IPSC-derived organoid-sourced skin cells enable functional 3D skin modeling of recessive dystrophic epidermolysis bullosa

Abstract

Recessive dystrophic epidermolysis bullosa (RDEB) is a severe inherited skin disorder caused by mutations in COL7A1. Patient-derived induced pluripotent stem cells (iPSCs) enable the personalized study of RDEB pathogenesis and potential therapies. However, current skin cell differentiation protocols via 2D culture perform suboptimally when applied to engineered 3D skin constructs (ESC). Here, we present an approach to source fibroblasts (iFBs) and keratinocytes (iKCs) from iPSC-derived skin organoids using an optimized differentiation protocol, and utilize them to engineer ESCs modeling wild-type and RDEB phenotypes. The resulting iPSC-derived skin cells display marker expression consistent with primary counterparts and produce ESCs exhibiting significant extracellular matrix remodeling, protein deposition, and epidermal differentiation. RDEB constructs recapitulated hallmark disease features, including absence of collagen VII and reduced iFB proliferation. This work establishes a robust and scalable strategy for generating physiologically-relevant, iPSC-derived skin constructs, offering a powerful model for studying RDEB mechanisms and advancing personalized regenerative medicine.

Introduction

Recessive dystrophic epidermolysis bullosa (RDEB) is a genetic skin blistering disease caused by mutations in COL7A1, disrupting the functionality of collagen VII and leading to impaired binding of the epidermal basement membrane (BM) to the dermal extracellular matrix (ECM). Current treatment options focus on surgical release of the contractures and wound dressings, ¹ autografts, or engineered allogenic skin substitutes.^2–5 In recent years, gene replacement therapies have emerged, with two FDA-approved strategies: a topical COL7A1 delivery developed by Gurevich et al. ⁶ using a modified herpes simplex virus type 1 (HSV-1), and another using ex vivo gene correction of patient-derived cells followed by autologous skin grafting. ⁷ Despite these advances, further preclinical research is critical to develop and evaluate durable, curative treatments for RDEB.

RDEB is primarily studied using genetically modified murine models that carry commonly seen pathogenic variants among patients.^8,9 While these knockout models can exhibit some pathological features and provide valuable information on the interaction between multiple cell types in a relevant environment, they do not fully recapitulate disease phenotypes due to interspecies differences in skin development of mice and humans. ⁸ Additionally, RDEB patients suffer from a large variety of pathogenic variants in COL7A1, each often resulting in variable phenotypes, and thus requiring the establishment of multiple genetically modified murine models, which increases the cost, time, and labor. ⁸ Another common method is to use primary RDEB skin cells and evaluate their phenotypes in 2D cultures. ¹⁰ 2D characterization of primary cells offers a simplified and scalable method for high-throughput screening of therapeutics, but the lack of mechanical and cell-cell interactions, as well as tissue complexity, results in a less relevant model that does not properly recapitulate RDEB characteristics.

3D models offer a physiologically-relevant platform for the study of disease-specific features,^11–13 especially since these allow for development of the BM, where collagen VII is found. However, primary cells are limited by their reduced proliferative capacity, requiring large quantities of cells from patient biopsies. To overcome this limitation, iPSCs are used to develop 2D models, ¹⁴ 3D skin constructs, especially in combination with primary cells, ¹⁵ and complex skin organoids. ¹⁶ Additionally, iPSCs can be genetically modified using CRISPR/Cas9 or other gene editing tools to correct COL7A1 mutations,^17,18 providing a promising tool for developing and validating gene therapies. While iPSC-derived skin offers a powerful pre-clinical testing platform and patient-specific disease modeling, the current iPSC-derived skin models suffer from poor reproducibility, and limited epidermal differentiation and maturation.

In this study, we present a 3D-differentiation and cell-sourcing strategy that relies on isolating keratinocytes (KCs) and fibroblasts (FBs) from iPSC-derived complex skin organoids and expanding them in 2D culture. With this approach, ready-to-use functional KCs and FBs can be obtained and banked from wild-type and RDEB patient-derived iPSCs. Both iKCs and iFBs generated this way are phenotypically and functionally similar to their primary counterparts and can produce a functional dermis and epidermis, yielding a patient-specific 3D skin model. Our strategy enables us to generate fully iPSC-derived models and study RDEB disease mechanisms, with the potential to be adapted for other skin conditions, thus filling an important gap in pre-clinical testing of RDEB therapies in a human-relevant context.

Results

Wild-type iPSC-derived keratinocytes and fibroblasts resemble primary counterparts

As an improved alternative method to 2D differentiation, we adapted a 3D differentiation approach based on pluripotent stem cell derived organoids (PSO) recently developed by Lee et al., ¹⁶ and used these complex organoids as the cell source for KCs and FBs (Figure 1(a)). We confirmed that PSOs can advance to hair follicle generation at day 75 (Supplemental Figure 1a). However, as early as day 45 of PSO culture, we observed expression of epidermal, dermal, and basal membrane (BM) markers (K14, CD90, COL VII), as well as dermal papilla precursor markers (LEF1 and LEPr; Figure 1(b) and Supplemental Figure 1). Thus, for cell isolation, we chose this early timepoint where our skin cells of interest are developed and express relevant markers. At day 45, we sorted the PSOs using magnetic-activated cell sorting (MACS) and selected for KCs and FBs before subculturing them in vitro for two passages.

Figure 1.

Isolation of iPSC-derived cells from skin organoids: (a) Diagram of iFB and iKC generation protocol. (b) Wholemount IF staining of wild-type PSOs at day 45, expressing K14, COL VII, and CD90. (c) IF staining of 2D-cultured iKCs and primary KCs at passage 2. (d) Quantification of TAp63 and DNp63 intensity from IF staining (mean ± SD). (e) Quantification of ITGA6 expression from flow cytometry reveals an increase in expression after each passage, with passage 2 iKCs expressing similar levels as passage 2 primary KCs. (f) Flow cytometry of passage 2 iFBs and primary FBs. (g) IF staining of 2D-cultured iFBs and primary FBs at passage 2. Scale bars: 200 µm for (b), 100 µm for (c) and (g).

