Fascia progenitors differentiate into specialized fibroblasts during wound healing
Fibroblasts inhabit different connective tissue compartments in the skin. Papillary and reticular fibroblasts reside in the upper skin regions while scar-producing fibroblasts are present in the deepest connective tissue layer, termed fascia27. To get a complete understanding of how any wound fibroblast type arise during skin wound healing, we performed scRNAseq of connective tissue-enriched cell fractions taken from full-thickness wounds on the back skin of mice including both upper and deep skin layers.
We have previously shown that back skin wound myofibroblasts originate from a single embryonic cell lineage that expresses the Engrailed-1 (En1) transcription factor during embryonic development (termed as EPFs)28-29. However, in adult skin, EPFs are found in all skin connective tissue layers: papillary, reticular, and fascia. Therefore, it remains unclear what makes fascia fibroblasts prone to generate scars.
We therefore combined our single cell transcriptomics with genetic lineage tracing by crossing transgenic mice expressing the Cre recombinase under the control of the promoter of En1 (En1Cre) with the fluorescent Rosa26mTmG (R26mTmG) reporter mouse line (extended data figure 1a). The resulting En1CreR26mTmG double transgenic mice have EPFs permanently expressing the green fluorescent protein gene (GFP). We sequenced 29,383 individual cells from En1CreR26mTmG uninjured skin and from skin wounds at 1-, 3-, 5-, 7-, 14-, and 27-days post injury (dpi) representing the inflammation (1, 3 dpi), proliferation (5, 7 dpi), and remodeling phases of wound healing (14, 27 dpi, extended data figure 1b). Our initial cell sequencing was unbiased with regards to EPFs, and therefore included both GFP+ (belonging to EPFs that we will return to later), as well as GFP- fibroblasts.
Our computational analysis spotted 7 distinct fibroblast clusters (extended data figure 1c-d). Clusters 0-2 were present in all samples, both injured and uninjured. Whereas clusters 3-6 were only found in injured tissue. To annotate previously known fibroblast populations on our 7 clusters, we curated a set of published markers for papillary, reticular, fascia, and myofibroblasts. We then annotated the identity of our clusters based on the gene expression scores from our curated sets. This analysis identified three main fibroblast populations at homeostatic conditions residing in papillary, reticular, and fascia connective tissues (clusters 0-2, Extended data figure 1e-f). Cluster 0 scored highest for the papillary profile, marked by the expression of the previously defined papillary markers Sparc14 and Decorin30 (Dcn), as well as the proteoglycan Lumican (Lum), which has been associated with keratinocyte-supportive functions31. Cluster 2 scored highest for the reticular profile and was marked by the expression of the reticular markers Stromal cell-derived factor-114 (Sdf1/Cxcl12), Cytoglobin14 (Cygb), and Matrix gla protein (Mgp)32. Cluster 1 scored highest for the fascia profile and expressed the known fascia markers Lymphocyte antigen-6a27,33 (Ly6a/Sca1), Placenta-specific gene-814 (Plac8), as well as Peptidase inhibitor-16 (Pi16) and Dermopontin (Dpt), which were recently found to also mark a fibroblast population present across many organs34. This is consistent with fascia connective tissue enclosing many organs35. We also identified CD201 (Procr), a cell surface receptor that is classically involved in controlling coagulation36, as a novel marker that distinguishes fascia fibroblasts.
