Allele-specific expression reveals genetic drivers of tissue regeneration in mice

Summary:

In adult mammals, skin wounds typically heal by scarring rather than through regeneration. In contrast, “super-healer” MRL mice have the unusual ability to regenerate ear punch wounds, yet the molecular basis for this regeneration remains elusive. Here, in hybrid crosses between MRL and non-regenerating mice, we used allele-specific gene expression to identify cis-regulatory variation associated with ear regeneration. Analyzing three major cell populations (immune, fibroblast, and endothelial), we found that genes with cis-regulatory differences specifically in fibroblasts were associated with wound healing pathways and also co-localized with quantitative trait loci for ear wound healing. Ectopic treatment with one of these proteins, CFH, accelerated wound repair and induced regeneration in typically fibrotic wounds. Through single-cell RNA-seq, we observed that CFH treatment dramatically reduced immune cell recruitment to wounds, suggesting a potential mechanism for CFH’s effect. Overall, our results provide insight into the molecular drivers of regeneration with potential clinical implications.

eTOC Blurb

In adult mammals, skin wounds typically heal by scarring. Mack, Talbott, Griffin and colleagues identify cis-regulatory changes associated with ear wound regeneration in a murine model. They implicate the gene CFH in accelerated wound repair and regeneration in typically fibrotic wounds, providing insight into the molecular drivers of regeneration.

Graphical Abstract:

Introduction

Fibrosis, or the replacement of functional tissue with non-functional connective tissue, can result from tissue damage to any organ in the human body. In the skin, fibrosis occurs as scarring and has major consequences for skin form and function. Scars lack the structures (e.g., hair, glands) of normal skin, compromising skin’s normal barrier system and its ability to thermoregulate, and are weaker and less flexible than uninjured skin. Healing via scarring has major consequences for human health: scarring can cause disfigurement, functional loss, and reduced quality of life.^1,2 Despite the substantial clinical burden scars impose, there are no current therapies that induce scar-free healing in humans.

In contrast to humans, some other species possess the ability to regenerate skin after injury without scar formation. The Murphy Roths Large (MRL) mouse represents a valuable biomedical model for studying mammalian wound regeneration. While injuries to mammalian skin and other organs typically heal via formation of fibrotic scar tissue, MRL and its progenitor strain, the Large (LG/J) mouse, have been reported to regenerate multiple tissue types without fibrosis.^3,4 The most well-studied example of MRL regeneration is that of ear punch wounds: while through-and-through ear wounds remain open and fail to regenerate the excised tissue in most mouse strains, MRL mice fully heal these wounds via initial formation of a blastema-like structure and subsequent regeneration of key tissue types including cartilage and hair-bearing skin.³ However, the molecular mechanisms underlying enhanced wound healing in the MRL ear remain poorly understood. To date, nine quantitative trait locus (QTL) mapping studies have been performed to identify associations between genomic regions and the ear closure phenotype^5–11, but these studies so far have failed to identify specific genes or pathways driving regenerative healing, with ear closure-associated QTL spanning dozens or hundreds of individual genes.

Identification of tissue- and behavior-specific cis-regulatory divergence, through analyses of allele-specific gene expression in hybrids, has previously revealed genes and pathways underlying complex traits.^12–14 MRL regeneration is wound site-specific – while ear wounds regenerate, dorsal wounds form fibrotic scars similar to other mouse strains¹⁵– providing the opportunity to apply a similar approach to elucidate genes driving MRL ear regeneration. Here, we capitalize on mouse strain- and site-specific differences in healing to identify divergence in cis-regulation of gene expression associated with MRL ear regeneration through allele-specific expression (ASE) analysis. Using this approach, we implicate one gene, Cfh, as a candidate for genetic differences in wound healing between strains. Through single-cell RNA-seq (scRNA-seq) we demonstrated that CFH treatment reduces the recruitment of several immune cell types to wound sites, suggesting a reduction in inflammation consistent with previous work.¹⁶ Interestingly, CFH treatment down-regulates the inflammation-associated chemokine CXCL2 across all cell types present at wounds, suggesting a potential mechanistic link. Lastly, chemical inhibition of the CXCL2 receptor CXCR2 in dorsal wounds mimicked the regenerative phenotype of MRL ear wounds. Collectively, our results highlight the power of this approach in dissecting complex phenotypes in mammals and implicate the complement pathway as a possible therapeutic target for improving wound healing and reducing scarring.

Results

Tissue- and strain- specific differences in wound healing

We first sought to robustly establish differences in healing phenotypes between two mouse strains: MRL, which regenerate ear wounds but heal dorsal wounds via scarring^3,15,17; and CAST/EiJ (CAST), which do not possess any known strain-specific regenerative ability.^7,11 Ear wounds were generated using a 2 mm punch tool to create a through-and-through wound in the center of the pinnae. Full-thickness dorsal excisional wounds were created via a previously published protocol¹⁸; in this model, silicone splints are applied around wounds to prevent the rapid contraction that typically occurs in mice and instead yield healing via granulation and re-epithelialization with human-like kinetics. Consistent with previous work^3,11, MRL ear wounds had largely closed by 3–4 weeks after wounding, with regeneration of normal-appearing skin grossly and cartilage histologically, while CAST ear wounds failed to close to an appreciable extent and instead formed scar tissue over the exposed wound edge (Fig. 1A–B). In contrast, dorsal wounds healed at a comparable rate in the two strains, with re-epithelialization complete by postoperative day (POD) 14 (Fig. 1C–D). Both MRL and CAST dorsal wounds healed by forming fibrotic scars, which grossly and histologically appeared as “bare areas” devoid of dermal appendages (e.g., hair follicles) and with dense connective tissue (Fig. 1C). Analysis of wound ECM ultrastructure using a previously published image analysis pipeline¹⁹ confirmed that dorsal wounds in both CAST and MRL healed with ECM architecture that was quantitatively distinct from that of unwounded skin (consistent with fibrotic scar ECM); while CAST ear wounds also had a distinct ECM pattern, MRL ear wounds had ECM features that had a higher overlap with unwounded ear tissue, suggesting regeneration at the tissue ultrastructural level (Fig. S1).

Figure 1: MRL ear wounds uniquely heal in an accelerated and regenerative fashion.

A. Gross photographs (first two columns) at postoperative day (POD) 2 and 90 of CAST (“normal healer”) and MRL (“super healer”) mice. Hematoxylin and eosin (H&E) histology (third column) of ear wounds at POD 90. Green dotted lines indicate border of wound site; blue overlay indicates cartilage; blue arrow highlights regenerating cartilage in MRL ear wounds. See also Fig. S1. B. Wound curves for CAST and MRL ear wounds showing closure over time. C. Gross photographs (first two columns) of splinted excisional dorsal wounds in CAST and MRL mice; white dotted lines indicate fibrotic scar “bare area.” H&E staining (third column) of POD 14 wounds and unwounded skin. D. Wound curves for CAST and MRL dorsal wounds reflecting rate of re-epithelialization over time. Points represent mean values and error bars represent standard error of the mean. B, D. *P < 0.05, **P < 0.01 (Student’s t-test). n = 6 wounds in three biological replicates.

Extensive cis-regulatory divergence during wound healing between regenerative and non-regenerative mouse strains

Having confirmed that enhanced/regenerative healing is specific to both the strain (MRL) and anatomic site (ear), we sought to leverage this unique phenotypic pattern to identify genes responsible for regeneration versus fibrosis. As wound repair involves a series of transcriptional cascades triggered by injury²⁰, we hypothesized that MRL ear closure may be driven by wound site-specific cis-regulatory activity. To identify cis-regulatory variation associated with wound healing, we crossed MRL with CAST mice to generate CAST × MRL F1 hybrids. In these first-generation hybrid offspring, alleles from both parents are present in the same cellular environment (i.e., are subject to the same trans-regulatory influences); so, differences in expression between the two alleles can only be the result of cis-regulatory differences.^21–23 Thus, assaying allele-specific gene expression in F1 hybrid wounds allows for identification of injury-relevant cis-regulatory differences between these two mouse strains. Further, comparing allele-specific expression between wounds in the ear – which exhibits regenerative healing specifically in MRL mice – and the dorsum – which heals with a scar in both MRL and CAST – may allow us to pinpoint causal cis-regulatory differences driving the unique MRL ear regeneration phenotype.