After isolation and expansion in 2D culture, immunofluorescence (IF) staining showed comparable expression of keratin 14 (K14) and desmoglein-3 (DSG3), proteins mainly expressed in the basal and immediate suprabasal layers of the epidermis, to primary KCs (Figure 1(c)). Additionally, quantification of the fluorescent signal revealed that iKCs express significantly more DeltaNp63 than TAp63 (Figure 1(d)), indicating a basal KC phenotype. The basal-like phenotype of iKCs was further confirmed through expression of Ki67, a marker of proliferation, lack of expression of keratin 10 (K10) and decreased expression of desmoglein 1 (DSG1), markers of the stratum spinosum. This data demonstrates that our skin organoid differentiation approach yields basal-like iPSC-derived KCs similar to primary cells.

Next, we identified that iKCs at passage 2 are the most suitable for 3D skin engineering. Flow cytometry analysis of integrin alpha 6 (ITGA6), which aids in anchoring KCs to the BM, in iKCs showed increased expression at passage 2, compared to passage 1, reaching a percent expression most similar to primary KCs (Figure 1(e)). Initial ITGA6 percent expression at passage 0 was low due to contamination by other cell types that lose ITGA6 expression after culturing, thus we re-sorted the cells after passage 1 to obtain an enriched iKC population. IF also showed a greater number of K14 negative cells at passage 1, confirming the presence of other cell types positive for ITGA6, while passage 3 iKCs showed decreased Ki67 expression, suggesting loss of proliferative capabilities (Supplemental Figure 2).

IF and flow cytometry analysis of iFBs revealed similar expression patterns to primary FBs (Figure 1(f)–(g)). By using flow cytometry, we observed similar expression of PDGFRα between iFBs and primary FBs, whereas CD90 expression was slightly lower in iFBs, potentially indicating a decreased representation of reticular FBs as a subpopulation. IF staining of vimentin (VIM) and fibroblast activated protein (FAP), two well-known FB markers, showed comparable expression in both iFBs and primary FBs, as shown in Figure 1(g). Finally, we confirmed the ability of iFBs to secrete ECM and BM proteins including fibronectin (FBN), collagen type IV (COL IV), and laminin α5 (LAMA5), which were found comparable to primary FBs. Overall, we concluded that 45 days of PSO culture followed by isolation and 2D culture for two passages yielded iFBs and basal iKCs phenotypically similar to their primary counterparts.

Wild-type iPSC-derived skin cells exhibit functional competence in stratified epidermis formation and dermal matrix remodeling

After 2D characterization of iKCs and iFBs, we tested their ability to generate the dermal and epidermal compartments of 3D engineered skin constructs (ESCs) following our previously established protocols (Figure 2(a)).^19,20 We first produced dermis-only ESCs using iFBs suspended in a collagen type I hydrogel without adding KCs. We demonstrated using IF that iFBs are capable of depositing common ECM proteins, including FBN, collagen III (COL III), and elastin (ELN; Figure 2(b) and Supplemental Figure 3), thus indicating their ability to remodel the initial collagen matrix to form the dermis. Maintenance of CD90 and FAP, with overlapping and non-overlapping expression, also suggested the presence of multiple iFB subpopulations. Quantification of contractility over a two-week period demonstrated a similar profile to primary constructs, with over 35% contraction (Figure 2(c)), further supporting the iFBs resemblance to primary FBs in a 3D context. Further functional assessment was done based on the capability of iFB-containing dermis to induce epidermal formation when seeding primary KCs on top. As shown in Figure 2(d) (also see Supplemental Figure 3), primary KCs formed a thick, differentiated epidermis when combined with iFBs, with the dermal compartment showing proper deposition and organization of common ECM (COL I-III, FBN, and ELN) and BM proteins (LAMA5 and COL IV). This data confirms the identity of iFBs as a functional stromal cell type in 3D skin.

Figure 2.

Functional characterization of wild-type iPSC-derived cells: (a) diagram of the ESC generation protocol. (b) IF staining shows iFBs can remodel the ECM and maintain expression of multiple FB subpopulation markers in dermis-only iESCs. (c) Quantification of contraction percent from iESCs and primary ESCs across three different time points (mean ± SD). (d) Functional capabilities of iFBs were further characterized by seeding primary KCs on an iFB dermis. (e) Seeding iKCs on top of a primary FB dermis shows functional capabilities of iKCs through proper epidermis formation and maturation. K14 and Ki67 show presence of a proliferative basal layer, DSG1, K10, and INV mark the stratum spinosum, and LOR shows a successful terminal maturation with epidermal barrier formation. Scale bars: 100 µm.

To characterize iKC function, we seeded iKCs on a dermis containing primary FBs and confirmed the formation of the four epidermal layers through IF staining (Figure 2e). K14 marked the presence of the basal layer while K10, involucrin, and DSG1 demonstrated proper differentiation of the iKCs into the stratum spinosum. In addition, loricrin (LOR) indicated terminal differentiation. Ki67 staining showed that basal iKCs retain their proliferative capabilities. Overall, this data showed the function of iFBs and iKCs in a 3D context.

Line-dependent optimization of BMP-4 concentration is required to successfully generate RDEB skin organoids

Next, we aimed to generate iFBs and iKCs from RDEB patient-derived iPSCs using the same organoid approach. At day 0 of the differentiation process, the medium was supplemented with BMP-4 (Figure 3(a)), a factor that drives the development of the surface ectoderm. While the wild-type iPSC line we used (namely Desmoplakin-GFP) has been previously optimized for suitable organoid development, ¹⁶ adjusting BMP-4 levels was necessary for our RDEB line to compensate for the cell line-specific endogenous expression. We implemented a morphology-based approach using wild-type iPSCs as a control to monitor the development of appropriate organoid morphology during each stage of the differentiation. We identified three distinct morphologies at the two-week mark of differentiation: cystic, which is characterized by a lack of a dense core; dense, which lacks the cystic element; and mixed-type, which displays both characteristics (Figure 3(b)).

