PDGF signaling regulates the adipocyte-to-fibroblast transition in wound healing

SUMMARY

Adipocytes are plastic cells with the ability to transition into fibroblast-like cells during tissue repair or physiological remodeling. It is unclear how such transitions are regulated and whether they exemplify cell plasticity or merely a transient suppression of adipocyte features. Fibroblasts express platelet-derived growth factor receptors (Pdgfra and Pdgfrb), while adipocytes express adiponectin (Adipoq). Here, we use two-step lineage tracing to label Pdgfra⁺ and Pdgfrb⁺ cells originating from Adipoq⁺ adipocytes. Adipocyte-derived Pdgfr⁺ cells (ADPCs) are induced adjacent to skin wounds and contribute to the wound fibroblast population. ADPCs lose adipocyte characteristics and acquire fibroblast characteristics, including collagen expression, smooth muscle actin expression, and cell proliferation. Active PDGFRα signaling in ADPCs increases cell proliferation and improves wound angiogenesis and re-epithelialization, while deletion of Pdgfra has the opposite effect. Wound ADPCs persist as fibroblast-like cells after completion of wound healing, indicating an enduring change rather than a transient downregulation of adipocyte characteristics.

In brief

Adipocytes exhibit surprising plasticity, but the regulation of adipocyte-to-fibroblast transition is not well characterized. Yao et al. show that PDGF signaling in adipocytes promotes adipocyte-to-fibroblast transitions in wound healing and supports angiogenesis and re-epithelialization. Transdifferentiated adipocytes persist as fibroblasts after wound healing, in contrast with transient dedifferentiation during physiological remodeling.

Graphical abstract

INTRODUCTION

Adult cell plasticity is a widespread and important process in tissue repair. Differentiated cells exhibit persistent cell phenotypes maintained by stable gene regulatory networks. Cell plasticity is defined as the ability of differentiated cells to change phenotype and function, sometimes in an enduring or permanent way.¹ This may include dedifferentiation of mature cells into less specialized progenitor cells, or transdifferentiation from one cell type directly into a different type.

Adipocyte progenitor cells (APCs) differentiate into adipocytes through a well-understood process of lineage progression.^2–8 APCs express platelet-derived growth factor receptors (Pdgfra and Pdgfrb), similar to fibroblasts in all tissues.^9–11 Indeed, APCs cluster as a subset of the fibroblast population in single cell surveys.¹² During the differentiation process from APCs to mature adipocyte, progenitors downregulate Pdgfra/Pdgfrb and express Adipoq, Plin1, and Pparg to specialize in lipid storage. Thus, fibroblasts and adipocytes are distinct cell types that arise from a common mesenchymal lineage.^13,14

Recent studies have demonstrated, however, that adipocytes transition into fibroblast-like cells in response to tissue damage and physiological cycles.^15–20 Given the close relationship between APCs and other fibroblasts, it is unclear whether these transitions are merely the transient suppression of adipocyte features as the cells deplete lipid stores, or a type of cell plasticity akin to transdifferentiation.^21,22 Clarifying the nature and limits of adipocyte plasticity is important for understanding stromal lineage homeostasis and cellular mechanisms of tissue repair. This informs our understanding of fibrotic disease mechanisms and translational paradigms for regenerative medicine involving tissue grafts or harnessing cell plasticity in situ.²³

Standard lineage tracing approaches involve Adipoq-CreER or Adipoq-rtTA/TRE-Cre (AdipoChaser) to label mature adipocytes with high efficiency and track their fate over time.^24,25 A downside is the challenge of identifying, among all the adipocytes labeled, the small subset of cells that exhibit plasticity. In this context, we sought to use intersectional lineage tracing to selectively label the subset of Adipoq-CreER⁺ adipocytes that induce Pdgfra or Pdgfrb. We show that skin wounding induces the two-step reporter system specifically in adipocyte-derived cells at the perimeter of the wound. Like fibroblasts, these cells replicate their DNA, express type I collagen, and functionally respond to changes in PDGF signaling. They persistently populate the scar tissue without reverting to Plin1⁺ adipocytes, arguing for an enduring transdifferentiation process rather than transient suppression of adipogenic features.

RESULTS

Intersectional lineage tracing labels adipocyte-derived Pdgfr⁺ cells at the wound edge

In adult tissues, Adipoq is expressed by mature adipocytes but not by fibroblasts or APCs, and conversely, Pdgfra and Pdgfrb are expressed by fibroblasts and APCs but not by mature adipocytes.^5,12,26–28 We reasoned that the intersection of Adipoq and Pdgfra or Pdgfrb promoter activities could be used to direct lineage tracing to a subset of adipocytes that acquire a fibroblast-like phenotype. To create a two-step lineage tracing system, we combined AdipoqCreER with a Flp reporter (R26-FSF-tdTomato) and a Cre-inducible Flp^o regulated by the Pdgfra or Pdgfrb gene. We generated three kinds of Flp^o alleles: Pdgfra^Flp, Pdgfra^K.Flp, or Pdgfrb^Flp (Figure 1A). Previously, we characterized Pdgfra^Flp and Pdgfrb^Flp alleles, each with a floxed cDNA encoding Pdgfra or Pdgfrb followed by Flp^o, which was inserted to replace the endogenous Pdgfra or Pdgfrb gene.²⁹ Cre recombination in these mice leads to deletion of the cDNA and expression of Flp^o controlled by the endogenous Pdgfr genes. Pdgfra^K.Flp is like Pdgfra^Flp except that Cre recombination leads to the deletion of a STOP cassette and expression of Flp^o and constitutively active Pdgfra^K.^29–31 This form of PDGFRα carries a D842V mutation in the kinase domain that disrupts receptor autoinhibition and generates constitutive signaling.³² We used these genetic tools to detect and lineage trace adipocytes that induce Pdgfra or Pdgfrb promoter activity, or in other words, to detect adipocyte-derived Pdgfr⁺ cells (ADPCs). After tamoxifen treatment, excisional wounding of the dorsal skin, and 7 days of healing (Figure 1B), Tomato⁺ cells formed a ring at the wound edge of all three genotypes aRFlp^Adipoq, aKFlp^Adipoq, and bRFlp^Adipoq (Figure 1C). Few Tomato⁺ cells were seen inside the wound bed, and none were seen in unwounded skin >1 mm from the wound edge. In comparison, standard lineage tracing with AdipoqCreER and R26-LSL-tdTomato labeled all adipocytes (Figure 1D). Therefore, the intersection of Adipoq and Pdgfr promoter activities specifically labeled ADPCs at the wound edge.

Figure 1. Genetic approach to label adipocytes that express Pdgfra or Pdgfrb.

(A) Intersectional lineage tracing approach. Components 1 and 3 are always the same. Component 2 can be three different alleles for Flp^° expression.

(B) Timeline of wound experiments.

(C) Whole mount day 7 wounds with intersected lineage tracing showing Tomato⁺ cells at wound edge (dotted line); scale bars, 500 μm.

(D) Day 7 wound with AdipoqCreER lineage tracing showing all adipocytes as Tomato⁺; scale bars, 500 μm (n = 7 aRFlp^Adipoq, 8 aKFlp^Adipoq, 7 bRFlp^Adipoq, AdipoqCreER).

To better visualize the ADPCs, we prepared wound cross sections and stained for the mature adipocyte marker Plin1. We found Tomato⁺ cells in aRFlp^Adipoq, aKFlp^Adipoq, and bRFlp^Adipoq mice located at the excised edge of the dermal fat layer where it meets the wound bed (Figures 2A–2C). For comparison, AdipoqCreER labeled cells throughout the dermal fat (Figure 2D) and PdgfraCreER labeled fibroblasts throughout the wound (Figure S1A). There was no adipocyte labeling with PdgfraCreER because skin adipogenesis is not active at 7 weeks. In most aRFlp^Adipoq, aKFlp^Adipoq, and bRFlp^Adipoq wounds, ADPCs were associated with the edge of the dermal fat layer but not with the wound center or the base. However, the variable formation of Tomato⁺ cells at some wound centers seemed to correlate with subcutaneous adipocytes adhering to the wound base. Subcutaneous tissue contributes to healing large wounds, and this may include adipocytes in varying quantities depending on anatomical location.^33,34 To maximize the involvement of subcutaneous adipocytes, we created skin wounds overlying large inguinal fat depots. Such wounds consistently formed Tomato⁺ cells at the wound base and extended into the wound bed (Figures S1C–S1E). These results show that aRFlp^Adipoq, aKFlp^Adipoq, and bRFlp^Adipoq mice label adipocyte-derived Pdgfra⁺ or Pdgfrb⁺ cells that form specifically at the wounded edge of fat.