In order to assess allele-specific gene expression across wound contexts, we performed bulk RNA-seq of key cell populations associated with wound healing. Adult F1 female mice were subjected to dorsal splinted excisional and ear punch wounding. On POD 7, all wounds were harvested. Ear and dorsal wound tissue were separately digested and subjected to fluorescence-activated cell sorting (FACS) to isolate three cell populations: immune cells (CD45⁺); endothelial cells (CD31⁺); and fibroblasts (Lin⁻, per published sorting strategy²⁴; see Methods for details). Due to cell number limitations, wounds from three individual mice were pooled per biological replicate. Cell samples were then subjected to bulk RNA-sequencing (Fig. 2A). At least three biological replicates were sequenced and analyzed for each cell type.

Figure 2: RNA-seq of key wound cell types from CAST x MRL hybrid mice cluster by wound type and allele.

A. Sampling scheme for RNA-seq libraries. MRL and CAST were crossed to produce F1 hybrids for allele-specific expression analysis. Each adult F1 mouse underwent both dorsal excisional and ear punch wounding. On POD 7, wound tissue was harvested, cell populations were isolated via fluorescence-activated cell sorting (FACS), and RNA was extracted for bulk RNA-seq. B. Heatmap of the most variable genes (1,000) following regularized log₂ transformation of allele-specific read counts. Hierarchical clustering groups samples by cell population (immune [n=8 libraries], endothelial [n=6], fibroblast [n=6]), allele (MRL [‘M’] vs. CAST [‘C’] and wound site (ear [‘E’] vs. dorsal [‘D’]). C. Principal component analysis of allele-specific read counts. Allele-specific samples separated into distinct clusters by wound site (ear vs. dorsal) and allele (MRL vs. CAST) for each cell population (see also Fig. S2A).

Across all tissue samples, we obtained a total of ~3.6 billion reads (Table S1). To enable the allele-specific assignment of RNA-seq reads, we performed whole-genome re-sequencing of MRL (~25X depth of coverage). Genetic variants differing between MRL and CAST strains were used to preferentially assign RNA-seq reads to either the MRL or the CAST allele (Table S1). After filtering for genes with low coverage in either genotype or wound context, we were able to analyze >10,000 genes in each cell population (Table S2). Hierarchical clustering and principal component (PC) analysis of allele-specific expression data grouped samples strongly by cell type (fibroblasts, immune, endothelial) (Fig 2B, Fig. S2A). PC analysis of individual cell populations clearly separated samples by wound site (dorsal vs. ear: PC1 for fibroblasts, 71% of variance explained; PC2 for immune and endothelial; 23% and 25% of variance explained, respectively) and allele (CAST vs. MRL: PC2 for fibroblasts, 14% of variance explained, PC1 for immune and endothelial, 43% and 42% of variance explained respectively; Fig. 2C).

We next sought to identify genes with expression patterns reflecting MRL ear-specific cis-regulatory divergence, which could reflect functional involvement of these genes in driving regeneration rather than fibrosis for wounds in this tissue (Fig. 3A). Across all cell types, a greater number of genes exhibited significant allele-specific expression (ASE; FDR < 0.05 for MRL vs. CAST allelic expression with DESeq2) in ear wounds (5,121 genes; 32.7%) compared to dorsal wounds (2,655 genes; 17%), consistent with phenotypic divergence restricted to the ear (Fig. 3B, Table S3). This was most apparent in immune cells, where over four times as many genes had evidence of ASE in ear compared to dorsal wounds. Differences in expression between CAST and MRL alleles (i.e., |log₂ fold change|) were also larger on average in ear wounds in each cell population (Fig. S2B; Wilcoxon rank sum test, all ear vs. dorsal pairwise comparisons p < 2.2 × 10⁻¹⁶).

Figure 3: Analysis of differential allele-specific expression (diffASE) reveals cis-regulatory divergence unique to MRL ear wounds.

A. Schematic example of diffASE between MRL and CAST in ear and dorsal wounds. Blue and green solid boxes represent gene regulatory regions affecting transcription of the MRL or CAST allele, respectively, of a given gene; transcription levels from each allele are represented by blue and green wavy lines. In the context of a dorsal wound (where MRL and CAST phenotypes are similar), expression is the same from the MRL vs. CAST allele. In contrast, in ear wounds, the presence of a context-specific (i.e., wound-related) transcription factor (TF; grey circle) reveals ASE through differences in the sequence of the MRL vs. CAST regulatory elements (which respond differentially to the TF). Overall, this results in a pattern of diffASE, where allele-specific expression is unique to ear wounds (exemplified in bottom panel bar graphs). B. Venn diagrams showing number of genes with ASE in ear wounds (blue region), dorsal wounds (yellow region), or both (overlapping region) in each analyzed wound cell type (see also Fig. S2C for overlap between cell types). C. Scatterplots for each cell type comparing distribution of allelic ratios between dorsal and ear wounds. Colored points represent genes with diffASE (gold points are genes with a larger difference between CAST and MRL alleles in the ear; blue points are genes with a larger difference between CAST and MRL alleles in the dorsum). See also Fig. S2B. D. Gene set enrichment analysis for genes with evidence of diffASE in fibroblasts, which are highly enriched for gene ontology (GO) categories (left) and mutant phenotypes (right) related to wound healing and injury responses. Such enrichment patterns were unique to fibroblasts (the end cellular mediators of scarring/fibrosis) and not seen in endothelial or immune cells. E. Specific genes associated with mutant phenotypes or GO terms related to responses to injury and wound healing with diffASE in fibroblasts (full gene list in Tables S4, S5). Yellow circles represent fold changes between alleles in the dorsum; blue circles represent fold changes in the ear.

Considering both ear and dorsal wounds, approximately 23% of genes demonstrated ASE in more than one cell population (Fig. S2C). Within each cell population, while a substantial proportion of genes showed ASE in both wound contexts, many exhibited ASE unique to either ear or dorsal wounds (Fig. 3B), suggesting the existence of both general and tissue-specific regulatory divergence between CAST and MRL during wound repair. For genes with ASE in both dorsal and ear wounds, the vast majority maintained the same directionality of allelic expression across wound sites (i.e., same allele up-/down-regulated in both ear and dorsal wounds; Table S4). Further, we found that allelic ratios were correlated between wound sites (i.e., log₂(CAST ear/MRL ear) vs. log2(CAST dorsal/MRL dorsal); Pearson’s correlation, all comparisons p < 2.2 × 10⁻¹⁶; Fig. 3C). Taken together, our findings of ASE were consistent with greater context-dependent regulatory divergence in ear wounds relative to dorsal wounds.

Wound context- and strain- specific cis-regulatory activity identifies genes involved in wound healing and injury response pathways

As MRL mice demonstrate a regenerative phenotype unique to the ear wound context and not seen in dorsal wounds¹⁵(Fig. 1), we reasoned that the subset of genes with differential allele-specific expression (“diffASE”) between ear and dorsal wounds could include genes driving the regenerative healing phenotype specifically in MRL ear wounds. Across different wound settings, context-specific ASE may reflect wound site-specific activity of genes controlled by cis-regulatory elements with sequence differences between MRL and non-regenerating (e.g., CAST) mice (Fig. 3A). Comparing genes’ ASE measurements in ear and dorsal wounds, we identified 432 genes in immune cells, 91 in endothelial cells, and 235 in fibroblasts with diffASE between wound healing contexts (DESeq2 Wald test, [CAST/MRL ear counts] vs. [CAST/MRL dorsal counts], FDR < 0.05; see Methods; Fig. 3C, Table S3). The majority of genes with diffASE were unique to a single cell population (732/745 genes total).

