Patient informed consent was obtained for the skin sampling as well as further investigations and experimental use. The study was authorized by the ethic committee of the Canton Zurich (KEK-ZH-Nr. 2015–0247 and BASEC No. PB_2020–00066) and conducted according to the Declaration of Helsinki Principles. Fetal skin biopsies were sampled during spina bifida aperta fetal surgery procedures at gestational week 24–25 for cell isolation and embedment in OCT compound (Sakura Finetek, Switzerland) with subsequent storage at - 20°C for later histological examination.

Skin cell isolation and the following cultivation was performed as previously described in Böttcher-Haberzeth et al. (Böttcher-Haberzeth et al., 2013). Briefly, skin samples were reduced to small pieces (about 3 mm³) and digested in 12 U/mL dispase (BD Biosciences, Switzerland) combined with Hank’s balanced salt solution containing 5 mg/ml gentamicin (all from Invitrogen, Switzerland) at 4°C overnight. Thereafter, forceps where used to separate the epidermis and dermis for subsequent cell isolation. Epidermal cells were extracted from the epidermis by further digestion in 1% trypsin and 5 mM EDTA (both from Invitrogen, Switzerland) for 10 min at 37°C and afterwards cultured in serum-free keratinocyte medium containing 25 mg/ml bovine pituitary extract, 0.2 ng/ml epidermal growth factor, and 5 mg/ml gentamicin (all from Invitrogen, Switzerland). Further digestion of the dermal tissue in 2 mg/ml collagenase blend F (Sigma, Switzerland) for 4 h at 37°C maintained to the extraction of fibroblasts, which were further cultivated in Dulbecco’s modified Eagle’s medium (DMEM) containing 10% fetal calf serum (FCS), 4 mM L-alanyl-l- glutamine, 1 mM sodium pyruvate, and 5 mg/ml gentamicin (all from Invitrogen, Switzerland).

Preparation of the dermo-epidermal skin analogues in six-well cell culture inserts with 3.0 μm pore size membranes, in a transwell system (BD Falcon, Switzerland), was performed as described in Biedermann et al. (Biedermann et al., 2015). In brief, the corresponding dermal compartment consisting of a membrane covered with collagen type I hydrogel containing human fetal fibroblasts was prepared first. Therefore, collagen type I (Symatese, France) was mixed with neutralization buffer containing NaOH and 1 × 10⁵ fibroblasts (passage 2) for polymerization. Those dermal analogues were further cultivated for 7 days in DMEM containing 10% FCS. After cultivation, 5 × 10⁵ keratinocytes (passage 2) were distributed onto each dermal equivalent. After one additional week of cultivation in SFM (Invitrogen, Switzerland), with medium change every second day, the dermo-epidermal skin substitutes were ready for transplantation.

The approval for the surgical procedure, performed on immuno-incompetent female nu/nu rats, 8–10 weeks old (Harlan Laboratories, Netherlands), was obtained from the local committee for Experimental Animal Research (permission number ZH 182/2016). The operation and transplantation procedure were conducted as described in Mazzone et al. (Mazzone et al., 2014). In brief, a full-thickness skin defect was prepared for the placement of the dermo-epidermal skin grafts. Additionally, to prevent wound closure from the surrounding rat tissue and therefore to protect the skin analogues, a surgical steel ring (diameter 2.6 cm) was sutured into the skin defect, using non-absorbable polyester sutures (Ethibond^®, Ethicon, United States). Coverage with a silicone foil (Silon-SES, BMS, United States ) was used for further protection of the transplants, and gauze (Sincohaft, Puras, Switzerland) and tape (Leukoplast, BSN Medical, Germany) as standard wound dressing. Photographic documentation and dressing changes were conducted weekly. After 4 weeks in vivo, the skin analogues were excised and embedded in OCT compound for further analysis.

