Large cutaneous ulcers are, in severe cases, life threatening1,2. As the global population ages, non-healing ulcers are becoming increasingly common1,2. Treatment currently requires the transplantation of pre-existing epithelial components, such as skin grafts, or therapy using cultured cells2. Here we develop alternative supplies of epidermal coverage for the treatment of these kinds of wounds. We generated expandable epithelial tissues using in vivo reprogramming of wound-resident mesenchymal cells. Transduction of four transcription factors that specify the skin-cell lineage enabled efficient and rapid de novo epithelialization from the surface of cutaneous ulcers in mice. Our findings may provide a new therapeutic avenue for treating skin wounds and could be extended to other disease situations in which tissue homeostasis and repair are impaired.
The epidermis is the outermost layer of the body, helping to maintain organismal homeostasis by protecting against environmental insults and preventing water loss. The multi-layered epidermis is maintained by stem and progenitor cells within the basal layer1 (that is, basal keratinocytes). An important step in repairing damaged skin is the migration of keratinocytes from adjacent epidermis into the wound to promote re-epithelialization3. For large wounds, this process is inefficient. To improve patient outcomes, more rapid and efficient methods for regenerating epidermal coverage must be developed, preferably non-surgical interventions. After recent advances in cellular reprogramming4,5, we reasoned that wound-resident cells could be reprogrammed towards an epidermal progenitor cell fate to generate new epithelial cells, thereby promoting de novo epithelialization from the surface of a cutaneous ulcer.
We initially sought to reprogram mesenchymal cells (human dermal fibroblasts (hDFs) and adipose-derived stromal cells (hADSCs)), as they participate in wound healing3,6. By comparing gene-expression profiles of human keratinocytes and primary hDFs (Extended Data Fig. 1a, b), upstream-promoter analyses (Supplementary Tables 1–3) and gene-expression reversal analysis (Extended Data Fig. 1c, Supplementary Table 4), we identified 55 transcription factors and 31 microRNAs that are potentially involved in keratinocyte specification (Supplementary Table 5). These candidates were further tested by assessing expression levels after calcium-induced differentiation of keratinocytes (Extended Data Fig. 1d), and by transducing each candidate into hDFs and measuring the levels of the keratinocyte markers, keratin 14 (KRT14) and cadherin 1 (Extended Data Fig. 1e). Selected factors were then co-transduced (via lentivirus) in different combinations into hDFs and the generation of keratinocyte-like cells was assessed (Extended Data Fig. 1f–k, Supplementary Table 6). A combination of 28 transcription factors generated keratinocyte-like cells in vitro. When these cells were grown in 3D organotypic culture, they formed a stratified epithelium (Extended Data Fig. 1l). We therefore named these cells 28TF-induced stratified epithelial progenitors (28TF-iSEPs).
To exclude the possibility that keratinocytes were contaminating the cell-isolation process, we switched to working with hADSCs. To improve transduction efficiency and lower cytotoxicity, we switched to retroviruses (Extended Data Fig. 1n, o). Using this system, we systematically eliminated redundant factors by: (1) excluding factors not integrated into the genome of 28TF-iSEPs (Extended Data Fig. 1m); (2) removing one factor at a time4; and (3) measuring keratinocyte phenotypes (Supplementary Table 7). Transduction of DNP63A (deltaNp63alpha, an isoform of TP63) and GRHL2 reprogrammed hADSCs into cells similar to 28TF-iSEPs (Extended Data Fig. 1p–s). An additional round of screening revealed that: (1) MYC (also known as c-MYC) enhanced reprogramming efficiency, cell proliferation and epithelial stratification, and (2) TFAP2A quickened the emergence of colonies (Supplementary Table 8, Extended Data Fig. 2a–m). Thus, the optimal combination of reprogramming factors was DNP63A, GRHL2, TFAP2A and MYC (DGTM factors). It is important to note that these in vitro-generated epithelia lacked cornification and expressed keratin 13 in the suprabasal layer (characteristic of mucosal epithelium7, foreskin epidermis8, and hyper-proliferative skin9), not keratin 10 (characteristic of adult skin epidermis7) (Extended Data Fig. 2h, m). As epithelia are potently influenced by their niche10, we hypothesized that applying DGTM factors in vivo would more effectively generate skin-like epithelial tissue.