Examining genes with diffASE in each cell population, we found that those in fibroblasts were uniquely enriched for known mutant phenotypes and gene ontology (GO) terms associated with injury and wound repair (Fig. 3D). In contrast, wound healing-related terms were not enriched for either immune or endothelial cell diffASE genes, suggesting that cis-regulatory divergence in fibroblasts may play a particularly important role in driving divergent wound healing phenotypes in the MRL ear versus dorsum. In fibroblasts, genes with diffASE were most highly enriched for the mouse mutant phenotype “abnormal response to injury” (MP:0005164, FDR-adjusted p-value = 1.85 × 10⁻⁵) and were also enriched for the phenotypes “abnormal wound healing” (MP:0005164, FDR = 0.00086) and “abnormal blood vessel physiology” (MP:0000249, FDR = 0.00023). Additionally, the GO terms “response to wounding” (GO:0009611, FDR = 1.37 × 10⁻⁴) and “wound healing” (GO:0042060, FDR = 2.37 × 10⁻³) showed greater than four-fold enrichment compared to a background set. Fibroblast diffASE genes were also enriched for GO and Reactome Pathway terms related to processes involved in scarring and regeneration. These included cell adhesion (GO:0007155, FDR = 1.34 × 10⁻⁶) and integrin cell surface interactions (R-MMU-216083, FDR = 1.95 × 10⁻²), which are implicated in activated mechanotransduction and pro-fibrotic changes in fibroblasts^19,25; and extracellular matrix organization (GO:0030198, FDR = 3.50 × 10⁻³; and R-MMU-1474244, FDR = 9.52 × 10⁻³), a critical determinant of scarring versus regenerative wound properties, among others (Fig. 3D).

Further, we identified several genes associated with wound repair phenotypes or known pathways with large differences in allelic ratio between ear and dorsal wound fibroblasts (Fig. 3E, Table S5, Table S6). Some of these genes with MRL-specific upregulation in ear wounds had known functions consistent with promoting wound healing and/or decreasing scarring. For instance, Slpi (Secretory leukocyte protease inhibitor; ear wounds: log₂(CAST/MRL) = −2.98, q = 5.74 × 10⁻⁶; dorsal wounds: log₂(CAST/MRL) = −0.036, q = 0.58) has important functions in wound healing, in part via regulating transforming growth factor-beta (TGF-β) activity, and Slpi-null mice exhibit delayed wound repair and increased inflammation.^26,27 Spp1 (Secreted phosphoprotein 1, which encodes the protein osteopontin; ear: log₂(CAST/MRL) = −1.91 FDR = 1.083 × 10⁻³⁶; dorsal: log₂(CAST/MRL) = −0.18, FDR = 0.57) has been implicated in resolution of inflammation as well as matrix remodeling following skin injury^28,29, the latter being especially critical in determining scarring versus regenerative healing outcomes.^19,30 Additionally, osteopontin knockout mice have impaired wound closure.³¹ Thbs4 (Thrombospondin 4, an ECM protein; ear: log₂(CAST/MRL) = −1.24, FDR = 6.55 × 10⁻³⁸; dorsal: log₂(CAST/MRL) = −0.23, FDR = 0.048) has previously been shown to promote wound healing by stimulating fibroblast migration and keratinocyte proliferation³² and is reported to promote angiogenesis and reduce fibrosis³³, with mouse Thbs4 knockout associated with damaging cardiac inflammation and fibrosis.³⁴

We also identified several genes with reported functions that promote fibrosis and/or impair injury repair, which exhibited upregulation from the CAST allele in ear wounds. For instance, Jaml (Junctional adhesion molecule-like; ear: log₂(CAST/MRL) = 1.78, q = 4.24 × 10⁻⁷; dorsal: log₂(CAST/MRL) = −0.04, q = 0.87) has been associated with inflammation and impaired injury repair in the intestine.^35,36 Adora2b (Adenosine A2b receptor; ear: log₂(CAST/MRL) = 0.82, FDR = 0.00057; dorsal: log₂(CAST/MRL) = −0.12, FDR = 0.71) inhibition has been associated with reduced dermal fibrosis³⁷, consistent with a pro-scarring role for this gene. Collectively, many genes exhibiting diffASE preferentially in ear wounds had known functions consistent with the phenotypes observed in CAST versus MRL ear wounds (i.e., pro-fibrotic genes enriched from the CAST allele; pro-regenerative genes and genes promoting efficient wound repair enriched from the MRL allele).

Overlap with healing quantitative trait loci identifies candidate genes for regeneration

Next, we sought to integrate our results with prior functional studies of MRL ear regeneration. Specifically, having identified genes with wound site-specific cis-regulatory differences, we next asked whether those genes were located within previously mapped genomic intervals associated with enhanced ear punch closure. We capitalized on a recent QTL fine-mapping study for ear wound closure in LG/J (LG), the MRL progenitor line (LG × SM, F32 generation.³⁸ LG shares ~75% of its genome with MRL and exhibits similar regenerative healing of ear punch wounds^38,39. Consequently, overlap between these studies will be restricted to causal loci for regenerative healing that are shared between lines. To test whether our diffASE gene sets were enriched in genomic regions driving ear wound closure, we compared the LOD scores of genetic markers closest to genes with diffASE (Fig. 4A) against those of randomly permuted gene sets of the same size. While the presence of a significant LOD score proximal to a single gene does not necessarily implicate that gene, a shift towards a higher average LOD score for a group of genes suggests that this set of genes is collectively more likely to be associated with differences in wound phenotypes (i.e., regenerative versus non-regenerative). Our analysis revealed that genes with diffASE in fibroblasts had significantly higher average LOD scores compared to random sets (20,000 permutations, diffASE genes FDR < 0.05, p = 0.0094; diffASE FDR < 0.1, p = 0.026; Fig. 4B). In contrast, genes with diffASE in immune or endothelial cells did not exhibit higher LOD scores on average (p > 0.05 for each comparison; 20,000 permutations). Further, genes with ASE in both ear and dorsal wound fibroblasts were not associated with higher LOD scores (p > 0.05), suggesting that diffASE was uniquely useful in pinpointing causal wound healing genes.

Figure 4: Integration of diffASE with QTL fine-mapping study identifies Cfh as a candidate gene for driving the MRL regenerative healing phenotype.

A. LOD scores vs. chromosome position for ear hole closure from Cheverud et al. 2014. Red circles indicate the positions of genetic markers closest to genes identified as having diffASE in fibroblasts. B. Distribution of mean LOD scores of permuted gene sets (20,000 permutations). Red line indicates the mean LOD score of genetic markers closest to the fibroblast diffASE gene set. C. Cfh, which is associated with the gene ontology term for wound healing (GO:0042060) and falls within a fine-mapped region for ear closure, shows ear wound-specific ASE specific to fibroblasts in CAST × MRL hybrids. In fibroblasts, we see significant upregulation of the MRL allele relative to the CAST allele in ear wounds, in contrast to dorsal wounds where the expression of these alleles are similar (n=10 libraries). Point are log₂ fold changes from individual libraries. *diffASE q < 0.05 (DESeq2 Wald test).

Next, to identify specific candidates for driving regenerative wound healing, we searched for genes with diffASE located within support intervals of significant wound closure QTL. Across cell types, we identified a total of 27 genes with diffASE overlapping these QTL intervals (diffASE FDR < 0.05; fibroblasts, 9 genes; endothelial cells, 6 genes; immune cells, 12 genes; diffASE FDR < 0.1, 40 genes). Genes with diffASE in fine-mapped intervals were enriched for the GO term “wound healing” (GO:0042060; Fisher’s exact test, p = 0.0032, 9.45-fold enrichment). Several genes within these intervals could be promising candidates based on their known functions or mutant phenotypes (see Supplemental Material⁴⁰). Of these genes, Cfh (Complement factor H) had the greatest difference in allelic ratios between ear and dorsal wounds, with expression from the MRL allele over four times that of the CAST allele on average in ear wounds, but no significant difference between alleles in dorsal wounds (Fig. 4C). The complement cascade is a part of the innate immune system that is involved in clearing microbes, immune complexes, and damaged self cells and is activated in response to tissue injury.^41,42 In addition to Cfh, ear-specific ASE in wound fibroblasts was also observed in two other genes encoding complement components, C3 and C1qb, both of which have been directly implicated in wound repair.^43–45 While the complement system plays a vital role in injury repair – for example, C1qb is important for angiogenesis⁴³ – inappropriate or prolonged complement activation is also known to perpetuate damaging inflammation and cause cell death.⁴¹ However, prior studies conflict on whether complement activation or inhibition may promote wound healing^45,46, and the effects of complement modulation on scarring have not been investigated.

