Hair follicle stem cell progeny heal blisters while pausing skin development

Abstract

Injury in adult tissue generally reactivates developmental programs to foster regeneration, but it is not known whether this paradigm applies to growing tissue. Here, by employing blisters, we show that epidermal wounds heal at the expense of skin development. The regenerated epidermis suppresses the expression of tissue morphogenesis genes accompanied by delayed hair follicle (HF) growth. Lineage tracing experiments, cell proliferation dynamics, and mathematical modeling reveal that the progeny of HF junctional zone stem cells, which undergo a morphological transformation, repair the blisters while not promoting HF development. In contrast, the contribution of interfollicular stem cell progeny to blister healing is small. These findings demonstrate that HF development can be sacrificed for the sake of epidermal wound regeneration. Our study elucidates the key cellular mechanism of wound healing in skin blistering diseases.

Subepidermal blisters (epidermal wounds) heal at the expense of hair follicle development. The progeny of hair follicle junctional zone stem cells mainly contributes to epidermal wound regeneration.

Introduction

Tissue responds to injury by transforming its cellular components and extracellular matrix from homeostasis into a regenerative state. Damaged tissue typically reactivates an embryonic gene program in epithelia to accelerate tissue regeneration (Fernandez Vallone et al, 2016; Nusse et al, 2018; Yui et al, 2018; Miao et al, 2019). However, it is unknown whether this phenomenon also applies to injuries in developing tissue, in which the embryonic gene expression program is switched on before damage.

The epidermis is a stratified epithelium of the skin and is located on the surface of the body, where it serves as a barrier against external stimuli and microorganisms (Natsuga, 2014). Cellular proliferation and differentiation in the epidermal basal layer, where epidermal stem cells (SCs) are present, are fine‐tuned to maintain the integrity of the epidermis (Donati & Watt, 2015). The epidermis attaches to the dermis through proteins in the epidermal basement membrane zone (BMZ) (McMillan et al, 2003). Epidermal BMZ proteins function as a niche for epidermal SCs (Watt & Fujiwara, 2011), and the loss of these proteins, such as α6 integrin (ITGA6), β1 integrin, and collagen XVII (COL17), leads to transient epidermal proliferation (Brakebusch et al, 2000; Niculescu et al, 2011; Watanabe et al, 2017).

Skin wounding causes pain and carries a significant risk of bacterial infection. The sources of skin wound healing have been extensively investigated in experimental animals (Rognoni & Watt, 2018; Dekoninck & Blanpain, 2019). Hair follicles (HFs), epidermal appendages, fibroblasts, and immune cells coordinate to heal the wound, and the contribution of each component can vary depending on the assay (Garcin et al, 2016). Wounding in adult skin induces the expression of genes regulating epidermal development, including SOX11 and SOX4 (Miao et al, 2019).

Conventional skin wounding assays have employed full‐thickness skin wounds, in which all skin components are removed, including the epidermis, epidermal appendages, dermis, and subcutaneous fat tissue. In contrast to conventional full‐thickness skin wounds, epidermal detachment, as exemplified by subepidermal blisters, is distinctive because it does not affect the structures below the epidermis per se. The epidermis is detached from the dermis in several pathological conditions, such as burns (Chetty et al, 1992), congenital defects in epidermal BMZ proteins (epidermolysis bullosa (EB)) (Fine et al, 2014; Vahidnezhad et al, 2019), autoimmunity to these proteins (pemphigoid diseases) (Schmidt & Zillikens, 2013), and severe drug reactions, such as Stevens–Johnson syndrome/toxic epidermal necrolysis (White et al, 2018). Although the cells that contribute to the repair of full‐thickness skin wounds have been identified (Ito et al, 2005; Page et al, 2013; Sada et al, 2016; Aragona et al, 2017; Donati et al, 2017; Gonzales & Fuchs, 2017; Park et al, 2017; Dekoninck & Blanpain, 2019; Kang et al, 2020), the cellular dynamics of subepidermal blister healing are completely unknown. In addition, full‐thickness skin wounding, when applied to developmental skin, is unsuitable for distinguishing tissue regeneration and development. In contrast, blistering injury allows us to monitor both skin regeneration (re‐epithelization of the epidermis) and morphogenesis (HF development) within the same wound bed.

Here, by taking advantage of subepidermal blisters, we explore the effects of injury on developmental tissue. Unexpectedly, blistering injury is found to reduce the expression of tissue morphogenesis genes in the healed epidermis and to direct HFSCs, rather than epidermal SCs, to provide progeny to heal the wound and to suspend HF development.

Results

Subepidermal blister formation and its healing process

The suction‐blister technique was developed more than a half‐century ago to selectively remove the epidermis from the dermis (Kiistala & Mustakallio, 1964, 1967), and it has been utilized to harvest epidermal pieces for transplant to repair human skin defects. We reasoned that suction blisters on neonatal mice, in which the epidermis is removed while HFs, the dermis, and subcutaneous fat tissues are maintained in the wounds, enable us to examine the direct relationship between tissue injury and skin development. Therefore, we applied constant negative pressure to the dorsal skin of C57BL/6 wild‐type (WT) neonates to produce subepidermal blisters (postnatal day 1 (P1), Fig 1A). Histologically, skin separation occurred at the level of the dermoepidermal junction (DEJ) (Figs 1B and EV1A–C). HFs, as shown by alkaline phosphatase (AP)‐positive dermal papillae, remained on the dermal side of the blisters (Figs 1B and EV1D). Dermis and subcutaneous tissues were intact after blistering (Fig 1B). α6 integrin (ITGA6), a hemidesmosome protein, was seen at the blister roof, whereas type IV collagen (COL4), a major component of the epidermal basement membrane, and laminin 332 (L332) were present at the base of the blister (Figs 1C and EV1B). In line with the immunofluorescence data, hemidesmosomes localized on the blister roof and lamina densa (basement membrane) were observed at the blister base by electron microscopy (Figs 1D and EV1C), as seen in human suction blisters (Kiistala & Mustakallio, 1967) and their murine counterparts (Krawczyk, 1971).

Figure 1. Healing of subepidermal blisters in neonatal mice.

• A
  Schematic diagram of suction blistering and sample collection. A blister produced on C57BL/6 wild‐type (WT) mouse dorsal skin at P1. BM: basement membrane.
• B–D
  Blistered samples at P1. (B) Hematoxylin and eosin (H&E, top) and alkaline phosphatase (AP, bottom) staining. Hair follicles (HFs) remaining in the dermis (indicated by arrows). Scale bar: 500 μm. (C) α6 Integrin (ITGA6, indicated by arrowheads) and type IV collagen (COL4, arrows) labeling (left). Laminin 332 (L332, arrows) staining (right). Scale bar: 100 μm. (D) Ultrastructural findings of blistered skin (left image: blister roof, right image: blister bottom). Hemidesmosomes (white arrowheads) and lamina densa (arrows) are indicated. Scale bar: 1 μm.
• E
  H&E (left), pan‐cytokeratin (PCK, middle), and ITGA6 staining (right) at P2. The regenerated epidermis is indicated by arrowheads. Scale bar: 200 μm.
• F
  Keratin 14 (K14) and keratin 10 (K10) staining of the nonlesional (intact) and lesional (blistered) skin at P2 (upper images and inlets: sections, lower images: whole‐mount imaging). HFs are indicated by arrows in the whole‐mount images. Scale bar: 30 μm.
• G
  H&E (top) and loricrin (LOR, bottom) staining at P3. Scale bar: 200 μm.
• H
  H&E (left), PCK (middle), and LOR staining (right) at P4. Scale bar: 200 μm.