Figure 3.

PSO development and optimization through morphology-based early screening: (a) overview of PSO differentiation protocol with key growth factors. (b) Identification of PSO distinct morphology two weeks into the differentiation protocol. Flow cytometry analysis of the organoid cell suspension collected at day 45 shows a correlation between ITGA6/PDGFRα expression and epidermis development. H&E shows proper development of an epidermis when dense iKCs sorted at day 45 and cultured to passage 2 are seeded on top of a primary FB dermis, and a lack of epidermis formation when cyst iKCs are used. (c) The morphology-based screening approach was used to select a narrow range of BMP-4 concentrations for optimization of RDEB PSO differentiation, and flow cytometry and functional characterization by seeding iKCs on a primary FB dermis show 4 ng/ml yields iKCs that form a thicker epidermis in vitro. Scale bars: 100 µm.

While the cystic organoids seemed to produce iKCs that more closely resembled primary KCs in 2D (Supplemental Figure 4), they failed to produce an epidermis when seeded on dermis made with primary FBs. We observed that the iKCs obtained from dense organoids produced a thicker and well-stratified epidermis compared to the mixed-type and cystic groups (Figure 3(b)). At day 45 of culture, we dissociated the organoids to a single cell suspension and analyzed the expression of ITGA6 and PDGFRα using flow cytometry (Figure 3(b)). This revealed a potential inverse relationship between the percentage of the organoid cell population expressing these markers and the ability of the isolated iKCs to develop a proper epidermis, although further analysis is needed to show a statistical correlation. Moving from this observation, we decided to exclude cystic organoids, and combine mixed-type and dense organoids to isolate functional iKCs.

Leveraging the morphology-based screening method, we tested multiple BMP-4 concentrations at day 0 for differentiation of the RDEB-iPSC line and identified a morphogen range likely to produce functional iKCs. We initially tested a BMP-4 concentration of 7.5 and 3 ng/ml. The high concentration yielded oddly-shaped organoids throughout the entire differentiation process, which failed to form an epidermis when the iKCs were seeded on 3D primary dermis constructs (Supplemental Figure 5). The low concentration (3 ng/ml) yielded organoids with the favorable dense morphology, which was maintained over multiple timepoints, and the isolated iKCs were able to form an epidermis (Figure 3(c)). Nonetheless, we were not fully satisfied with the level of stratification in the epidermis and slightly increased to the concentration to 4 ng/ml. Both conditions yielded the desired dense morphology of organoids, but as shown in Figure 3(c), the 4 ng/ml condition resulted in a thicker and better stratified epidermis, comparable to epidermis developed by wild-type dense iKCs. Additionally, flow cytometry revealed the same inverse relationship observed between wild-type iPSCs, ITGA6 and PDGFRα expression, and epidermis differentiation. Thus, we continued our experiments using a BMP-4 concentration of 4 ng/ml for the RDEB-iPSC line.

RDEB iPSC-derived FBs and KCs can produce 3D engineered skin constructs

Using the protocol described above, we developed iPSC-derived ESCs (iESCs) using both iKCs and iFBs from wild-type and an RDEB patient. As expected, there was a lack of COL VII expression in the BM of RDEB iESCs (Figure 4(a)), since RDEB cells present COL VII pathogenic variants in both alleles. Other BM proteins, LAMA5 and COL IV, were properly secreted and aided in the anchoring of the epidermis to the dermis (Figure 4(b)).

Figure 4.

Development of iESCs with wild-type and RDEB-derived iFBs and iKCs: (a) COL VII IF staining of wild-type and RDEB iESCs. (b) RDEB iESCs express other BM proteins (LAMA5 and COL IV). (c) Dermis-only RDEB iESCs still contract and express common ECM proteins, but VIM shows sparse iFB organization. (d) Quantification of wild-type and RDEB dermis-only iESC proteins (mean ± SD). (e) Elastic modulus (kPa) measured through microindentation of dermis-only wild-type and RDEB iESCs and primary ESCs (median ± SD). (f) Permeability (m2) measured through microindentation of dermis-only wild-type and RDEB iESCs and primary ESCs (median ± SD). (g) H&E of wild-type and RDEB iESCs made with both iKCs and iFBs. (h) IF staining for epidermis markers in wild-type and RDEB iESCs. *p < 0.05. **p < 0.01. ***p < 0.001. ****p < 0.0001. Scale bars: 50 µm.

Next, we characterized the formation of the dermis by using RDEB iFBs to generate constructs without adding iKCs (i.e. dermis-only iESCs; Figure 4(c)). While the constructs contracted and remodeled the matrix to an extent, H&E and IF staining (Figure 4(c)) revealed that RDEB iFBs were sparsely dispersed in the dermis. Quantification of deposition of common ECM and BM proteins in wild-type and RDEB dermis (Figure 4(d)) showed a significant difference in COL III, COL IV, and LAMA5, with an overall decrease of ECM proteins in RDEB except FBN. Additionally, the quantification showed an overall significant decrease in BM proteins (COL IV, LAMA5, COL VII) in RDEB iFBs compared to wild-type iFBs.

Moreover, we measured the elastic modulus of both groups using microindentation and compared it to primary dermis-only ESCs (Figure 4(e)). The measurements showed a significant reduction in the stiffness of RDEB dermis-only iESCs when compared to those achieved with either wild-type iFBs or primary FBs. The lower stiffness of the RDEB constructs could be a result of the reduced remodeling of the dermis by these cells. Unexpectedly, constructs made with wild-type iFBs exhibited higher permeability compared to RDEB, but no significant difference was found (Figure 4(f)).