Next, we examined the clusters appearing exclusively during injury (clusters 3-6). Samples taken during the early inflammation phase (1-3 dpi) were enriched in cluster 6 (Extended data figure 1c-d), which showed transcriptional patterns of proinflammatory fibroblasts. This wound-exclusive cluster specifically expressed the markers Podoplanin (Pdpn), and the chemokines Ccl2 and Cxcl1 (Extended data figure 1f). Notably, Pdpn and chemokines are also upregulated in inflammatory synovial fibroblasts during rheumatoid arthritis, a chronic inflammatory disease37. Gene-ontology (GO)-overrepresentation analysis revealed that, proinflammatory fibroblasts specialized in immunomodulatory functions with GO-terms such as “response to interferon-beta”, “neutrophil/lymphocyte/monocyte chemotaxis”, and “chemokine-mediated signaling pathway” (Extended data figure 1g). Clusters 3 and 5 appeared during the following proliferation phase of wound healing (5-7 dpi, Extended data figure 1d). Using our curated list of fibroblast markers, we determined that these two populations have a myofibroblast profile, differing only by the expression of the marker Alpha smooth muscle actin (αSMA/Acta2, Extended data figure 1e-f). Indeed, cluster 5, representing the Acta2high myofibroblasts also expressed other reported myofibroblast markers such as Periostin38 (Postn) and Leucine-rich repeat-containing protein-1539 (Lrrc15), as well as Runt-related transcription factor-2 (Runx2), a marker associated with keloid lesions40. Whereas cluster 3, representing the immature proto-myofibroblast with Acta2low, expressed Tenascin41 (Tnc), Signal transducer and activator of transcription-3 (Stat3), whose activity has been linked to dermal fibrosis42, as well as Proprotein convertase subtilisin/kexin type-543 (Pcsk5, Extended data figure 1e-f). Of note, proto-myofibroblasts (cluster 3), but not mature myofibroblasts (cluster 5), showed the GO-terms “cellular response to hypoxia” (Extended data figure 1h), indicating that hypoxia signaling is active in proto-myofibroblasts. Cluster 4 uniquely expressed the marker Secreted frizzled-related protein-2 (Sfrp2, henceforth named as SFRP2+ fibroblasts), which has been previously detected in human scRNAseq datasets44. Cluster 4 also uniquely expressed the keloid fibroblast marker Collagen triple helix repeat-containing protein-145 (Cthrc1) and the fibrosis-related marker Follistatin-related protein-146 (Fstl1, Extended data figure 1f). Clusters 3-5 specialized in matrix production and tissue contraction, presented GO-terms such as “collagen biosynthetic process”, “collagen fibril organization”, “extracellular matrix assembly”, “positive regulation of focal adhesion assembly”, and “actin filament bundle assembly”. On the other hand, only cluster 5 (Acta2high myofibroblasts) expressed terminal differentiation programs including “regulation of cell proliferation” and “endochondral ossification” (Extended data figure 1g).
Next, we sought to trace the most likely cellular origin of mature myofibroblasts by using the most robust algorithm for trajectory inference47, Partition-based graph abstraction (PAGA), which predicts the connections (e.g. differentiation steps) between defined clusters (Extended data figure 1h). Surprisingly, we found that myofibroblasts, proto-myofibroblasts and proinflammatory fibroblasts (clusters 3, 5, 6) represent a single cell lineage that emerges exclusively from homeostatic fascia fibroblasts (cluster 1). PAGA connectivity analysis revealed that fascia fibroblasts (cluster 1) give rise to proinflammatory fibroblasts (cluster 6), that in turn differentiate into proto-myofibroblasts (cluster 3) and finally into myofibroblasts (cluster 5) (Extended data figure 1h-i).
As EPFs are a predominant cell lineage in wounds, we wanted to place the previous EPF fibrogenic lineage in the context of our new differentiation trajectory during wound healing. We sorted the EPFs in the scRNAseq dataset based on their GPF expression and reanalyzed them in a higher clustering resolution (Extended data figure 2a). GFP+ cells (EPFs) were present across all homeostatic populations as well as in the proinflammatory, proto-myofibroblast, and myofibroblast clusters (Extended data figure 2b-d).
Our analysis also revealed an additional intermediate fascia EPF subcluster that is present in very early injured samples but is absent from uninjured skin (Extended data figure 2c-d). This intermediate injury cluster has a transcriptional signature in between fascia and proinflammatory fibroblasts. These injured fascia EPFs express the naïve fascia markers: CD201/Procr, Pi16 and Plac8, while simultaneously upregulating the proinflammatory fibroblast markers Pdpn, Ccl2 and Cxcl1 (Extended data figure 2e). This indicates that all three injury-related cell clusters (clusters 3,5,6) are most likely derived from fascia-resident EPFs.
