Short-Term Treatment with Rho-Associated Kinase Inhibitor Preserves Keratinocyte Stem Cell Characteristics In Vitro

Primary keratinocytes including keratinocyte stem cells (KSCs) can be cultured as epidermal sheets in vitro and are attractive for cell and gene therapies for genetic skin disorders. However, the initial slow growth of freshly isolated keratinocytes hinders clinical applications. Rho-associated kinase inhibitor (ROCKi) has been used to overcome this obstacle, but its influence on the characteristics of KSC and its safety for clinical application remains unknown. In this study, primary keratinocytes were treated with ROCKi Y-27632 for six days (short-term). Significant increases in colony formation and cell proliferation during the six-day ROCKi treatment were observed and confirmed by related protein markers and single-cell transcriptomic analysis. In addition, short-term ROCKi-treated cells maintained their differentiation ability as examined by 3D-organotypic culture. However, these changes could be reversed and became indistinguishable between treated and untreated cells once ROCKi treatment was withdrawn. Further, the short-term ROCKi treatment did not reduce the number of KSCs. In addition, AKT and ERK pathways were rapidly activated upon ROCKi treatment. In conclusion, short-term ROCKi treatment can transiently and reversibly accelerate initial primary keratinocyte expansion while preserving the holoclone-forming cell population (KSCs), providing a safe avenue for clinical applications.


Introduction
Cultured epidermal sheets generated from primary keratinocytes have been used in autologous epidermal sheet graft therapy for severe burns and chronic ulcers for more than three decades [1]. This approach has recently been combined with gene therapies for rare genetic skin diseases, in which ex vivo gene-corrected epidermal sheets generated from patients' keratinocytes, including keratinocyte stem cells (KSCs), are cultured and grafted onto patients with severe skin damage [2][3][4]. Ex vivo gene-corrected epidermal sheet therapy has been applied for junctional epidermolysis bullosa (JEB), recessive dystrophic epidermolysis bullosa (RDEB), and Netherton syndrome (NS) [2][3][4][5][6][7]. However, the success We investigated these questions by culturing human primary keratinocytes including KSCs obtained from human interfollicular skin with Y-27632 (hereafter referred as 'ROCKi'). Cultured primary keratinocytes were treated with ROCKi for six days and then compared to non-treated cells to examine the changes in cell growth, protein markers and transcriptomics at single-cell resolution. We subsequently withdrew ROCKi in cultures for six days and compared these cultures against non-treated cells in the same way. In addition, we investigated the possible signalling pathways associated with cell proliferation and differentiation following ROCKi treatment in primary keratinocytes.
Freshly isolated primary keratinocytes (P0) were seeded in 10 cm dishes and cultured in the Green's medium containing 10 µM of ROCKi for six days. The culture medium was changed at day 3, replenished with fresh ROCKi. Cells cultured in Green's medium without ROCKi but containing 0.1% DMSO were used for controls and run with treated cells in parallel ( Figure 1). Briefly, cells were seeded in triplicates for each group for assays including cell growth rate and morphology, mitochondrial mass, single cell RNA sequencing (scRNAseq), and immunoblotting. Six days after seeding, cultured cells from both ROCKi-treated (ROCKi-6D + ) and non-treated (Control-6D − ) groups were harvested using 0.05% trypsin in DPBS. Before trypsinization, feeder layer was removed using 0.01% EDTA in DPBS for 30 s at room temperature. Trypsinized cells were neutralized with culture media and centrifuged at 500× g for 5 min at 4 • C. Cell pellets were resuspended in 0.5% fetal calf serum (FCS) in DPBS, and an aliquot was immediately proceeded for cell counting and cell size determination using the automated cell counter (CellDrop BF, DeNovix Inc, Wilmington, DE, USA). Half of the cells were harvested at day 6 (ROCKi-6D − ) and divided into three portions for mitochondrial mass assay, single-cell RNA sequencing 0.5% fetal calf serum (FCS) in DPBS, and an aliquot was immediately proceeded for cell counting and cell size determination using the automated cell counter (CellDrop BF, DeNovix Inc, Wilmington, DE, USA). Half of the cells were harvested at day 6 (ROCKi-6D − ) and divided into three portions for mitochondrial mass assay, single-cell RNA sequencing analysis and immunoblotting. The remaining half cells were seeded back in triplicates at a density of 1 × 10 6 cells per 10 cm dish and grown for a further six days without addition of ROCKi, and then were harvested for analysis. Cells without ROCKi (Control-6D − 6D − and ROCKi-6D + 6D − ) were processed in the same way.