Cryosections of OCT embedded fetal skin and skin analogue explants as well as freshly isolated and cultured fetal keratinocytes were further analysed by immunofluorescence staining (Pontiggia et al., 2009). Therefore, freshly isolated or cultivated keratinocytes (passages 0–2) were centrifuged to glass microscope slides, using attached cytofunnels (Shandon CytoSpin 4, Thermo Fisher Scientific, Switzerland). Per slide 200.000 cells were seeded, air dried, fixed in acetone/methanol 1:1 for 5 min at −20°C and stored at 4°C prior to the staining. Sliced cryosections (12 μm) were fixed and permeabilised using acetone/methanol 1:1 for 5 min at −20°C, subsequently air-dried, washed in phosphate-buffered saline (PBS, Invitrogen, Switzerland) and finally blocked in PBS containing 2% BSA (Sigma, Switzerland) for 30 min. Following primary antibodies were used for immunofluorescence analysis: CK20 (1:50, clone K20.8, Dako, Switzerland), Laminin 5 (Lam5, 1:50, clone P3H9-2, Santa Cruz, United States ), NF200 (200 kDa neurofilament, clone NF01, 1:50, Abcam, Germany), Ki67 (1:50, clone B56, BD Pharmingen, Switzerland), CK8 (1:50, K8.8, Abcam, Switzerland), Chromogranin A (1:50, clone C-12, Santa Cruz, United States ), Synaptophysin (1:50, clone SY38, Progen, Germany), CK15 (1:50, clone LHK15, Chemicon, Switzerland), CK10 (1:50, clone DE-K10, Dako, Switzerland), Piezo2 (1:50, NBP1-78624, Novus Biologicals, Switzerland).

After incubation with primary antibodies for 45 min in 2% BSA in PBS, slides were washed 3 times for 5 min. Thereafter the staining was continued by blocking with 2% BSA in PBS for 15 min and incubation with secondary antibody for 1 h in 2% BSA in PBS, followed by 5 min washing of the slides for three times and additional blocking with 2% BSA in PBS for another 15 min before incubation with additional primary antibody.

Secondary TRITC- or FITC- (Abcam, Germany) conjugated polyclonal goat F (abʹ)2 antibodies were used to visualize the primary antibody or Alexa 555-conjugated polyclonal goat F (abʹ)2 fragments were used to pre-label the primary antibodies according to the manufacturer’s instructions (Zenon Mouse IgG Labelling Kit, Molecular Probes/Invitrogen, Switzerland). For additional visualization of the cell nuclei, all slides were incubated for 5 min in PBS containing 1 μg/ml Hoechst 33342 (Sigma, Switzerland), washed twice for 5 min in PBS, and finally mounted with Dako mounting solution (Dako, Switzerland).

Microscopic evaluation of the immunohistochemical staining was performed using a DXM1200F digital camera connected to a Nikon Eclipse TE2000-U inverted microscope equipped with filter sets for Hoechst 33342-, FITC-, and TRITC (Nikon AG, Switzerland; Software: Nikon ACT-1 vers. 2.70). The elaborated graphical material was further processed with Photoshop 11.0 (Adobe Systems Inc., Germany).

The fraction of Merkel cells in primary freshly isolated keratinocytes (freshly isolated) and during in vitro culturing at the point of passaging (0P and P1) was determined by analyzing 100′000 cells using a Shandon CytoSpin 4 (Thermo Fisher Scientific, Switzerland). To control for donor variability, several primary cultures were analyzed (n = 6). After depositing cells on coated DoubleCytoslide slides (Thermo Fisher Scientific, Switzerland), cells were fixed for 5 min in ice-cold acetone/methanol and stained for CK20. Fluoroshield with DAPI (Sigma, Switzerland) was used to mount coverslips and counterstain nuclei. CK20 positive cells were manually identified and counted using a Nikon Eclipse TE-2000-U confocal microscope (Nikon AG, Switzerland; Software: Nikon ACT-1 vers. 2.70). Statistical analysis was used to compare CK20 positive fractions. Ordinary one-way ANOVA with Tukey’s multiple comparison test was used to analyze differences in population means (Prism 8, Vers. 8.3.1). A p-value <0.05 was considered statistically significant.