Following the validation of dermal constructs, we characterized the function of RDEB iKCs to produce an epidermis. Figure 4(g) shows the ability of both wild-type and RDEB iKCs to differentiate and form a stratified epidermis, with wild-type iKCs displaying a higher degree of cornification. IF staining (Figure 4) demonstrates the formation of basal, spinous, and granular layers for both groups. However, RDEB iESCs show fragmented and less uniform K10 expression, as well as intermittent LOR expression.

Maturing iFBs through keratinocyte co-culture enhances their phenotype and function in iESCs

While we successfully developed iESCs, their epidermis was thinner and less uniform than mixed ESC containing iKCs/primary FBs or from iFBs/primary KCs (shown in Figure 2(d) and (e)). Quantification of epidermal thickness showed a significant difference between these three groups, with primary KCs/iFBs displaying the thickest epidermis, followed by iKCs/primary FBs, and finally iKCs/iFBs (i.e. iESCs; Supplemental Figure 6).

Lack of iFB maturity, a common issue with iPSC-derived cells, ²¹ could negatively affect the ability of the iKCs to form a thick epidermis. Previous work showed that removal of FBs in ESCs reduces the differentiation of the epidermis, resulting in a significantly thinner epidermis, due to the lack of epithelial-mesenchymal cross-talk. ²² Additionally, previous in vitro studies have shown the importance of KC-FB crosstalk for secretion of FB growth factors (KGF, FGF7, IL6) that promote KC proliferation.^22,23 Therefore, we hypothesized that further maturation of iFBs through FB-KC crosstalk may be needed to ensure proper differentiation of the epidermis in iESCs.

To test our hypothesis, we established a co-culture of wild-type iFBs with primary KCs by seeding the latter on a collagen-coated transwell and placing it on top of an iFB culture. After 5 days of co-culture, we used the resulting matured iFBs and iKCs to generate iESCs (Figure 5(a)). When we cultured control and matured iFBs in 2D and quantified the expression of common FB and ECM markers (Figure 5(b) and (c)), matured iFBs showed decreased expression of CD90, a marker typically associated with reticular fibroblasts. Notably, expression of BM proteins, LAMA5 and COL IV, was higher in matured iFBs than control iFBs. Additionally, the increase of PDGFRα, VIM, and ELN in matured iFBs, reaching comparable expression to primary FBs, indicates that priming by KCs enhances iFB maturation.

Figure 5.

Maturation of iFBs through primary KC co-culture: (a) diagram of the co-culture protocol. (b) IF staining of 2D control and matured wild-type iFBs. (c) Quantification of IF intensity of primary FBs, control iFBs, and matured iFBs (mean ± SD). (d) H&E of ESCs made with primary FBs, control iFBs, matured iFBs, and conditioned matured iFBs. iKCs were seeded on top of all groups. (e) IF staining of control and matured iESCs shows a better localization of K14 to the basal layer and COL IV to the BM in matured iESCs. (f) Quantification of epidermis thickness (mean ± SD). (g) Quantification of epidermal thickness variability (mean SD between thickness measurements across each construct). *p < 0.05. **p < 0.01. ***p < 0.001. ****p < 0.0001. Scale bars: 50 µm.

iESCs generated with matured iFBs developed a thicker epidermis compared to ESCs made with either primary FBs or control iFBs (Figure 5(d) and (f)–(g)). We quantified the epidermal thickness variability across each construct as a measure of consistent epidermis formation using the mean standard deviation (SD) of epidermis thickness measurements. iESCs with primary FBs exhibited the lowest SD, followed by iESCs with control iFBs and lastly iESCs with matured iFBs, yet without a significant difference. Thus, we concluded that maturing iFBs had a positive effect on epidermis formation.

Characterization of the epidermal layers by IF (Figure 5(e)) revealed that K14 was more localized to the basal layer in iESCs made with matured iFBs, while control iESCs showed a diffuse K14 signal throughout the entire epidermis, suggesting immature development marked by disorganized differentiation. While there was no apparent difference in the expression pattern of FBN, COL IV expression in the dermis of matured iESCs was increased and highly localized in the BM, whereas in control iESCs COL IV was diffusely expressed by the epidermis. The lack of a defined COL IV band suggests abnormal deposition by iKCs and iFBs. Overall, maturation of iFBs prior to iESC development seemed to improve the instructive role of iFBs and their crosstalk with iKCs, driving better epidermis differentiation into its respective layers, as well as organized deposition of BM components.

To understand whether soluble factors secreted by primary versus iFBs or iKCs could contribute to further improve the development of iESCs, we established three separate co-cultures in a transwell format (described in Supplemental Figure 7a): primary FBs/primary KCs (condition A), iFBs/primary KCs (condition B), and primary FBs/iKCs (condition C). We collected the conditioned medium from these co-cultures and used it to feed three different groups of matured iESCs (Supplemental Figure 7). While conditioning the medium resulted in an overall increase in epidermis thickness compared to primary, control, and matured iFBs, the SD between thickness measurements across conditioned constructs was significantly higher, suggesting that a thicker epidermis formed in some areas of the construct. Additionally, conditioned groups exhibited irregular iKC morphology and inconsistently thick cornified areas compared to non-conditioned groups. Considering both epidermis thickness and thickness variability, we selected condition C to compare with our control and matured iFB groups (Figure 5(d), (f), and (g); Supplemental Figure 7), and showed that medium conditioning can improve epidermal thickness, albeit with compromised epidermal consistency.

Maturation of RDEB iESCs enhances ECM remodeling but has no effect on epidermis differentiation

Following these results, we applied the same maturation method to RDEB-iFBs. In 2D, matured RDEB iFBs showed significantly increased expression of COL III, COL IV, and LAMA5, and decreased expression of PDGFRα and CD90, compared to control RDEB-iFBs (Figure 6(a) and (b)). Additionally, matured iFBs showed a positive trend in the expression of the ECM protein ELN. Thus, maturation could be driving iFBs to a more structural supportive role by increasing the deposition of ECM and BM proteins, similarly to maturation of wild-type iFBs.