Figure 1.
Workflow of experiments and sample collections. Freshly isolated keratinocytes (passage 0) were divided into two populations, either with ROCKi (10 μM Y-27632) or without ROCKi treatment (0.1% DMSO) for six days. Half of the cells were harvested on day 6 and the remaining cells were passaged and cultured without ROCKi for additional six days and then harvested. Image created with BioRender.com.
For short-time point experiments, P0 keratinocytes were seeded at a density of 1 × 10 6 cells per 60 mm dish in the Green's medium and cultured for two days. On day 2, culture media were changed with 3 mL of fresh Green's medium. On day 4, 1 mL of 40 μM ROCKi in Green's medium was added to the keratinocytes culture already containing 3 mL of culture media so that the final concentration of ROCKi was 10 μM in the culture. Control groups were treated with DMSO in parallel. These cells were incubated for either 15, 30, 60, or 120 min and were immediately rinsed with ice-cold PBS at the end of each time point and lysed directly on the culture plate using 1x Laemmli sample buffer supplemented with fresh 50mM DTT for the detection of AKT1/2 and ERK1/2 signalling.

Keratinocyte Growth Rate
Growth rates of keratinocyte cultures were calculated as previously described [29]. Growth rate was calculated using the equation, growth rate = ln (total number of cells harvested/total number of cells seeded)/duration of culture. For short-time point experiments, P0 keratinocytes were seeded at a density of 1 × 10 6 cells per 60 mm dish in the Green's medium and cultured for two days. On day 2, culture media were changed with 3 mL of fresh Green's medium. On day 4, 1 mL of 40 µM ROCKi in Green's medium was added to the keratinocytes culture already containing 3 mL of culture media so that the final concentration of ROCKi was 10 µM in the culture. Control groups were treated with DMSO in parallel. These cells were incubated for either 15, 30, 60, or 120 min and were immediately rinsed with ice-cold PBS at the end of each time point and lysed directly on the culture plate using 1x Laemmli sample buffer supplemented with fresh 50mM DTT for the detection of AKT1/2 and ERK1/2 signalling.

Keratinocyte Growth Rate
Growth rates of keratinocyte cultures were calculated as previously described [30]. Growth rate was calculated using the equation, growth rate = ln (total number of cells harvested/total number of cells seeded)/duration of culture.

Colony-Formation Assay
The colony forming efficiency (CFE) of the keratinocytes was determined by a colony formation assay as previously described [31]. Keratinocytes were seeded in 10 cm dishes at a density of 800 cells per dish. On day 6 and day 12, dishes cultured with cells were washed twice with DPBS and stained with 0.25% rhodanile blue (Merck Life Science, Poole, UK, #121495) in 1% sulphuric acid/DPBS solution for 20 min at room temperature. Dishes were washed several times with DPBS and air dried. Colony staining image were scanned at the setting of 600 DPI using an HP Envy 4500 (HP, Bracknell, UK) scanner. A ruler was scanned at 600 DPI to measure the number of pixels per millimetre. Colony numbers and areas were counted and measured using Fiji software (https://imagej.net/software/fiji/, accessed Cells 2023, 12, 346 5 of 21 on 22 April 2022). Briefly, images were converted to 8-bit grayscale. Uneven brightness of the staining was adjust using adaptive median filter (r = 2), the threshold was set to default, hollow colonies were filled by using binary fill holes, and joined colonies were separated using binary watershed. Colonies ≥ 1mm 2 in area with the circularity of ≥0.5 were counted.