CD201+ progenitors choreograph wound healing phases
We next focused on linking the stepwise differentiation of fascia progenitors with the wound healing phases in vivo. We first confirmed the temporal expression patterns of our fibroblast state-specific markers (Figure 1a) across mouse skin wounds at 3 and 7 dpi: two critical time points marking the transition from the inflammation to proliferation phases and from the proliferation to remodeling phases of wound healing, respectively. Consistent with our bioinformatics analysis, PDPN+ (proinflammatory) and phospho-activated STAT3+ (pSTAT3, proto-myofibroblast) cells peaked during the first transition at 3 dpi, and declined at later timepoints (Extended data figure 3a-b). On the other hand, RUNX2+ (myofibroblast) cells were prominent at the second transition into the remodeling phases of wound healing at 7 dpi (Extended data figure 3c). These data confirm a temporal linkage between fascia progenitor cell differentiation and wound healing phase transitions in vivo.
As spatial distribution of fibroblast populations is another extremely important facet of wound healing phases26, we next analyzed the spatial locations of proinflammatory, proto-myofibroblast, and myofibroblast from across three distinct wound compartments: upper wound, wound bed, and underlying fascia areas (Extended data figure 3d). At 3 dpi, PDPN+ proinflammatory fibroblasts were confined to the wound bed and upper wounds (Extended data figure 3e), while pSTAT3+ proto-myofibroblasts showed no preference among these areas (Extended data figure 3f). In contrast, at 7 dpi, RUNX2+ myofibroblasts clearly localized in upper wound compartments, with minimal presence in deep wound areas (Extended data figure 3g). This gradual spatial bottom-up differentiation of fibroblasts towards upper wound compartments, indicates a spatial distribution of guiding signals that promote the sequential differentiation of fascia progenitors (Figure 1c).
To test if fascia progenitors undergo spatiotemporal differentiation into all three distinct fibroblast states in vivo, we performed anatomic fate mapping of the fascia cells during wound healing coupled with immunolabeling for our fibroblast state-specific markers. First, we exclusively tagged fascia cells by delivering TAT-Cre recombinase into the fascia layer of R26mTmG fluorescent reporter mice (Extended data figure 4a). This allowed us to trace the fates of fascia fibroblasts during wound healing without labeling upper dermal (reticular, papillary) fibroblasts (Extended data figure 4b-c). At 3 dpi, close to half of all GFP+ fascia-derived cells acquired the PDPN+ proinflammatory state, and at 7 dpi half of all labeled cells acquired either pSTAT3+ proto-myofibroblast, or RUNX2+ or αSMA/ACTA2+ mature myofibroblast characteristics (Extended data figure 4d-e). These results confirm that fascia fibroblasts indeed give rise to all 3 fibroblast types during wound healing.
As our anatomical fate mapping above is not fibroblast specific, we then sought for a restricted marker to better study this progenitor population. We compared levels of cell enrichment for the classical fascia fibroblast marker Ly6a/Sca1, Pdgfra, and our newly discovered marker CD201. Ly6a/Sca1 mRNA was 1.24-fold enriched in fascia fibroblasts as compared to dermal fibroblasts (44.6 %, Ly6a/Sca1+ cells in naïve fascia). Whereas the pan-fibroblast marker Platelet-derived growth factor alpha (Pdgfra) showed no enrichment. In contrast, CD201 mRNA was 1.7-fold enriched in fascia fibroblasts as compared to dermal fibroblasts, indicating that, transcriptionally, CD201 expression enriches fascia fibroblasts (Extended data figure 5a).
We then generated a genetic reporter system by crossing a mouse line expressing the tamoxifen-inducible Cre-ER recombinase under the control of CD201 (CD201CreER) with the reporter line R26Ai14. In the CD201CreERR26Ai14 double transgenic offspring, tamoxifen administration induces the enduring expression of TdTomato in cells that express CD201, thereby specifically labeling fascia fibroblasts prior to injury (Figure 1d). Consistent with our previous anatomic fate mapping strategy, skin from CD201CreERR26Ai14 mice labeled with TdTomato specifically fibroblasts in the fascia compartment (78.74 % of total traced cells). TdTomato showed minor expression in epidermal and endothelial cells (Extended data figure 5b and Figure 1e), and TdTomato was absent from dermal fibroblasts. Together, our transcriptomics analysis and genetic reporter system indicate that CD201 is a highly selective marker for fascia progenitors.
