day wound beds that was observed only when newly generated CD301b+ macrophages were ablated during the proliferative phase of repair (Fig. 5D). Consistent with this model, cell transplantation of CD301b+ macrophages, and not that of other immune cells, increased AP proliferation (from 18 to 28%), whereas CD29High and CD29Low cells were unaltered (Fig. 5, E and F). Further, cultured CD301b+ macrophages doubled AP proliferation in vitro (Fig. 5G), demonstrating direct signaling between CD301b+ macrophages and APs.

To identify signaling molecules that activate AP proliferation during repair, we compared the transcriptomes of CD301b+ macrophages with those of F4/80 immune cells isolated from day 5 wounds (Fig. 6A and fig. S18, B and C) (n = 2 profiles per group). We identified ligands enriched in CD301b+ macrophages that bind to receptors on APs (Fig. 6B and fig. S18D) and validated these results by quantifying protein secretion (fig. S18E). Cultured APs were treated with candidate molecules, and only PDGFC and IGF1 induced proliferation (Fig. 6C). To determine whether PDGFC and IGF1 signaling pathways contribute to AP proliferation in vivo, we administered ligand-neutralizing antibodies or receptor antagonists after injury (Fig. 6D). Local injection of PDGFC- or IGF1-neutralizing antibodies in vivo reduced AP proliferation; however, no change in the proliferation of other cells was detectable. Additionally, the inhibition of PDGFRα and IGF1 receptor (IGF1R) or downstream phosphatidylinositol 3-kinase (PI 3-kinase) signaling selectively reduced AP proliferation (Fig. 6D). We did not observe spatial biasing of CD301b+ macrophages in wound beds (Fig. 6E), and the expression of wound healing–associated genes changed minimally in myofibroblasts from 5-day wound beds of Mgl2DTR mice relative to controls (fig. S19). These data suggest that the distinct gene expression profile of each myofibroblast subset results from interactions with other tissue resident cells, such as keratinocytes (49). As a result, the delayed re-epithelialization and revascularization observed in Mgl2DTR mice (41) may result from CD301b+ macrophages interacting with keratinocytes and endothelial cells. IGF1 can stimulate repair, potentially by promoting the migration and proliferation of keratinocytes and fibroblasts (35, 5053), yet the contribution of PDGFC to healing has not previously been explored. To examine the contribution of PDGFC signaling to wound healing, we locally injected a PDGFC-neutralizing antibody at the periphery of wound beds and examined skin repair. We did not observe gross changes in re-revascularization and myofibroblast repopulation in 5-day wound beds compared with controls; however, we observed a slight decrease in re-epithelialization (fig. S20). These data demonstrate that multiple ligands produced by CD301b+ macrophages activate the proliferation of APs, and not other myofibroblast subsets, during wound healing.

Fig. 6 CD301b+ macrophage–derived ligands activate AP proliferation.

(A) Cell populations isolated from 5-day wound beds for RNA-seq (left) and FPKM (fragments per kilobase per million) scatter-plotting (right). (B) Table of ligands enriched in CD301b+ macrophages that bind to receptors on APs. (C) Quantification of AP proliferation after the administration of ligands. Fetal bovine serum (FBS) at 10% is a positive control (n = 5 samples, ***P < 0.001). BSA, bovine serum albumin. (D) Quantification of in vivo cellular proliferation after the administration of PDGFC (n = 6 samples, *P = 0.0337)– and IGF1 (n = 6 samples, *P = 0.0436)–neutralizing antibodies (nAbs) or antagonists against PDGFRα (crenolanib) (n = 6 samples, ***P = 0.0001), IGFR1 (linsitinib) (n = 4 samples, *P = 0.01017), or PI 3-kinase (wortmannin) (n = 4 samples, **P = 0.0028). DMSO, dimethyl sulfoxide. (E) Pipeline for processing immunostained wound bed sections to infer the distribution of CD301b+ macrophages in day 5 wounds (n = 6 samples). Yellow lines delineate wound edges. Scale bar, 250 μm. Error bars indicate the mean ± SEM.

Numbers of CD301b+ macrophages increase in wounds as AP abundance increases (41). However, wound beds from aged mice contain fewer CD301b+ cells than those from young controls, and human keloid scars, which have been shown to contain many CD26+ fibroblasts (54), are enriched with CD301+ cells (fig. S21). Thus, the interaction between CD301b+ macrophages and mesenchymal cells may provide a therapeutic target for fibrosis-related diseases.