Immunoblotting
All antibodies used in this study are listed in Table S1. Keratinocytes were lysed directly on the culture plate using 1x Laemmli sample buffer (BioRad, Herts, UK, #1610747) supplemented with fresh 50mM DTT at room temperature. Cell lysates were passed through 20 gauge needle several times and then heated at 96 • C for 5 min. Total protein concentration in each sample was quantified using Pierce 660nM protein assay reagent (Thermo Fisher Scientific, Woolwich, UK, #22662) supplemented with ionic detergent compatibility reagent (IDCR, Thermo Fisher Scientific, Hampshire, UK, #22663). Equal amounts of total proteins were loaded in each lane, separated on Any kD Mini-PROTEAN TGX Stain-Free Protein Gels (BioRad, Hertfordshire, UK, #4568126) and transferred on to 0.22µm Nitrocellulose membrane using Trans-Blot Turbo semi dry transfer system (BioRad, Herts UK). The transferred membranes were blocked for 1 h with 5% w/v non-fat milk in tris-buffered saline containing Tween 20 (TBST, 19mM tris base, 137mM NaCl, 3mM KCl, 0.2% Tween 20, pH 7.4) for 1 h and then incubated with primary antibodies diluted in blocking solution overnight at 4 • C. Membranes were washed three times (ten minutes each) in TBST before the addition of HRP-conjugated secondary antibodies for 2 h at room temperature. Membranes were washed three times (ten minutes each) in TBST, and target proteins were detected by using ECL Prime Western Blotting detection kit (SLS, Wilford, UK, # RPN2232). Signal was visualized and recorded using ChemiDoc MP imaging system (BioRad, Hertfordshire, UK). The membranes were re-used by stripping the previous antibodies using Restore Plus western blot stripping buffer (Thermo Fisher Scientific, Horsham, UK, #46430) and re-probed for the house-keeping protein glyceraldehyde 3-phosphate dehydrogenase (GAPDH) as a loading control. Protein bands were quantified against GAPDH by optical densitometry using Image Lab 6.1 software (BioRad, Wilford, UK).

MitoTracker Green Staining
Primary keratinocytes were loaded with 200nM MitoTracker Green FM (Thermo Fisher Scientific, Horsham, UK, #M7514) and incubated at 37 • C for 30 min. Cells were then washed with DPBS, and feeder cells were removed using 0.01% EDTA in DPBS for 30 s at room temperature. Cells were then harvested by 0.05% of trypsin in DPBS, pelleted and then re-suspended in ice-cold 0.2% FCS in DPBS for flow cytometry analysis using CytoFLEX S (Beckman Coulter, Bucks, UK). Cells were excited at 488 nm and the emission fluorescence was collected at 535 nm. A total of 40,000 live cells were analysed per sample.

3D Organotypic Culture
A 3D organotypic culture was performed as described by Di, Larcher [32]. A deepidermalised dermis (DED, Euro Tissue Bank, BEVERWIJK, Netherlands) was used as a scaffold for generating 3D organotypic culture. A metal ring with 1 cm diameter was placed on the dermal side of the DED in a 35 mm dish and 2 × 10 5 primary human dermal fibroblasts were seeded on the reticular side and cultured in the Green's medium. After 24 h, the metal ring was removed, DED was flipped over (in the papillary side), and the metal ring was placed again on the papillary side. Primary keratinocytes sized 5 × 10 5 were then seeded on the papillary side of the DED in Green's medium. After 48 h, the metal ring was removed, and the DED culture was air-lifted by transferring it on to a mesh so that reticular side of the DED was in contact with the culture medium and the papillary side was in contact with air. Air-lifted DED with cells were cultured for 14 days, allowing the keratinocytes to proliferate and differentiate. Primary keratinocytes from Cells 2023, 12, 346 6 of 21 three different donors were used to generate three biological replicates. Primary human fibroblast (passage 3 to 4) from same donor was used for all 3D cultures.

Cryosectioning and Immunostaining
Supplementary Information.

Single-Cell Library Preparation and RNA Sequencing
Cultured keratinocytes were harvested and resuspended in ice-cold buffer containing 0.5% FCS in DPBS. A small aliquot of the cell suspension was used to determine the percentage of live/dead cells by stained with Acridine Orange/Propidium Iodide (AO/PI) Cell Viability Kit (Vita Scientific, Beltsville, MD, USA, #LGBD10012) and counted by using LUNA-FL automated cell counter (Logos Biosystems, Annandale, VA, USA). Approximately 3000 live cells from each group were loaded into one channel of the Chromium Chip G using the Single Cell reagent kit v3.1 (10X Genomics, Pleasanton, CA, USA) for capturing single-cell Gel Bead Emulsion using the Chromium controller. Following single-cell capture, cells were lysed, and cDNA of individual cells were synthesized within the Gel Bead Emulsion and were used for library preparation. cDNA was amplified and size-selected using SPRISelect (Beckman Coulter, Krefeld, Germany, #B23317) following the manufacturer's protocol. In all, 25 ng of the size-selected cDNA of each sample was used to construct Illumina sequencing libraries. Libraries were pooled at equal concentration and sequenced on the NovaSeq SP100 or NextSeq2000 Illumina sequencing platform at 50,000 reads per cell.