Merkel cells in human fetal skin and in tissue-engineered skin sections were identified and counted manually based on CK20 expression using ImageJ software. The data are presented as the number of CK20⁺ Merkel cells per 1 mm of basal cell layer. In each case counts were performed using five microscopic fields (×40 magnification) from two independent biological donors. Data are presented as mean ± SD.

Merkel cells in human fetal skin sections were identified and counted manually based on CK20 expression using ImageJ software. Subsequently, oval and dendritic Merkel cells were quantified over total number of CK20⁺ Merkel cells. Five microscopic fields at × 40 magnification were used in each group (n = 5 per condition; from three independent donors). Data are presented as a mean ± SD.

We performed combined immunofluorescence co-staining against cytokeratin 8 (CK8) and CK20 in order to visualize Merkel cells in skin tissue sections from 24 to 25-week old human fetuses (Figure 1A–A’). In addition, the epidermal basal layer localization of CK20 positive Merkel cells was further confirmed by an immunofluorescence co-staining with Laminin 5, a basement membrane marker (Figure 1B). In fetal skin, migrating Merkel cells still exhibited dendritic protrusions (white arrow, Figure 1B), whereas a more oval shaped cells were visible only when they reached maturation at their target location in the basal cell layer (white arrow, Figure 1C). Quantification revealed that the number of oval and dendritic Merkel cells in human fetal skin is similar (50.9 ± 6.8% vs. 49.1 ± 6.8%, respectively, p > 0.05, not significant) (Supplementary Figure S1A).

Furthermore, CK20-expressing Merkel cells were found in close proximity to CK15 positive basal layer keratinocytes (Figure 1C). The different localization of basal Merkel cells from the suprabasal terminally differentiated keratinocytes was verified by CK20 and CK10 co-staining (Figure 1D).

Moreover, we confirmed the neuroendocrine nature of Merkel cells by the expression of Chromogranin A in all CK20⁺ cells (Figures 1E–E’’). Further, the presence of the mechanosensitive and stretch-gated ion channel Piezo 2 (Piezo-type mechanosensitive ion channel component 2) was also detected in CK20⁺ Merkel cells (Figures 1F–F’’). Interestingly, distinct shapes of Merkel cells were observed in the skin sections of human fetuses, particularly oval and dendritic cells. The innervation of matured oval-shaped CK20⁺ Merkel cells was further visualized by presence of neurofilament 200 (NF200) positive nerve endings (Figure 1G; Supplementary Figures S3A–C), as NF200 is a late-stage marker for mature afferents. In contrast, non-matured Merkel cells were not associated with NF200⁺ peripheral nerve endings (Supplementary Figures S2A–C) and exhibited dendritic phenotype.

Importantly, we did not detect Ki67 expression in CK20⁺ Merkel cells (Figures 1H–H’), demonstrating that Merkel cells were not in a cell cycle.

Next, we isolated epithelial cells from human fetal back skin biopsies and we cultured them on tissue culture plastic. We investigated the percentage of CK20-expressing cells directly after isolation and in the cultured keratinocytes at passage 0 (P0) and 1 (P1) (Figures 2A–D).

Immunofluorescence staining of cytospinned keratinocytes directly after isolation revealed that 0.037 ± 0.012% of all cells were positive for CK20. In culture, the percentage of CK20-expressing Merkel cells significantly decreased to 0.011% ± 0.007% already at P0 and even further to 0.006 ± 0.006% at P1. The decrease was significant between freshly isolated cells and P0 (p < 0.001), as well as between freshly isolated cells and P1 (p < 0.001). Bright field picture in Supplementary Figure S4 represents a typical morphology of human fetal keratinocytes at P0 and the positivity of Merkel cells for CK20 (green) and CK19 (red) (Supplementary Figures S3A–B’).