To further test the origins of proinflammatory, proto-myofibroblasts and mature myofibroblasts from CD201+ fascia progenitors, full thickness wounds were then performed on back skin of CD201CreERR26Ai14 double transgenic mice. CD201+ fascia progenitors-derived fibroblasts (Tdtomato+) populated the wounds and contributed to 83% of all αSMA+ myofibroblasts (Figure 1f and Extended data figure 5c-d). Consistent with our computational analysis, CD201+ fascia progenitors underwent a sequential conversion into proinflammatory (PDPN+) and proto-myofibroblasts (pSTAT3+) between 3 and 7 dpi. During the progression from inflammation/proliferation to proliferation/remodeling phases of wound healing, CD201+ progenitors consistently downregulated proinflammatory and proto-myofibroblast markers: 71.5 down to 46.4 % for PDPN and 88.2 down to 38.4 % for pSTAT3 from 3 to 7 dpi. On the other hand, the expression of the mature myofibroblast marker, RUNX2, increased up to 73.5 % in the CD201 fibroblast lineage (Figure 1g and Extended data figure 5e). The temporal shift in marker gene expression in CD201+ fascia lineage cells indicates that the transition from naive progenitors to proinflammatory and fully mature myofibroblasts occurs in synchrony to the progression of the wound healing phases.
To further test if CD201+ progenitors transition to myofibroblasts, through a proinflammatory intermediate state, we generated a second transgenic mouse system where the expression of the tamoxifen inducible Cre-ER recombinase is expressed under the PDPN promoter (PdpnCreER). These mice were crossed with the R26mTmG reporter mouse line, enabling (in PdpnCreERR26mTmG double transgenic mice) the lineage tracing from proinflammatory states onwards (Figure 1h). In contrast to our CD201CreER fascia progenitor tracing system, uninjured skin of PdpnCreERR26mTmG mice had no GFP expression in any skin fibroblast population under homeostatic conditions (Figure 1i). At 7 dpi, GFP+ αSMA+ proinflammatory-derived myofibroblasts were detected abundantly in upper wounds (Figure 1i), confirming that proinflammatory fibroblasts transition into myofibroblasts during skin wound healing and that our CD201CreER and PDPNCreER transgenic systems faithfully trace their conversion steps.
Taken together, our in-silico trajectory analysis, anatomic fate mapping and CD201CreERR26Ai14 and PdpnCreERR26mTmG genetic lineage tracing methods, detail a spatiotemporal coordination of fibroblast differentiation taking place during wound healing, that arises from CD201+ progenitors differentiating into proinflammatory fibroblasts within wound beds that finally maturate into myofibroblasts in the upper wound region.
Sequential checkpoints control CD201+ progenitor differentiation
Having identified CD201+ progenitors of proinflammatory fibroblasts, proto-myofibroblasts, and myofibroblasts, we next sought to explore the molecular programs that regulate the spatiotemporal emergence of each fibroblast state. For this, we analyzed the coalitions of transcription factors (‘termed as “program”) and the gene targets that regulate each cellular state (termed as “regulon”; Extended data figure 6a). Our analysis revealed programs and regulons specific to each fibroblastic state.
1) The naïve CD201+ progenitor state was modulated by a program comprising Fos, Jun, and Early growth response protein-1 (Egr1) transcription factors, which oversaw a regulon consisting of 263 different genes. This regulon promotes cell survival and growth with terms including “insulin-like growth factor signaling” as well as “negative regulation of apoptotic process” (Figure 1b and Extended data figure 6b).
2) The PDPN+ proinflammatory state was separately controlled by Nuclear factor erythroid 2-related factor-2 (Nfe2l2) and Bach1 transcription factors that oversaw a regulon consisting of 116 different genes. This regulon directs processes related to oxygen level sensing such as “response to oxygen levels” and “response to hyperoxia” (Extended data figure 6c).