Data information: Blisters are indicated by stars. Representative images are shown from three or more replicates in each group.

Figure EV1. Healing processes of subepidermal blisters.

1. H&E (left) and PCK labeling (right) of blistered skin at P1 (WT). Scale bar: 100 µm.
2. ITGA6 (arrowheads in the blister roof) and COL4 (arrows in the blister bottom) labeling at P1 (WT). HFs that express ITGA6 on the dermal side are indicated by hashtags. Scale bar: 100 µm (left) and 50 µm (right).
3. Electron microscopy of blistered skin at P1 (WT). Hemidesmosomes (white arrowheads) and the lamina densa (arrows) are indicated. Scale bar: 10 µm (left) and 1 µm (right).
4. AP staining of WT (left) and Col17a1 ^−/− (right) blistered skin at P1. Scale bar: 100 µm.
5. H&E (left), ITGA6 (middle), and ITGA5 (right) staining at P2 (WT). The regenerated epidermis is indicated by arrowheads. ITGA5⁺ cells at the tip of the HFs are indicated by arrows. Blisters are indicated by stars. Scale bar: 100 µm.
6. Schematic of the suction‐blister experiments and HF development/cycles.
7. Quantification of immune cells (CD3, F4‐80, and Ly6G) in the dermis of blistered WT and unaffected littermate control skin at P2 (n = 4 biological replicates) and P4 (n = 3 biological replicates). The data are shown as the mean ± SE. *0.01 < P < 0.05, Student’s t‐test. NS, no significance.

Data information: Blisters are indicated by stars. Representative images are shown from three or more replicates in each group.

We then characterized the healing processes of the subepidermal blisters. One to two layers of the regenerated epidermis, marked with pan‐cytokeratin, were found 1 day after blister formation (P2, Figs 1E and EV1E). The regenerated epidermis restored ITGA6 expression at the DEJ (Figs 1E and EV1E). The shape of the basal keratinocytes in the intact skin was cuboidal or columnar (P2, keratin 14 (K14)‐positive cells in the nonlesional area, Fig 1F). In contrast, the regenerated keratinocytes in the blistered skin transformed from cuboidal to a wedge/flattened shape (P2, K14‐positive cells in the lesional area, Fig 1F) (Krawczyk, 1971). Two days after blistering (P3), the stratified epidermal layers were mostly restored but still lacked loricrin‐positive granular layers, a hallmark of proper epidermal differentiation, in the lesional area (Fig 1G). The final step in epidermal differentiation was completed by the formation of loricrin‐positive granular layers 3 days after blister formation (P4, Fig 1H). The immunofluorescence data for subepidermal healing are summarized in Table EV1. These results demonstrate that subepidermal blisters on neonates can serve a model for visualizing wound healing without damaging HFs and other dermal components at the developmental stage.

Epidermal restoration at the expense of skin development

To elucidate the effects of the blistering injury on the neonatal skin, we performed RNA‐seq profiling of the wounded tissue above the dermis (the regenerated epidermis and the blister roof) 1 day after blistering (P2, Figs 2A and EV2A). Unexpectedly, the expression of genes involved in HF morphogenesis, such as Wnt signaling, melanogenesis, and Hedgehog signaling, was significantly downregulated (Figs 2B and C, and EV2B–D, Dataset EV1). The HF undergoes morphogenesis in utero and after birth (Fig EV1F) (Paus et al, 1999; Saxena et al, 2019). In murine skin, HF morphogenesis is classified into nine stages: the accumulation of nuclei in the epidermis without downward growth of HFs (stage 0), HFs with the most proximal part in the dermis (stages 1–5), and HFs with the most proximal part in the subcutaneous tissue (stages 6–8) (Paus et al, 1999). The downregulation of HF morphogenesis genes (Fig 2B and C) led us to hypothesize that epidermal wounding tunes down tissue development to accelerate blister healing. In agreement with this hypothesis, the number of hair canals, which are tube‐like connections between the epidermal surface and the most distal part of the inner root sheath (IRS) and are present in only developed HFs (HF morphogenesis, stage 6–8) (Paus et al, 1999), was reduced in the regenerated epidermis (Fig 2D and E). HF growth under the regenerated epidermis at P4 was delayed at stages 5 and 6 where the IRS is halfway up to the HF or contains the hair shaft up to the level of the hair canal. In contrast, the surrounding intact skin of the blisters or the normal skin of the littermate controls had stage 7 HFs, in which the tip of the hair shaft leaves the IRS and enters the hair canal (Fig 2F). The observation of smaller HFs in the skin lesion when compared to the surrounding intact skin is accompanied by diminished Wnt signaling indicated by the LacZ‐positive area in Wnt reporter mice (ins‐Topgal⁺) (P2, Fig 2G).

Figure 2. Delayed HF growth during subepidermal blister healing.

1. Heat map (Pearson’s correlation) of differentially expressed genes between the blistered (regenerated) and control WT dorsal skin epidermis at P2 (n = 3).
2. GO analysis of differentially expressed genes in the regenerated epidermis.
3. Scatter plots of differentially expressed genes in the regenerated epidermis. The red dots represent upregulated genes, and the blue dots represent downregulated genes. The gray dotted lines indicate |logFC| > 1.
4. Hair canals in the regenerated (lesional) and nonlesional epidermis at P4 (indicated by asterisks). Scale bar: 300 μm.
5. Quantification of hair canals in the lesional, nonlesional, and unaffected littermate control epidermis at P4 (n = 5 biological replicates). The data are shown as the mean ± SE (littermate control) or connected with lines showing individual mice. *0.01 < P < 0.05, one‐way ANOVA test, followed by Tukey’s test.
6. HF morphogenesis stages at P4 in lesional, nonlesional, and unaffected littermate control skin (n = 5 biological replicates).
7. Whole‐mount imaging of the blistered skin of ins‐Topgal⁺ mice at P2. Scale bar: 500 μm.

Data information: Representative images are shown from three or more replicates in each group.

Figure EV2. RNA‐seq data on subepidermal blister healing.