Figure 2. Specific lineage tracing of ADPCs at the wound edge.

(A–D) Plin1 immunostaining, cross sectioned wounds, day 7. Whole-wound images are stitched together from three pictures; scale bars, 500 μm. Zoomed-in areas shown on the right; scale bars, 100 μm.

(E and F) Flow cytometry gating strategy (E) and quantification of Tomato⁺ nuclei (F), day 7. Error bars indicate mean ± SD, one-way ANOVA (n = 11 no reporter, 9 aRFlp^Adipoq, 17 aKFlp^Adipoq, 6 bRFlp^Adipoq, and 14 AdipoqCreER).

We also used flow cytometry to quantify ADPCs in wounds that were dissected to include the ring of labeled cells at the wound edge. Because lipid-filled adipocytes are difficult to pass through a flow cytometer, we isolated nuclei after brief tissue fixation to preserve nuclear Tomato after cell lysis (Figure 2E). Thus, PdgfraCreER and AdipoqCreER wound tissues registered 43% and 3% of nuclei, respectively, as Tomato⁺ (Figure S1B). Among the wounds with intersected tracing, aKFlp^Adipoq had the most labeling with an average of 4.29% Tomato⁺ nuclei. Wounds with aRFlp^Adipoq and bRFlp^Adipoq averaged 1%–2% Tomato⁺ (Figures 2F and S1B). Therefore, intersectional tracing with aRFlp^Adipoq labels similar proportion of cells in wounds as AdipoqCreER, while aKFlp^Adipoq labels significantly more and bRFlp^Adipoq labels less. A potential reason for increased labeling with aKFlp^Adipoq is that constitutively active PDGFRα increases cell proliferation.³²

To screen for overlapping AdipoqCreER and Pdgfr promoter activity in unwounded tissue, we examined skin and other fat depots in aRFlp^Adipoq, aKFlp^Adipoq, and bRFlp^Adipoq mice for tamoxifen-induced Tomato. However, as observed by flow cytometry, unwounded skin and subcutaneous fat tissue samples from tamoxifen-induced mice were indistinguishable from the samples from mice with no reporter. We also could not see labeling using fluorescent microscopy to examine unwounded skin or the major subcutaneous, perigonadal, or interscapular depots. In contrast, unwounded skin and subcutaneous fat of PdgfraCreER and AdipoqCreER mice registered between 3% and 50% Tomato⁺ depending on the tissue and lineage (Figure S2).

Out of curiosity, we also crossed the non-inducible AdipoqCre line, which is active during development, with Pdgfra^Flp mice. This approach detected ADPCs at the wound edge, but it lacked selectivity because there was labeling in adipocytes in unwounded skin (Figure S3A). Labeling was already present in skin and subcutaneous fat of AdipoqCre;Pdgfra^Flp pups at postnatal day 5 but not in mice with Pdgfra^Flp and R26-FSF-tdTomato but no Cre (Figures S3B–S3D). This indicates that Pdgfra expression in embryonic mesenchymal cells overlaps with initial Adipoq expression³⁵ (Figure S3E). Therefore, to specifically identify ADPCs, AdipoqCreER must be used.

Hair cycle and mammary gland remodeling induce less dramatic adipocyte transitions

Dermal adipocytes were reported to transition into PDGFRα⁺ APCs during murine hair follicle regression, which occurs in cycles throughout life.¹⁹ This transition is temporary, as the dedifferentiated adipocytes redifferentiate into mature adipocytes when hair follicles re-enter the growth phase. Reversible transitions have also been reported in mammary gland remodeling, with mature adipocytes becoming PDGFRα⁺ APCs during lactation and then reverting to adipocytes at the end of lactation.¹⁸ We reasoned that intersectional lineage tracing may record such transitions even if PDGFRα⁺ APCs revert to adipocytes because Tomato expression is indelible after the removal of the STOP cassette.

In pilot experiments examining hair cycle and mammary remodeling separately, we tested all three of the reporters (Figure 1A), but we decided to focus on aKFlp^Adipoq because it labeled the most ADPCs in wounds (Figure 2F). We also tested various times of tamoxifen treatment before deciding to administer three doses to young female aKFlp^Adipoq mice during their first anagen (P28–P35) because dermal fat is fully developed at this time to maximize labeling of mature adipocytes.¹⁹ Then we devised an experimental scheme to examine the effects of multiple hair cycles, lactation, and wound healing in the same mice (Figure 3A). We aged the mice until P120, which is enough time for three complete hair growth cycles with accompanying adipocyte transitions (Figure 3B). Near the endpoint, we also allowed the females to become pregnant to induce mammary gland remodeling, during which, mammary adipocyte transitions should occur. As positive controls, we also created excisional wounds in the dorsal skin (Figure 3A). Thus, after three complete adult hair cycles, one cycle of lactation-induced mammary remodeling, and 7 days of wound healing, we isolated skin, mammary tissue, and wounds for analysis. Wounds formed Tomato⁺ ADPCs (Figures 3C and 3D), which confirms the persistence of Cre-recombined adipocytes months after tamoxifen treatment. However, we were surprised by the complete absence of Tomato⁺ cells in lactating or involuted mammary tissue and in unwounded skin from aKFlp^Adipoq mice (Figures 3F–3H). We used flow cytometry to quantify Tomato⁺ nuclei, which registered ~2% of wound nuclei as Tomato⁺ but negligible numbers in skin and mammary tissues (Figure 3E). In previous studies, transiently de-differentiated adipocytes in skin and mammary tissue were reported to express Pdgfra.^18,19 But here, intersectional lineage tracing failed to detect ADPCs in skin and mammary fat while in the same mouse, we could detect ADPCs in wounds.

Figure 3. Intersectional lineage tracing does not label ADPCs in hair cycle and mammary gland remodeling.

(A) Experimental timeline using female aKFlp^Adipoq mice: 3× tamoxifen was administered between postnatal days 28–35; mating was conducted on ~P85; after birth and nursing for 1 week, the pups were removed and female was wounded; and after 1 week of healing and mammary gland involution, the wound, skin, and mammary gland were analyzed on ~P120.

(B) Hair cycle timeline.

(C) Whole mount aKFlp^Adipoq day 7 wound at the end of the timeline. Dotted line indicates wound edge; scale bars, 500 μm.

(D) Plin1 immunostaining, flat sectioned aKFlp^Adipoq day 7 wound, P120. Dotted line indicates wound edge; scale bars, 250 μm.

(E) Flow cytometry quantification of Tomato⁺ nuclei, P120. Error bars indicate mean ± SD, one-way ANOVA (n = 4 mice supplying mammary, skin, and wound tissue samples).

(F) Plin1 and Tomato, mammary tissue, lactation day 5. From a pilot experiment with 3× tamoxifen administered at P28–P35 and analysis during lactation instead of wounding; scale bars, 250 μm.

(G) Plin1 and Tomato, mammary tissue, involution day 7; scale bars, 250 μm.

(H) Plin1 and Tomato on skin at P120; scale bars, 250 μm.

We also used AdipoqCreER;R26-LSL-Tomato mice to verify that adipocytes had transitioned into Plin1^neg fibroblast-like cells during hair cycle and lactation. Tamoxifen was administered to mice from P28–P35. During the first telogen (postnatal day 42), small Tomato⁺ adipocytes were seen in the dermal fat, but unlike wound adipocytes (Figure S4A), telogen adipocytes remained Plin1⁺ with adipocyte-like morphology (Figure S4B). Mammary tissue was analyzed at day 10 of lactation when most of the fat was replaced by milk-producing glands.¹⁸ Small Tomato⁺ adipocytes persisted in the space between milk glands, but they remained Plin1⁺ despite complete or near-complete loss of adipocyte morphology (Figure S4C). Adipocytes in unwounded adult skin were Plin1⁺ at homeostasis (Figure S4D). Thus, we could not confirm a clear adipocyte-to-fibroblast transition during the hair cycle or lactation. As an alternative approach, we introduced a PdgfraH2BEGFP reporter line. This mouse expresses EGFP fused to histone H2B under the control of the endogenous Pdgfra gene³⁶ and is useful for labeling PDGFRα⁺ cells. We reasoned that the lower induction threshold for labeling would mark PDGFRα⁺ cells more permissively and more quickly than the Flp^o system. We confirmed that AdipoqCreER;R26-LSL-tdTomato;PdgfraH2BEGFP mice induced ADPCs with EGFP⁺ nuclei at the wound edge (Figure S4E). Small and multilocular Tomato⁺ adipocytes with EGFP⁺ nuclei were seen in the dermal fat during telogen (Figure S4F), confirming ADPCs in the hair cycle. However, Tomato⁺ adipocytes with EGFP⁺ nuclei could not be visualized in mammary tissue on lactation day 10 (Figure S4G), which was like adipocytes in unwounded adult skin (Figure S4H).