To investigate whether in vivo reprogramming via DGTM factors could induce de novo generation of epithelial tissue from the surface of a skin ulcer, we developed an assay of isolated skin ulcers that simulated the central portion of a large cutaneous ulcer. We surgically removed skin from the back of mice to generate an ulcer and isolated the resulting wound from the surrounding skin using a skin chamber sutured to the deep fascia. This serves to prevent migration of keratinocytes into the wound, and closure of the wound by contraction (Fig. 1a). The absence of epithelial cells within these wounds was confirmed using Krt14cre;Rosa-CAG-loxP-stop-loxP (LSL)-tdTomato (hereafter Krt14cre;LSLtdTomato) mice, in which all cells that express Krt14 sometime during their lifetime are labelled (Extended Data Fig. 3a–h). In the absence of treatment, isolated wounds did not re-epithelialize (Fig. 1b). To test the DGTM factors in vivo, we switched to adeno-associated viruses (AAVs). As AAV cell tropisms differ between serotypes, we determined which AAV could be effective by subcutaneously injecting or directly applying green fluorescent protein (GFP)-expressing AAVs (GFP-AAV) to isolated ulcers. AAVDJ (a serotype of AAV capsid generated by DNA family shuffling11) resulted in the highest levels of GFP (Extended Data Fig. 3i–l). Furthermore, we assessed AAVDJ tissue organ distribution via injection of luciferase-AAVDJs into tail veins (that is, systemically), subcutaneous injection, or inoculation into the wound chamber. Regardless of the delivery method, luciferase expression distal to the injections site was primarily confined to the liver (Extended Data Fig. 3m).
We next administered DGTM-AAVs (DNP63A-AAV, GRHL2-AAV, TFAP2A-AAV, and MYC-AAV) in our in vivo ulcer assay. Within 18 days, we observed epithelia-like tissue inside the chamber (Fig. 1c). Histological analysis of day-18 samples resulting from different titres of AAV (Fig. 1d, Extended Data Fig. 4a), revealed that 5.0 × 1011 gene copies of each factor was the minimal titre needed to efficiently generate epithelial tissue. This titre was used for subsequent in vivo experiments. Twenty-eight days after transducing isolated wounds with DGTM-AAVs, de novo epithelial tissues were histologically very similar to the skin adjacent to the wound edge (Fig. 1e).
To determine the proportion of transduced cells contributing to the healing process, we repeated these experiments using Pdgfracre;R26Rconfetti mice12, in which mesenchymal cells are randomly labelled with GFP, YFP, or RFP, although not all mesenchymal cells were labelled. Each epithelial cluster was derived from a single mesenchymal cell and as these individual clones grew, they intermingled with one another to form a single epithelium. The efficiency of in vivo reprogramming was estimated to be approximately 0.1% (Extended Data Fig. 4b–i).
To assess the role of each DGTM factor in reprogramming mesenchymal cells to epithelial cells, we tested different combinations of DGTM AAVs (for example, D, G, T, but not M). First, a lineage tracing system was introduced using PdgfracreER;LSLtdTomato mice6. Preoperative administration of tamoxifen labelled a broad spectrum of mesenchymal cells (but not epithelial cells) with tdTomato (Fig. 2a). Different combinations of DGTM AAVs were tested in vitro using Pdgfra+ mouse ADSCs (mADSCs) sorted from PdgfracreER;LSLtdTomato mice on the basis of tdTomato fluorescence (Extended Data Fig. 5a–c). After confirmation of in vitro reprogramming of mouse mesenchymal cells towards iSEPs with DGTM-AAVs (Extended Data Fig. 5d–j), different combinations of DGTM factors were tested at relatively lower titre, and the number of epithelial-like colonies counted on day 14 (Extended Data Fig. 5k–n). Epithelial-like colonies arose for all combinations that contained DNP63A (Fig. 2b). All DNP63A-containing combinations resulted in cells similar to primary keratinocytes (Extended Data Fig. 5o, p). Cumulative evidence suggests that the clonogenicity of cultured keratinocytes reflects their stemness, and thus their potential for repairing and maintaining epidermal tissue13,14. We therefore assessed the clonogenicity of epithelial cells generated using each of the DNP63A-containing combinations. On the basis of the expansion of single-cell clones (Extended Data Fig. 5q), clonogenicity was enhanced as the number of transduced factors increased (Fig. 2c). Next, we analysed the efficacy of the eight DNP63A-containing combinations (as well as GTM) in generating epithelial tissue in vivo using isolated skin ulcers, analysing the frequency and size of generated epithelia on day 28. These analyses also suggested an indispensable role for DNP63A in reprogramming (Fig. 2d–f, Extended Data Fig. 5r, s). For both frequency and size of the generated epithelia, GTM factors worked collaboratively with DNP63A to generate epithelial tissues. Notably, the oncogene MYC was dispensable for the de novo generation of epithelial tissues. Considering the in vitro emergence of epithelial cells and clonal ability, in vivo ulcer conditions (such as inflammation) may further constrain the generation and survival of epithelial tissue, thus leaving a relatively limited number of factor combinations that can effectively generate epithelial tissues in vivo.