1. Volcano plot showing differentially expressed genes (DEG) between the blistered (regenerated) and control skin epidermis at P2. Significantly (|LogFC| > 1; FDR < 0.05) upregulated and downregulated DEG are shown in red and blue, respectively.
2. Bar plot summarizing GSEA enrichment results for selected upregulated and downregulated KEGG pathways. The normalized enrichment score (NES), P‐value, and FDR are shown.
3. GSEA enrichment plots of “Wnt signaling pathway”, “Hedgehog signaling pathway” and “Melanogenesis” KEGG gene sets.
4. Network visualization of the top ten downregulated and upregulated (FDR < 0.05) GO term clusters for different GO categories (Biological process, Molecular Function, and Cellular Component). Node size reports the number of enriched genes in each GO term (gene numbers in bottom panels). Nodes are colored as a pie chart depicting the proportion of downregulated (blue) and upregulated (red) genes in each GO term. Edge thickness depicts the number of shared genes between GO terms.

The expression of genes involved in cytokine–cytokine receptor interactions and chemokine, TNF, IL‐17, and JAK‐STAT signaling pathways was increased in the regenerated epidermis (Figs 2B and C, and EV2B), and these pathways are implicated in the recruitment of immune cells. However, the number of neutrophils, lymphocytes, and macrophages was not increased in the lesional dermis (P2, Fig EV1G), in which 1–2 layers of the regenerated epidermis covered the wound 1 day after blistering. There was no apparent increase in these immune cells either at P4 (Fig EV1G), in which the whole epidermis was restored. These results suggest that the immune cells might not play a significant role in blister healing, although the involvement of immune cells or of molecules that they secrete—such as γδ T cell‐derived Fgf9, which induces HF neogenesis (Gay et al, 2013)—cannot be fully excluded or may serve as a confounding factor affecting HF growth.

These data indicate that, in the context of skin morphogenesis where the immune system is not yet fully defined, subepidermal blisters heal at the expense of HF growth.

Progeny of junctional zone SCs represent the main cellular contribution to blister healing

Wounded lesions require epithelial cell proliferation and migration to restore skin integrity. We then investigated the dynamics of epidermal and HF keratinocytes during blister healing. One day after blister formation (P2), BrdU⁺ cells were abundant in the HFs and the intact epidermis adjacent to the blisters (Fig 3A). Cells positive for α5 integrin (ITGA5), a marker of migrating keratinocytes (Aragona et al, 2017), were seen in the HFs within the lesional area and in the epidermal boundary between the blister and the nonlesional area (epidermal tongue) (Figs 3B and EV1E). As HF growth was delayed in the regenerated epidermis (Fig 2D–G) and proliferative cells were abundant in HFs of the lesional area (Fig 3A), HF keratinocytes were deduced to participate in epidermal regeneration rather than in HF development.

Figure 3. Predominant contribution of HF‐derived keratinocytes to subepidermal blister healing.

1. (Top) BrdU labeling of blistered samples at P2. Scale bar: 100 μm. BrdU‐positive cells are indicated by arrows. Blisters are indicated by stars. (Bottom) Quantification of BrdU‐positive cells in the epidermis (left) and HFs (right) (n = 4 biological replicates). The data are shown as the mean ± SE. *0.01 < P < 0.05, one‐way ANOVA test, followed by Tukey’s test. NS, no significance.
2. α5 integrin (ITGA5) labeling at P2 (left image: section, right image: whole‐mount). Scale bar: 100 μm. Blister edges (epidermal tongue) and HFs are indicated by arrowheads and arrows, respectively. Blisters are indicated by stars.
3. Lineage tracing strategy.
4. (Top) Sections of K14CreER:H2B‐mCherry mice at P4. Scale bar: 100 μm. (Bottom) Quantification of mCherry‐positive cells (n = 3). The data from individual mice are connected by lines. Student’s t‐test. NS, no significance.
5. Whole‐mount imaging of K14CreER:R26R‐confetti samples at P4. Scale bar: 200 μm.
6. (Top) Sections of Lrig1CreER:H2B‐mCherry mouse skin at P4. Scale bar: 100 μm. (Bottom) Quantification of mCherry‐positive cells (n = 3). The data from individual mice are connected by lines. *0.01 < P < 0.05, Student’s t‐test.
7. Whole‐mount imaging of Lrig1CreER:R26R‐confetti mouse samples at P4. Scale bar: 200 μm.

Data information: Representative images are shown from three or more replicates in each group.

To confirm this hypothesis, we employed a short‐term lineage tracing strategy with suction blistering (Fig 3C). K14‐lineage‐labeled cells (K14CreER:R26R‐H2B‐mCherry or K14CreER:R26R‐confetti), mainly progeny of SCs in the interfollicular epidermis (IFE), were sparse in the regenerated epidermis (Figs 3D and E, and EV3A). In contrast, most of the cells in the regenerated epidermis were Lrig1 (leucine‐rich repeat and immunoglobulin‐like domain protein 1)‐lineage‐labeled cells (Lrig1CreER:R26R‐H2B‐mCherry or Lrig1CreER:R26R‐confetti), which are the progeny of junctional zone SCs (Figs 3F and G, and EV3A). In line with this, phospho‐Histone H3 (PH3)‐positive cells were observed in Lrig1‐lineage‐labeled cells at P2 (Figs EV3B and 3C). K14‐ and Lrig1‐lineage‐labeled cells were increased from P1 (at the time of suction blistering) to P4 (sampling), but the expansion of Lrig1 lineage‐labeled cells was more evident in the regenerated epidermis (Fig EV3D). The expression of the Lrig1 gene was not upregulated in the regenerated epidermis at P2 in our RNA‐seq. These data indicate that the HF junctional zone on the dermal side of the blister is the main pool for the keratinocytes that heal subepidermal blisters while halting HF development, although other hair follicle populations might also be involved in blister healing.

Figure EV3. Lineage tracing of subepidermal blister healing.

1. Blistered area of K14CreER:R26R‐confetti (left) and Lrig1CreER:R26R‐confetti (right) mouse skin samples at P4. Scale bar: 100 µm.
2. Phospho‐Histone H3 (PH3) staining of blistered skin at P2 (WT). Scale bar: 100 µm.
3. PH3 staining (arrowhead) of blistered skin at P2 (Lrig1CreER:R26R‐H2B‐mCherry). Scale bar: 100 µm.
4. Quantification of mCherry‐positive cells during lineage tracing (n = 3 biological replicates). The data are shown as the mean ± SE.

Data information: Blisters are indicated by stars. Representative images are shown from three or more replicates in each group.