Together, these results suggest a significant difference between the cell transitions induced by tissue damage vs. physiological remodeling. Damage induced transdifferentiation to Plin1^neg fibroblast-like cells with high Pdgfra expression that triggered Flp^o-mediated labeling of wound ADPCs. Physiological remodeling, by contrast, did not trigger Flp^o-mediated labeling, and Plin1 persisted despite morphological changes. Hair cycling induced ADPCs detectable with PdgfraH2BEGFP. Lactation induced very dramatic tissue remodeling and loss of adipocyte area, but ADPCs were not visualized by any method. Thus, physiological remodeling involves less dramatic changes in cell phenotype, consistent with transient and reversible processes.^18,19

Skin explant culture does not induce ADPCs

Explant culture of excised mouse skin has been used to mimic wound healing and scarring.³⁷ In a recent study, the explant system was used to model adipocyte transitions, but the data did not support conversion of adipocytes into fibroblast/myofibroblasts.³⁸ We decided to ask whether explant conditions mimic in vivo wound healing with the ability to induce ADPCs along the cut edge of dermal fat. Following the published methods,^37,38 2 mm disks of full thickness skin were cultured in a standard cell culture setting with medium changes every other day until day 5 (see STAR Methods).

First, we made explants from 1-day-old tamoxifen-treated aKFlp^Adipoq mice and cultured them for 5 days (Figure S5A). As positive controls, we also made explants from AdipoqCreER and PdgfraCreER reporter mice. As expected, AdipoqCreER- and PdgfraCreER-regulated reporters labeled adipocytes and fibroblasts at day 0 (Figure S5B). Curiously, fluorescence was rapidly quenched in AdipoqCreER and only the outer perimeter of the explant retained strong Tomato expression by day 3. Tomato also decreased somewhat in PdgfraCreER over time. Yet, Tomato⁺ ADPCs were never induced in aKFlp^Adipoq explants (Figure S5B). Suspecting that cell death might play a role in quenched fluorescence of controls, we sectioned explants during the first day of culture and discovered abundant TUNEL⁺ cells (Figure S5C). To investigate whether age could be a confounder, we repeated the experiment with skin of 7-week-old mice after tamoxifen treatment (Figure S5D). We encountered the same loss of fluorescence in PdgfraCreER and AdipoqCreER explants (Figure S5E). Again, ADPCs were not induced in aKFlp^Adipoq explants. We confirmed that 5 days was sufficient time for two-step labeling in vivo (Figure S5F).

We conclude that skin explants have a limitation for detection of adipocyte transitions because of cell death and loss of fluorescent signal for certain cell types like adipocytes, which needs to be considered in ex vivo assays. The explant setting may also lack specific molecular, metabolic, and/or mechanical cues that trigger the transition.

Wound ADPCs exhibit fibroblast characteristics that are increased by PDGF signaling

Tomato⁺ ADPCs were heterogeneous, with some exhibiting round-shaped adipocyte morphology and others exhibiting spindle-shaped fibroblast morphology. Immunostaining for Plin1 was also heterogeneous, with adipocyte-like ADPCs having Plin1 and fibroblast-like ADPCs lacking it. ADPCs with a transitional morphology were usually Plin1⁺. Interestingly, morphology and Plin1 expression differed between genotypes, with those in aKFlp^Adipoq wounds having a preponderance of Plin1^neg fibroblastic ADPCs compared to mostly Plin1⁺ adipocyte-shaped ADPCs in aRFlp^Adipoq and bRFlp^Adipoq wounds (Figures 4A–4C). We calculated the percentage of Plin1⁺ ADPCs within the total ADPC population. By this measurement, ~25% of ADPCs were Plin1^neg in aRFlp^Adipoq and bRFlp^Adipoq wounds, but ~90% were Plin1^neg in aKFlp^Adipoq wounds (Figure 4D). These data suggest that some wound ADPCs retain adipocyte characteristics while others undergo morphological and gene expression changes. This transition is PDGF-sensitive, as elevated PDGFRα signaling in aKFlp^Adipoq adipocytes increased Plin^neg fibroblast-like phenotypes. Since aRFlp^Adipoq and bRFlp^Adipoq ADPCs were similar, we focused on comparing aRFlp^Adipoq and aKFlp^Adipoq.

Figure 4. ADPCs exhibit fibroblast characteristics that are increased by PDGF signaling.

(A–C) Plin1 immunostaining, flat sectioned wounds, day 7. Dotted line indicates wound edge; scale bars, 500 μm. Zoomed-in view shown below. Arrowheads indicate Plin1^neg ADPCs; scale bars, 250 μm.

(D) Quantification of Plin1^neg cells within the ADPC population. Most ADPCs are Plin1^neg in aKFlp^Adipoq. Error bars indicate mean ± SD, one-way ANOVA (n = 7 aRFlp^Adipoq, 7 aKFlp^Adipoq, and 6 bRFlp^Adipoq).

(E) Tomato and EGFP, aRFlp^Adipoq wound edge, day 7. Filled arrowheads indicate EGFP⁺ ADPCs. Empty arrowheads indicate Tomato^neg adipocytes with EGFP; scale bars, 100 μm.

(F) Tomato and EGFP, aKFlp^Adipoq wound edge, day 7. Filled arrowheads indicate EGFP⁺ ADPCs; scale bars, 100 μm.

(G) Tomato and EGFP, AdipoqCreER wound edge, day 7. Many adipocytes are EGFP⁺. Filled arrowheads indicate EGFP⁺ adipocytes or adipocyte-derived cells; scale bars, 100 μm.

(H and I) Flow cytometry plot of wound cell nuclei from aKFlp^Adipoq Col1a1-EGFP or with no fluorescence (H), and quantification of EGFP in Tomato⁺ nuclei (I) on day 7. Error bars indicate mean ± SD (n = 3 AdipoqCreER, 4 aRFlp^Adipoq, and 6 aKFlp^Adipoq).

(J) Schematic of EdU labeling experiment.

(K and L) Flow cytometry plot of wound cell nuclei from aKFlp^Adipoq mouse treated with EdU and labeling reaction or from a non-fluorescent mouse without EdU treatment but with labeling reaction (K) and quantification of EdU in Tomato⁺ nuclei (L) on day 7. Error bars indicate mean ± SD, one-way ANOVA (n = 5 aRFlp^Adipoq, 4 aKFlp^Adipoq, and 3 PdgfraCreER).