Figure 6.

Maturation of RDEB iFBs: (a) 2D staining of control and matured RDEB iFBs. (b) Quantification of 2D IF intensity for RDEB iFBs and primary FBs (mean ± SD). (c) H&E and IF staining of iESCs made from either control or matured iFBs (both with iKCs seeded on top). (d) Quantification of ECM and BM protein deposition in control and matured iESCs (mean ± SD). (e) Quantification of epidermis thickness and uniformity across the construct (mean ± SD; SD). *p < 0.05. **p < 0.01. ***p < 0.001. ****p < 0.0001. Scale bars: 50 µm.

Next, we generated iESCs with matured RDEB iFBs and quantified the deposition of ECM and BM proteins compared to control iESCs. Overall, matured RDEB iESCs showed slightly increased expression of FBN, COL III, LAMA5, and COL IV, but only the latter showed statistical significance (Figure 6(d)). In accordance with the 2D results, as well as the pattern observed in wild-type matured iFBs, maturation seems to drive iFBs away from a reticular phenotype and to express increased levels of BM proteins.

While quantification showed that maturation resulted in a significantly thicker epidermis than control iESCs (Figure 6(e)), thickness variability was higher in matured iESCs and we did not observe significant differences in terminal differentiation. IF staining revealed expression of K14, K10, and INV throughout the entire epidermis (Figure 6(c)), suggesting disorganized epidermal stratification. Ki67 staining was localized to the basal layer and present in both groups. This data indicates that the iFB maturation approach enhances dermal remodeling and induces epidermal thickening in the RDEB model, however it is not sufficient to enhance the differentiation of the epidermis.

Discussion

In this study, we developed a fully iPSC-derived 3D skin model reconstructed entirely in an in vitro context, overcoming a key limitation in patient-specific skin disease modeling. By optimizing an organoid differentiation protocol followed by 2D expansion of cells, we generated functional iKCs and iFBs capable of forming a stratified epidermis and depositing essential ECM and BM proteins within engineered 3D skin constructs. Although our work primarily focuses on RDEB, this platform offers a scalable and patient-specific system broadly applicable to other genetic skin disorders and regenerative medicine applications. iPSC-derived skin cells enable high-fidelity modeling of disease mechanisms and provide a powerful, personalized tool for preclinical therapeutic screening in a controlled in vitro environment.

The conventional method of iPSC differentiation into KCs involves the addition of factors, including BMP-4 and retinoic acid, to a 2D culture of iPSCs.^14,24–27 This method produces iKCs that express basal markers, such as K14, and that can generate an epidermis when grafted on mice.^25,28 iFBs can also be obtained through 2D culture of iPSCs with differentiation factors, such as TGF-b and ascorbic acid,^14,24,28 and have been shown to express canonical FB markers including VIM and CD90.^24,28 Although iPSC-derived 3D skin constructs composed of both 2D-differentiated iFBs and iKCs show some success when applied as grafts on mice, they do not translate optimally in vitro, often exhibiting poor stratification, organization, or ECM deposition. This may be due to the limitations of the 2D differentiation approach, which derives each cell type individually in the absence of ectoderm-mesoderm crosstalk (e.g. generation of KCs without FB signals). In contrast, 3D differentiation in PSOs provides a cellular microenvironment more similar to in vivo skin morphogenesis. Beyond cell-cell interactions, recent studies suggest that the 3D microenvironment may promote cytoskeletal tension-mediated changes in chromatin organization, enhancing epigenetic plasticity and lineage commitment,^29–31 resulting in sustained transcriptional flexibility and efficient differentiation. The combination of enhanced intracellular communication and more permissive epigenetic landscape could explain the success of our approach of sourcing cells from PSOs over the previous 2D differentiation methods.

The efficiency of differentiation is important for in vitro modeling and clinical translation. Our approach produces around 2 million iKCs and 3.5 million iFBs at passage 0 from an initial 10⁵ iPSCs in 45 days. The cells can be frozen and banked at this stage for further use. After expansion to passage 2, in 14 days, these cells can yield about 80 replicates of iESCs in 12-well format for disease modeling and a skin graft with a surface area of ~0.5 cm²/10⁵ iPSCs for skin transplantation applications. This could not be achieved using primary RDEB cells due to their limited proliferation in vitro and limited availability of donor skin from RDEB patients.

During organoid development, we observed three distinct morphological phenotypes – cystic, mixed-type, and dense – each resulting in a different level of epidermal development. Both mixed-type and dense organoids yielded iKCs that can produce an epidermis, with the latter developing into a thicker and stratified epidermis, suggesting that, at this stage of organoid formation, a tight core of potentially epithelial cells may be producing stronger signaling cues that translate into successful in vitro functionality, but further studies should be conducted to identify the exact features of this relationship.

Endogenous levels of BMP-4 differ across cell lines, and fine-tuning the concentration during skin organoid development is pivotal for proper development of skin cells. Using our morphology-based assessment method, we were able to adjust the differentiation of our RDEB line through careful evaluation of the percentage of organoids in each category produced by varying concentrations of BMP-4. Since BMP-4 is responsible for driving the embryonic ectoderm away from neural fate and toward epidermal commitment, ³² high concentrations are often preferred for KC differentiation. However, increasing the concentration of BMP-4 in our RDEB organoids produced poor outcomes, including disorganized morphology and lack of epidermis formation, and a drastic decrease to 3 ng/ml improved the organoid structure but still produced suboptimal epidermis. Thus, our results emphasize the importance of optimizing the concentration of BMP-4 for proper lineage specification. While applying our morphology-based analysis to identify the best BMP-4 concentration decreases the optimization timeline and resources, it is critical to perform a functional analysis (e.g. epidermis formation or collagen contraction) for fine-tuning adjustments once a small range of concentrations has been identified.