Ectopic application of CFH reduces scarring and drives partial regeneration after wounding

Complement factor H is a central regulatory protein in this pathway that inhibits complement activation⁴² and which has previously been shown to prevent inflammation and fibrosis in the mouse kidney.¹⁶ Given these known functions, strong ear wound-specific ASE of Cfh, and its presence within a QTL interval for ear closure, we hypothesized that Cfh could be a driver of wound regeneration. We first sought to verify that our gene expression findings corresponded to differences at the protein level. First, we cultured fibroblasts from the ear and dorsum of both CAST and MRL mice, then performed immunostaining for CFH (Fig. 5A). CFH protein expression was absent in both ear and dorsal CAST fibroblasts, and was significantly greater in MRL ear than MRL dorsal fibroblasts (Fig. 5B), consistent with our finding of MRL ear-specific upregulation from RNA-seq. We next performed CFH staining on sections from POD 14 ear and dorsal wounds from CAST and MRL (Fig. 5C), which revealed that CFH protein expression was markedly upregulated in MRL ear wounds compared to all other conditions (Fig. 5D).

Figure 5: CFH treatment leads to partial regeneration and enhanced healing of dorsal wounds in wildtype mice.

A. Schematic of MRL and CAST dermal fibroblast culture from dorsal and ear skin. B. Left, fluorescent histology of cultured MRL and CAST dorsal and ear fibroblasts with immunohistochemical (IHC) staining for complement factor H (CFH) and DAPI nuclear counterstain. Right, quantification of CFH expression across in vitro conditions. C. Schematic of MRL and CAST dorsal and ear wounding for histology. D. Fluorescent histology (left) and quantification (right) of IHC staining of wounds for CFH (with DAPI nuclear counterstain). E. Schematic of wildtype mouse dorsal splinted wounding with local wound treatment with either recombinant CFH protein or phosphate-buffered saline (PBS; vehicle control) (see Methods for full details and dosing). F. Left, gross photographs of control (−CFH) and CFH-treated (+CFH) wounds; black dotted outline indicates healed wound region. Right, wound curve reflecting rate of re-epithelialization of −CFH vs. +CFH wounds over time. G. Picrosirius red connective tissue histology of −CFH and +CFH wounds and unwounded skin (UW). H. T-distributed stochastic neighbor embedding (t-SNE) plot of quantified extracellular matrix (ECM) ultrastructural parameters, based on picrosirius red histology of unwounded skin and POD 14 wounds (G), showing overall similarities/differences in ECM ultrastructure between conditions. Each dot represents quantified parameters from one histologic image. I. Hematoxylin and eosin (H&E) histology of POD 14 wounds and skin (n=3). Yellow dotted lines denote borders of healed wounds; white arrows indicate putative regenerating dermal appendages (hair follicles or glands) in +CFH wounds. J. Dermal thickness quantified from histology of wounds and skin. B, C, F, J. *P < 0.05 (Student’s t-test). Scale bar; B, 40μm, D, 150μm, G, 40μm, I, 250μm. See also Figs. S3, S4.

Given that Cfh expression was strongly associated with regenerating conditions (MRL ear wounds), we next evaluated whether modulating CFH signaling could drive wound regeneration and reduce scarring. We treated dorsal wounds in another scarring wildtype mouse strain (C57BL/6J) with recombinant CFH protein (5μg/μl and 10μg/μl doses) or vehicle control (phosphate-buffered saline [PBS]), then evaluated wound outcomes (Fig. 5E; see Methods for full details and dosing). We found that the reduction in scar thickness with CFH treatment was dose-dependent, with more significant scar prevention observed at higher doses (Fig. S3; see Methods for details). Therefore, we selected the higher dose of 10μg/μl for the CFH subsequent in vivo experiments.

Grossly, CFH-treated wounds had reduced scarring and more robust re-epithelialization by POD 14 compared to control wounds (Fig. 5F and Fig. S4A–B). Quantification of ECM ultrastructural parameters from picrosirius red histology (Fig. 5G and Fig. S4C) showed that CFH treatment yielded ECM intermediate between that of control scars and unwounded skin (Fig. 5H), indicating partial regeneration at POD 3, 7, and 14. On hematoxylin and eosin (H&E) histology, CFH-treated wounds had more complete re-epithelialization (confirming gross observations), significantly reduced dermal thickness (consistent with reduced scarring), and structures morphologically resembling early invaginating neogenic hair follicles, in contrast with control scars which remained “bare areas” devoid of any dermal appendages (Fig. 5I–J and Fig. S4). Immunostaining further demonstrated a decrease in Collagen type 1 expression in CFH compared to PBS-treated wounds (Fig. S4D). Finally, we also found that the reduction in scar thickness with CFH treatment was dose-dependent, with more significant scar prevention observed at higher doses (Fig. S3; see Methods for details). Collectively, these findings were consistent with CFH being sufficient to drive partial wound regeneration and significantly reduce scarring in mouse dorsal wounds, and suggest that this gene may play a similar, pro-regenerative role in MRL ear wounds.

Single-cell RNA-sequencing suggests that CXCL2 may contribute to the effect of CFH on wound healing

Given that CFH treatment effectively reduced scar thickness in wildtype mouse wounds, we questioned how CFH treatment induced regeneration in dermal wounds. We treated dorsal wounds in a wildtype mouse strain (C57BL/6J) with recombinant CFH protein or vehicle control (PBS), and subjected wounds at POD 3, 7, and 14 to single-cell RNA-seq (scRNA-seq; Fig. 6A).

Figure 6: CFH treatment reduces scarring through Cxcl2 inhibition in dorsal skin wounds.

A. Schematic of CFH and PBS treatment scRNA-seq experiment (n=12 wounds in six biological replicates for each condition and timepoint). B. UMAP of all cells captured from scRNA-seq experiments colored by cell type. C. Bar graphs showing proportions of Neutrophils (top), Macrophages & Monocytes (middle) and T cells (bottom) across all timepoints and treatment groups (Blue: Unwounded; Red: PBS; Black: CFH). D. Violin plot of Cxcl2 expression by postoperative day and treatment group in fibroblasts. E. Immunostaining of CXCL2 in unwounded, PBS-, and CFH-treated wounds at POD 7 with quantification right (*p < 0.05). F. Schematic of CXCL2 receptor inhibitor (CXCR2i) treatment experiment. G. H&E analysis of unwounded, PBS-, and CXCR2i-treated wounds at POD 7 (n=3 wounds). (Yellow dotted lines show wound borders) H. Representative Picrosirius red analysis of unwounded, PBS-, and CXCR2i-treated wounds at POD 7 and UMAP quantification right. Scale bar; A, 150 μm, B, 250 μm, C, 40 μm. See also Figs. S5,S6.

Eighteen transcriptionally defined clusters were identified by Louvain-based (Seurat) clustering including immune, endothelial, fibroblast, and epithelial cells (Fig. 6B). We first sought to understand to what degree CFH treatment might affect the recruitment of immune cells involved in inflammation, as this has been linked to CFH previously.¹⁶ We found that CFH-treated wounds had a significantly lower proportions of T cells, macrophages, monocytes, and neutrophils (Permutation test P < 0.008 at POD 7 for each cell type) (Fig. 6C). Immunostaining confirmed decreased abundance of macrophages and T cells in CFH-treated wounds at POD 7 (Fig. S5A). Consistent with this, CFH-treated wounds showed lower activation of inflammatory signaling pathways (e.g., annexin and IL-10 pathways; see Supplemental Material⁴⁰). Genes with differential expression between CFH- and PBS-treated wounds were also enriched for ontogeny terms related to inflammation and immune responses (e.g., Inflammatory Response, adjusted p=6.25 × 10⁻¹³; Cytokine-Mediated Signaling Pathway, p=1.58 × 10⁻¹⁰; Neutrophil Migration, p=2.20 × 10⁻⁴).