Collagen type I expression is a property of fibroblasts but not adipocytes at homeostasis. Using a Col1a1-EGFP transgene, we verified EGFP expression in fibroblasts but not adipocytes from unwounded skin (Figures S6A–S6C). By flow cytometry, we observed that 24% of nuclei were EGFP⁺Tomato⁺ in unwounded PdgfraCreER mice but only 1.73% were EGFP⁺Tomato⁺ in AdipoqCreER mice (Figure S6B). To see if wounding induced collagen expression in ADPCs, we introduced Col1a1-EGFP into aRFlp^Adipoq and aKFlp^Adipoq mice. At the edge of aRFlp^Adipoq and aKFlp^Adipoq wounds, EGFP was expressed in Tomato⁺ ADPCs, as observed by fluorescence microscopy (Figures 4E and 4F). However, it was also expressed in wound-adjacent adipocytes that did not express Tomato, suggesting that Col1a1 is upregulated more broadly than Pdgfra. Even in AdipoqCreER wounds, EGFP was expressed in adipocytes at the wound edge (Figure 4G). By flow cytometry, we observed that EGFP was expressed in ~95% of Tomato⁺ nuclei from aRFlp^Adipoq and aKFlp^Adipoq wounds (Figures 4H and 4I). In AdipoqCreER wounds, the rate of EGFP expression in Tomato⁺ nuclei was lower (72%), indicating that intersectional tracing enriches for adipocytes that induce Col1a1. When fibroblasts acquire contractile properties, they express alpha smooth muscle actin (αSMA) and transition into myofibroblasts that populate the wound bed, peaking at day 7 of healing. By immunostaining, we noted that fibroblast-like ADPCs outside the wound bed were αSMA^neg in aRFlp^Adipoq and aKFlp^Adipoq wounds but those in the wound bed were mostly αSMA⁺ (Figures S7A and S7B). To compare aRFlp^Adipoq and aKFlp^Adipoq wounds, we counted fibro-blasts-like ADPCs for αSMA expression. Considering all fibroblast-like ADPCs regardless of location relative to the wound bed, ~70% were αSMA⁺ in aRFlp^Adipoq compared to 40% in aKFlp^Adipoq (Figure S7C). Considering only the wound bed, 90% of ADPCs were αSMA⁺ in aRFlp^Adipoq compared to ~70% in aKFlp^Adipoq (Figure S7D). Therefore, wound adipocytes that transition to fibroblasts can further transition into αSMA⁺ myofibroblasts if they localize to the wound bed. This transition is reduced in aKFlp^Adipoq because elevated PDGFRα signaling opposes the contractile state.³⁰ Together, these data characterize wound-associated ADPCs as collagen-expressing cells that contribute to the larger population of fibroblasts and myofibroblasts in the wound bed.

We profiled the adipocyte translatome from day-3 or day-7 flank wounds compared to unwounded dermal adipocytes. To do this, we combined AdipoqCreER with the NuTRAP system, which expresses EGFP-tagged ribosomes for translating ribosome affinity purification (TRAP) of mRNA from Cre-dependent cell types.²⁷ By TRAP sequencing, we identified differentially expressed genes (DEGs) indicative of a time-dependent adipocyte-to-fibroblast transition, including progressive upregulation of fibroblast/myofibroblast markers Pdgfra, Pdgfrb, Col1a1, Col1a2, Postn, Acta2, Tagln, Trpc6, Cthrc1, and the proliferation marker Mki67 (Figure S7G). The APC marker Dlk1 was significantly upregulated, while adipocyte markers Adipoq, Plin1, Cebpa, and Pparg were downregulated in day-7 wound adipocytes (Figure S7H). Interestingly, wound adipocytes upregulated transcription factors involved in fibrosis and cell plasticity (Sox9, Tead2, Gli1, Snai1, and Snai2) and secreted cytokines and ECM-modifying proteins (Adam12, Bmp1, Spp1, Mmp14, Ccn2, Ccl2, Angpt2, Pdgfb, Areg, Fgf7, and Vegfc) (Figures S7I and S7J).

Activated fibroblasts replicate DNA and proliferate but adipocytes at homeostasis should be non-dividing. To visualize DNA replication, we injected EdU during wound healing and harvested nuclei on day 7 for flow cytometry (Figure 4J). In aRFlp^Adipoq and aKFlp^Adipoq wounds, the raw percentage of EdU⁺Tomato⁺ nuclei was very low (0.2% and 0.62%, respectively), reflecting a low number of Tomato⁺ cells overall (Figures S7E and S7F). Therefore, we normalized the EdU⁺Tomato⁺ nuclei by the total number of Tomato⁺ nuclei. After normalization, 6.8%, and 12.0% of Tomato⁺ nuclei were EdU⁺ in aRFlp^Adipoq and aKFlp^Adipoq wounds, compared to 17.81% in PdgfraCreER (Figures 4K and 4L). The percentage of EdU⁺ ADPCs was significantly higher in aKFlp^Adipoq compared to aRFlp^Adipoq wounds.

Together, these findings of changes in ADPC morphology, gene expression, and DNA replication agree with previous studies characterizing wound adipocyte plasticity.^17,34 More fibroblastic ADPCs in aKFlp^Adipoq demonstrate that ADPCs are sensitive to PDGF signaling and respond with cell phenotype transitions and increased DNA replication. Notably, this extends our previous work showing that PDGFRα signaling increases the proliferation of wound fibroblasts and APCs.^29,30,39

PDGFRα knockout reduces fibroblast characteristics of ADPCs

To investigate whether PDGFRα signaling is required for the acquisition of fibroblast characteristics, we generated aRFlp^Adipoq mice with two aRFlp alleles (aRFlp^Adipoq-iKO) for inducible Pdgfra knockout that replaces PDGFRα protein with Flp. As expected, Tomato⁺ cells appeared at the edge of day 7 aRFlp^Adipoq-iKO wounds (Figure 5A). As observed by flow cytometry, the percentage of Tomato⁺ nuclei was lower in aRFlp^Adipoq-iKO wounds compared to aRFlp^Adipoq (Figure 5B). This shows the Pdgfra promoter is active and Flp^o can label ADPCs even if PDGFRα signaling per se does not occur, but the lower percentage of Tomato⁺ nuclei suggests that the transition was less efficient without signaling. The ADPCs were sensitive to Pdgfra knockout, as they exhibited a round adipocyte morphology in aRFlp^Adipoq-iKO (Figure 5D), and the percentage of Plin1^neg ADPCs in aRFlp^Adipoq-iKO wounds was lower than in aRFlp^Adipoq wounds (Figure 5E). This is the opposite result from aKFlp^Adipoq where the percentage of Plin1^neg ADPCs was increased and most ADPCs exhibited a ramified fibroblast-like morphology (Figure 4B).

Figure 5. PDGFRα knockout reduces fibroblast characteristics of ADPCs.

(A) Plin1 immunostaining, cross sectioned wound, day 7. Whole-wound image is stitched together from three pictures; scale bars, 500 μm. Zoomed-in view shown on the right; scale bars, 100 μm.

(B) Flow cytometry of Tomato⁺ nuclei, day 7. Error bars indicate mean ± SD, Student’s t test (aRFlp data are same as Figure 2F; n = 9 aRFlp^Adipoq and 5 aRFlp-iKO).

(C and D) Plin1 immunostaining, flat sectioned wounds, day 7. Dotted line indicates wound edge; scale bars, 500 μm. Zoomed-in view shown below. Arrowheads indicate Plin1^neg ADPCs; scale bars, 100 μm.

(E) Quantification of Plin1^neg cells within the ADPC population. Error bars indicate mean ± SD, Student’s t test (aRFlp data are the same as Figure 4D; n = 7 aRFlp^Adipoq and 8 aRFlp-iKO).

(F and G) Tomato and Col1a1-EGFP, flat sectioned wound edge, day 7. Arrowheads indicate EGFP⁺ cells; scale bars, 100 μm.

(H) Quantification of EGFP in Tomato⁺ nuclei, day 7. Error bars indicate mean ± SD, Student’s t test (aRFlp data are the same as Figure 4I; n = 4 aRFlp^Adipoq and 4 aRFlp-iKO).

(I) Quantification of EdU in Tomato⁺ nuclei, day 7. Error bars indicate mean ± SD, Student’s t test (aRFlp data are the same as Figure 4L; n = 5 aRFlp^Adipoq and 3 aRFlp-iKO).

ADPCs in aRFlp^Adipoq-iKO wounds expressed Col1a1-EGFP like aRFlp^Adipoq wounds, consistent with Col1a1 being independent from PDGF signaling (Figures 5F–5H). However, DNA synthesis was significantly downregulated in aRFlp^Adipoq-iKO wounds compared to that in aRFlp^Adipoq (Figure 5I). Again, this is the opposite result from aKFlp^Adipoq wounds (Figure 4L). Collectively, these gain-of-function and loss-of-function data show that PDGF signaling promotes Plin1 downregulation, acquisition of the ramified or spindle-shaped fibroblastic phenotype, and DNA replication.

PDGFRα signaling in ADPCs promotes wound repair

We wondered whether aRFlp^Adipoq-iKO wounds or aKFlp^Adipoq wounds heal differently compared to control aRFlp^Adipoq wounds. Adipocytes per se, and their lipolytic function, are required for efficient skin repair, as adipocyte ablation or inhibition of lipolysis caused delayed wound re-epithelialization.¹⁷ We previously showed that PDGFRα signaling in fibroblasts could significantly impact cell proliferation, myofibroblast differentiation, and wound contraction.³⁰ However, PDGFRα signaling in ADPCs might have subtle impacts because of their smaller number. Therefore, it was pertinent to examine the wound phenotype here.