To determine whether DGTM-generated epithelia are maintained over extended periods of time and whether it could promote the healing of large wounds, we subjected mice to multiple rounds of wounding. We used PdgfracreER;LSLtdTomato mice for these tests to ensure that DGTM-generated epithelia originated solely from mesenchymal cells (Fig. 3a, Extended Data Fig. 6a–e). After the generation of epithelial tissue within the isolated ulcer, the skin chamber was replaced with a larger one. Generated tissues laterally expanded within the large chamber (Extended Data Fig. 6f–h). After the large chamber was removed, generated tissue successfully connected to the surrounding epidermis while retaining a stratified epithelial structure (Fig. 3b, Extended Data Fig. 6i). As AAVDJ showed a specific tropism towards liver, we investigated possible histological alterations of the liver, complete blood count, and blood chemistry 3 months after AAV administration, with no evidence of pathology (Extended Data Fig. 6j–l).
To further assess the regenerative potential of DGTM-generated epithelial tissue, we generated ulcers within a large chamber, but left a small patch of skin intact in the centre of the ulcer. The ability of this skin island to expand and heal the ulcer could then be tested. We tested four types of islands: (1) intact skin, (2) epithelialized skin, (3) cell sheets and (4) DGTM-generated epithelia (Extended Data Fig. 6m–q). On average, epithelialization kinetics of intact skin, epithelialized skin, and DGTM-generated epithelia were similar, whereas the cell sheet group exhibited early contraction of the skin island and delayed epithelialization (Extended Data Fig. 6r, s).
To compare the histological properties of the generated epithelia and differentiated epidermis, samples were collected at day 18, days 28–30 and days 90–110 and analysed via immunohistochemistry. Similar to normal skin, epithelial tissues generated in vivo formed a cornifying envelope and expressed loricrin in the most superficial layers (Fig. 4a). Regarding keratins expressed in the suprabasal layer, generated epithelia could be classified into three types: keratin-10 (K10)+keratin-13 (K13)−, K10+K13+ and K10−K13+ (Fig. 4a, Extended Data Fig. 6t). K10 is expressed in the natural skin epidermis, whereas K13 is expressed in mucosal epithelium7 (Fig. 4a) and hyperproliferative epidermis (for example, fetal skin9) (Extended Data Fig. 6u). Epithelial tissue that first emerged had sporadic K13 expression with or without K10, but this progressively shifted towards a K13−K10+ pattern, thereby moving closer to the keratin pattern of natural skin epidermis (Fig. 4b). Thus, if allowed to mature for a sufficient amount of time within the wound niche, DGTM-generated epithelia exhibited histological characteristics of normal skin.
One of the most important functions of the skin is to serve as a barrier against environmental insults. To assess the outside–in barrier function of DGTM-generated epithelia, a toluidine blue dye penetration assay15 was performed. On day 28, samples including the ulcer, DGTM-generated epithelia and surrounding skin were dissected and their external surfaces were exposed to toluidine blue dye. Similar to the surrounding skin, generated epithelial tissue effectively blocked dye penetration (Fig. 4c). To assess dye penetration after complete epithelialization, we used day-40 PdgfracreER;LSLtdTomato mice in which the skin chamber was removed on day 30. The generated epithelia and surrounding skin exhibited similar abilities to block penetration of lucifer yellow (after 1 h immersion)15 (Fig. 4d). To measure inside–out barrier function, transepidermal water loss (TEWL)15 was compared between generated epithelium (days 90–110), intact skin and a freshly created ulcer on the same animal. DGTM-generated tissues showed TEWL values equivalent to those of intact skin (Fig. 4e).
To assess the clinical relevance of this technique, we characterized the effect of DGTM-AAV on vascularity of the wound bed in our chamber model. DGTM-AAV showed no significant influence on vascularity of the wound bed (Extended Data Fig. 7a–d). Next, we investigated whether in vivo reprogramming could be used to heal an older wound. DGTM-AAVs were applied 7 days after creating the ulcer, and generation of an epithelium was observed in all ten animals tested (18 days after DGTM-AAV application). Notably, generated epithelia tended to be larger than seen with fresh ulcers (compare Extended Data Fig. 7e–g and Extended Data Fig. 4a). To improve the clinical relevance of this potential therapy, we sought to enhance the AAV delivery by administering AAVs via collagen gel. A GFP-AAV solution was mixed with an equal amount of collagen gel and then applied to ulcers in a chamber. The collagen gel increased both GFP expression and residual AAV copy number in ulcer tissue, and reduced AAV copy number in the liver (Extended Data Fig. 7h–k). Thus, collagen gel increased both the potency and specificity of the AAV system.