As Cfh diffASE was found specifically in fibroblasts via bulk RNA-seq, to investigate potential molecular mechanisms we examined gene expression levels in fibroblasts. Comparing fibroblasts between CFH- and PBS- treated wounds, the most significantly differentially expressed gene was the chemokine Cxcl2, which was strongly down-regulated by CFH treatment, particularly at POD 7 (fold-change = 6.68 at POD 7; Wilcoxon test, adjusted p = 6.68 × 10⁻²⁸⁵) (Fig. 6D). Immunostaining confirmed down-regulation of CXCL2 protein in CFH-treated wounds at POD 7 (Fig. 6E and Fig. S5B). We also observed down-regulation of downstream targets of CXCL2 such as Ccl3, Ccl4, and Tnf (fold changes of 12.21, 8.75, 5.13; adjusted p-values of 5.39 × 10⁻¹⁰³, 7.98 × 10⁻¹¹⁴, 1.18 × 10⁻²², respectively), as well as its upstream regulator Il1b47 (fold change=7.93. adjusted p=3.49 × 10⁻¹⁷³) in CFH-treated wounds. As CXCL2 drives neutrophils and inflammatory cells to wound sites⁴⁸ and could therefore potentially play a role in the reduced immune cell recruitment we observed, we sought to further explore its contribution to the effects of CFH treatment.

Inhibition of CXCR2 mimics the regenerative capacity of CFH treatment in dorsal wounds

Given that Cxcl2 gene expression was strongly down-regulated by CFH, we next evaluated whether inhibiting the CXCL2 receptor (CXCR2) may reduce scarring in a similar manner to CFH treatment. We treated dorsal wounds in a wildtype mouse strain (C57BL/6J) with a CXCR2 small molecule inhibitor (CXC2Ri) or vehicle control (PBS), then evaluated wound outcomes (Fig. 6F). Gross images showed CXC2Ri-treated wounds healed more rapidly than controls (Fig. S6A). On H&E histology, CXCR2 inhibition increased re-epithelization at POD 7 (Fig. 6G) and reduced scar thickness at POD 14 (Fig. S6B), resembling CFH-treated wounds. Furthermore, ECM analysis demonstrated that CXCR2i-treated wounds displayed a more similar ECM architecture to unwounded skin than that of PBS-treated wounds (Fig. 6H). Collectively, these data suggest that CXCR2 inhibition may mimic the regenerative effect of CFH treatment in skin wound healing. An important caveat is that CXCR2 is the receptor for several other chemokines in addition to CXCL2, so we cannot rule out their potential contribution to the effects of inhibition; however, CXCL2 was the only ligand for CXCR2 that showed any differential expression in response to CFH treatment.

Discussion

Healing via fibrosis, rather than through regeneration, is a major cause of morbidity and an immense burden for healthcare systems worldwide, with over $20 billion spent annually on the treatment and management of scars in the United States alone.⁴⁹ Instances of regeneration in nature, such as the striking example of regenerative mammalian wound repair that occurs in the ears of MRL mice, may provide valuable insights for therapeutically promoting regeneration and preventing fibrosis.

While QTL mapping has previously been used to identify genomic intervals associated with regenerative healing in the MRL strain, identifying specific candidates for functional follow-up has proven challenging, in part due to the large size of intervals identified. Our approach, using allele-specific gene expression to probe for site-specific cis-regulatory divergence, offers the advantage of not only interrogating potential drivers of regeneration at the single-gene level, but also being substantially less resource-intensive (e.g., requiring fewer than 20 mice, compared to multiple hundreds in prior studies) and thus more accessible. The strength of our methodology is supported by multiple interesting findings. Our approach found greater ASE in ear compared to dorsal wounds across all cell types studied, as well as an enrichment of wound repair pathways and genes associated with differential allele-specific expression in fibroblasts. This would not be expected if the genetic changes leading to MRL ear regeneration were entirely protein-coding, and instead suggests that the strain-specific phenotypic difference seen in ear wounds may reflect cis-regulatory divergence between CAST and MRL. Furthermore, our findings were supported by integrating our results with the QTL fine-mapping study by Cheverud et al. 2014.⁶ The overlap observed with this orthogonal method suggests that at least some of the cis-regulatory changes we identified underlie the MRL ear regeneration phenotype.

Of note, while our findings highlighted fibroblasts as major drivers of MRL ear wound regeneration, our study and others have implicated alterations in diverse cell types (e.g., reduced inflammation mediated by immune cells, more rapid re-epithelialization mediated by keratinocytes) in MRL ear regeneration. These could result from tissue-specific cis-regulatory differences directly affecting other cell types (such as immune or epithelial cells). However, they also likely result, at least in part, from cell-cell crosstalk mediated by fibroblasts. For instance, intimate fibroblast-keratinocyte crosstalk is critical for wound repair⁵⁰, and we have previously found that modulating pro-fibrotic fibroblast molecular processes can also induce changes in the overlying epidermis.¹⁹ Fibroblast-immune and immune-epithelial interactions have also been extensively reported in the context of injury repair.^51,52 Thus, it is feasible that altered fibroblast phenotype in wounds could fundamentally drive many of the differences observed in regenerating MRL ear wounds; we found that fibroblast diffASE genes were enriched for pathways involved in modulating cartilage development and immune cell activity, further supporting this hypothesis.

Through diffASE analysis, we have identified several genes that may be associated with regeneration in ear wounds compared to dorsal wounds. These included Spp1, Thbs4, and Clic4 (Fig. 3E). These genes are associated with extracellular matrix organization and cell matrix signaling, suggesting that regeneration in MRL ear wounds may be associated with alterations in ECM-related processes.^53,33,54 Finally, our genomic findings are also supported by the results of our experiments with CFH, which was identified by diffASE analysis and subsequently shown via in vivo wounding experiments to promote regeneration and reduce scarring. In addition to providing important functional validation for our methodological approach, this finding could have important clinical implications, as no targeted molecular therapies currently exist to prevent human scarring.

Our scRNA-seq data further revealed that CFH treatment down-regulates CXCL2 and reduces recruitment of immune cells (Fig. 6), suggesting a potential mechanism for CFH’s effect. Interestingly, CXCL2 secretion is increased in CFH KO mice, supporting the relationship observed in our scRNA-seq data.⁵⁵

We also found that inhibition of CXCR2, the receptor for CXCL2, mimicked the regenerative characteristics of CFH-treated dorsal wounds (Fig. 6). Mice ubiquitously lacking CXCR2 have been found to display delayed wound healing with reduced epithelization and neutrophil recruitment. While this might seem to be at odds with our chemical inhibition of CXCR2, it should be noted that CXCR2 null mice also have thinner skin due to fewer and smaller subcutaneous adipocytes, suggesting that transient inhibition of CXCR2 may be a more faithful model of transient reduction in CXCR2 signaling than a ubiquitous knockout mouse.⁵⁶ Consistent with this, selectively blocking CXCR2 was able to reverse the delayed wound healing in diabetic mice and accelerate ex vivo human skin wound healing.⁵⁷ Taken together, previous studies suggest the role of CXCR2 in wound healing is complex and context-dependent. Given that small molecule inhibitors have potential off-target effects and limited specificity, and that CXCR2 is a receptor for several chemokines in addition to CXCL2, further studies are needed to fully ascertain the relationship between CFH, CXCL2, and wound healing.

Limitations of study

While the pro-regenerative effects of recombinant CFH treatment were substantial, they also fell short of complete regeneration. Future studies may seek to determine whether other genes identified by our diffASE analysis also have similar pro-regenerative effects.

STAR Methods

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Michael T. Longaker (longaker@stanford.edu).

Materials availability

This study did not generate new unique reagents.

Data and code availability

Sequencing data (bulk and scRNA-seq) have been deposited at NCBI Sequence Read Archive Raw and are publicly available as of the date of publication. Accession numbers are listed in the key resources table. Read counts, gene lists for each cell type, enrichment lists and additional supplemental figures and tables have been deposited at Figshare and Mendeley Data and are publicly available as of the date of publication. DOIs are listed in the key resources table.