HF reduction from the wound bed of the blisters promotes the contribution of IFESC progeny to blister healing

The contribution of junctional zone HFSC progeny to blister healing led us to investigate how the epidermis regenerates in the absence of HFs at the blister base (dermis). Type XVII collagen (COL17) is expressed not only in the IFE but also in the bulge region of the HFs (Fig 4A) (Tanimura et al, 2011; Matsumura et al, 2016; Watanabe et al, 2017; Liu et al, 2019; Natsuga et al, 2019). COL17 is encoded by the COL17A1 gene, and its deficiency leads to junctional EB (McGrath et al, 1995). The splitting of neonatal Col17a1 ^−/− (Nishie et al, 2007) dorsal skin upon suction blistering was observed between ITGA6 and COL4/L332 (Fig 4B and C), as was the case in wild‐type neonates (Fig 1C and D). Intriguingly, suction blistering (Fig 1A) of Col17a1 ^−/− dorsal skin detached most, but not all, of the HFs from the dermis (P1, Fig 4D). In agreement with this finding, dermal papilla cells (AP⁺) were observed on the roof side of the blisters of Col17a1 ^−/− mice, whereas the blister roofs of control mice did not have these cells (P1, Figs 4E and EV1D). Epidermal regeneration was not apparent in Col17a1 ^−/− mice 1 day after suction blistering (P2), whereas the control mice showed regeneration of the epithelial layers (Fig 4F, Table EV1). Col17a1 ^−/− mice had delayed expression of loricrin in the regenerated epidermis 3 days after blister formation (P4, Fig 4F, Table EV1). BrdU⁺ cells were abundant in the Col17a1 ^−/− mouse epidermis surrounding blisters as was the case for controls (Figs 3A and 4G). Lineage tracing experiments (Fig 3C) revealed that IFESC progeny covered most of the regenerated area (K14CreER:R26R‐H2B‐mCherry:Col17a1 ^−/−) at P4 (Fig 4H and I). The transgenic rescue of Col17a1 ^−/− by overexpressing human COL17 (hCOL17⁺; Col17a1 ^−/−) (Nishie et al, 2007) ameliorated blister healing (P4, Appendix Fig S1, Table EV1). These data demonstrate that upon detachment of most HFs from the dermis, the IFE can compensate for the lack of junctional and HF SCs and repair defects in the IFE.

Figure 4. Effects of HF reduction on subepidermal blister healing.

• A
  Type XVII collagen (COL17, arrowheads indicate the hair bulge) and laminin β1 (LAMB1) labeling in WT dorsal skin sections (P1). Scale bar: 100 μm.
• B, C
  Blistered samples of Col17a1 ^−/− mouse dorsal skin at P1. ITGA6 (indicated by arrowheads) and COL4 (arrows) labeling (B). L332 staining (C, arrows). Scale bar: 100 μm.
• D
  H&E staining of blistered skin from Col17a1 ^−/− mice at P1. HFs detached from the dermis in Col17a1 ^−/− skin are indicated by arrowheads. Scale bar: 500 μm.
• E
  Whole‐mount AP staining of the blister roof epidermis from Col17a1 ^−/− mice (right) and littermate controls (left) at P1. Scale bar: 500 μm.
• F
  H&E (P2, left), ITGA6 (P2, middle), and LOR staining (P4, right) of Col17a1 ^−/− mice (top) and littermate controls (bottom). The regenerated epidermis is indicated by arrowheads. Scale bar: 200 μm.
• G
  (Top) BrdU labeling of Col17a1 ^−/− skin at P2. Scale bar: 100 μm. (Bottom) Quantification of BrdU‐positive cells in the epidermis surrounding blisters (n = 3 (control) and 4 (Col17a1 ^−/−) biological replicates). The data are shown as the mean ± SE. Student’s t‐test. NS, no significance.
• H, I
  (H) Lineage tracing of K14CreER:R26R‐mCherry:Col17a1 ^−/− at P4. Scale bar: 100 μm. (I) Quantification of mCherry‐positive cells in the regenerated epidermis (n = 3 biological replicates). The data from individual mice are connected by lines. *0.01 < P < 0.05, Student’s t‐test.

Data information: Blisters are indicated by stars. Representative images are shown from three or more replicates in each group.

Impaired flattening of regenerated keratinocytes accompanies slower blister healing

We further sought to identify other modulators of subepidermal blister healing. We first focused on collagen VII (COL7), encoded by Col7a1. COL7 forms anchoring fibrils and is located at the DEJ (Fig 5A) but just below the basement membrane (Shimizu et al, 1997; Watanabe et al, 2018), and its deficiency leads to dystrophic EB (Christiano et al, 1993; Hilal et al, 1993). As conventional wound healing is delayed in COL7‐hypomorphic mice (Nystrom et al, 2013), we applied the suction‐blister method to Col7a1 ^−/− mice (Heinonen et al, 1999) (Fig 5B–H, Appendix Fig S2). In contrast to that in WT and Col17a1 ^−/− suction blisters (Figs 1C and D, and 4B and C), skin splitting occurred at the level below the basement membrane in Col7a1 ^−/− mouse dorsal skin, as shown by the presence of L332 and COL4 on the blister roof epidermis (P1, Fig 5B and C). The epidermal defects were not repaired in Col7a1 ^−/− mice, whereas the epidermis of the control blistered skin regenerated 1 day after blistering (P2, Fig 5F, Appendix Fig S2, Table EV1), which is consistent with the delayed healing of full‐thickness skin wounds in COL7‐hypomorphic mice (Nystrom et al, 2013). This finding is contrasted by the fact that COL7‐depleted keratinocytes migrate faster than WT keratinocytes in vitro (Chen et al, 2000; Chen et al, 2002). We examined the HFs of Col7a1 ^−/− mice to explain the slowed epidermal regeneration because HFs are the main contributor to blister healing (Fig 3A–G). However, HFs were present in the Col7a1 ^−/− mouse wound bed (blister base) (P1, Fig 5D and E) as opposed to that of Col17a1 ^−/− mouse (P1, Fig 4D and E). Moreover, the number of BrdU⁺ cells in HFs was comparable between Col7a1 ^−/− and control mice (P2, Fig 5G, Appendix Fig S2), suggesting that the proliferation of HF keratinocytes does not account for the delayed blister healing of Col7a1 ^−/− mice.

Figure 5. Involvement of keratinocyte shape transformation in subepidermal blister healing.