We reasoned that DGTM-AAV effectiveness could be further improved by adding bioactive molecules, such as FGF216,17 and/or a Rock inhibitor18. To test this, DGTM-AAVs in a collagen gel were applied to an isolated wound. Subsequently, from day 4, wounds were treated with Rock inhibitor and/or FGF2 until day 17. When applied alone, Rock inhibitor and FGF2 both enhanced healing. For wounds treated with both Rock inhibitor and FGF2, the wound surface was almost completely epithelialized within 2–2.5 weeks (Extended Data Fig. 7l–m). Thus, these two factors, in combination with a collagen scaffold, greatly enhanced in vivo reprogramming efficiency, resulting in concomitant de novo epithelialization from different regions within the ulcer leading ultimately to rapid wound healing.
To characterize in vivo reprogrammed cells, iSEPs were sorted from generated epithelia on the basis of tdTomato fluorescence 1, 3 and 6 months after AAV-DGTM administration. To investigate stemness13,14, clonogenic abilities of 3- and 6-month in vivo iSEPs were compared with primary keratinocytes from adult animals and one-day-old pups. In vivo iSEPs consistently showed higher clonogenic abilities than primary keratinocytes (Extended Data Fig. 8a–d). To determine whether this high clonogenic potential resulted from malignant transformation, we cultured in vivo iSEPs in soft agar and confirmed that they lacked anchorage-independent proliferative capacity (Extended Data Fig. 8e–f). To reveal their kinetics in vivo, 3-month in vivo iSEPs were subcutaneously transplanted into immunodeficient mice. For comparison, we also transplanted mouse embryonic stem cells (mES cells) that ubiquitously expressed GFP, HeLa cells that could be detected with a human-specific antibody, and mouse primary keratinocytes from KRT14cre;LSLtdTomato mice. In vivo iSEPs formed smooth, soft, round nodules that were larger than those formed by primary keratinocytes. Histologically, in vivo iSEPs nodules consisted of a large epithelial cyst and a small number of scattered cells within the transplant capsule, consistent with the histological characteristic of primary keratinocytes. Teratogenecity, as observed for mES cells, and transfascial invasive properties, as observed for HeLa cells, were not detected for in vivo iSEPs (Extended Data Fig. 8g–n). As a final test, we generated iSEPs that expressed luciferase and subcutaneously transplantated them into mice. Fifty-six days later, the iSEPs had remained in place, revealing low motility (Extended Data Fig. 8o–q). In summary, in vivo iSEPs formed large epithelial cysts without signs of malignant transformation. They instead behaved like non-malignant epithelial cells with a high proliferative potential.
To molecularly characterize DGTM-reprogrammed cells, RNA sequencing (RNA-seq) analysis was performed on: (1) in vivo iSEPs, (2) in vitro iSEPs (iSEPs generated from Pdgfra+ ADSCs with DGTM-AAV in vitro), (3) Pdgfra+ ADSCs (that is, mesenchymal cells), and (4) primary keratinocytes. When using AAV technologies in dividing cells, the typical goal is to transiently express a transgene in the targeted cells, but AAVs often integrate into the genome19. As our DGTM-AAVs express human versions of the DGTM factors, we could compare levels of AAV-derived gene expression (exogenous/human) to the endogenous/mouse genes. In vitro and in vivo iSEPs had high levels of human DNP63A expression and variable levels of human-GTM factors (Extended Data Fig. 9a). Levels of endogenous Trp63 expression were almost completely suppressed in iSEPs, whereas levels of endogenous GTM factors were similar to those of primary keratinocytes. Consistent expression of exogenous factors in iSEPs at different time points suggested the genomic integration of AAV-derived genes.
Clustering analysis of these RNA-seq data indicated that in vivo iSEPs, in vitro iSEPs, and keratinocytes were similar to one another (for example, they expressed high levels of keratinocyte marker genes), but very distinct from Pdgfra+ ADSCs. Transcriptomic analysis revealed that in vivo iSEPs were more similar to primary keratinocytes than in vitro iSEPs. The top gene ontology terms associated with both primary keratinocytes and in vivo iSEPs involved developmental and/or differentiation processes (Extended Data Fig. 9b–h). We compared a time course of gene expression changes for in vivo iSEPs to Pdgfra+ ADSCs and identified five clusters of genes (Extended Data Fig. 9i–l). The time course of gene expression changes may be influenced by the selective survival of clones with stem-like properties, diminishing effects of the wounding itself and resulting inflammation, and epithelial–mesenchymal interactions8,10,13.