Key resources table.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
PE anti-CD45	BioLegend	Cat#103105
APC anti-CD31	Invitrogen	Cat#17-0311-80
eFluor 450-conjugated Lineage (Lin) antibodies anti-CD45	ThermoFisher Scientific	Cat#48-0451-82
anti-Ter-119	ThermoFisher Scientific	Cat#48-5921-82
anti-CD31	BioLegend	Cat#303114
anti-Tie-2	ThermoFisher Scientific	Cat#13-5987-82
anti-CD326	ThermoFisher Scientific	Cat#48-5791-82
anti-CD324	ThermoFisher Scientific	Cat#13-3249-82
anti-CFH primary antibody	LS bio	Cat#LS-C819285
Alexa Fluor 488 anti-rabbit secondary antibody	Invitrogen	Cat#A-11008
CXCR2 antagonist AZ 10397767	Tocris	Cat#5872
Critical commercial assays
SMART-Seq v4 Ultra Low Input Kit	Takara Bio	Cat#634894
KAPA Hyper Prep Next Generation Sequencing	Roche	N/A
miRNeasy Micro Kit	QIAGEN	Cat#217084
Fibroblast culture media (DMEM + Glutamax media)	ThermoFisher	Cat#10569010
Deposited data
Raw RNAseq data	This paper	NCBI Sequence Read Archive; BioProject PRJNA839777
Raw scRNAseq data	This paper	NCBI Sequence Read Archive; BioProject PRJNA839777
Allele-specific count data (MRLxCAST)	This paper; Figshare	doi: https://doi.org/10.6084/m9.figshare.19952513.v1
CAST/EiJ SNP calls	Wellcome Trust Mouse Genome Project	https://www.sanger.ac.uk/science/data/mouse-genomes-project
Ear closure QTL mapping data	Cheverud et al.6	N/A
Mouse genome (GRCm38)	Ensembl	http://ensembl.org/
Experimental models: Organisms/strains
Mouse: CAST/EiJ	The Jackson Laboratory	Strain #:000928
Mouse: MRL/MpJ	The Jackson Laboratory	Strain #:000486
Mouse: C57BL/6J	The Jackson Laboratory	Strain #:000664
Software and algorithms
Bowtie2	Langmead and Salzberg58	http://bowtie-bio.sourceforge.net/bowtie2/index.shtml
DESeq2	Love et al.64	https://bioconductor.org/packages/release/bioc/html/DESeq2.html
DESeq2 analysis code	This paper; Figshare	doi: https://doi.org/10.6084/m9.figshare.c.6025157
Cutadapt	Martin 201162	https://cutadapt.readthedocs.io/
Genome Analysis Toolkit (GATK)	McKenna et al.60	https://gatk.broadinstitute.org/
BEDTools	Quinlan and Hall66	https://bedtools.readthedocs.io/
Seurat	Satija et al.69	https://satijalab.org/seurat/
scProportionTest	Miller et al.72	https://github.com/rpolicastro/scProportionTest
CellChat	Jin et al.73	http://www.cellchat.org/
Enrichr	Chen et al.71	https://maayanlab.cloud/Enrichr/
PANTHER	Mi et al.67	https://pantherdb.org
modPhea	Weng and Liao68	https://evol.nhri.org.tw
Fiber quantification pipeline	Mascharak et al.19	https://github.com/shamikmascharak/Mascharak-et-al-ENF
Other
Supplemental figures and tables	This paper; Mendeley Data	doi: 10.17632/9jhfh79b8d.1

DESeq2 code has been deposited on Figshare and is publicly available as of the date of publication. DOIs are listed in the key resources table.

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

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Animals

CAST/EiJ (CAST), and C57BL/6J mice were obtained from The Jackson Laboratory (Bar Harbor, ME). Male MRL and female CAST mice were bred to produce CAST × MRL F1 hybrid (F1) offspring. Adult (postnatal day [P]60) female F1 mice were used for RNA-seq experiments, and both female and male P60 mice were used for all other experiments.

Mice were housed and maintained in sterile micro-insulators at the Stanford University Comparative Medicine Pavilion in accordance with Stanford University Administrative Panel on Laboratory Animal Care (APLAC) guidelines (APLAC-21308). Food and water were provided ad libitum.

Primary cell culture

Fibroblast cells were isolated from female and male 6- to 8- week-old MRL and CAST mouse dorsal and ear skin. Mice were housed and maintained in sterile micro-insulators at Stanford University Comparative Medicine Pavilion in accordance with APLAC guidance (APLAC-21308). Food and water were provided ad libitum. Details regarding primary culture methods can be found under the “Fibroblast cell culture” subheading.

METHOD DETAILS

Dorsal and ear wounding

Mice underwent ear punch wounding and dorsal splinted excisional wounding following established protocols without modification.^3,18 Briefly, mice were anesthetized using 1.5–2% isoflurane. All surgical tools were autoclaved prior to the procedure. For ear wounds, the skin was prepped using alcohol wipes. Punch wounds 2mm in diameter were created using a thumb-type metal ear punch (Fisherbrand). For mice wounded for histologic and wound curve analysis, one wound was created per ear, roughly in the middle of the pinna. For mice wounded for RNA-seq analysis, in order to obtain sufficient cells for analysis while minimizing the number of mice required, three wounds were created per ear, spaced at least 2mm apart. For dorsal wounds, hair was removed from the entire dorsum using an electric shaver followed by depilatory cream. The dorsal skin was then prepped using three sequential alternating swabs of betadine and 70% ethanol. Using sharp surgical scissors, two 6mm-diameter full-thickness excisional wounds were made per mouse, roughly at the level of the scapulae and 4mm lateral to midline. Wounds were stented open by affixing silicone rings (1cm internal diameter) around the wound using adhesive and eight simple interrupted sutures (6–0 nylon, Ethicon). Wounds were dressed using Tegaderm (3M) and dressings were changed every other day until harvest.

For dose response experiments with CFH-treated dorsal wounds (see Fig. S3), which were performed in CAST mice, recombinant mouse complement factor H protein (R&D Systems) was resuspended in phosphate-buffered saline (PBS) at a concentration of either 5 or 10 μg/mL, then 50 μL of CFH at these concentrations, or PBS (vehicle control), were injected locally into the wound base and surrounding dermis immediately following wounding (POD 0) and again at POD 7. For CFH scar prevention experiments (Fig. 5), which were done in C57BL/6J mice (due to poor anesthesia tolerance and excessive fighting leading to high morbidity/mortality with initial wounding experiments in CAST mice), recombinant CFH was resuspended at 30 μg/mL and 30 μL of CFH solution or PBS were injected on POD 0, 2, 4, 6. For wound curve analysis, wounds were photographed every other day for the first two weeks of healing, then (for ear wounds) weekly for two additional weeks. For ear wounds, a circular stencil was placed over wounds prior to photographing in order to provide a consistently sized reference for measurements; for dorsal wounds, the silicone splints served as a size reference. Area of the stencil/splint and remaining wound area at each timepoint were measured in Photoshop (Adobe) and used to calculate remaining wound area as a percentage of original wound size (normalized to size of stencil/splint for each photograph).

For CXCR2i scar prevention experiments (see Fig. 6), which were done in C57BL/6J mice, the CXCR2 antagonist AZ 10397767 (Tocris) was resuspended in DMSO at 802 ng/mL and 30 μL of CXCR2 solution or PBS were injected on POD 0, 2, 4, and 6.

