Human skin samples were obtained from healthy donors at Severance Hospital, Yonsei University. All human samples were collected with informed consent for research use, and the study design was approved by the Institutional Review Board (4-2019-0582). Keraskin [14] is a commercially available RHE model, which was obtained from Biosolution Co. (Seoul, Korea). Ex vivo pig skin obtained from Apures Co. (Gyeonggi, Korea) was prepared as previously described [15]. Rabbit skin was obtained from Biosolution Co., Ltd. (Seoul, Korea). The human, rabbit and pig skin were obtained from the chest, back and ear, respectively.

The epidermis of the skin can be sub-stratified into the stratum basale, stratum spinosum, stratum granulae, stratum lucideum (only in the foot pad), and stratum corneum [16]. The thickness of the stratum, and the granular, spinous, and basal layers constituting the epidermis was measured by IHC staining with specific marker proteins [14], including loricrin (LOR) for the granular layer, keratin 10 (KRT10) for the spinous layer, and keratin 5 (KRT5) for the basal layer. In addition, filaggrin (FLG), claudin (CLDN1), and E-cadherin (CDH1) expressions were compared among the models, which are considered important proteins for the manifestation of skin barrier function in the epidermis. For IHC, the skin samples were sequentially rehydrated with a graded series of ethanol. Next, pH 6.0 antigen retrieval (DAKO, S1699, Santa Clara, CA, USA) was conducted using a high-pressure cooker for 15 min, followed by a cooling phase over 1 h until the solution was fully transparent. After two washes in phosphate-buffered saline (PBS), the sections were incubated in 3% H2O2 for 30 min to block endogenous peroxidase. After another three washes in PBS, the sections were incubated with a protein block (DAKO, X0909, Santa Clara, CA, USA) for 1–2 h at room temperature in a humidity-controlled chamber. The sections were then incubated with the following primary antibodies overnight at 4 °C: anti-LOR (Abcam, ab24722, 1:2000), anti-KRT10 (Abcam, ab76318, 1:3000), anti-KRT5 (Abcam, ab52635, 1:100,000), anti-MKI67 (Abcam, ab16667, 1:1000), anti-FLG (Novus, NBP1-87,528, 1:1000), anti-CLDN1 (Cell Signaling Technology, 13,255, 1:200), and anti-CDH1 (Cell Signaling Technology, 3195, 1:400). After three washes in PBS, the sections were incubated in horseradish peroxidase (HRP)-labeled anti-rabbit antibody (DAKO, K4003, Santa Clara, CA, USA) for 15 min at room temperature. For development of the HRP-labeled antibody on the sections, DAB (DAKO, K3468, Santa Clara, CA, USA) was diluted and placed on each section for the same period of time. Mayer’s hematoxylin (DAKO, S3309; Santa Clara, CA, USA) was used for counterstaining.

All slides were scanned using a virtual microscope scanner (MoticEasyScan One, Hong Kong) at 40 × optical magnification. Thickness was measured in randomly selected fields of view for each model using QuPath software. Epidermal thickness was measured from the beginning of the epidermis to the end of the cornified layer on hematoxylin and eosin (H&E)-stained slides. The thickness of the epidermal layers (cornified, granular, spinous, and basal layers) was measured from the bottom to top direction consistently. The proliferation index was measured in fields divided into four zones for each model using QuPath software. The DAB-positive nucleus percentage was normalized to the total nucleus (hematoxylin + DAB) counts.

The deposition percentage of the epidermal layer was calculated as the sum of the average lengths of the granular, spinous, and basal layers of each skin model, and the proportion occupied by each epidermal layer to the sum was calculated.

Statistical analyses were performed using GraphPad Prism v7.0. Statistical significance was determined using unpaired two-tailed Student’s t tests. Data are presented as mean ± standard error of the mean; p < 0.05 was considered statistically significant.