Previous studies have shown decreased COL VII expression in PSOs generated with wild-type iPSCs. ³³ After sorting and expanding the wild-type iKCs and iFBs in 2D culture using our method, COL VII was robustly expressed, unlike the RDEB-derived iESCs where COL VII was missing, demonstrating the disease relevance of our organoid-sourcing strategy. Despite the improved functional characteristics of our cells, we found limitations to their performance compared to their primary counterparts. Inferior terminal differentiation, as evidenced by a decrease of LOR expression despite the presence of early differentiation markers, suggests a degree of immaturity commonly associated with iPSC-derived cells.

These limitations were more pronounced in RDEB iPSC-derived cells. RDEB iKCs, potentially due in part to the lack of COL VII, usually exhibit a hyperproliferative phenotype and an upregulation of intermediate differentiation markers like K10, ³⁴ consistent with our findings. iFBs exhibited reduced proliferative and ECM deposition capacity, compared to wild-type iFBs, likely reflecting both the effects of the disease phenotype and incomplete cellular maturation during differentiation. The reduced iFB density could be due to an increase in the senescent profile of these cells, a characteristic that has been observed in FBs from RDEB patients and that results from an increase in the release of pro-inflammatory molecules.^35–37 While RDEB fibroblasts tend to secrete increased levels of structural proteins,^10,38 impaired proliferation caused by an increased senescent population and a lack of maturation could result in a lower deposition than wild-type iESCs, which show a greater number of iFBs present in the dermis and thus higher secretion of proteins.

Additionally, the RDEB phenotype could have an effect on the mechanical properties of the constructs, where RDEB dermis-only iESCs were measured to be significantly softer than wild-type or primary constructs, as well as less permeable than wild-type dermis-only iESCs. This suggests that wild-type iFBs can form a dermis that allows for molecular diffusion comparable to primary cells, while a decrease in permeability from RDEB constructs could be a result of a pathological state of the dermis, rendering it less structurally robust and less permissive to molecular exchange. Further studies should be performed to draw any conclusions on the mechanical properties of RDEB constructs.

Maturation of iFBs by primary KC co-culture resulted in improved dermal matrix composition and epidermal architecture, especially in wild-type iESCs. Due to the crosstalk with primary KCs, iFBs could have been driven to express growth factors associated with FBs responsible for the secretion of BM proteins and epidermis morphogenesis, ²³ thus explaining the decrease in CD90 expression (usually expressed by FBs in the reticular layer) and increase in BM protein expression. Despite the enhanced dermal ECM and BM protein deposition observed in matured RDEB iESCs, particularly a significant increase in COL IV, this did not translate into improved epidermal maturation. While epidermal thickness and uniformity across the constructs were increased, IF staining revealed broad expression of early and intermediate differentiation markers (K14, K10, and involucrin) throughout the epidermis in both control and matured iESCs. This suggests that, while maturation of RDEB iFBs improves the structural quality of the dermal matrix and potentially promotes basal cell attachment through increased BM protein deposition, key epidermal differentiation cues remain impaired. This data implies that enhanced ECM is insufficient to fully drive epidermis stratification and terminal differentiation in the absence of COL VII or other critical crosstalk signals in iESCs.

The fully iPSC-derived 3D skin model of RDEB generated in this study recapitulates key structural and molecular features of the disease and provides a customizable system for disease modeling and therapeutic testing. The ability to recreate RDEB tissue in 3D also has the potential for customization of tissue complexity, from equivalents containing only dermal or epidermal cells to the addition of vasculature and immune cells, which could better recapitulate in vivo mechanisms. Additionally, previous studies have successfully generated gene-corrected iPSCs form RDEB patients,^17,18 which in combination with our optimized differentiation approach and scalable iESC development would provide a relevant therapeutic strategy for clinical applications. While limitations such as incomplete epidermal maturation and reduced functionality of RDEB-derived cells remain, these can be addressed through further optimization of differentiation factors, co-culture conditions, and medium supplementation. Continued optimization of this system holds great promise for transforming preclinical testing in RDEB and advancing personalized approaches for a broad range of inherited skin disorders.

Materials and methods

Cell culture

Primary human KCs and FBs were isolated from neonatal foreskin and cultured up to passage 3 in KC medium (EpiLife; Gibco #MEPI500CA supplemented with S7, Gibco #S0175 and Antibiotic-Antimycotic Gibco #15240112) and FB medium (DMEM, Gibco #10566016 with 10% fetal bovine serum, Gibco #16000069 and Antibiotic-Antimycotic Gibco #15240112) respectively. KCs were cultured in collagen I-coated plates (Corning #354450). FBs were dissociated using Trypsin-EDTA 0.05% (Gibco # 25300054) and KCs with Accutase (Gibco # A1110501). Wild-type iPSCs (Coriell Institute #AICS-0017) and RDEB patient iPSCs (kindly provided by Dr. Angela Christiano) were cultured in vitronectin-coated plates with mTeSR Plus (STEMCELL Technologies #100-0276). After isolation from organoids, iKCs were plated in collagen IV and laminin-coated plates and cultured with iKC medium (1% GlutaMax, 1% Anti-Anti, 5% Human Serum Albumin, 2% B27 without vitamin A, 0.02% isoproterenol hydrochloride, 0.04% hydrocortisone, 0.01% EGF, 0.01% KGF, 0.1% adenine, in Advanced DMEM F12) and rock inhibitor (Y-27632; STEMCELL Technologies #72304) until passage 2, before being transitioned to KC medium. iFBs were cultured in FB medium until passage 2.

For maturation experiments, 7.5 × 10⁴ primary KCs were seeded on collagen I-coated transwells, allowed to attach for 1–2 h, and placed on top of a well of a 6-well plate containing 15K iFBs. Cells were co-cultured for 5 days before iFBs were detached and added to a salt-balanced neutralized collagen I solution following our ESC protocol.