Dermal cells, including fibroblasts and adipocytes, support epidermal functions and integrity (11, 12). Although a common embryonic precursor for SMA+ wound bed myofibroblasts exists (14, 15, 20, 33, 55, 56), the diversity of mesenchymal cells during adult tissue repair is ill defined. In this study, we discovered that after injury, skin wound beds contain tremendous mesenchymal heterogeneity, similar to what is observed during lung fibrosis (57). We identified two major classes of SMA+ and Col I+ myofibroblasts that arise from different cellular origins: cells with a cell surface marker profile of APs, and CD29High cells. During tissue repair, a greater percentage of APs express profibrotic cell surface proteins CD26High and CD9, with reduced adipogenic potential. Spatially, these two myofibroblast populations are distinct, with APs evenly distributed within wounds and CD29High cells biased toward superficial, outer regions of wound beds. RNA-seq and functional analysis of these myofibroblast subsets revealed that each subset has distinct transcriptomes with some functional overlap.

With age, the abundance of APs decreases and CD29High cells are more prevalent, as differential gene expression between myofibroblast subsets is reduced. Although myofibroblasts are dynamic after injury in mouse skin, human dermal fibroblasts express profibrotic cell surface proteins in uninjured skin, possibly resulting in stronger fibrotic biasing in humans. These studies illuminate distinct functional subsets of fibrotic cells, providing a stepping stone to develop therapeutic strategies that promote efficient wound healing and treat fibrosis.

Regulation of functional myofibroblast diversity

In this study, we have shown that CD26High myofibroblasts are largely CD34+;CD29+ APs that function as myofibroblasts in regenerating mouse skin. Although previous reports did not observe the same degree of CD34+ and CD29+ colocalization on myofibroblasts (14, 16), these differences likely result from changes in fibroblast surface marker expression associated with different ages and hair follicle stages (29). Our data reveal that biased proliferation and plasticity of fibroblast subsets promote myofibroblast heterogeneity in skin wounds. Our lineage tracing data suggest that a combination of proliferation and plasticity supports fibroblast heterogeneity within regenerating skin.

Although multiple signals likely influence myofibroblast heterogeneity, our study highlights the importance of myofibroblast-macrophage interactions and, particularly, of PDGFC and IGF1 in promoting myofibroblast heterogeneity and repair. These data correspond with the function of macrophages in tissue fibrosis (1, 58), the ability of exogenous PDGFC to rescue delayed skin wound healing in diabetic mice (59), and the promotion of fibroblast proliferation and repair by IGF1 (35, 5053). In various tissues, macrophages express Pdgfc and Igf1 after injury or under pathological conditions (6064). PDGF signaling and IGF signaling cooperate synergistically to promote fibroblast proliferation and enhance wound healing without increasing scarring (6567). Treatments aimed at fine-tuning the number of CD301b+ macrophages could be of tremendous clinical value, as reduction in CD301b+ macrophages and CD26-expressing fibroblasts is associated with aging and defective wound healing and keloid scars contain excessive ECM, CD26-expressing fibroblasts, and CD301+ cells. Further understanding of how myofibroblast subsets function and are influenced by the microenvironment during fibrosis and pathologies with irregular ECM homeostasis will allow optimization of treatments for these encumbering diseases.

Supplementary Materials

Materials and Methods

Figs. S1 to S21

Tables S1 to S3

References (6878)

References and Notes

Acknowledgments: We thank members of the Horsley lab, R. Atit, and A. MacLeod for critical reading of the manuscript. We thank B. Hogan and C. Barkauskas for sharing PdgfraCreER mice and C. Wolfrum for sharing Dlk1CreER mice before the initial mouse line publications and R. Atit for En1Cre;Rosa26-LSL-tdTomato mice. Funding: This work was supported in part by NIH grants to V.H. from NIAMS (AR060295 and AR069550) and NIA through pilot project grants from the Claude D. Pepper Older Americans Independence Center at Yale (NIA P30AG21342) awarded to V.H. This research was supported in part by a New Investigator award to H.C.H. from the Yale Department of Surgery Ohse Fund. B.A.S. is a New York Stem Cell Foundation–Druckenmiller Fellow. This research was supported by the New York Stem Cell Foundation. Author contributions: B.A.S. conceived, designed, and oversaw the experiments with suggestions from V.H., J.L.A., K.M.-J., D.A.C., and H.C.H., R.R.W, G.C.R.-G., and A.R.M.-R. performed experiments and analyzed the data. B.C.D., E.S.-G., K.D.A., R.K.Z., and V.L. assisted with collection and analysis of data. F.L.-G. analyzed the RNA-seq data. B.A.S. and V.H. wrote the manuscript. Competing interests: The authors have no competing interests. Data and materials availability: RNA-seq data are available at the Gene Expression Omnibus (GEO) under accession numbers GSE105788, GSE105789, and GSE105790.
View Abstract

Stay Connected to Science

Navigate This Article