1. Type VII collagen (COL7) labeling in WT dorsal skin sections (P1). Scale bar: 200 μm.
2. ITGA6/COL4 (left, indicated by arrows) and L332 labeling (right, indicated by arrows) in the blistered skin of Col7a1 ^−/− mice at P1. Scale bar: 100 μm.
3. Schematic diagram of control, Col17a1 ^−/−, and Col7a1 ^−/− mouse skin splits. BM: basement membrane.
4. H&E staining of blistered skin from Col7a1 ^−/− mice (right) and their littermate controls (left) at P1. Scale bar: 200 μm.
5. Whole‐mount AP staining of the blister roof epidermis from Col7a1 ^−/− mice (right) and their littermate controls (left) at P1. Scale bar: 500 μm.
6. K10/K14 (low magnification, left) and K14 (high magnification, middle and right) labeling of Col7a1 ^−/− mouse (bottom) and littermate control (top) blistered skin at P2. Scale bar: 30 μm.
7. Quantification of BrdU‐positive cells per μm HF length (n = 55 HFs from three control and 143 HFs from four Col7a1 ^−/− mice). The data are shown as violin plots. Student’s t‐test. NS, no significance.
8. Length of the major axis of keratinocytes in the regenerated epidermis (n = 244 (control, L), and 132 (Col7a1 ^−/−, L) cells from four mice, respectively) and in the surrounding intact epidermis (basal cells; n = 200 (control, NL), 299 (Col7a1 ^−/−, NL), from four mice, respectively). NL: nonlesional area. L: lesional area. The data are shown as violin plots. ****P < 0.0001, one‐way ANOVA test, followed by Tukey’s test. NS, no significance.
9. K10/K14 (left) and K14 (high magnification, right) labeling of WT blistered skin treated with CaCl₂ (middle and bottom) or PBS (top) at P2. Scale bar: 30 μm.
10. Quantification of BrdU‐positive cells per μm HF length (n = 83 (PBS), 95 (1.8 mM CaCl₂), and 97 (9.0 mM CaCl₂) HFs from four mice). One‐way ANOVA test, followed by Tukey’s test. NS, no significance.
11. Length of the major axis of keratinocytes in the regenerated epidermis (n = 433 (PBS, L), 451 (1.8 mM CaCl_2, L), and 425 (9.0 mM CaCl₂, L) cells from four mice) and in the surrounding intact epidermis (basal cells; n = 311 (PBS, NL), 279 (1.8 mM CaCl_2, NL), 302 (9.0 mM CaCl₂, NL) cells from four mice). NL: nonlesional area. L: lesional area. The data are shown as violin plots. ****P < 0.0001, one‐way ANOVA test, followed by Tukey’s test. NS, no significance.

Data information: Blisters are indicated by stars. Representative images are shown from three or more replicates in each group. The dashed and dotted lines in the violin plots show the median and quartiles, respectively.

Second, we treated the blistered skin in wild‐type mice with extracellular calcium (Fig 5I–K, Appendix Fig S2). Extracellular calcium is a potent inhibitor of proliferation and migration in cultured keratinocytes as well as an inducer of differentiation (Hennings et al, 1980; Magee et al, 1987). Consistent with previous in vitro assays, the intrablister administration of CaCl₂ (1.8 or 9.0 mM) just after suction blistering delayed epidermal regeneration in vivo (P2, Fig 5I, Table EV1). Premature differentiation, which might hinder wound healing, was not apparent in the CaCl₂‐treated blisters, as K10 labeling was seen only at the blister roof but not in the keratinocytes on the wound bed (P2, Fig 5I). Similar to that in Col7a1 ^−/− mouse blisters, the number of BrdU⁺ cells in HFs was not reduced in CaCl₂‐treated blisters (P2, Fig 5J, Appendix Fig S2).

These two examples strongly suggest that there are factors other than HF keratinocyte proliferation that modulate blister healing. During blister healing, keratinocytes reshape into a wedge‐shaped morphology (Fig 1F), which is mirrored by the RNA‐seq data showing that the expression of genes involved in the regulation of the actin cytoskeleton is decreased in the regenerated epidermis (Fig 2B and C). Wedge‐shaped/flattened keratinocytes are believed to be superior to cuboidal/columnar keratinocytes for covering epidermal defects. These data led us to wonder whether the cell morphology was altered in settings of delayed blister healing. In the intact (nonblistered) skin, the morphology of Col7a1 ^−/− or Ca‐treated basal keratinocytes was similar to that of control cells (P2, the nonlesional area in Fig 5F, H, I and K). The regenerated keratinocytes became wedge‐shaped/flattened in the control group, as shown in Fig 1F. However, the keratinocytes in the regenerated epidermis of Col7a1 ^−/− or Ca‐treated mice were not as flat as those of control mice but were still rather cuboidal (P2, the lesional area in Fig 5F, H, I and K), which correlates with the delayed blister healing in these mice.

Mathematical modeling reproduces blister healing

These in vivo experiments led us to speculate that HF/IFE cell proliferation during paused HF development and the morphological changes of the regenerated keratinocytes might simply account for the dynamics of subepidermal blister healing. To answer this question, we employed mathematical modeling. We adopted an agent‐based model (Kobayashi et al, 2016; Kobayashi et al, 2018), where keratinocytes were modeled by spheroids and cell division was described as replication of the spheroids. Such a model allowed us to visualize the dynamics of the epidermal basal layer and to establish the epidermal defects on the basement membrane. We utilized the data on the number of BrdU⁺ cells among HF versus IFE keratinocytes (approximately 7:1 per unit of epidermal length; Fig 3A) and on the shape of the regenerated versus normal keratinocytes (2:1 the length of the major cell axis; Fig 5H and K). SC progeny within the epidermal defects (colored in red), simulating HF‐derived cells, had a more substantial contribution to wound healing than IFE SCs (colored in yellow; Fig 6A and B, Movie EV1), as seen in the lineage tracing experiments (Fig 3C–G). The absence of SCs within the epidermal defects, simulating HF reduction in the wound bed, showed delayed healing (Fig 6C and D, Movie EV2), in agreement with Col17a1 ^−/− epidermal regeneration results (Fig 4E–G). A less flattened morphology of the regenerated keratinocytes, simulating Col7a1 ^−/− and Ca‐treated blister healing (regenerated versus normal keratinocytes, 1.3‐1.6:1 in length; Fig 5H and K), slowed epidermal regeneration (Fig 6E and F, Movie EV3). By systematically changing the shape of regenerated versus normal keratinocytes (from 1:1 to 2:1 in length) for different initial SC distributions within epidermal defects (Fig EV4A), we observed the same tendency: Less flattened morphology led to more delayed healing (Fig EV4B and C), which suggests that the morphology affects healing dynamics irrespective of the initial SC distribution. These in silico data demonstrate that the contribution of HFSC progeny and the morphological change in the regenerated keratinocytes is sufficient to recapitulate the in vivo subepidermal blister healing.

Figure 6. Mathematical modeling of subepidermal blister healing.

1. A particle‐based model of subepidermal blister healing at the basal layer. Epidermal basal cells (colored in blue), which do not divide, are placed on the basement membrane (gray). Stem cells (SCs, green) give rise to progeny (simulating HF‐derived cells; red) within epidermal defects or in the surrounding epidermis (IFE‐derived cells; yellow). t: arbitrary time. See Movie EV1.
2. Contribution of each progeny cell within the epidermal defect or of surrounding epidermis to subepidermal blister healing, measured as the ratio of the area occupied by each progeny to the area of the initial epidermal defect.
3. A model of subepidermal blister healing without SCs within epidermal defects. See Movie EV2.
4. Time course of subepidermal blister healing in control (A) and SC‐depleted epidermal defects (C).
5. Effects of the impaired flattening of keratinocytes upon epidermal regeneration. The diameter of basal keratinocytes (long axis of the spheroid) in the regenerated versus surrounding epidermis was calculated as 1.5:1 (in contrast to 2:1 in Fig 6A). See Movie EV3.
6. Time course of wound healing for control (A) and less flattened keratinocytes in the regenerated epidermis (E).