Tissue histologic analyses

Tissue for histologic analysis was harvested and fixed by incubation in 10% neutral buffered formalin for 16–18 hours at 4 °C. Following fixation, tissue was processed for paraffin or OCT embedding by standard procedures. Briefly, for paraffin embedding, tissue underwent sequential dehydration (ethanol), clearing (xylene), and infiltration by paraffin wax. For OCT embedding, tissue was incubated in 30% sucrose/PBS for two weeks at 4 °C, OCT for 1 day at 4 °C, then embedded in OCT blocks by freezing in a dry ice/tert-butanol bath. All wounds were bisected and embedded cut-side-down. Tissue sections were cut using a microtome (paraffin) or cryostat (OCT) at 8 μm thickness. Hematoxylin and eosin (H&E) and picrosirius red staining (using Picro Sirius Red Stain Kit, Abcam) were performed on paraffin sections, using standard protocols without modification. Dermal thickness was measured from H&E histology images; Photoshop (Adobe) was used to measure the dermis (from the bottom of the epidermis to the top of the subcutaneous tissue), and a minimum of nine measurements (three measurements per image from three individual histology images/sections) were averaged per wound. Machine learning analysis of ECM ultrastructure was performed as previously described using Matlab.¹⁹ Briefly, picrosirius red histology images were normalized, color deconvoluted, noise reduced, then binarized. Binarized images were filtered to select for fiber-shaped objects and the fiber network was skeletonized. Finally, 294 parameters of the digitized map (including fiber length, width, persistence, alignment, etc.) were measured. Dimensionality reduction of quantified fiber network properties by t-distributed stochastic neighbor embedding (t-SNE) was used to plot parameters for each image. Comparisons between conditions were based on visual assessment of t-SNE clustering from calculated ECM parameters. Matlab scripts containing the fiber quantification pipeline are available at the following Github repository: https://github.com/shamikmascharak/Mascharak-et-al-ENF.

FACS isolation of wound cell populations

To harvest a consistent region for RNA-seq analysis, wounds were excised with a 1mm ring of tissue around each wound using a biopsy punch (4 mm punch for ear wounds; 8 mm punch for dorsal wounds). Wound tissue was incubated in ammonium thiocyanate (3.8% in Hank’s balanced salt solution [HBSS]) for 20 minutes at room temperature to dissociate the epidermis, then dermal tissue was separated from overlying epidermis and underlying cartilage (for ear wounds) under a surgical microscope. For each biological replicate, ear or dorsal wounds from three individual mice were pooled to obtain sufficient cell numbers for sequencing. Wounds were finely minced with sharp surgical scissors then enzymatically digested in collagenase type IV (1500 U/mL in Dulbecco’s Modified Eagle Medium [DMEM]) at 37 °C, with agitation at 150 rpm, for 1 hour. After 1 hour, digestion was quenched by addition of equal volume of DMEM with 10% heat-inactivated fetal bovine serum (FBS), filtered through 70 μm followed by 40 μm nylon filters, then pelleted (200 × g, 5 min, 4 °C). Cell pellets were resuspended in FACS buffer (PBS with 1% FBS and 1% penicillin-streptomycin) then stained with the following antibodies: PE anti-CD45 (BioLegend #103105); APC anti-CD31 (Invitrogen #17-0311-80); and eFluor 450-conjugated Lineage (Lin) antibodies anti-CD45 (ThermoFisher Scientific #48-0451-82), anti-Ter-119 (ThermoFisher Scientific #48-5921-82), anti-CD31 (BioLegend #303114), anti-Tie-2 (ThermoFisher Scientific #13-5987-82), anti-CD326 (ThermoFisher Scientific #48-5791-82), and anti-CD324 (ThermoFisher Scientific #13-3249-82), for isolation of fibroblasts via lineage depletion as per previously published protocol.²⁴ DAPI (BioLegend; 1:1000) was added as a viability stain. Live (DAPI⁻) singlet cells were sorted to obtain immune cells (PE/CD45⁺), endothelial cells (APC/CD31⁺), and fibroblasts (PB/Lin⁻), which were sorted directly into lysis reagent (QIAzol, QIAGEN), then stored at −20 °C until RNA purification.

Bulk RNA-sequencing of wound cell populations

RNA was purified from each cell sample using the miRNeasy Micro Kit (QIAGEN), then kept at −20 °C until sequencing. Samples were shipped on dry ice and library preparation and sequencing were performed by Admera Health (South Plainfield, NJ). Library preparation was with the SMART-Seq v4 Ultra Low Input Kit (Takara Bio) with PolyA Selection. Sequencing was done using Illumina HiSeq (2×150).

Sequencing of MRL and variant calling

To identify variant calls for allele-specific expression, we generated whole-genome data for the MRL inbred line. MRL tail DNA was extracted and purified using the Invitrogen PureLink Genomic DNA Mini Kit. The library was prepared using the KAPA Hyper Prep kit. The MRL genome was sequenced to moderate coverage (average of 25x for sites with at least one read) on the Illumina HISeq X platform (2×150 reads). A plot of read coverage can be found on Mendeley Data⁴⁰. Genomic reads were then mapped to the M. m. domesticus mm10 (GRCm38) reference genome using bowtie2 v2.3.4 (argument: --very-sensitive)⁵⁸. CAST/EiJ sequence data was obtained from Wellcome Trust Mouse Genome Project (https://www.sanger.ac.uk/science/data/mouse-genomes-project)⁵⁹ in bam format, mapped to mm10. SNP calling was performed using the Genome Analysis Toolkit v4.1 (GATK).⁶⁰ Duplicate reads were marked with the Picard tool MarkDuplicates. GATK HaplotypeCaller was used to call variants between CAST/EiJ, MRL, and the mouse reference genome (mm10). We filtered variants for low quality calls (SNPs: QD < 2.0 || FS > 60.0 || MQ < 40.0 || MQRankSum < −12.5 || ReadPosRankSum < −8.0; Indels: QD < 2.0 || FS > 200.0 || ReadPosRankSum < −20.0), sites with a read depth of less than 5, and for any sites with heterozygous calls using GATK SelectVariants and bcftools⁶¹. This resulted in 19,472,153 SNPs, 1,171,415 of which were in exons. Variant calls where CAST and MRL differed from each other or both differed from the mm10 reference were then used to create alternative references for MRL and CAST for mapping. SNP calls were inserted in mm10 and indels were masked. The concatenated MRL-CAST genome was used to identify reads mapping uniquely to each parental genome in F1 hybrids.

Mapping and allele-specific assignment

Raw reads were trimmed for adapter contamination with the TrimGalore (v0.5) wrapper for Cutadapt v1.18.⁶² Trimmed reads were mapped to a concatenated MRL-CAST genome using STAR v2.5.4b⁶³, discarding any reads that did not map uniquely to one of the reference genomes (arguments: --outFilterMultimapNmax 1 --outFilterMultimapScoreRange 1). Requiring that reads map uniquely to one genome ensures we only consider reads overlapping a heterozygous site in F1 individuals. This resulted in an average of 41,215,441, 60,938,290, 76,012,313 uniquely mapped reads for each immune, endothelial, and fibroblast library, respectively (see Table S1). Reads overlapping exonic regions were summed to generate a total count for each gene based on the Ensembl GRCm38 annotation. Cases where reads only mapped to one allele were discarded as they likely reflect SNP calling errors or genomic imprinting. Approximately 50% of reads mapped uniquely to both CAST and MRL, indicating little evidence of mapping bias (see Supplemental Data on Mendeley Data⁴⁰). DESeq2⁶⁴ (v1.34.0; R computing environment, v4.1.2) was used to perform a variance stabilizing transformation for principal component analysis and perform regularized log₂ transformation of the count data (which minimizes differences between samples for rows with small counts and normalizes with respect to library size) for visual comparison of read count data in Fig. 2.

Identifying allele-specific expression

DESeq2 was used to identify allele-specific expression and condition-specific ASE (i.e., diffASE).^64,65 Allele-specific expression analyses were restricted to genes with at least 30 reads in each condition (wound type, allele) and non-zero values for >4 alleles across individuals (see Table S2). We analyzed allele-specific reads from each cell population separately with DESeq2 with the model “~tissue + tissue:sample + tissue:allele” (where “:” denotes an interaction term between two variables in DESeq2)(Table S3). Here, the term “tissue” is the wound site, for differences between ear and dorsal wounds. “Sample” refers to the sample pool the allele-specific sample pertains to, and the term accounts for variation among the different sample pools within wound site groups. “Allele” refers whether allele-specific reads are mapped preferentially to MRL or CAST, and the interaction between allele and tissue is used to estimate the MRL vs CAST allele ratios separately between wound sites. “DiffASE” genes are identified via a contrast between CAST/MRL ratios in ear and dorsal (DESeq2, Wald test). Consequently, significant cases represent scenarios in which the log₂ fold change of CAST/MRL differ between wound types (Table S3,S4). As read counts come from MRL and CAST come from the same sequencing library, library size factor normalization was disabled by setting SizeFactors = 1⁶⁵. A false discovery rate correction was applied using the Benjamini-Hochberg method for each comparison.