We analyzed day 5 wounds to quantify re-epithelialization, angiogenesis, and myofibroblasts. Histological sections across the wound bed were measured for percentage of the wound surface covered by newly generated epidermis (a higher percentage indicates more epithelialization). Control aRFlp^Adipoq wounds displayed ~60% re-epithelialization, but aKFlp^Adipoq wounds displayed complete or nearly complete re-epithelialization. There was decreased re-epithelialization in aRFlp^Adipoq-iKO wounds compared to that in controls (Figures 6A and 6B). Sections were stained for CD31, NG2, and αSMA to analyze angiogenesis, pericytes, and myofibroblasts in the wound bed. The area of CD31⁺ stain was significantly increased in aKFlp^Adipoq wounds and decreased in aRFlp^Adipoq-iKO wounds compared to that in control wounds (Figures 6C and 6D). NG2⁺ pericytes were present on wound capillaries (Figures S7K and S7L). The area of αSMA had not significantly changed between genotypes (Figures 6E and 6F). These results indicate that PDGFRα signaling in ADPCs promotes wound healing by cell non-autonomous mechanisms influencing keratinocytes and endothelial cells. Potential mediators include the secreted factors upregulated in wound adipocytes (Figure S7K), and although wound adipocytes can become myofibroblasts, most myofibroblasts are derived from tissue-resident fibroblasts. Our finding that the αSMA⁺ population was not changed in aRFlp^Adipoq-iKO and aKFlp^Adipoq wounds is consistent with ADPCs being a relatively minor myofibroblast source.

Figure 6. PDGFRα signaling in ADPCs promotes wound repair.

(A and B) H&E staining, cross sectioned wounds, day 5. Dotted line indicates wound bed area. Solid black line indicates length of re-epithelialization and black arrows indicate end of epithelial tongues; scale bars, 500 μm. Dotted boxes (1)–(7) are re-imaged at high magnification, showing end of epithelial tongues. aKFlp^Adipoq is 100% re-epithelialized; scale bars, 100 μm. (B) Quantification of re-epithelialization normalized to wound width.

(C) CD31 immunostaining, cross sectioned wounds, day 5; scale bars, 300 μm. Zoomed-in view shown in lower right; scale bars, 100 μm.

(D) Quantification of CD31⁺ area.

(E) αSMA immunostaining, cross sectioned wounds, day 5; scale bars, 500 μm.

(F) Quantification of αSMA⁺ area.

(G and H) Flow cytometry plots (G) and quantification of macrophages in wounds or unwounded skin (H). Error bars indicate mean ± SD (n = 2 1.5 days after wounding, 2 3 days after wounding, 4 unwounded skin).

(I and J) Quantification of macrophages at 1.5 days normalized to total CD45⁺ cells (I) or raw number per wound (J). Error bars indicate mean ± SD, one-way ANOVA (n = 3 of each genotype).

In (B), (D), and, (F), error bars indicate mean ± SD, one-way ANOVA (n = 6 aRFlp^Adipoq, 5 aKFlp^Adipoq, and 6 aRFlp-iKO).

We also performed flow cytometry to quantify macrophages. Macrophage recruitment was maximum at day 1.5 compared to day 3 wounds or unwounded skin, as shown previously¹⁷ (Figures 6G and 6H), but macrophage recruitment at day 1.5 was not changed in aRFlp^Adipoq-iKO and aKFlp^Adipoq wounds compared to controls (Figures 6I and 6J). This differs from the results of Shook et al., where suppressing adipocyte survival and lipolytic activity reduced macrophage recruitment.¹⁷ Thus, the changes wrought by modulating PDGFRα signaling in ADPCs are distinct from the consequences of adipocyte depletion and lipolysis suppression.

Wound-induced ADPCs persist as fibroblasts after wound healing

To investigate whether wound healing induces enduring changes in adipocytes, we examined ADPCs at 21 days after wounding. By this time, the scar is fully established and well past the dynamic period of cell migration, matrix synthesis, and myofibroblast contraction. The edges of day 21 aRFlp^Adipoq scars contained Plin1⁺ adipocyte-like ADPCs but also Plin1^neg fibroblast-like ADPCs that clearly had not redifferentiated into adipocytes (Figure 7A). The history of Plin1⁺ ADPCs is ambiguous: they could have redifferentiated into adipocytes, or they could have expressed Pdgfra initially without fully losing adipocyte characteristics, as they were also seen at day 7 (Figure 4D). However, the percentage of Plin1^neg ADPCs increased from day 7 to day 21, implying loss of adipocyte identity over time rather than redifferentiation (Figure 7C). Notably, day 21 aKFlp^Adipoq scars had abundant Plin1^neg ADPCs at day 21, the same as day 7 (Figures 7B and 7C). These results show that Plin1^neg ADPCs do not redifferentiate into adipocytes after wound healing. This is most obvious in aKFlp^Adipoq ADPCs, but it also applies to aRFlp^Adipoq ADPCs with physiological signaling.

Figure 7. Wound-induced ADPCs persist as fibroblasts after wound healing.

(A and B) Plin1 immunostaining, flat sectioned wounds, day 21. Dotted line indicates wound edge; scale bars, 500 μm. Zoomed-in view shown to the right. Arrowheads indicate Plin1^neg ADPCs; scale bars, 250 μm.

(C) Quantification of Plin1^neg cells within the ADPC population, day 7 or day 21. Error bars indicate mean ± SD, one-way ANOVA (day 7 data are the same as that in Figure 4D; n = 5 aRFlp^Adipoq and 5 aKFlp^Adipoq for day 21).

(D) Plin1 immunostaining, cross sectioned wound, day 21. Dotted line indicates wound edge. Large image is stitched together from three pictures; scale bars, 250 μm. Zoomed-in view shown below. Arrowheads indicate Plin1^neg ADPCs; scale bars, 100 μm.

(E) Schematic for interpreting red cells (adipocyte-derived Plin^neg fibroblasts), yellow cells (Tomato⁺Plin1⁺ adipocytes), and green cells (newly generated Tomato^negPlin1⁺ adipocytes) in (D).

Similar experiments with day 21 AdipoqCreER scars revealed three interesting populations (Figures 7D and 7E). First was of Tomato⁺Plin1^neg cells with fibroblast-like morphology, presumably ADPCs, mainly localized to the scar edge (Figure 7D’). Second was of Tomato⁺Plin1⁺ adipocytes located more distant from the wound edge (Figure 7D). These cells could represent transiently dedifferentiated cells that reverted to the adipocyte phenotype, but their location away from the wound edge was more consistent with naive adipocytes that did not transition. Interestingly, a third population of Tomato^negPlin1⁺ adipocytes was seen between the two Tomato⁺ populations (Figure 7D). Tomato^neg indicates that these adipocytes were not targeted by AdipoqCreER when tamoxifen was given prior to wounding and were, therefore, newly generated from APCs during healing. It is known that anagen hair follicles, which are present at the scar edge on day 21, secrete signals that promote adipogenesis.^6,40 Thus, the scar edge environment is pro-adipogenic, but even so, adipocyte-derived Tomato⁺Plin1^neg cells did not redifferentiate into adipocytes. This leads us to conclude that wound healing causes adipocytes to transdifferentiate into fibroblasts.

DISCUSSION

We developed genetic tools to lineage trace mature adipocytes that acquire Pdgfra or Pdgfrb expression. Wound healing strongly induced ADPCs at the excised front of dermal and subcutaneous fat. In 3-week-old scars, Plin1^neg ADPCs were still present at the scar edge, abutting anagen hair follicles and newly generated adipocytes. We addressed the role of PDGFRα signaling in the ADPC transition. By genetic gain-of-function and loss-of-function approaches, DNA replication was sensitive to PDGFRα signaling. Col1a1 was induced in adipocytes and ADPCs near the wound, but it was not sensitive to PDGFRα signaling. This may seem surprising because PDGF is functionally important for collagen-intensive processes, including connective tissue development and organ fibrosis.^{32,39,41–44} But PDGF’s principal role may be to increase fibroblast populations via proliferation and survival rather than driving collagen transcription directly.³⁰

Adipocytes contribute to wound healing by undergoing lipolysis to release stored lipid and by transitioning to matrix-producing fibroblasts. Under experimental conditions that depleted adipocytes or blocked their lipolytic function, there was a reduction in macrophage recruitment, angiogenesis, and re-epithelialization.¹⁷ Adding to this, we found that PDGFRα signaling in ADPCs was needed for optimal healing, and furthermore, constitutively active PDGFRα in ADPCs increased angiogenesis and re-epithelialization. We did not observe changes in macrophage recruitment, however. It has been proposed that fibroblasts and macrophages communicate in a stable circuit with reciprocal signaling by PDGF and colony stimulating factor 1 (CSF1).⁴⁵ Thus, a more detailed examination of immune cell behaviors could still reveal effects due to modulated PDGFRα signaling. Our observation that Pdgfra gene manipulation resulted in measurable effects on wound angiogenesis and re-epithelialization implies paracrine contributions of ADPCs to wound healing. This could occur through upregulated cytokine or ECM secretion, including angiogenic factors or epithelial-supporting factors suggested by TRAP sequencing.