3) The pSTAT3+ proto-myofibroblast state was controlled by a third independent program comprising Stat3, Hypoxia-inducible factor-1-alpha (Hif1α), En1, Ets2, CCAAT/enhancer-binding protein beta (Cebpb), Receptor ROR-alpha (Rora), and Fosl1 transcription factors, which together oversaw a regulon comprising 251 different genes. This regulon directs processes that promoted migration (“positive regulation of migration”) oxygen sensing (“oxygen homeostasis”), as well as processes such as adhesion (“regulation of cell-matrix adhesion” and “positive regulation of homotypic cell-cell adhesion”), and collagen fibers production and processing (“peptidyl-proline hydroxylation”, Figure 1b and Extended data figure 6d).
4) Finally, the mature myofibroblast state was led by a fourth program comprising the Transcription factor 4 (Tcf4), Runx1, and Runx2 that oversaw a regulon comprising 21 different genes. This fourth regulon directs processes such as terminal differentiation and ossification (“regulation of cell differentiation” and “ossification”, Extended data figure 6e).
Taken together, our bioinformatic analysis and in-vivo genetic lineage tracing data reveal sequential genetic programs that control the stepwise differentiation of CD201+ progenitors through proinflammatory, proto-myofibroblast, and myofibroblast states that likely control wound healing phases.
To delve into this idea, we first tested the link between fascia fibroblast differentiation and progression of wound healing ex-vivo. For this, we excised fascia explants from mouse back skin and cultured them in a free-floating system (Figure 1j). Thin, translucent fascia explants underwent tissue contraction akin to that seen in mammalian wounds until forming an opaque sphere of scar tissue that has contracted down to 50% of its original surface area (Figure 1k).
Immunohistochemistry confirmed that fascia progenitors ex-vivo transitioned into proinflammatory fibroblasts within 1-3 days post culture (22.6 % and 80.9 % PDPN+ from total cells at 1- and 3-day post culture, respectively). Contrastingly, pSTAT3+ proto-myofibroblasts and RUNX2+ myofibroblasts increased in numbers from 6.8 % and 19.8 % respectively at day 1 to 64.3 % and 76.1 % of total cells at 6 days of culture. Remarkably, the shift from PDPN+ proinflammatory fibroblasts to pSTAT3+ proto-myofibroblasts and RUNX2+ myofibroblasts coincided with initiation of fascia tissue contraction at day 3 ex-vivo (Figure 1k-l), suggesting that the classic proliferation and remodeling phases of wound healing, characterized by tissue contraction and scar formation, are directly regulated by the differentiation of CD201+ progenitors.
The proinflammatory transition is gated by retinoic acid
Progression of wound healing phases is associated with dynamic biochemical and biomechanical signals such as TGFβ and Wnt48, cytokines such as Interleukins and Tumor necrosis factor49, cell-cell interactions50-51, and extracellular matrix mechano-transduction4, 52.
To explore the connections between these and more signaling pathways and CD201+ progenitor differentiation, we scored the expression of genes from several signaling pathways in our scRNAseq dataset throughout the fascia-to-myofibroblast trajectory. As expected, we observed expression of genes associated with classical myofibroblast-promoting signaling pathways: e.g. TGFβ, Wnt, and extracellular matrix mechano-transduction, peaking in the myofibroblast state. Whereas, chemokine production peaked much earlier in the middle of the proinflammatory state (Figure 2a).
Interestingly, we observed that retinoic acid (RA) pathway components peaked even earlier, in the very first transition from CD201+ progenitors into proinflammatory fibroblasts, suggesting that RA might act as a very early signal promoting differentiation of CD201+ progenitors into the proinflammatory state. Consistent with this notion, the transcript levels of the RA-synthetizing enzymes, aldehyde dehydrogenase family-1 member-a3 (Aldh1a3) and Retinol dehydrogenase 10 (Rdh10) were elevated in proinflammatory fibroblasts (Figure 2b). Furthermore, the signal repressor RA-degrading enzyme cytochrome P450-26-b1 (Cyp26b1) peaked in the subsequent proto-myofibroblast state (Figure 2b), suggesting that exit from the proinflammatory state into proto-myofibroblasts requires the downregulation of RA signaling. Chemokines Ccl11, Ccl2, Ccl7, Ccl8, and Cxcl1 expression, were co-elevated in proinflammatory fibroblasts (Figure 2b) indicating they are linked with RA signaling.