In summary, reprogramming of wound-resident mesenchymal cells enables all regions of the wound to re-epithelialize, relieves the spatial constraints observed during normal healing (Extended Data Fig. 10) and leads to a regenerative functional response in the endogenous skin. Before clinical applications of this potentially transformative, non-surgical technology can be realized, further improvements must be made. These include improving reprogramming efficiency for more prompt epithelialization, and optimizing gene delivery methods to target specific human cell types20. Long-term safety studies must also be performed to evaluate off-target effects and tumorigenic potential, especially becuase DNP63A is a potential oncogene21. The system must also be tested using more clinically relevant animal models of injury. For example, preexisting immunity to AAV reduces transduction efficiencies in humans22. Thus, the inflammatory status of the wound, or other clinical features of the patient may affect reprogramming. Finally, a mechanistic understanding of adult stem cell dynamics and plasticity during wound repair is needed, including the extent of dedifferentiation and the inherent heterogeneity of the reprogrammed cells. Our observations constitute an initial proof of principle for functional in vivo regeneration of not only specific and individual cell types, but also of a three dimensional functional tissue in mice. This knowledge might not only be useful for enhancing skin repair, but could also serve to guide in vivo regenerative strategies in other human pathological situations in which tissue or organ homeostasis and repair are impaired.
Human iSEPs generation
Dermal fibroblasts or adipose-derived stromal cells were seeded at 3,000–5,000 cells/cm2 in 6-, 12- and 24-well cell culture plates. The next day, retroviral supernatant of DGBM factors was obtained and mixed with complete DMEM medium in a ratio of 4:1 v/v. Polybrene (Sigma) was added at a final concentration of 8 μg/ml. The plate was centrifuged twice at 800g for 30 min at 33 °C, then washed with PBS and changed to fresh medium. The medium was changed on days 1, 2 and 4. Cells were reseeded onto mitomycin C-treated 3T3-J2 feeder cells with F medium containing Y27632 as well as vitamin C at a final concentration of 50 μg/ml (Fisher). The medium was changed daily. Typically, epithelial shaped colonies could be identified after 7 to 9 days. When the colonies were ~1,000 cells23, the cells were passaged by removing the feeder cells and non-converted cells with an initial 1 to 2 min of trypsinization and PBS wash, and a further 5 to 10 min of trypsinization to dissociate the cells within the clones. Passaged cells were maintained in the same way as primary keratinocytes.
Mouse in vitro iSEPs generation
Adipose-derived stromal cells were seeded at 20,000 cells per well in 24-well culture plates. The next day, AAVs were mixed with complete DMEM medium. The medium was changed on days 1, 2 and 4. The medium was changed to F medium containing Y27632 and vitamin C from day 4 or 5. The medium was changed daily. After the emergence of colonies, cells were treated with the same protocols as human iSEPs.
Organotypic cultures were prepared using a previously reported method with some modifications10. Epithelial cells were cultured in 3D, at the air–liquid interface, on top of a dermal equivalent. Dermal equivalents were constructed by casting 1.0 × 106 BJ cells (ATCC CRL-2522) per 6 ml pyruvate (-) complete DMEM medium and 3 ml of bovine dermis-derived atelocollagen, I-PC50 (KOKEN) into a 60-mm Petri dish. The solution was allowed to form a gel and contract for 7 days. Final concentrations of collagen and fibroblasts were 1.7 mg/ml and 1.2 × 105 cells/ml, respectively. After loading the dermal equivalents onto an aluminium mount using nylon mesh, an acrylic ring (10 mm in diameter) was set on top and the cells were seeded at more than 4 × 105 cells/cm2 inside the ring with Y27632 containing F-medium. Skin equivalent cultures were maintained in a 60-mm Petri dish or in one well of a six-well plate. For the first 6 days, the concentration of Ca2+ inside and outside of the ring was gradually increased to 1.8 mM. On day 7, the volume of the medium inside the ring was reduced to the level of the epithelial cell sheet, so that the epithelial cells were grown at the air–liquid interface. The specimens were harvested on day 14 for histological and immunohistochemical examinations. The schedule for changing the medium during 3D organotypic culture is summarized in Supplementary Table 9.
The clonogenic ability of primary keratinocytes and epithelial cells generated using DGTM-AAVs was assessed by measuring the proliferative ability of single cell clones. First, the cells were inoculated in 96-well plates at a density of 1 cell per well after being passed through a 70-μm filter. For primary keratinocytes, 3–4 plates were required to obtain around 10 expandable clones, whereas for iSEPs, 1 plate is sufficient. Seven days later, only the wells containing a visually identifiable single colony were passaged to 12-well plates. Seven days later, the plates were stained with rhodanile blue (Alfa Aesar), photographed using a Gel Doc XR+ Imager (Bio-Rad), and analysed for the percentage areas of colonies for each well using ImageJ and Photoshop (Adobe).