Transdifferentiation is direct conversion from one specialized cell type to a different cell type without a progenitor intermediate. Dedifferentiation is when a mature cell type reverts to a progenitor state, which is implicitly temporary and reversible.^22,46 In the context of the hair growth cycle and mammary gland remodeling, transient loss of adipocyte characteristics followed by reacquisition of said characteristics has been termed dedifferentiation.^18,19 A potential reason for why intersectional lineage tracing did not label ADPCs in the hair cycle and mammary gland is because temporary dedifferentiation may not raise Pdgfra levels high enough to trigger Flp-recombination. We could detect ADPCs in the hair cycle with PdgfraH2BEGFP, requiring no recombination, which implies lower expression of Pdgfra and possibly a lesser role for PDGF signaling in reversible dedifferentiation. We do not know why this approach still failed to detect ADPCs in the mammary gland; more extensive studies are required. In the future, it will be important to survey chromatin landscapes in ADPCs because transdifferentiation to a fibroblast should bring changes in chromatin structure.

We compared adipocyte-to-fibroblast transitions in pathological and physiological conditions, but this raised more questions about the generalizability of the process. Our results suggest that dermal adipocytes are more plastic than mammary gland adipocytes, and pathological conditions induce more dramatic transitions than physiological conditions. What other tissues and conditions support adipocyte plasticity? How plastic are different kinds of fat cells, including perivascular, mesenteric, intramuscular, and bone marrow adipocytes? Adipocyte transitions have been implicated in diseases including scleroderma, tumor stroma formation, and creeping fat associated with bowel inflammation.^20,47,48 The high stringency of intersectional lineage tracing makes it suitable for screening drugs, mutations, or in vivo conditions that promote transdifferentiation, but keeping in mind that subtle or reversible transitions will likely be missed. In scarring and fibrosis, how relevant is adipocyte transdifferentiation compared to conventional sources of fibroblasts/myofibroblasts? This likely depends on the tissue and disease context, but it is important to resolve so that proper emphasis can be placed on therapy to block the transition.

Single-cell sequencing is ideal for describing cell heterogeneity, which obviously exists in wound adipocytes. However, buoyant, fragile, lipid-laden adipocytes are not ideal for microfluidics approaches, resulting in adipocytes missing from single cell datasets. Shook et al. sequenced 13 adipocytes in their study, and from these data, they reported upregulation of collagens, Acta2, Pdgfra, and Pdgfrb.¹⁷ Kalgudde Gopal et al. challenged the notion of adipocyte-to-fibroblast plasticity. Their study primarily used in vitro skin explants and non-wound adipocytes placed in cell culture.³⁸ These choices assumed that basic in vitro conditions were sufficiently close to in vivo wounds as to induce adipocyte plasticity, but with no positive evidence or control. We did not find ADPCs in skin explants, at least partly due to adipocyte death but also possibly because the adipocyte-to-fibroblast transition requires specific conditions that are not replicated by routine culture conditions. The molecular, metabolic, and/or mechanical cues needed to induce adipocyte transitions are still unknown, but they at least exist in wounds. Kalgudde Gopal et al. performed bulk sequencing of FACS-en-riched wound adipocytes and fibroblasts, but they could not discern a convergence of lineages at the bulk transcriptomic level.³⁸ Bulk sequencing measures average gene expression in the population, which would mask the relatively small number of ADPCs. We used TRAP with AdipoqCreER to isolate ribosome-bound mRNA from adipocytes, which identified progressive upregulation of fibroblast/myofibroblast genes and downregulation of adipocyte genes in wound adipocytes. Interestingly, ADPCs that entered the wound were usually αSMA⁺, while those that remained at the wound perimeter were αSMA^neg. This suggests that ADPCs, unsurprisingly, like many other cell types, acquire myofibroblast gene expression when they experience stiff surfaces, fibrotic ECM, and TGFβ signaling.^47–49

Fibroblast identity of ADPCs remains underdefined at the level of chromatin or genome-wide transcription. Single-cell and single-nucleus RNA-seq and ATAC-seq would be ideal for resolving how ADPCs are transcriptionally aligned with fibroblast populations and the degree of retained adipocyte features. Because the ADPC population includes adipocyte-like cells that present challenges for microfluidics and sorting, single-nucleus approaches will be required to capture an unbiased picture of adipocyte and ADPC heterogeneity. Our intersectional lineage tracing model could improve nuclei isolation by allowing enrichment of transitioning adipocytes, which could then be compared to PDGFRα⁺ wound fibroblasts and Adipoq⁺ dermal adipocytes. The NuTRAP model facilitates nuclei isolation with its inducible biotinylation component, but we did not employ this tool in the current study. Despite technical challenges, the field would benefit from improved datasets to resolve adipocyte heterogeneity in wound healing.

In summary, our data highlight the unique characteristic of wound repair to induce robust Pdgfr expression in adipocytes and enduring loss of adipocyte features, consistent with transdifferentiation. This is in contrast with reversible adipocyte dedifferentiation in the hair cycle and mammary gland remodeling. These results reinforce the role of adipocytes in skin wound repair by upregulating PDGF signaling and transdifferentiating into matrix secreting fibroblasts. The ADPC population is small compared to other fibroblasts in wound healing, but we found that modulating PDGFRα signaling in ADPCs influenced angiogenesis and re-epithelialization. The specific cues that trigger Pdgfr expression and the adipocyte-to-fibroblast transition remain to be identified. It is important to know that this transition is a form of plasticity because it implies that adipocytes are not a dead-end cell type. Instead, they can produce an expanded cell lineage for healing and regeneration, and potentially as a mechanism of organ fibrosis.

Limitations of the study

In most lineage tracing experiments with tamoxifen- or doxycycline-regulated Cre recombinase, a reporter gene can be detected within 24 h of drug treatment or earlier. The high specificity of our two-step system has a limitation in that it takes more time for the reporter to develop. Tamoxifen primes the reporter by allowing Flp^o to be expressed by Pdgfra or Pdgfrb, but Flp^o cannot begin to express until the Pdgfr genes are induced in adipocytes. A major determinant of the reporter’s development is how quickly Flp^o expression reaches a threshold that efficiently removes the STOP cassette. We used codon-optimized Flp^o that is equivalent to Cre in its activity in mammalian cells.⁵⁰ As we hoped, there was no off-target labeling of Pdgfr^neg adipocytes with two recombinases acting in sequence. But the earliest we could detect Tomato was day 4 after wounding, so the first days of the ADPC transition cannot be visualized. As a trade-off for stringent labeling, adipocytes with low or transient Pdgfr expression are likely to be missed. As a less stringent approach, we used a PdgfraH2BEGFP reporter to validate ADPCs induced by hair cycle, but this labeling is not indelible.

RESOURCE AVAILABILITY

Lead contact

All data supporting the findings of this study are available from the lead contact Lorin E. Olson (lorin-olson@omrf.org) upon reasonable request.

Materials availability

This study did not generate new, unique reagents.