Overlap with previous QTL mapping

Marker locations and LOD scores for a model with additive and dominance values for ear wound closure are as described by Cheverud et al. 2014 (LOD and marker scores for analysis provided by J. Cheverud).⁶ Marker locations were converted from mm9 to mm10 using LiftOver. Autosomal genes were annotated to their closest genetic marker using BEDTools⁶⁶ (tool: “closest”) based on Ensembl mm10 gene start and end coordinates. Liftover coordinates of QTL support intervals defined by Cheverud et al.⁶ were used to identify overlap with diffASE genes within QTL using the BEDTools (tool: “intersect”). QTL mapping was performed using LG, the progenitor of MRL. MRL and LG mice share ~75% of their genome, and both shared and unique QTL from these lines contribute to advanced wound healing.^6,38 Consequently, comparisons with this study will be restricted to identifying QTL shared between the lines. However, this should only make our enrichment tests more conservative and for shared causal regions.

Enrichment analyses

GO enrichment analyses were performed with PANTHER⁶⁷, using a foreground list of genes of interest vs. a background list of all genes with sufficient expression to be tested in a cell population (GO Ontology database released 2019-12-09). Mutant phenotype enrichment tests were performed with modPhEA⁶⁸, also using a foreground list of genes and a background list of all genes with sufficient expression to be tested in a cell population.

Enrichment of GO terms and mutant phenotypes for diffASE genes are available on Mendeley Data⁴⁰. Enrichment for wound healing terms for fibroblasts was found for diffASE genes at both the FDR<0.05 and FDR<0.1 cut-offs.

Fibroblast cell culture

Fibroblast cells were isolated from MRL and CAST dorsal and ear skin for in vitro analysis. Following dissection of the dermis from the dorsum and ear, tissue was washed in PBS and finely minced using sterile scissors. Tissue was then digested in collagenase type IV (1500 U/mL in DMEM) at 37 °C, with agitation at 150 rpm, for 1 hour. Enzyme activity was quenched by addition of FBS-enriched media, and digested tissue was successively strained through 300μm followed by 100μm cell strainers. Filtered samples were then centrifuged at 1500 rpm for 5 minutes at 4 °C to obtain a cell pellet. Pelleted cells were resuspended and plated in fibroblast culture media (DMEM + Glutamax media [ThermoFisher, Cat: 10569010] enriched with 10% fetal bovine serum [ThermoFisher, Cat: 10082147] and 1% Antibiotic-Antimycotic [ThermoFisher, Cat: 15240062]) then grown until confluency in tissue culture incubators kept at 37 °C and 5% CO₂. Cells used for experiments were between passages 2–4. For in vitro analysis of CFH expression, fibroblasts were seeded onto coverslips at 15,000 cells/coverslip for immunostaining (see section below).

Immunostaining of cells and wounds

For both OCT wound section slides and cell-seeded coverslips, immunofluorescent staining was performed as follows. Samples were washed twice in Tween 20 (Sigma-Aldrich, St. Louis, MO) followed by one wash in PBS. Samples were then blocked for 1 hour with Power Block (Biogenex, Fremont, CA) prior to addition of anti-CFH primary antibody (LS bio, LS-C819285, 1:200). Samples were washed then incubated for 1 hour with Alexa Fluor 488 anti-rabbit secondary antibody (Invitrogen, Waltham, MA). Finally, samples were mounted in Fluoromount-G mounting solution with or without DAPI (ThermoFisher Scientific, Waltham, MA). Fluorescent images were acquired with a LSM880 inverted confocal, Airyscan, AiryscanFAST, GaAsP detector upright confocal microscope.

Single Cell RNA-sequencing of wound cells

Dorsal dermal wounds from C57BL/6J mice treated with CFH or PBS control were harvested at POD 3, 7, and 14, and mechanically digested with sharp surgical scissors (6 wounds pooled from 3 mice per condition). Two independent unwounded tissue samples were taken as a control for comparison from 3 mice each. Tissue then underwent enzymatic digestion with Collagenase II (ThermoFisher, Cat: 17101015) and IV (ThermoFisher, Cat: 17104019) in DMEM-F12 (GIBCOTM, Fisher Scientific, Hampton, NH). These samples were placed on an orbital shaker for 90 minutes at 150 rpm at 37°C. After adding FACS buffer to quench the digest, samples were passed through 70 μm cell strainers and centrifuged at 1500 rpm at 4°C for 5 minutes. Samples were then passed through a second filtration step using 40μm cell strainers and then resuspended with 0.04% UltraPure BSA (Thermo Fisher, Waltham, MA). Following this cell counting was followed by scRNA-seq using the 10x Chromium Single Cell platform (Single Cell 3’ v3, 10x Genomics, USA) as previously described.¹⁹

Using the Cell Ranger (10X Genomics; version 3.1) implementation mkfastq, base calls were converted to reads. These reads were then aligned against the mm10 reference genome (http://cf.10xgenomics.com/supp/cell-exp/), applying Cell Ranger’s count function with SC3Pv3 chemistry and 10,000 expected cells per sample, as previously described.¹⁹ Cutoffs of 2,500 maximum unique genes and a maximum percent mitochondrial RNA of 15% were employed. This resulted in 12995 cells, of which 3185 were annotated as fibroblasts.

For downstream analysis, each cell’s unique molecular identifiers (UMIs) were normalized with a scale factor of 10,000 UMIs per cell, and, using the Seurat R package (version 4.3.0)⁶⁹, and the first 15 principal components used for uniform manifold approximation and projection (UMAP). Using SingleR (version 3.11), cell annotations were assigned to each cell using the Mouse-RNA-seq reference dataset (https://rdrr.io/github/dviraran/SingleR/man/mouse.rnaseq.html). The neutrophil cluster was identified based on genes that are highly expressed in neutrophils or specific to neutrophils.⁷⁰ Using Seurat’s FindMarkers function, cell-type marker lists were created using a log fold change threshold of 0.25. Using the 200 most highly ranked genes within each cluster, EnrichR (version 2.1) was used to conduct gene set enrichment analysis.⁷¹ To identify differences in expression between CFH- and PBS-treated wound clusters, a Wilcoxon rank-sum test was implemented in Seurat. To test for differences in cell proportions between PBS- and CFH-treated clusters, we implemented a permutation test using the R library scProportionTest.⁷² In brief, cells from CFH- and PBS-treated are pooled together and then randomly segregated into one of the two conditions while maintaining the original sample sizes. Proportional differences are then calculated and compared to the observed proportional difference.

CellChat Receptor-Ligand Analysis

The CellChat platform was used to assess possible interactions between cell types in our scRNA-seq dataset.⁷³ Using our scRNA-seq Seurat object in R, we implement this in conjunction with standalone CellChat Shiny App for its Cell-Cell Communication Atlas Explorer. SingleR-defined cell types were used to bin cells. Secreted Signaling, ECM-Receptor, and Cell-Cell Contact relationships were considered, and default parameterizations used throughout.

QUANTIFICATION AND STATISTICAL ANALYSIS

All statistical analyses and quantification were performed in R, Matlab, or Prism as described in the Results and Methods Details. Details regarding the exact value of and identify of n, definition of center, and dispersion measures are defined in the figure legends. Significance was defined as p < 0.05 for mouse (non-sequencing) experiments. For expression comparisons in scRNA-seq and RNA-seq experiments, p-values were corrected for multiple testing using the Benjamini-Hochberg procedure implemented in DESeq2 (bulk RNA-seq; Wald tests) and a Bonferroni correction in Seurat (scRNA-seq; Wilcoxon rank-sum test). Significance was defined as adjusted p-values (q-values) < 0.05. Exact p-values reported where possible or relevant. No data were excluded from the analysis.

Highlights.

• “Super-healer” MRL mice regenerate multiple tissue types without fibrosis

• Cis-regulatory differences associated with regeneration are identified

• CFH applied ectopically accelerates wound repair and induces regeneration

• CFH treatment dramatically reduced immune cell recruitment to wounds