Soft-agar colony formation assay
Soft-agar colony formation assay was performed as described elsewhere24,25. In brief, 2 ml of 0.5% base layer SeaPlaque agarose (Lonza) in DMEM containing 10% FBS, Keratinocyte feeder medium (KFM) or 3T3-J2 cell-conditioned keratinocyte feeder medium (CKFM) was plated in each well of six-well plates, and 5,000 cells suspended in 1.5 ml 0.5% SeaPlaque agarose (Lonza) in the same medium were placed over the base layer. Medium was replaced every 3 days. Three weeks later the clones were counted and photographed.
Tumorigenicity and teratogenicity assay
In our protocol approved by IACUC, a maximum tumour diameter of <20 mm was permitted and none of the experiments exceeded these limits. mES cells established from C57BL/6-TfN(act-EGFP) mice26 were a gift from F. Sugiyama. HeLa cells were obtained from ATCC (ATCC CCL-2). Primary keratinocytes were cultured from the back skin of Krt14cre;LSLtdTomato mice. In vivo iSEPs were sorted from primary culture of generated epithelium 3 months after the administration of DGTM-AAVs in PdgfracreER;LSLtdTomato mice. Cells culture medium (5.0 × 106 in 100 μm) were mixed with an equal amount of Matrigel (BD Biosciences) and subcutaneously injected into the back of NOD/SCID IL-2Rnull (NSG) mice. Twenty-eight days later, after hair removal, the size of the nodule was estimated with the equation; tumour volume (mm3) = (major axis) × (minor axis)2/2. Each nodule was collected with overlying skin and histologically investigated. For the assessment of motility and metastatic profile of iSEPs, luciferase-expressing cells were prepared by transduction of MSCV Luciferase PGK-hygro retrovirus (a gift from S. Lowe (Addgene plasmid 18782)) and drug selection. Fifty-six days after injection, the systemic distribution of the cells was investigated by the detection of luciferase.
Mice of both genders were used in this study. All animal procedures were approved by the IACUC committee of the Salk Institute for Biological Studies or the Animal Research Committee of Kyorin University.
Lineage tracing animal
PdgfracreER;LSLtdTomato, PdgfracreER;RosamT/mG, Pdgfracre;LSLtdTomato, Pdgfracre;RosamT/mG, and Fsp1cre;LSLtdTomato mice were used for labelling a broad spectrum of mesenchymal cells to screen for the leakage of signal expression to the epidermis. All strains show aberrant epidermal labelling, but we noticed that a portion of PdgfracreER;LSLtdTomato mice was free from epidermal leakage. In other words, epidermal leakage with PdgfracreER;LSLtdTomato mice was specific to individual animals and at different anatomical locations. For a more stringent lineage-tracing system and to eliminate animals with epidermal leakage, we used PdgfracreER;LSLtdTomato mice. This was accomplished via a two-step selection process: (1) histological investigation of ear biopsies and (2) investigation with primary keratinocytes cultures of skin removed to generate the wound. All lineage-tracing animals were prepared by intraperitoneal injection of 200 μg/day tamoxifen (postnatal day (P)28–32), histological investigation of ear biopsies (P33), and investigation of primary keratinocytes cultures from the skin portion obtained during attachment of the chamber (P35).
Attachment of the chamber and rubber ring
To evaluate the de novo generation of epithelial tissues from the bottom of cutaneous ulcers, we aimed to induce epithelial tissues in skin ulcers that were isolated from the surrounding skin by a skin chamber. To avoid possible contamination from existing epithelial tissues, we performed these assessments using Krt14cre;LSLtdTomato mice. In these mice, cells developmentally expressing Krt14 are labelled. With the surgical removal of skin and panniculus calnosus, tdTomato signals were detected at the site of the subcutaneously located glands, such as the thyroid and mammary glands. Accordingly, to avoid these structures, two anatomical locations (interscapular area and lateral thoracic area) were chosen as optimal sites for chamber attachment.
Chambers were made up by cutting 1.5-ml Eppendorf tubes. A section was smoothened and broadened by warming with a flame. Two layers of holes were made using 22-gauge needles warmed using a flame. After tanning the internal surface by electric rasp, the chambers were autoclaved. Under general anaesthesia with isoflurane, the site of the chamber attachment was shaved and sterilized. Afterwards, a circle of skin and sub-cutaneous tissue of 1-cm diameter was removed beneath the panniculus carnosus, the chamber was sutured down to the deep fascia with 4–6 horizontal mattress sutures using 4-0 Ethilon (Ethicon Inc.). Then, 200 μl of Cell Matrix Type-1A collagen (Nitta Gelatin) was poured into the gap between the external surface of the chamber and the surrounding skin flap to ensure sealing of the chamber. After suturing the chamber to the skin flap with 4 horizontal mattress sutures, AAVs were administrated and the lid was closed. The edge of the skin flap was glued to the chamber when necessary. Large chambers were made by cutting 5-ml syringes (Becton, Dickinson and Company) using a metal knife warmed with a flame. Basic procedures for manufacturing and fixation were the same as used for the small chamber, except collagen sealing was not used. For the lid, a rubber disk from a syringe was used with a hole in the centre. Fixative rubber rings were made using rubber disks from 5- or 10-ml syringes. After making a hole in the central part, the ring was sutured to the skin and deep fascia to ensure fixation.