Immunolabeling of skin wounds from CD201CreERR26Ai14 double transgenic animals at 3 dpi confirmed the co-expression of ALDH1A3, CCL2 and CXCL1 proteins in CD201+-derived PDPN+ proinflammatory fibroblasts in the wound bed (Figure 2c-e), indicating that RA signaling is linked to chemokine expression. However, expression of the RA degrading enzyme, CYP26B1, was compartmentalized to the upper wound margins where differentiation into myofibroblasts occurs (Figure 2f). This spatial link, between RA synthesis/degradation and fibroblast differentiation, suggests a RA gradient coming from the wound bed and receding in the upper wound regions that acts on fascia fibroblasts and enable their myofibroblast differentiation.
To test the potential role of RA signaling in proinflammatory fibroblast differentiation from CD201+ progenitors, we added exogenous RA to fascia explants. RA administration impaired myofibroblast differentiation and concomitant tissue contraction in a dose-dependent manner: delaying tissue contraction at low concentration and completely blocking contraction at higher concentrations (Figure 2g). RA signaling acts canonically through the activation of three nuclear RA receptors, RARα, RARβ, and RARγ, which trigger transcriptional expression changes in the nucleus53. Treatments with specific activators for each of the three RA receptors phenocopied the contraction blockages seen with RA administration (Extended data figure 7a). This indicates that RA signaling in fascia progenitors is mediated by its canonical transcriptional activity and not through any individual RA receptor. Interestingly, increasing endogenous RA levels, by inhibiting CYP26B1, had the same contraction blocking effect as adding exogenous RA (Extended data figure 7a). These experiments indicate that RA is actively produced and degraded in fascia cells during tissue contraction, and that high-RA concentration promotes the entry into proinflammatory states, whereas RA-signaling downregulation grants the exit and progression into myofibroblasts.
As RA signaling activation preceded classic myofibroblast signals, such as TGFβ and YAP-TAZ mechano-transduction in our bioinformatic analysis (Figure 2a), we sought to establish the precise signaling pathway order-of-function for myofibroblasts. We treated fascia explants with exogenous RA in combination with activators for TGFβ and YAP-TAZ pathways. In the presence of exogenous RA, YAP-TAZ mechano-transduction signaling activation failed to induce tissue contraction. On the other hand, TGFβ pathway activation rescued normal tissue contraction in the presence of exogenous RA (Extended data figure 7b-c). This indicates that RA acts upstream of YAP-TAZ mechano-transduction pathway.
As RARγ was the highest expressing RA receptor on fascia fibroblasts in our scRNAseq dataset (Figure 2b), we specifically analyzed the involvement of RARγ activity on fascia fibroblast differentiation. Treatments of fascia explants with the specific RARγ activator increased the number of PDPN+ proinflammatory fibroblasts, and CCL2 production, while significantly decreasing RUNX2+ myofibroblast numbers after 6 days of culture (Figure 2h). This confirms that RA activation promotes and sustains the proinflammatory fibroblast state and, thus, prevents the subsequent differentiation into myofibroblasts.
To explore the ability of RA overactivation to limit the exit from the proinflammatory state and sustain wound healing in an inflammation phase, we treated skin wounds with the RARγ activator. Treated wounds had prolonged inflammation, revealed by a significant increase in leukocyte numbers at 3 dpi wounds (Figure 2i). Consistent with RA limiting myofibroblast differentiation, RA-treated wounds developed scars at 14 dpi that were significantly smaller than controls, with significantly reduced numbers of mature αSMA+ myofibroblasts (Figure 2j). These data confirm that RA activity controls the entry of the proinflammatory state and its overactivation limits the subsequent conversion into myofibroblasts, reducing scar formation as a result.