STAR★METHODS

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

The mouse strains B6.129-Tg(Adipoq-cre/Esr1*)1Evdr/J (AdipoqCreER, #024671), B6.Cg-Gt(ROSA)26Sor^{tm14(CAG-tdTomato)Hze}/J (R26-LSL-Tomato, Ai14, #007914), B6.FVB-Tg(Adipoq-Cre)1Evdr/J (AdipoqCre, #028020), B6.129S4-Pdgfra^{tm11(EGFP)Sor}/J (PdgfraH2BEGFP, #007669), and B6;129S6-Gt(ROSA)26Sor^{tm2(CAG-NuTRAP)Evdr}/J (R26-NuTRAP, #029899) were purchased from Jackson Laboratories.^{24,27,36,56,57} PdgfraCreER mice were obtained from Brigid Hogan.⁵¹ Col1a1-EGFP mice were obtained from David Brenner.⁵² Pdgfra^Flp, Pdgfrb^Flp and Pdgfra^K.Flp mice were developed in the Olson laboratory, and R26^{Frt-STOP-Frt-tdTomato} mice were derived from JAX#021875 as described previously.²⁹ All animal experiments were performed according to procedures approved by the Institutional Animal Care and Use Committee at the Oklahoma Medical Research Foundation. Mice were maintained on a 12 h light/dark cycle and housed in groups of two to five with unlimited access to food and water. All strains were maintained on a mixed C57BL6/129 genetic background at room temperature. Both males and females were analyzed, and no association was observed between the results and male vs. female sex.

For explant culture of P1 mouse skin, pregnant females (E18.5) were administered 50 mg/kg tamoxifen by gavage. After delivery, the skin of the P1 pups was excised using a sterile 2mm biopsy punch. For explant culture of adult skin, 6–8 week female or male mice were administered 100 mg/kg tamoxifen by gavage three times. Skin was harvested using a sterile 2 mm biopsy punch after hair removal with depilatory cream. After two washes with PBS, the skin was transferred to DMEM/F12 media with 10% FBS, 0.5% Penicillin and streptomycin, 1X GlutaMAX, 1X MEM non-essential amino acid. Media was changed every 2 days. During the 5-day culture, explant tissue was collected on the indicated day. Image analysis was performed using Nikon AX R confocal microscope. For TUNEL stain, the harvested skin was fixed with 4% PFA, dehydrated with 20% sucrose and then embedded in OCT compound and sectioned into 10-μm sections. TUNEL stain was performed according to the manufacturer’s instructions (Cat# C10245).

METHOD DETAILS

Wounding

Tamoxifen was prepared as a 20mg/mL stock in corn oil and mice were gavaged three times on alternating days with 100 mg tamoxifen/kg body weight. Wounds were created 4 days after the last tamoxifen treatment. To create excisional wounds, 7-9 weeks-old mice were wounded during the telogen phase of hair cycling. Mice were administered analgesic (ketoprofen 5mg/kg) followed by inhaled anesthesia (5% isoflurane/1% oxygen). Dorsal hair was shaved and then completely removed using depilatory cream. The exposed skin was sterilized with 70% ethanol. Excisional wounding was performed using a 5 mm biopsy punch to create 4 full-thickness dermal wounds. Most wounds were created in the mid-dorsal back skin. Some wounds were created in the flank region to involve subcutaneous fat. Mice were then single housed, and wounds were left uncovered during healing. At the time of harvest, wound areas and a margin of unwounded skin were harvested and imaged from the underside with a fluorescent dissecting microscope. Wounds and fat depots were then fixed overnight in 4% paraformaldehyde (PFA) for microscopy or processed for flow cytometry.

Fat remodeling with lactation and hair cycle

Pilot experiments were performed on males (hair cycle) and females (lactation + hair cycle) with 3–5 tamoxifen treatments before 6-weeks-old, chase times of 3–15 weeks, and unfixed tissues screened for Tomato⁺ cells using a fluorescent dissecting microscope. aRFlp^Adipoq, aKFlp^Adipoq, and bRFlp^Adipoq mice were used for pilot experiments. When no Tomato⁺ cells could be observed in any fat depot, formal experiments were planned with wounding as a positive control. Female aKFlp^Adipoq mice between 28 and 35 days old were gavaged three times on alternating days with 100 mg tamoxifen/kg body weight. At 12 weeks of age, the females were mated and allowed to deliver their pups. After 1 week of nursing, the pups were removed to induce mammary gland involution, and the females were wounded. After 1 week of wound healing and mammary involution, the females were sacrificed and mammary fat, unwounded skin, and wound tissue were harvested for analysis by fluorescent microscopy and flow cytometry. For experiments combining AdipoqCreER and PdgfraH2BEGFP, mice of either sex between 28 and 35 days old were gavaged three times on alternating days with 100 mg tamoxifen/kg body weight and telogen skin was analyzed on day 42. Alternatively, female mice were mated, allowed to deliver their pups, and mammary glands were analyzed on day 10 of nursing. EGFP was detected by antibody staining with a Cy5-conjugated secondary to avoid Tomato background in the green channel.

Staining and imaging

Mouse wounds and fat depots were dehydrated in 20% sucrose, embedded in optimum cutting temperature (OCT) compound, and frozen sections were cut. Some wounds were sectioned in a cross-section (or full thickness) orientation at 10–20 μm thickness. Some wounds were sectioned in a flat-section orientation, parallel to the plane of the skin at 20 μm thickness. Frozen subcutaneous fat was sectioned at 50 μm. Sections were stained with antibodies listed below, counterstained with DAPI, and mounted in Fluoro Gel. Paraffin sections were cut at 10 μm and stained with hematoxylin and eosin. Images were captured on a Nikon Eclipse 80i microscope or Nikon AX R confocal microscope. Wound-spanning fluorescent images (like Figures 2A–2D) were spliced together from 2 to 3 sequential images.

Image quantification

Quantification of Plin1^neg ADPCs was performed on flat sections. Between 70 and 220 Tomato⁺ cells were scored at the wound edge for each mouse. For each mouse the number of Plin1^neg ADPCs was divided by the total number of ADPCs counted to obtain the % Plin1^neg ADPCs. Quantification was performed manually. Experiments were performed blind when there were not obvious differences in phenotype that revealed the genotype. Quantification of CD31⁺ and αSMA⁺ area was performed using ImageJ and digital microscope images of fluorescent antibody staining to measure stained area and total wound bed area, then converted to a percentage. Wound bed area was defined as the area bounded by the intact dermis flanking the wound, the scab or new epidermis above, and the fascia or bottom of the tissue sample below. Quantification of re-epithelialization was performed using ImageJ and digital microscope images of wound sections stained with hematoxylin and eosin. The two epithelial tongues were measured, and their combined length was divided by the wound width. Quantification of αSMA⁺Tomato⁺ fibroblasts was performed on flat sections. Between 29 and 129 Tomato⁺ cells with a fibroblast morphology were scored for each mouse including cells inside and outside the wound bed. For each mouse the number of αSMA⁺Tomato⁺ fibroblasts was divided by the total number counted to obtain the % αSMA⁺Tomato⁺ fibroblasts in the total population of Tomato⁺ fibroblasts. A separate calculation was performed only considering cells inside the wound bed to obtain the % αSMA⁺Tomato⁺ fibroblasts in the wound population of Tomato⁺ fibroblasts. Quantification was performed manually. Blinding was not possible because the genotype aRFlp^Adioq or aKFlp^Adipoq was obvious based on cell morphology.

Flow cytometry

Quantification of nuclei: Small pieces of tissue were fixed with 0.5% PFA for 30 min at room temperature to prevent loss of fluorescence during nuclei isolation. Then the tissue was washed with PBS, mechanically disrupted with a scalpel, and added to 500 μL nuclei isolation buffer (NIB, 250 mM sucrose, 10 mM HEPES, 1.5 mM MgCl2, 10 mM KCL, 0.001% IGEPAL CA-630, 0.2 mM DTT). The minced tissues were transferred to a pre-cooled 2 mL glass Dounce homogenizer and homogenized to release nuclei.⁵⁸ Then the nuclei were filtered through a 70 μm strainer and the Dounce homogenizer was washed with 1mL NIB and added to the filter. The filtrate was centrifuged at 1000 x g for 10 min at 4°C, the pellet was washed with 1mL NIB, and spun again at 500 x g for 10 min at 4°C. The nuclei pellet was resuspended in 200 μL nuclei resuspension buffer (NRB, PBS with 1% BSA, 2mM MgCl2) and filtered with a 40 μm strainer. After staining with DAPI the nuclei were checked under the microscope and then used for flow cytometry. Compensation was determined using nuclei from mice having no fluorescence compared to ubiquitous Tomato expression or Col1a1-EGFP. Compensation tissue type was matched to the experimental tissue type (e.g., skin to skin, wound to wound, fat to fat). Analysis of EdU was performed following the manufacturer’s instructions (Cat# C10634) plus an extra wash to remove EdU background. EdU was administered on day 2, 4, and 6 of wound healing (200 μL of 2mM EdU each time). Background fluorescence was determined by performing the Click-iT reaction on nuclei from mice that were not treated with EdU. Quantification of macrophages: Wounds or skin tissue was digested with Liberase TH for 1 h at 37°C with shaking and then centrifuged at 1000 x g for 5 min. Cell pellets were resuspended in FACS buffer (PBS with 1% BSA and 2mM EDTA) and filtered with 70 μm and 40 μm strainers. Cells were then incubated with Fc blocker (anti-CD16/32) for 10 min on ice and then stained with antibodies for CD45 (1:200), CD11b (1:200) and F4/80 (1:100) for 30 min on ice. Then cells were resuspended in FACS buffer. Zombie Aqua was used to exclude dead cells. Compensation was determined with unstained cells or cells single-stained with each antibody. Flow cytometry was performed using the BD FACSCelesta, and data analysis was performed with FlowJo software.