Figure EV4. Dependence of subepidermal blister healing on flattening ratio and SC arrangement.

1. Arrangements of the SCs within epidermal defects (dispersed, 2 cells grouped, 4 cells grouped, and 8 cells grouped) at t = 0.
2. Time course of subepidermal blister healing for different keratinocyte flattening ratios (1.0–2.0), each with four SC arrangements.
3. Time to full recovery (100% healing rate) for different keratinocyte flattening ratios, averaged over four SC arrangements. The data are shown as the mean ± SD.

We finally asked whether this HF contribution to wound healing could be applied to a human setting, in which HFs are larger but much more sparsely distributed than their murine counterparts (Table EV2). We examined human subepidermal blister samples and found that epidermal regeneration by HFs was observed in the samples with the re‐epithelized area (Fig EV5). These findings from human samples are consistent with the mice data.

Figure EV5. Subepidermal blister healing in humans.

H&E staining of human subepidermal blister samples with re‐epithelized areas (blisters 1, 2, and 3). Blisters are indicated by stars. The regenerated epidermis from HFs is indicated by arrows. Scale bar: 300 μm.

Discussion

Although recent studies have reported that injury causes damaged tissue to shift to an embryonic‐like state, there is a poor understanding of how regeneration affects development at the damaged tissues. Here, we applied the blistering injury to neonatal mouse dorsal skin and showed the skewed contribution of HFSC progeny to wound healing rather than HF development.

Previous studies on skin wounding combined with the fate mapping of murine skin delineated the involvement of epithelial, mesenchymal, and immune cells in wound healing, depending on different settings (Rognoni & Watt, 2018; Dekoninck & Blanpain, 2019). However, as full‐thickness skin wounds, even when they are applied to neonatal skin, remove all skin components, it is challenging to see the effects of development on injury or vice versa. Our blistering injury has an advantage over conventional skin wounding studies in that only the epidermis is removed by constant negative pressure, with the other skin components and basement membrane being retained in the wounds, which allowed us to examine “pure” epidermal wound healing processes during skin development. Our study contrasts with wound‐induced embryonic gene expression (Miao et al, 2019) and follicular neogenesis upon full‐thickness skin wounding in adult (Ito et al, 2007; Osaka et al, 2007) and neonatal mice (Rognoni et al, 2016). So far, our study has not distinguished whether the waning Wnt signaling was the primary cause, or the result, of the delayed HF morphogenesis. Further studies are needed to clarify this.

Previous studies have suggested the possible involvement of HFs in epidermal regeneration in human suction blisters (Lane et al, 1991) and extracellular matrix alterations in the skin‐split area (Hertle et al, 1992; Leivo et al, 2000). We show that the progeny of HF junctional zone SCs mainly repair subepidermal blisters (Figs 3F and G, and 6A and B), although the involvement of other HF populations cannot be excluded. However, the IFE can also serve as a reservoir of keratinocytes to repair epidermal defects when most HFs are detached from the dermis (Fig 4H). How the contribution of two sources of keratinocytes to blister healing is regulated is unknown, but the significant contribution of HF junctional zone progeny is reasonable because HFs are densely located in the wound bed (blister base). In contrast, the progeny of IFESCs could recover only from the blister edge, as demonstrated by mathematical modeling (Figs 3D–G and 6A and B). The role of HFSCs is also highlighted by delayed HF growth in the regenerated epidermis, as corroborated by the downregulated Wnt signaling (Fig 2A–G). These findings indicate that there is a coordinated balance between tissue development and wound healing, which has not been well recognized. In addition, the expression of IL‐17 signaling pathway genes was increased, although the recruitment of immune cells was not evident 1 day or 3 days after blistering (Fig EV1G). Recently, IL‐17 signaling has been shown to drive Lrig1‐lineage cell recruitment in wound healing and tumorigenesis (Chen et al, 2019). Therefore, IL‐17 signaling might also help Lrig1‐lineage cells translocate from HFs to repair epidermal defects in our study (Fig 3F and G).

The dynamics of cytoskeletal changes directly affect cellular morphology and migration potential (Tang & Gerlach, 2017). Previous studies have shown that cells undergo a morphological transformation into a wedge/flattened shape at the leading edge of migrating cells (Uroz et al, 2019) and in regenerated keratinocytes during wound healing (Krawczyk, 1971; Paladini et al, 1996). Our study has shed further light on the significant impact of the keratinocyte morphological changes on in vivo wound healing through blistering experiments in Col7a1 ^−/− and Ca‐treated mice and mathematical modeling (Figs 5F, H, I and K, and 6A, E and F). However, the morphological changes of keratinocytes might not directly foster blister healing but might be simply correlated with other primary causes that help regenerate the epidermis. The loss of a functional basement membrane in the Col7a1 ^−/− blister bottom might slow blister healing due to the lack of substrates for cell migration. Pressure during blister induction could be a factor that changes the cellular morphology. Additional mechanistic studies are needed to verify the hypothesis we raise in our study.

Our in vivo blistering experiments can mimic and replace the in vitro cultured cell wound healing assay (e.g., scratch wounding), which has been used in the field of cell biology for decades because subepidermal blisters are epidermal wounds. In vivo intrablister administration of drugs, as exemplified by extracellular calcium administration (Fig 5I), could be an alternative to in vitro chemical treatment of scratch‐wounded cultured cells to develop new therapeutic options for wound healing, especially of subepidermal blisters.

Our in vivo suction‐blister model recapitulates the human pathological epidermal detachment seen in EB, pemphigoid diseases, burns, and severe drug reactions such as Stevens–Johnson syndrome/toxic epidermal necrolysis. Loss‐of‐function mutations in COL17A1 (McGrath et al, 1995) and COL7A1 (Christiano et al, 1993; Hilal et al, 1993) lead to junctional and recessive dystrophic EB in humans, respectively. Therefore, Col17a1 ^−/− and Col7a1 ^−/− mouse blistering also serves as an EB wound model. The prominent hair loss in human COL17A1‐mutated junctional EB might be reflected by the reduction in HFs from the wound bed in Col17a1 ^−/− blisters (Fig 4D and E), whereas the hair loss in recessive dystrophic EB is not as severe as that in the junctional subtype (Tosti et al, 2010), consistent with the maintenance of HFs in the dermis of Col7a1 ^−/− mouse dorsal skin upon suction blistering (Fig 5D and E). It is plausible that recurrent blistering can exhaust the pool of HFSCs, leading to delayed blister healing and scarring, especially in recessive dystrophic EB. Taken together, the processes of subepidermal blister healing highlight HFs as a target for treating the wounds of EB and other blistering diseases.