The (proto)myofibroblast transition is gated by Hif1α
Our bioinformatic analysis reveals that the proinflammatory fibroblast precedes the proto- and myofibroblast in the differentiation trajectory of CD201+ progenitors. To further prove this previously unknown connection between these functionally distinct wound fibroblast populations, we performed genetic ablation of proinflammatory fibroblasts in vivo by crossing our PdpnCreER mice with the R26DTA line, in which Cre-mediated recombination activates expression of diphtheria toxin, causing cell death exclusively in proinflammatory fibroblasts. Wounds from two separate control animal groups (PdpnCreERR26WT and PdpnCreERR26mTmG) showed normal wound contraction and closure rates that fully closed between 7 and 14 dpi. Ablation of proinflammatory fibroblasts in PdpnCreERR26DTA animals, caused a significant delay in wound closure in which wounds failed to fully close by 14 dpi (Figure 3a). 7 dpi wounds in PdpnCreERR26DTA animals showed a minimal wound bed and limited re-epithelialization compared to control animals (Figure 3b) indicating tissue repair is significantly impaired. Importantly, ablation of proinflammatory fibroblasts caused a significant decrease in the amount of αSMA+ myofibroblast at 7 dpi (Figure 3c), further proving that proinflammatory fibroblast give rise to myofibroblasts in vivo.
Next, our signaling pathway analysis revealed that hypoxia signaling is induced during the subsequent transition from proinflammatory into proto-myofibroblasts (Figure 2a). Moreover, GO term analysis indicated that biological processes linked to oxygen sensing are prevalent in the proinflammatory and proto-myofibroblast states (Extended data figure 1g). Notably, the transcription factor Hif1α was itself found to control the regulon governing proto-myofibroblast differentiation (Extended data figure 6d). This suggested that hypoxia signaling through Hif1α, regulates the exit from the proinflammatory state into proto-myofibroblasts.
Consistent with this, chemical inhibition of Hif1α activity in fascia explant cultures increased PDPN+ proinflammatory fibroblast numbers while blocking αSMA+ myofibroblast differentiation (Extended data figure 7d) and tissue contraction in a concentration-dependent manner (Extended data figure 7e).
We then analyzed Hif1α signaling pathway order-of-function by combining Hif1α inhibitor treatments with TGFβ1 or TAZ activators. In both cases there was persistently impaired tissue contraction. This indicates that, Hif1α activity acts upstream of both YAP-TAZ mechano-transduction and TGFβ pathways in the conversion to proto-myofibroblasts (Extended data figure 7d-e).
To verify the gatekeeping role of Hif1α in the conversion of proinflammatory fibroblasts into proto-myofibroblast states, we chemically inhibited Hif1α activity in full-thickness wounds in animals. This resulted in a significant delay in wound closure in vivo (Figure 3d), similar to the contraction blocking effects seen in fascia explants ex vivo. Wounds treated with the Hif1α inhibitor also mimicked wounds from PdpnCreERR26DTA animals, had delayed wound healing, a marginal wound bed, and the persistence of an immune infiltrate at 7 dpi. In comparison, control wounds were completely re-epithelialized, with a significant wound bed and no persistent inflammation (Figure 3e). Moreover, wounds treated with Hif1α inhibitor at 1-3 dpi showed similar leukocyte infiltration numbers to control wounds (Figure 3f), confirming that Hif1α controls proinflammatory exit, alone, without affecting early progenitor differentiation.
To unambiguously determine if Hif1α inhibition prevented the transition from proinflammatory fibroblasts into myofibroblast states, we treated full thickness wounds in PdpnCreERR26mTmG double transgenic mice. At 7 dpi, inhibition of Hif1α activity in wounds significantly decreased proinflammatory-derived (GFP+) proto-myofibroblast (pSTAT3+) and myofibroblast numbers (GFP+RUNX2+, Figure 3g). This confirms that Hif1α activity indeed controls the transition from proinflammatory into proto- and myofibroblast states needed for wound closure, tissue contraction, and scar formation.