Generation of Gene Expression Matrix of Counts and Quality Control for scRNAseq
The gene expression count matrix was generated from scRNAseq raw sequences using cellranger [33] with the standard cellranger pipeline consisting of the mkfastq, count, and aggr steps. Gene expression was analysed using Seurat [34]. Genes present in fewer than three cells and cells expressing fewer than 200 genes were excluded from the analysis. Low-quality cells and barcodes corresponding to multiple cells or empty droplets were removed by excluding cells associated with either an anomalous number of genes, number of molecules, or percentage of mitochondrial DNA. Preparation of the Gel Bead Emulsion for each sample in separate wells caused these measures to vary between each other. Hence, we used different thresholds for each sample as a universal threshold would include lowquality cells in some samples and exclude high-quality cells in others (Supplementary Figure S5). Non-keratinocytes were removed by excluding cells expressing (>1 count) of markers of melanocytes (MLANA, PMEL, MITF) and mesenchymal-like cells (ACTG2, DLK1, MTRNR2L6, MTRNR2L10, MTRNR2L7, MTRNR2L1, PRKAR2B, NR2F1) [35].

Normalisation and Cell Cycle Removal
Gene expression was normalised using sctransform [36]. The effect of cell cycle was mitigated by assigning scores to each cell representing the probability of its being in different stages of the cell cycle (using Seurat's CellCycleScoring function) and then regressing out the difference between these scores using sctransform (satijalab.org/seurat/ archive/v3.1/cell_cycle_vignette.html-"Alternative Workflow," accessed on 18 May 2021).

Dimensionality Reduction and Donor Effect Removal
Dimensionality reduction was conducted by principal components analysis (PCA) using the 3000 most highly variable genes as features. The 50 most significant components were used since there was no clear point at which the generated components became less significant (Supplementary Figure S6). Inter-donor variation was removed with Harmony [37].

Differential Expression and Differential Abundance Analyses
Seurat was used for differential expression analysis of treated and control cells with its standard Wilcoxon rank sum test. We used DAseq [38] to measure the proportion of two cell populations that cluster separately (detailed in Supplementary Materials).

Cell Type Proportion Comparison
SCINA [39] was used to compare the cell type proportions in the control and ROCKitreated cells. Cell type markers published by Enzo, Secone Seconetti [40] were used to classify cells as either holoclone-forming, clonogenic, or terminally differentiated. Bootstrap resampling was used to estimate confidence intervals around the total absolute differences between cell type proportions.

Trajectory Analysis
Slingshot [41] was used to infer the differentiation trajectory for each sample class with the Harmony reduction. The groups of non-treated and ROCKi-treated cells were combined for the inference procedure to ensure that the pseudotimes were directly comparable. Three clusters were used for each sample (approximately corresponding to holoclone-forming, clonogenic, and terminally differentiated cells) as the input to Slingshot. The clusters likely to contain mostly holoclone-forming cells were used as the starting point for each inferred trajectory. Trajectories were then compared by computing the 0.1th, 0.2th, . . . , 99.9th, 100th percentiles of the pseudotimes corresponding to each sample and calculating the mean absolute difference at each percentile. Bootstrap resampling was used to estimate confidence intervals around the mean absolute differences between the pseudotimes both during ROCKi treatment and following its withdrawal.

Statistics
Experiments were performed in three individual donors (n = 3). Data are presented as mean ± S.D. Statistical analysis was performed using a Student's t-test (two-sample, two-tailed, unequal variance, p < 0.05 was considered significant).
For the single-cell RNAseq analyses, the state-of-the-art statistical analysis tool DAseq was used to identify the overall similarities between samples, and the Wilcoxon Rank Sum test was used to identify differentially expressed genes. Bootstrap resampling was used to create confidence intervals around the differences in cell type proportions differentiation behaviour between treated and control samples, and confidence intervals was used to show that the difference is likely to be at least 9 times larger during treatment than afterwards, which is a statically stronger claim than simply testing for a significant p value.