Our study has some limitations, primarily due to the discrepancies between mice and humans. First, eccrine sweat glands have been reported to contribute to wound healing in human skin (Rittie et al, 2013). As murine back skin does not harbor sweat glands, our study was unable to estimate the contribution of the sweat glands in human settings. Second, although the skin‐split level of human recessive dystrophic EB is generally just beneath lamina densa (basement membrane), as was the case for the Col7a1 ^−/− blisters in our study, suction blistering on human recessive dystrophic EB induces skin detachment in the lamina lucida, that is, between hemidesmosomes and the lamina densa (Tidman & Eady, 1984). In addition, our study did not look into the contribution of mesenchymal cells to blister healing, which has been described in previous studies on epidermolysis bullosa (Chino et al, 2008; Tolar et al, 2009; Fujita et al, 2010; Tamai et al, 2011; Iinuma et al, 2015; Webber et al, 2017). Further studies are warranted to elucidate the role of mesenchymal cells in blister healing.

In closing, our study has revealed the imbalance between development and wound regeneration in the skin blisters. Our findings of the healing processes pave the way for tailored therapeutic interventions for epidermolysis bullosa, pemphigoid diseases, and other blistering diseases.

Materials and Methods

Animals

C57BL/6 strain mice were purchased from Clea (Tokyo, Japan). Ins‐Topgal⁺ mice were obtained from RIKEN BRC (Tsukuba, Japan) (Moriyama et al, 2007). K14CreER, Lrig1CreER, and R26R‐confetti mice were purchased from the Jackson Laboratory (Bar Harbor, Maine, USA). R26R‐H2B‐mCherry mice were provided by RIKEN (Kobe, Japan). Col17a1 ^−/− and hCOL17⁺;Col17a1 ^−/− mice were generated as previously described (Nishie et al, 2007). Col7a1 ^−/− mice were provided by Prof. Jouni Uitto (Heinonen et al, 1999). The institutional review board of the Hokkaido University Graduate School of Medicine approved all animal studies described below.

Suction blisters

Suction blisters were produced on the neonatal murine dorsal skin (P1) using a syringe and connector tubes. The negative pressure applied to the skin (generally for minutes) was 523.4 ± 1.3 mmHg (evaluated by an Ex Pocket Pressure Indicator PM‐281 (AS ONE, Osaka, Japan)). The diameter of the syringe attached to the skin was 4 mm. The size of the typical blister was 3 mm in diameter.

Histology

Mouse dorsal skin specimens were fixed in formalin and embedded in paraffin after dehydration or were frozen on dry ice in an optimal cutting temperature (OCT) compound. Frozen sections were fixed with 4% paraformaldehyde (PFA) or cold acetone or were stained without fixation. Antigen retrieval with pH 6.0 (citrate) or pH 9.0 (EDTA) buffer was performed on deparaffinized sections. Sections were incubated with primary antibodies overnight at 4°C. After being washed in phosphate‐buffered saline (PBS), the sections were incubated with secondary antibodies conjugated to FITC, Alexa 488, Alexa 647, or Alexa 680 for 1 h at room temperature (RT). The nuclei were stained with propidium iodide (PI) or 4′,6‐diamidino‐2‐phenylindole (DAPI). The stained immunofluorescent samples were observed using a confocal laser scanning microscope (FV‐1000 (Olympus, Tokyo, Japan) or LSM‐710 (Zeiss, Oberkochen, Germany)).

For immunohistochemistry, horseradish peroxidase (HRP)‐tagged secondary antibodies were used. Sections were blocked with hydrogen peroxide, labeled with antibodies, and counterstained with hematoxylin. For morphological analysis, deparaffinized sections were stained with hematoxylin and eosin (H&E) by conventional methods. Alkaline phosphatase staining was performed using a StemAb Alkaline Phosphatase Staining Kit II (Stemgent, San Diego, California, USA). Images of immunohistochemistry, and H&E‐ and alkaline phosphatase‐stained sections were captured with a BZ‐9000 microscope (Keyence, Tokyo, Japan).

For whole‐mount staining, mouse dorsal skin samples were fixed with 4% PFA and immunolabeled or stained with the Alkaline Phosphatase Staining Kit II. For X‐gal staining of ins‐Topgal⁺ mouse skin, a beta‐galactosidase staining kit (Takara‐bio, Shiga, Japan) was used according to the provider’s protocol. Briefly, dorsal skin samples were fixed with 4% PFA for 1 h at 4°C and soaked in staining solution overnight at RT. Tissues were mounted in a Mowiol solution. Images were observed with LSM‐710, FV‐1000, or BZ‐9000 microscopes.

HF morphological stages were evaluated as previously described (Paus et al, 1999). The length of the major axis of keratinocytes in the intact and regenerated epidermis was measured using ImageJ (NIH, Bethesda, Maryland, USA) on K14‐stained sections. The quantification of the cells expressing a particular marker was performed as previously described (Natsuga et al, 2016).

Antibodies

The following antibodies were used: anti‐BrdU (Abcam, Cambridge, UK; BU1/75, Dako; M0744), anti‐phospho‐Histone H3 (Ser10) (Merck Millipore, Billerica, Massachusetts, USA), anti‐loricrin (Covance, Princeton, New Jersey, USA), FITC‐conjugated anti‐CD3e (BioLegend, San Diego, California, USA; 145‐2C11), Alexa Fluor 488‐conjugated anti‐F4/80 (Affymetrix, Santa Clara, California, USA; BM8), FITC (fluorescein isothiocyanate)‐conjugated anti‐Ly‐6G (Beckman Coulter, Brea, California, USA; RB6‐8C5), anti‐COL4 (Novus Biologicals, Centennial, Colorado; NB120‐6586), anti‐COL7 (homemade (Iwata et al, 2013)), anti‐COL17 (Abcam; ab186415), anti‐ITGA5 (Abcam; EPR7854), anti‐ITGA6 (BD Biosciences Pharmingen, San Diego, California, USA; GoH3), anti‐L332 (Abcam; ab14509), anti‐laminin β1 (Abcam; ab44941), anti‐pan‐cytokeratin (PROGEN, Wieblingen, Heidelberg, Germany; PRGN‐10550), anti‐cytokeratin 10 (Biolegend; Poly19054), and anti‐cytokeratin 14 (Thermo Fisher, Waltham, Massachusetts, USA; LL002).

BrdU labeling

For proliferation analysis, 10 μg of BrdU (BD Biosciences Pharmingen) per head was intraperitoneally administered 4 h before sacrifice.

Transmission electron microscopy

The samples were taken from C57BL/6 mouse dorsal skin (P1) just after suction blistering was performed. The samples were fixed in 5% glutaraldehyde solution, postfixed in 1% OsO₄, dehydrated, and embedded in Epon 812. The embedded samples were sectioned at 1 µm thickness for light microscopy and thin‐sectioned for electron microscopy (70 nm thick). The thin sections were stained with uranyl acetate and lead citrate and examined by transmission electron microscopy (H‐7100; Hitachi, Tokyo, Japan).