We have previously shown that fascia fibroblasts transfer pre-made extracellular matrix into wounds, and that this matrix transfer contributes to scar formation27. To understand how fascia progenitor differentiation impacts matrix transfer in-vivo, we labeled fascia extracellular matrix, with N-Hydroxy-Succinimidyl (NHS)-Pacific Blue, prior to injury (Extended data figure 8a). 3 dpi wounds showed transferred pre-made matrix fibers extending and covering the wound tissue. Interestingly, PDPN+ proinflammatory fibroblasts were associated with thinner more loose extracellular matrix fibers in the wound bed, while pSTAT3+ proto-myofibroblasts were associated with thicker more woven fiber bundles in upper wound regions (Extended data figure 8a). This indicates that fascia fibroblast differentiation, and extracellular matrix scar remodeling and maturation are interlinked. To test this idea, we labelled fascia extracellular matrix in animals, and generated full thickness wounds while blocking fascia progenitor differentiation with Hif1α inhibitor. Control 3 dpi wounds had matrix fibers bundles with a condensed fiber organization at the lateral borders and in the upper wound region where myofibroblast differentiation mainly takes place. Whereas, treated animals showed a frail and immature fiber organization along the entire wound (Figure 3h). Next, we assessed the extracellular matrix lattice organization in matrix labeled and Hif1α treated wounds, using fractal analysis29. This revealed a significant reduction in matrix organization (fractal dimension) and an increase in porosity (lacunarity) in the Hif1α-treated group compared to control (Figure 3i), indicating a relaxed immature fiber organization. Our data shows that Hif1α mediates the differentiation of proinflammatory fibroblast into proto- and myofibroblasts, and that this conversion step is needed for extracellular matrix maturation, tissue contraction and wound closure (Figure 3j).
Taken together, our comprehensive scRNAseq, in-vivo functional studies and genetic lineage tracing and ablation experiments identify a spatiotemporal pattern of progenitor>proinflammatory>proto->myofibroblast differentiation from CD201+ progenitors, that controls the timely progression of wound healing phases through checkpoint control by RA and Hif1α.
CD201+ fibroblast differentiation is conserved in mice and men
To determine if CD201+ lineage differentiation is conserved across mouse and human skin diseases, we analyzed publicly available scRNAseq datasets of human keloids18 and psoriasis54 (Extended data figure 9a-b), both of which are skin fibrotic pathologies characterized by the accumulation of myofibroblasts. Keloid and psoriatic fibroblast clusters were curated based on transcriptional signatures taken from our mouse wound healing dataset (Extended data figure 9c-f). Cellular census of diseased human skin showed clear naïve fascia, proinflammatory, and mature myofibroblast clusters, in both keloid and psoriasis datasets. Interestingly, the proto-myofibroblast signature was diffusely spread between the proinflammatory and myofibroblast clusters in the human datasets (Extended data figure 9e-f and Figure 4a), suggesting that in human skin diseases, fascia progenitors differentiate into myofibroblasts directly through the proinflammatory state (Extended data figure 9g-h and Figure 4b).
Just as in our mouse dataset, human fibroblast progenitors were characterized by the expression of the novel fascia marker CD201 in both keloid and psoriatic human datasets (Figure 4c). Furthermore, partition-based graph abstraction connectivity analysis showed differentiation trajectories that mirrored our mouse CD201+ lineage, from the naïve fascia progenitor into the proinflammatory states and ending in the myofibroblast state (Figure 4d). These analyses confirm that the connection between the proinflammatory and myofibroblast states is part of a common differentiation process undertaken by fascia progenitors.
Our data in mouse demonstrated that RA acts to promote entry of progenitors into the proinflammatory state while Hif1α promotes exit from the proinflammatory state into myofibroblasts, and that these two checkpoint regulators chronologically and functionally precede classical myofibroblast inducers such as TGFβ. Analysis of the chronology of these signaling pathways in our human datasets showed they mirrored the signaling order in mouse wound healing. Both in mice and human datasets, RA activity score showed a peak at the beginning of the proinflammatory state followed by hypoxia signal activity that peaked at the transition from the proinflammatory to the myofibroblast state, mirroring our findings from mouse skin wound healing. Lastly, TGFβ peaked at the final mature myofibroblast state in all datasets confirming it is preceded by RA and Hif1α signals in its order-of-function (Figure 4e).
Our analysis reveals, in mouse and human, a cellular lineage derived from CD201+ fascia progenitors that undergo a graduated differentiation into proinflammatory and myo-fibroblasts. Moreover, that CD201-cell lineage differentiation is regulated by RA and Hif1α; two early signals that time the transition into phenotypically distinct wound fibroblast populations to resolve the sequential phases of wound healing.