Organoid differentiation and sorting

On day −2, iPSCs were dissociated with Accutase (Gibco # A1110501) and strained to obtain a single cell solution. In mTeSR Plus with 20Y, 3500 cells in 100 µl were added per well of a 96-well U-bottom plate (Thermo Scientific #174925) and centrifuged to concentrate the cells at the bottom. The next day, 100 µl of mTeSR Plus were added per well to dilute the rock inhibitor. At day 0, the organoids were collected in a 50 ml tube to be washed with basal E6 medium (Gibco #A1516401) three times. Each organoid was transferred to a new well of a 96-well U-bottom plate with 100 µl of E6SBF (E6; Anti-Anti Gibco #15240112; 10 M SB431542 Reprocell #04001005; 4 ng/ml bFGF PeproTech #10018B; 4 or 5 ng/ml BMP-4, Thermo Scientific #12005ET10UG; 2% Matrigel Corning #354277). At day 3, 25 µL of 1 µM LDN (Selleck Chemicals #S2618; working concentration of 250 ng/ml) and 250 ng/ml of bFGF (working concentration of 50 ng/ml) in E6 were added to each well. At day 6, 75 µl of E6 were added per well to dilute the previous factors, and at day 9, half media change was performed. At day 12, the organoids were transferred to a 50 ml tube, washed with 3 ml of advanced DMEM F12 (Gibco #12634010), and individually placed in a well of a 24-well Ultra-Low Attachment plate (Corning #3473) with 1% Matrigel diluted in 500 µl of organoid maturation medium (OMM; 49% Advanced DMEM F12; 49% Neurobasal medium Gibco #21103049; 1% GlutaMax Supplement Gibco #35050061; 1% Anti-Anti; 1% B27 minus vitamin A Gibco #12587010; 0.5% N2 supplement Gibco #17502048; 0.182% MeOH Gibco #21985023) per well. The plates were kept at 37°C on a rocking platform. At day 15, half media change was performed with 1% Matrigel in OMM, and at day 18, half media change was performed with OMM. Until day 45, the plates were kept on a rocking platform and media was changed twice a week, with 1 day performing half media change and the other a full media change.

At day 45, the organoids were collected in a 50 ml tube (excluding any cyst organoids) and washed with PBS before being transferred to a 10 cm dish with 10 ml of accutase and placed on a rocking platform at 37°C for 30 min. Every 5 min, the organoid solution was pipetted to dissolve clumps and placed back in the incubator. After 30 min, the solution was strained with 30 µm filters (Miltenyi Biotec #130-041-407) and pelleted before starting MACS sorting. Briefly, the pellet was resuspended in 500 µl of MACS buffer (0.5% bovine serum albumin, 2 mM EDTA, 1% Anti-Anti, DPBS) and 10 µl of Integrin alpha 6 antibody (GeneTex #GTX76413) and incubated on ice for 20 min. After washing and centrifuging, the new pellet was resuspended in 80 µl of MACS buffer and 20 µl of anti-Rat Magnetic Beads (Miltenyi Biotec #130-048-502) for 15 min on ice. The solution was washed again and resuspended in 500 µl of MACS buffer and passed through a positive selection column (Miltenyi Biotec #130-042-401) on a magnetic stand (Miltenyi Biotec #130-042-303). The positive fraction (iKCs) was centrifuged and plated on collagen IV and laminin-coated plates with iKC medium. The negative fraction was pelleted and resuspended in anti-Fibroblast beads (Miltenyi Biotec #130-050-601) for 30 min at room temperature. Then, the solution was washed, pelleted, and passed through a new column. The positive fraction (iFBs) was pelleted and plated in 10 cm dishes with FB medium. The negative fraction was frozen and kept in the −80°C.

For iKC culture, the cells were passaged at 80% confluence. After reaching 80% confluence again, the iKCs were sorted with anti-ITGA6 once more to eliminate other cell types that seem to proliferate after culture, and thus obtain a pure iKC culture. For both iFBs and iKCs, cells were frozen to create a cell bank for future experiments.

Engineered skin construct development

To make ESCs, primary or iPSC-derived FBs were harvested at passage 3 and resuspended in a neutralized collagen type I (3 mg/ml; EMD Millipore, #08-155) solution at a density of 165 × 10³ cells/ml. First, 0.2 ml of an acellular collagen solution was first added to a transwell (Fisher Scientific #07200157) in a 12-well plate to promote uniform contraction, followed by 0.6 ml of the dermis solution on top. The dermis was incubated at 37°C for 30 min and was then submerged in medium tailored for either primary or iPSC-derived cells. After 2 days, the primary or iPSC-derived keratinocytes were resuspended in KC medium (1.5 × 10⁵ cells in 10 µl per ESC) and added on top of the dermis. The ESCs were incubated at 37°C for 15 min before submerging in its respective epidermalization medium. A week after keratinocyte seeding, the ESCs were transitioned to air-liquid-interface (ALI) by moving the transwells to a 12-well deep plate (VWR #665110) and adding 4 ml of cornification medium, enough to reach the bottom of the transwell but leaving the top of the ESCs in contact with the air. Five to 7 days after transitioning to ALI, the ESCs were fixed in 4% paraformaldehyde and processed for downstream assays. For dermis-only ESCs, the same process was followed except no keratinocytes were seeded and the constructs were kept submerged in dermis medium for 2 weeks.

For conditioned media co-cultures, three groups were established: a co-culture of primary FBs and primary KCs (condition A), a co-culture of iFBs and primary KCs (condition B), and a co-culture of primary FBs and iKCs (condition C). Media was collected every day and fed to iESCs made from matured iFBs.

Immunofluorescence and confocal imaging

The constructs were fixed in 4% paraformaldehyde overnight at 4°C and transferred to 30% sucrose the next day for at least 24 h at 4°C before being embedded in optimal cutting temperature compound and cryosectioned (16 µm). The resulting sections were either stained with hematoxylin and eosin (H&E) or with fluorescent antibodies.