Translating ribosome affinity purification

AdipoqCreER;R26-NuTRAP mice were treated with tamoxifen and wounded with two dorsal wounds and two flank wounds. On the designated day, mice were sacrificed, depilated, and immersed in 10% neutral buffered formalin for 10 min to eliminate RNAse activity from skin surface. Then skin was isolated and four wounded sites were combined for each mouse. Skin from an uninjured site was also collected. Samples were pulverized after snap-freezing in liquid nitrogen, then transferred to homogenization buffer (50 mM Tris (pH 7.4), 100 mM KCl, 12 mM MgCl2, 1% NP-40, 1 mM DTT, 100 μg/mL cycloheximide, 1 mg/mL sodium heparin, 600 U/mL RNase inhibitor (NEB, M0314), and 1x Complete EDTA-free protease inhibitor in DEPC-treated H2O) on ice. After centrifugation at 20,000 g for 10 min at 4°C, supernatant was collected without the top lipid layer and transferred to a new tube. Input was collected, and the remaining solution was incubated with anti-GFP antibody (ab290) for 1 h at 4°C. After two washes, Protein G dynabeads were added to the mixture of antibody and solution and incubated for 30 min at 4°C. The beads and immunoprecipitates were washed three times with high-salt wash buffer (50 mM Tris (pH 7.4), 300 mM KCl, 12 mM MgCl2, 1% NP-40, 1 mM DTT, 100 μg/mL cycloheximide in DEPC H2O). RLT buffer (Nucleospin RNA kit), supplemented with 2-mercaptoethanol, was added to the beads, and the mixture was vortexed vigorously to dissociate the mRNA from the GFP-tagged polysomes. Polysomal complexes were isolated using the Nucleospin RNA kit, including an on-column DNase digestion step.

Library generation and sequencing

After analyzing RNA integrity with Agilent TapeStation and quantifying RNA amount with the Quant-iT RiboGreen RNA assay (Thermo Fisher Scientific), libraries were generated using the Takara SMART-Seq mRNA LP kit according to the manufacturer’s instructions using 20 ng RNA. The quality of the final libraries was assessed on the Agilent Bioanalyzer and quantified using Quant-iT PicoGreen RNA assay (Thermo Fisher Scientific). 8 samples were pooled and then 10 0bp paired-end reads were generated on the NovaSeq platform (MedGenome US). Raw sequencing reads were aligned to the mouse reference genome (mm10) and quantified using Spliced Transcripts Alignments to a Reference (STAR, v 2.7.10b)⁵³ and featureCounts (v 2.0.8).⁵⁴ Read counts were imported into R (v 4.4.1) and normalized and processed using DESeq2.⁵⁵ Differentially expressed genes were defined as those with a p value ≤ 0.05 and an absolute log2 fold change ≥ 0.25. Heatmaps were generated using Pheatmap, an R package.

QUANTIFICATION AND STATISTICAL ANALYSIS

To determine significance between two groups, comparisons were made using Student’s t test. Analyses across multiple groups were made using a one-way ANOVA in GraphPad Prism. Each mouse was considered a biological replicate. Statistical details for each experiment can be found in the figure legends.

Supplementary Material

Supplemental information can be found online at https://doi.org/10.1016/j.celrep.2026.117421.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Monoclonal rabbit anti-αSMA	Cell Signaling	Cat# 19245; RRID: AB_2734735
Monoclonal rabbit anti-Plin1	Cell Signaling	Cat# 9349; RRID: AB_10829911
Polyclonal goat anti-CD31	R&D Systems	Cat# AF3628; RRID: AB_2161028
Polyclonal goat anti-GFP	Abcam	Cat# Ab6673; RRID: AB_305643
Polyclonal rabbit anti-GFP	Abcam	Cat# Ab290; RRID: AB_303395
Polyclonal rabbit anti-NG2	Millipore	Cat# AB5320; RRID: AB_91789
Monoclonal mouse anti-CD45.2 BV421 conjugate	BioLegend	Cat# 109832; RRID: AB_2565511
Monoclonal rat anti-CD11b PCP/Cy5.5 conjugate	BioLegend	Cat# 101277; RRID: AB_2876398
Monoclonal rat anti-F4/80 FITC conjugate	BioLegend	Cat# 123107; RRID: AB_893500
Polyclonal goat anti-PDGFRα	R&D Systems	Cat# AF1062; RRID: AB_2236897
Polyclonal donkey anti-rabbit IgG Cy5 conjugate	JacksonImmunoResearch	Cat# 711175152; RRID: AB_2340607
Polyclonal donkey anti-goat IgG Cy5 conjugate	JacksonImmunoResearch	Cat# 705175147; RRID: AB_2340415
Monoclonal rat anti-CD16/CD32 (FC blocker)	eBioscience	Cat# 14016182; RRID: AB_467133
Chemicals, peptides, and recombinant proteins
Tamoxifen	Sigma Aldrich	Cat# T5648
EdU	Invitrogen	Cat# A10044
Liberase TH	Sigma	Cat# 5401151001
Dynabeads protein G	Invitrogen	Cat# 10003D
Critical commercial assays
Click-iT TUNEL Assay for Microscopy	Invitrogen	Cat# C10245
Click-iT Plus EdU Assay for Flow Cytometry	Invitrogen	Cat# C10634
Zombie Aqua Fixable Viability Kit	BioLegend	Cat# 423101
SMART-Seq mRNA LP	Takara	Cat# 634768
Nucelospin RNA kit	Macherey-Nagel	Cat# 740955
Deposited data
Unwounded vs. day7 wound adipocyte TRAP-seq dataset	GEO	GEO: GSE327928
Experimental models: Organisms/strains
Mouse: Adipoq-CreER	The Jackson Laboratory	JAX: 024671
Mouse: R26-LSL-tdTomato, Ai14	The Jackson Laboratory	JAX: 007914
Mouse: Adipoq-Cre	The Jackson Laboratory	JAX: 028020
Mouse: Pdgfra-H2BEGFP	The Jackson Laboratory	JAX: 007669
Mouse: Pdgfra-CreER	Chung et al.51	N/A
Mouse: Col1a1-EGFP	Magness et al.52	N/A
Mouse: Pdgfra-Flp	Sun et al.29	N/A
Mouse: Pdgfrb-Flp	Sun et al.29	N/A
Mouse: Pdgfra-K.Flp	Sun et al.29	N/A
Mouse: R26-FSF-tdTomato	Sun et al.29	N/A
Mouse: R26-NuTRAP	The Jackson Laboratory	JAX:029899
Software and algorithms
ImageJ	N/A	ImageJ.net/ij
GraphPad Prism 10	N/A	Graphpad.com
Photoshop (2021)	N/A	Adobe.com
FlowJo	N/A	Flowjo.com
R (v 4.4.1)	N/A	RRID:SCR_001905
Pheatmap	N/A	github.com/raivokolde/pheatmap
STAR (v 2.7.10b)	Dobin et al.53	RRID:SCR_004463
featureCounts (v 2.0.8)	Liao et al.54	RRID:SCR_012919
DESeq2	Love et al.55	RRID:SCR_015687

Highlights.

• Intersectional lineage tracing identifies adipocyte-derived Pdgfr⁺ cells

• PDGF signaling promotes adipocyte transdifferentiation and optimal wound healing

• Wound environment induces adipocyte-to-fibroblast transdifferentiation

• Physiological tissue remodeling induces reversible adipocyte plasticity

Data and code availability

• RNA sequencing data have been deposited at the NCBI Gene Expression Omnibus (GEO: GSE327928) and are publicly available as of the date of publication.

• This paper does not report original code.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.