Maintenance of the wound
The wound in the chamber is prone to be contaminated because of the use of artificial implant and the closed nature of the wound. When chambers were first attached, in principle, the lids were kept closed to avoid contamination. When chambers were to be changed to larger ones, the lids were kept open for several days in advance, to dry up the wound and prevent carryover of contamination. After changing to the larger chamber, the wound was washed with PBS and photographed every two days under inhalation anaesthesia.
Skin island in chamber assay
An ulcer is generated within a large chamber, but a small patch of skin is left intact in the centre of this ulcer. Four animal groups are defined for comparison; (1) intact skin group, (2) epithelialized skin group, (3) cell sheet groups, and (4) generated epithelium group. For the intact skin group, an island of primary skin was created within a large chamber. For the epithelialized skin group, an ulcer was created and allowed to heal via epithelialization from the surrounding skin (a rubber ring was attached to prevent wound contraction). After epithelialization, the rubber ring was removed and an island of this epithelialized skin was created within a large chamber. For the cell sheet group, epithelial cell sheets were prepared by primary culture of keratinocytes from skin biopsies from each animal. Cell sheets were transplanted onto the backs of these mice via two operations: transplantation of the sheet beneath a silicone sheet, and removal of the overlying skin. An island of this cell sheet-generated epithelium was then created within a large chamber. We attempted to test pure grafted areas, but we found that skin purely composed of a cell-sheet transplant could not survive within a chamber for more than 4 weeks after transplantation. Therefore, we allowed the transplants to contract for another 1–2 weeks before subjecting the cell sheets to the skin island assay. For the DGTM-generated epithelium group, an ulcer was generated using PdgfracreER;LSLtdTomato mice and treated via DGTM-AAV administration. After removal of the chamber and confirmation of complete epithelialization, an island of DGTM-generated epithelium was created. For this group, we histologically confirmed that the tested skin island was composed of mesenchymal-reprogrammed epithelium by tdTomato fluorescence after completion of the assay. In each group, on experimental day 0, the skin island was created by circumferential removal of the surrounding skin and the large chamber was attached to make the skin island in an ulcer. Epithelialized areas from the skin island were measured by imaging analysis every 2 days until experimental day 14. For the DGTM-generated epithelium group, after completion of the study period, the tested skin island is subjected to histological analysis to test tdTomato expression in the epithelial layer and confirm that tested skin islands are entirely composed of mesenchymal-derived generated epithelium.
Estimation of nucleated cell number in deep fascia
Deep fascia was dissected out of the chamber, just after the attachment of chamber. Tissues beneath fascia were removed under dissection microscope. Dissected deep fascia was stained with DAPI and mounted on slide glass. The numbers of nucleated cells were counted in ten randomly selected visual fields in 3D images obtained using confocal microscopy. The area of deep fascia was measured in images obtained using a slide scanner. Using the areas and number of nucleated cells, total numbers of nucleated cells were calculated.
Estimation of efficiency of in vivo reprogramming
The efficiency of in vivo reprogramming was estimated by dividing the numbers of clones that contributed to each skin-chamber epithelium in Pdgfracre;R26Rconfetti mice by the numbers of estimated nucleated cells present in the deep fascia at the time of virus administration.
Blood samples were obtained from animals that underwent DGTM-AAVs administration and multiple wounding procedures (PdgfracreER;LSLtdTomato mice) 3 months after the administration of DGTM-AAVs. Control samples were obtained from age (±2 weeks)-, sex- and genotype-matched non-operated animals. Samples were subjected to test haematology and chemistry assessment at the UCSD murine haematology and coagulation core laboratory.
Four groups of animals (n = 5) underwent chamber attachment and administration of VEGF165a-AAV (1.0 × 1011 gene copies (GC) per animal), DGTM-AAVs (1.0 × 1011 GC for each factor per animal), GFPNLS-AAVs (4.0 × 1011 GC per animal) or PBS to the inside of the chamber. The volume of solutions are adjusted to 100 μl in all groups. Seven days after application, chambers were collected and immunohistologically investigated for microvessel densities31. Sections are stained with H&E and for CD31 and subjected to imaging analysis. Density of vessels is quantified by the proportion of areas of vessels in 6–8 sections of the central portion of the ulcer for each animal. Tissues above and below the fascia are analysed.