For IF, the slides were dried overnight in the dark at room temperature. Permeabilization was performed with 0.1% Triton X for 10 min, followed by blocking at room temperature with 5% donkey serum and 8% bovine serum albumin in PBS for 1 h. The following primary antibodies were used and incubated overnight at 4°C: keratin 14 (BioLegend #906004), keratin 10 (BioLegend #905404), keratin 5 (BioLegend #905904), Ki67 (Abcam #ab16667), involucrin (Abcam #ab27495), DSG1 (Thermo Fisher #MA191590), loricrin (Abcam #ab188994), filaggrin (Abcam #ab218395), vimentin (Santa Cruz #sc-6260), fibronectin (Santa Cruz #sc-73611), collagen I (Abcam #ab6308), collagen III (Abcam #ab6310), collagen IV (Rockland #6004011060.1), collagen VII (EMD Millipore #234192), CD90 (BioLegend #328110), FAP (R&D Systems #AF3715-SP), elastin (Abcam #ab21610), enactin (Abcam #ab254325), laminin α5 (EMD Millipore #MABT39).

For staining of 2D cells, the slides were fixed with 4% paraformaldehyde for 20 min at room temperature, permeabilized with 0.3% Triton X for 10 min, and blocked with 5% donkey serum and 8% bovine serum albumin for 1 h at room temperature before incubating with primary antibodies overnight at 4°C. The following primary antibodies were used: DeltaNp63 (Cell Signaling #67825S), TAp63 (BioLegend #938102), ITGA6 (GeneTex #GTX6413), DSG1 (Thermo Fisher #MA191590), keratin 10 (BioLegend #905404), DSG3 (Invitrogen #326300), Ki67 (Abcam #ab16667), keratin 14 (BioLegend #906004 ), laminin α5 (EMD Millipore #MABT39), FAP (R&D Systems #AF3715-SP), collagen IV (EMD Millipore #234192), fibronectin (Santa Cruz #sc-73611), elastin (Abcam #ab21610), CD90 (BioLegend #328110), vimentin (Santa Cruz #sc-6260).

For wholemount staining, the organoids were fixed in 4% paraformaldehyde overnight at 4°C, then incubated with 0.3% Triton X-100, 5% donkey serum, and 8% bovine serum albumin overnight at 4°C. Primary antibodies were incubated for 72 h at 4°C and secondary antibodies were incubated for 48 h on a rocking platform at 37°C. All washing steps were performed with 0.3% Triton X-100 and 0.5% 1-thioglycerol on a rocking platform at room temperature. The organoids were cleared with Ce3D Tissue Clearing solution (BioLegend #A12381). The antibodies used were: collagen VII (EMD Millipore #234192), keratin 14 (BioLegend #906004), LEPr (R&D Systems #AF497), LEF1(Santa Cruz #sc-374412), CD90 (BioLegend #328110).

All samples were incubated with the corresponding Alexa Fluor (ThermoFisher) secondary antibodies and imaged using Leica Stellaris 5 (Leica, Germany).

Flow cytometry

The organoids were collected at day 45 and dissociated with accutase for 30 min at 37°C on a rocking platform, strained with a 30 µm strainer (Miltenyi Biotec #130-041-407) and added to a V-bottom 96-well plate (200K cells per well). The cells were blocked with Human TrueStain FcX (BioLegend #422302) for 10 min and stained with 50× ITGA-PE (Invitrogen MA5-16885) and PDGFRα-APC (eBioscience #17140181) antibodies diluted in Cell Staining Buffer (BioLegend #420201) for 30 min on ice. The compensation was performed with UltraComp eBeads (eBioscience #01222242). The cells were stained with 1000× DAPI (BioLegend #422801) and analyzed right away on a Novocyte Penteon 5 Lasers Cytometer (Agligent Technologies) using the software NovoExpress (Agligent Technologies).

Mechanical testing

Spherical microindentation (Piuma, Optics11Life, Amsterdam, NE) was used to assess the time-dependent material properties of dermis-only iESCs and ESCs. A probe with a spherical 101 radius tip and 0.49 N/m cantilever stiffness were used for microindentation tests. Each test consisted of a 4 × 4 point indentation scan with an interpoint distance of 200 µm, thus obtaining 16 data points in a 0.6 mm × 0.6 mm region. Samples were tested in 1× PBS solution. For every indentation point, the indentation depth was 7 µm, which corresponded to 5% indentation strain. The probe was then held at that depth for 15 s. Stromal tissue exhibits poroelastic and viscoelastic behavior, and the aforementioned testing protocol yields load relaxation that approaches equilibrium. ³⁹ The stiffness and permeability of the samples were determined by fitting the load relaxation curves with an established poroviscoelastic model in Matlab.^40–42 Fitted data points were excluded if loads increased over time.

Quantification and statistical analysis

For 2D and 3D quantification of IF intensity signal, CellProfiler was used to analyze three images from different sections or wells of each marker per condition. DAPI was used to identify the nuclei as a primary object, and cells were identified as secondary objects based on the surrounding signal. For microindentation analysis, three replicates per condition were used. For epidermis thickness quantification, three replicates were analyzed, taking samples from the first slide, 10th slide, and twentieth slide to evaluate epidermis uniformity across the entire construct. ImageJ was used to measure epidermis thickness. For contraction quantification of iESCs, the diameter of four replicates per condition was measured at three different time points, and contraction percentage was calculated from the means relative to day 0. Results were imported into GraphPad Prism (Prism Software, Irvine, CA) and the data was presented as normalized integrated mean fluorescent intensity, contraction percentage, mean elastic modulus, mean permeability, mean epidermis thickness, and mean standard deviation of epidermis thickness. To establish statistical significance, the Shapiro-Wilk normality test, one-way analysis of variance (ANOVA) test, and unpaired t-test were applied using Prism, where *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.