Accelerated Colony Formation in the Culture with Six-Day ROCKi Treatment Stopped after ROCKi Withdrawal
Freshly isolated keratinocytes without passage (passage zero, P0) containing heterogenous cell populations including KSCs were cultured in Green's keratinocyte culture medium with feeder cells and 10 µM of ROCKi for six days. The culture media were changed at day 3 and replenished with fresh ROCKi. At day 6, half of the cells were harvested for evaluation (ROCKi-6D + ) while the remaining cells were continuously cultivated without ROCKi for further six days (ROCKi-6D + 6D − ). Cells without ROCKi treatment but cultured with 0.1% DMSO were used as negative controls (Control-6D − and Control-6D − 6D − , respectively) ( Figure 1).
Colony forming efficiency (CFE) was assessed based on the size and circularity of colony. A 15-fold increase in colony formation was detected in ROCKi-6D + cells compared to Control-6D − cells (Figure 2a,b, n = 3, p < 0.05). The growth rate of keratinocytes was calculated and there was a fourfold increase in ROCKi-6D + groups compared to Control-6D − groups (Figure 2d,e, p < 0.05). The size of ROCKi-6D + cells were smaller than those in Control-6D − (Figure 2c, p < 0.05). Control-6D − keratinocytes formed tightly packed colonies during the first two days of culture, whereas ROCKi-treated keratinocytes (ROCKi-6D + ) dispersed with loose connections in the first 2-3 days culture but formed tightly packed colonies at day 4 (Supplementary Figure S1). These morphological changes suggested that cells treated with ROCKi were in a proliferating state. Since the mitochondrial content in stem cells is lower when they are dormant, but increases upon activation and proliferation, mitochondrial mass/content were examined Since the mitochondrial content in stem cells is lower when they are dormant, but increases upon activation and proliferation, mitochondrial mass/content were examined using Mitotracker Green assay. Two distinct cell populations with low and high mitochondrial mass/content were observed in both ROCKi-6D + and Control-6D − cells. However, ROCKi-6D + cells had a larger population with high mitochondrial mass than Control-6D − cells (31.84 ± 1.99% v 3.56 ± 3.53%, p < 0.05) (Figure 2f,g). Live cell staining with Mitotracker green showed intensely stained cells located at the periphery of the colonies in which the proliferation marker keratin 14 (K14) was also highly expressed (Supplementary Figure S2), suggesting elevated proliferation of keratinocytes in the culture treated with ROCKi.
Importantly, in cells treated with ROCKi for six days and then passaged and cultured without ROCKi treatment for six additional days (ROCKi-6D + 6D − ), the colony formation efficiency, cell growth rate, cell size and mitochondrial mass returned to the levels comparable to those of non-treated cells (Control-6D − 6D − ) (Figure 3a-g). The markers related to cell proliferation and differentiation were also similar between the ROCKi-6D + 6D − and Control-6D − 6D − cultures (Figure 3h), indicating a transient effect of ROCKi on cells following short-term treatment.

Keratinocytes Treated with Short-Term ROCKi Maintained Differentiation Ability
The differentiation potential in cells with and without ROCKi treatment were evaluated in the organotypic culture in vitro. Fully differentiated multi-epidermal layers including a basal layer containing cube-shaped basal cells and suprabasal layers with stratum spinosum, granulosum, and anucleated cornium were detected in the cultures generated