In addition to CK20, CK8 is also considered as a marker for Merkel cells. Therefore, we also investigated the co-expression of CK8 and CK20 in the freshly isolated epidermal cell mixture as well as in fetal keratinocytes at different passages (Figures 2A–C). We observed that in the freshly isolated epidermal cell mixture all CK20⁺ Merkel cells were in addition positive for CK8 (Figure 2A). Interestingly, CK8 expression in fetal epidermal cells increased markedly from isolation to higher (P0 and P1) passage numbers, but almost all CK8 positive cells were CK20 negative (Figures 2B,C). Therefore, we concluded that CK8 was not a suitable marker for detecting Merkel cells in vitro and therefore, we continued to use only CK20 as a marker for Merkel cells.

The presence of the Piezo2 cation channel in Merkel cells has previously been demonstrated for human cells (García-Mesa et al., 2017; García-Piqueras et al., 2019). Here, we support this finding by revealing sustained co-expression of Piezo2 in CK20⁺ Merkel cells in freshly isolated epidermal cell mixture as well as over several passages (P0, P1 and P2) in vitro by immunofluorescence co-staining (Figures 3A–D). Moreover, neuroendocrine markers such as chromogranin A and synaptophysin (SYP) were continuously expressed in CK20 positive cells on cell culture plastic (Figures 4A–B’’). Taken together, these results show that Merkel cells retained their neuroendocrine character also in vitro after passaging.

Further, we analysed the expression of Ki67, a common proliferation marker indicating cells currently in the cell cycle (Sun and Kaufman, 2018). The majority of keratinocytes in vitro is actively dividing and hence positive for Ki67 (Figures 4C,C’). In contrast, we were not able to find any CK20⁺/Ki67⁺ double positive cell, implying that CK20 Merkel cells are not mitotically active during in vitro culturing.

We then used cultured fetal epithelial cells and dermis derived fetal fibroblasts both at P1 to prepare dermo-epidermal skin substitutes. Further, we characterized the expression of Merkel cell markers 4 weeks after transplantation of these skin grafts onto immune-incompetent nude rats.

Most Merkel cells showed a migrating phenotype with dendritic protrusions in the transplanted human skin substitutes (Figures 5A, C, E–G). Strikingly, the prevalent expression of CK8 among the CK20-negative keratinocytes on culture plastic was not detected anymore in vivo. Therefore, CK8 could be considered as a marker for Merkel cells in vivo (Figures 5A,B). CK20/CK8 double positive cells were exclusively localized either in close proximity to or in contact with the basal lamina (Figures 5C,D). CK20 expressing cells were not observed in any of the stratified CK10-positive suprabasal layers (Figures 5E,F). Furthermore, Merkel cells mostly appeared as single cells, and only very rarely two or more cells were found adjacent to each other (Figures 5C–F). The quantification revealed that in tissue-engineered dermo-epidermal skin substitutes the number of CK20⁺ Merkel cells per 1 mm of basal cell layer is 1.88 ± 0.41 (Supplementary Figure S1B). This number is comparable to the number of Merkel cells in native human fetal skin (2.96 ± 1.27, p > 0.05, not significant).

In addition, even though early innervation from the host animal into the bioengineered dermo-epidermal skin were demonstrated based on NF200 immuno-staining, these nerve fibers were not yet associated with Merkel cells (Figure 5G,G’).

To confirm further the non-proliferative nature of Merkel cells in vivo, we investigated the expression of Ki67 in all CK20⁺ cells (Figure 5H). As observed in fresh isolations and in vitro cultures, CK20⁺ Merkel cells also did not express Ki67 in the skin substitutes in vivo.

We further performed immunofluorescence staining to investigate the presence of mechanosensitive and neuroendocrine markers in Merkel cells in the human skin substitutes in vivo. We found that CK20-positive Merkel cells also displayed Chromogranin and synaptophysin, which are both considered as a marker for neuroendocrine cells (Figure 6A). In addition, Piezo2 was also identified in the CK20-expressing cells in the tissue-engineered skin substitutes confirming the identity of these cells (Figure 6C).