Lineage tracing

K14CreER:R26R‐H2B‐mCherry, Lrig1CreER:R26R‐H2B‐mCherry, K14CreER:R26R‐confetti, and Lrig1CreER:R26R‐confetti mice were intraperitoneally treated with 0.5 mg of tamoxifen (T5648; Sigma‐Aldrich, St. Louis, Missouri, USA) at P0. The dorsal skin samples were harvested 4 days later (P4).

RNA sequencing and analysis

Suction‐blistered and control samples were collected at P2 at the same time of day to exclude the effects of circadian oscillations on epidermal gene expression (Janich et al, 2013). The skin samples were treated with 0.25% trypsin EDTA overnight at 4°C. The blistered and regenerated epidermis was collected by separating it from the dermis, minced with a scalpel, and suspended in 10% FCS DMEM. The cell suspension was filtered through a 70 µm filter, and cell pellets were collected. Library preparation was performed using an Illumina TruSeq RNA prep kit by following the manufacturer’s instructions. Briefly, following TRIzol extraction and chemical fragmentation, mRNA was purified with oligo‐dT‐attached magnetic beads and reverse transcribed into cDNA. Following a second strand synthesis step with DNA polymerase I and RNAse H, the resulting cDNA was subjected to end repair, A‐tailing, and Illumina compatible adaptor ligation. Following purification and PCR‐mediated enrichment, libraries were purified with AMPure XP beads and sequenced on a NextSeq 500 Illumina sequencer.

After quality controls were performed, the raw reads were aligned to the NCBIm37 mouse reference genome (mm9) using HiSat2 (Kim et al, 2015) (version 2.0.0) using options ‐N 1 ‐L 20 ‐i S, 1, 0.5 ‐D 25 ‐R 5 ‐‐pen‐noncansplice 20 ‐‐mp 1, 0 ‐‐sp 3, 0 and providing a list of known splice sites. Expression levels were quantified using featureCounts (Liao et al, 2014) with RefSeq gene annotation and normalized as TPM using custom scripts. Differential expression analysis was performed using the edgeR (Robinson et al, 2010) software package. After lowly expressed genes (1 count per million in less than two samples) were filtered out, the treatment and control groups were compared using the exact test method (Robinson et al, 2010). Genes with an absolute log2‐fold change greater than 1 and false discovery rate (FDR) less than or equal to 0.05 were considered differentially expressed. Hierarchical clustering of gene expression profiles was performed on differentially expressed genes using only Euclidean distances and the complete linkage method. TPM values were normalized as Z‐scores across samples, and the distances were computed. GO term, KEGG pathway enrichment analysis on differentially expressed genes, and GO term network visualization were performed using the clusterProfiler R/Bioconductor package (Yu et al, 2012). To validate the enriched pathways, GSEA analysis was performed.

Intrablister administration

Ten microliters of 1.8 or 9.0 mM CaCl₂ in PBS was administered by syringe into the blisters just after the suction blistering procedure was performed.

Statistics

Statistical analyses were performed using GraphPad Prism (GraphPad Software, La Jolla, California, USA). P‐values were determined using Student’s t‐test or one‐way ANOVA followed by Tukey’s test. P‐values are indicated as *0.01 < P < 0.05, **0.001 < P < 0.01, ***0.0001 < P < 0.001, and ****P < 0.0001. The values were shown as the means ± standard errors (SE), violin plots or connected with lines showing individual mice.

Mathematical modeling

A mathematical model proposed for epidermal cell dynamics (Kobayashi et al, 2016; Kobayashi et al, 2018) was adapted to simulate epidermal wound healing. (For the detailed mathematical formulation, see the Appendix Supplementary Methods and Dataset EV2.) In this model, epidermal basal cells were represented as spherical particles, with the cell diameter set to 10 µm. Cells designated as epidermal SCs and their progeny could undergo division on the basement membrane. Cell division was described as a process of two initially completely overlapping particles gradually separating into two distinct particles. When a newly created cell was not fully surrounded by other cells, it was judged as being in a regeneration process and immediately underwent a transition to an oblate spheroid shape with the long axis increased by a factor of 2 (normal) or 1.5 (simulating Col7a1 ^−/− and Ca‐treated blisters) while its volume was kept constant. The same division rate was assigned to all proliferative cells, with an average division period of 57.6 [arb. unit]. Forces exerted on a cell came from adhesion and excluded‐volume interactions with other cells and with the basement membrane. SCs were tightly bound and unable to detach from the basement membrane, while the progeny were weakly bound so that they could detach from the membrane via the ambient pressure: The detached cells were removed from the system. The basement membrane was assumed to be a rigid flat surface, whose shape remained unchanged over time. These interactions were calculated to obtain the time evolution of the whole system by solving equations given in a previous report (Kobayashi et al, 2018). The simulation region was set to 600 × 600 µm horizontally with periodic boundary conditions. To prepare the initial conditions for the simulation of subepidermal blister healing, we first ran a simulation with SCs placed on the basement membrane until their progeny covered the whole surface. Then, we set the progeny to be nonproliferative and created epidermal defects by removing the cells that were inside a disk domain with a diameter of 480 µm.

Human samples

From the H&E‐stained skin of patients with congenital or autoimmune subepidermal blistering diseases (73 EB or 188 bullous pemphigoid (BP) samples, respectively), the samples that met the following histological criteria were selected: (i) subepidermal blisters or a skin split at the dermoepidermal junction; (ii) re‐epithelization in the area; and (iii) presence of HFs on the blister base. Three BP samples fulfilled all the criteria (blisters 1, 2, and 3) and were observed with a Keyence BZ‐9000 microscope. The institutional review board of the Hokkaido University Graduate School of Medicine approved all human studies described above (ID: 13‐043 and 15‐052). The study was conducted according to the Declaration of Helsinki Principles. Participants or their legal guardians provided written informed consent.

Author contributions

Experiment design, data analysis, result interpretation, and manuscript writing: YF; Experiments and data analysis: MW, ST, HN, HK, YW, YM, AL, VP, HU, HI, WN, and SO; Mathematical modeling and manuscript writing: KO, YK, and MN; Experiment design, data analysis, result interpretation, and manuscript writing: GD; Result interpretation and study supervision: HS; Experiment conception and design, data analysis, result interpretation, manuscript writing, and study supervision: KN.

Conflict of interest

The authors declare that they have no conflict of interest.

Supporting information

Appendix

Expanded View Figures PDF

Table EV1

Table EV2

Dataset EV1

Dataset EV2

Movie EV1

Movie EV2

Movie EV3

Review Process File

EMBO reports (2021) 22: e50882.

Data availability

The datasets produced in this study are available in the following databases:

• RNA‐Seq data: Gene Expression Omnibus GSE154871 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE154871)