Sample collection for RNA-seq analysis
For comparison of primary keratinocytes, iSEPs, and mesenchymal cells, three sets of (1) primary keratinocytes, (2) Pdgfra+ ADSCs, (3) in vitro iSEPs, and (4) in vivo iSEPs were prepared from three PdgfracreER;LSLtdTomato mice. On postnatal day 35, under general anaesthesia, groin adipose tissues and back skin was collected before chamber attachment and DGTM-AAV inoculation. From skin and adipose tissue, primary keratinocytes (1) and ADSCs were isolated and cultured, respectively. Pdgfra+ ADSCs (2) were sorted on the basis of tdTomato signal using a FACS Vantage SE DIVA. Using Pdgfra+ ADSCs, in vitro iSEPs (3) were generated with DGTM-AAVs. Mice underwent a stepwise operation to obtain generated epithelial tissues. On postnatal days 125–145, mice were euthanized, and generated epithelium tissues were collected. In vivo iSEPs were isolated with primary keratinocytes from surrounding skin. Pdgfra+ in vivo iSEPs (4) were sorted from contaminating primary keratinocytes on the basis of tdTomato signal using a FACS. Other in vivo iSEPs (1 month (1M), n = 3, 3M, n = 2, 6M, n = 5) were obtained from different animals with the same procedures described above. Total RNA was purified from each sample and subjected to RNA-seq analysis.
Toluidine blue penetration assay
Tissue samples were collected in a way that included generated epithelium, ulcer, and the surrounding skin. Samples were dehydrated by incubations (1 min each) in 25%, 50%, and 75% methanol in PBS with a further 1 min in 100% methanol. Then rehydrated with the same series of methanol solutions (1 min incubations) and washed in PBS. Soaking the whole sample caused staining from the backside of the sample and affected the results. Therefore, we attached the tube on the surface of the sample to partially include the surrounding skin, and added 0.1% toluidine blue O/PBS into the tube for 1 min before washing32. Samples were photographed. The frozen cross-section was also photographed. After sectioning, both non-stained sections and serial H&E sections were prepared to analyse the detailed location of toluidine blue stain.
Lucifer yellow dye penetration assay
Under general anaesthesia, mice were restrained in Petri dishes with generated tissues in contact with 1 mM Lucifer yellow in PBS (pH 7.4) at room temperature. After 1 h of incubation, mice were euthanized and tissues were dissected, frozen and sectioned at a thickness of 5 μm33. The sections were counterstained with 10 μg/ml DAPI and then analysed by fluorescence microscopy.
Transepidermal water loss (TEWL) measurement
On the day of measurement, hair was carefully shaved from the testing area in mice under general anaesthesia using isoflurane. Special effort was taken not to damage the skin. TEWL was measured using a VapoMeter (Delfin Technologies) with nail adapters at the area of generated epithelium and intact skin. Because TEWL measurements can be influenced by factors such as room and body temperatures and depth of anaesthesia, at least three rounds of measurements were taken. Subsequently, a portion of the skin was surgically removed to generate an ulcer (1-cm diameter) and the TEWL value was measured more than three times.
No statistical methods were used to predetermine sample size. The experiments were not randomized and the investigators were not blinded to allocation during experiments and outcome assessment.
Further information on research design is available in the Nature Research Reporting Summary linked to this paper.
Data from the microarray, RNA-seq and miRNA microarray analyses have been deposited in the Gene Expression Omnibus (GEO) and ArrayExpress under accession numbers GSE85803, GSE106419 and E-MTAB-5055, respectively. All other relevant data that support the findings of this study are available from the corresponding author upon reasonable request.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
This work was supported by MEXT KAKENHI Grant numbers JP26293381(Grant-in-Aid for Scientific Research (B) to M.K.), JP23689073 (Grant-in-Aid for Young Scientists (A) to M.K.), JP21689046 (Grant-in-Aid for Young Scientists (A) to M.K.), Kyorin University research promotion award to M.K. (2013), JSPS Overseas Research Fellowships (2015–17) to M.K., and the Uehara Memorial Foundation Research Fellowship for Research Abroad (2017–18) to M.K. M.K. thanks H. Green for support materials. T.H. thanks F. Sugiyama for support materials. M.N.S. is supported by NIH-NCI CCSG: P30 014195 and The Leona M. and Harry B. Helmsley Charitable Trust. Work in the laboratory of J.C.I.B. was supported by the G. Harold and Leila Y. Mathers Charitable Foundation, The Leona M. and Harry B. Helmsley Charitable Trust, The Moxie Foundation, The Evergreen Foundation, Fundacion Dr. Pedro Guillen and Universidad Católica San Antonio de Murcia (UCAM).
Nature thanks S. Akita, V. Horsley, A. Lombardo and the other anonymous reviewer(s) for their contribution to the peer review of this work.
Extended data figures and tables
About this article
Nature Medicine (2018)