Keratinocytes Treated with Short-Term ROCKi Maintained Differentiation Ability
The differentiation potential in cells with and without ROCKi treatment were evaluated in the organotypic culture in vitro. Fully differentiated multi-epidermal layers including a basal layer containing cube-shaped basal cells and suprabasal layers with stratum spinosum, granulosum, and anucleated cornium were detected in the cultures generated using Control-6D − , ROCKi-6D + , Control-6D − 6D − , and ROCKi-6D + 6D − cells (Figure 4). This suggested that short-term ROCKi treatment did not alter the differentiation ability of keratinocytes. The expression and distribution of protein markers related to basal cells (∆-P63α) and differentiated cells (K10, involucrin, and filaggrin) were examined. There were no differences in the expressions and distributions of markers of differentiated cells (K10, involucrin and filaggrin) (Figure 4k-y), but the expression of the basal cell marker ∆-P63α increased in ROCKi-treated cells (ROCKi-6D + ) compared to non-treated cells (Control-6D − ) (Figure 4g,h). However, the expression of ∆-P63α became indistinguishable between cultures withdrawal of ROCKi treatment (ROCKi-6D + 6D − ) and non-treated cells (Control-6D − 6D − ), indicating the enhanced proliferation cells did not change cell differentiation potential (Figure 4i,j). As ∆-P63α is a proliferation marker for basal keratinocytes (which include KSCs and transient amplifying cells [42]), the increase in ∆-P63α expression in the 3D organotypic culture generated using cells treated with ROCKi suggests that KSCs and transient amplifying cells were actively transitioning from quiescent to proliferative status in ROCKi-treated cells. ROCKi thus accelerates keratinocyte growth.
Individual keratinocytes that passed QC criteria were initially plotted in the uniform manifold approximation and projection (UMAP) based on the similarity of their transcriptomes. ROCKi-treated cells (ROCKi-6D + ) and non-treated cells (Control-6D − ) formed distinct clusters (Figure 5a), indicating large differences in transcriptomes between the two groups. In contrast, the cells following ROCKi withdrawal (ROCKi-6D + 6D − ) and nontreated cells (Control-6D − 6D − ) clustered together (Figure 5b), revealing the similarity of their transcriptomes. Quantification of the overlap of cell populations by DAseq [37] corroborated this observation, showing that only 10% of cells of ROCKi-6D + clustered similarly to the cells of Control-6D − , whereas 63% of ROCKi-6D + 6D − overlapped with Control-6D − 6D − (Supplementary Figure S4).  Fourteen differentially expressed genes were identified with most two-fold changes (log 2 >1 or <−1) between ROCKi-6D + and Control-6D − (Figure 5c). Eleven genes (VIM, KRT7, KRT19, IFI27, DCBLD2, F3, THBS1, IGFBP7, ARHGAP29, EMP1, FST) were downregulated and three genes (S100A8, FABP5, KRTDAP) were upregulated. VIM is a fibroblast marker. As fibroblasts were excluded from the transcriptomic analysis, it was surprising to observe VIM expression in keratinocyte clusters. To confirm that the analysed data set did not include fibroblasts, the co-expressions of VIM and K14, which is a unique marker for keratinocytes, were analysed in the clusters classified as keratinocytes (Supplementary Figure S4). The results confirmed that a proportion of keratinocytes co-express K14 and VIM. This result was also consistent with other studies in which keratinocytes expressed VIM during wound healing and re-epithelization [43,44]. KRT7 and KRT19 are normally absent in skin, but in cultured keratinocytes, these keratins can present dependent on the culture conditions used [45]. ARHGAP29 (which codes for Rho GTPase activating protein 29) regulates Rho GTPase signalling by interacting with Rap [46]. F3, also known as coagulation factor III, has been reported to highly express in the human epidermal skin during wound healing [47]. The genes S100A8, FABP5, and KRTDAP are associated with terminal differentiation of keratinocytes [48,49]. It is not clear why these gene expressions were changed in cells treated with ROCKi, but it was noticed that once short-term ROCKi treatment was withdrawn, there were no genes with at least a two-fold change between ROCKi withdrawn cells and non-treated cells (Figure 5d). This again indicated that the influence of short-term ROCKi treatment in cells was transient and reversible.

The Proportion of Holoclone-Forming Cells Was Reduced after 6-day ROCKi Treatment but Recovered following Its Withdrawal
As holoclone-forming cells have hallmarks of stem cells [12], the proportion of KSCs was evaluated based on the proportion of holoclone-forming cells. Single-cell transcriptomics was used to quantify the proportion of holoclone-forming cells, which were identified by the co-expression markers of AURKB, CCNA2, CKAP2L, FOXM1, and HMGB2 described by Enzo, Secone Seconetti [40]. Whilst ROCKi-6D + cells expressed higher levels of holoclone-forming cell markers, they had a reduced proportion of holoclone-forming cells compared to non-treated cells (Figure 6a,c, p < 0.05). On the other hand, in cells after the withdrawal of ROCKi treatment, the proportions of holoclone-forming cells were similar between ROCKi-6D + 6D − and Control-6D − 6D − (Figure 6b,c, p < 0.05).
The differentiation behaviour (from holoclone-forming cells to transient amplifying cells and finally to differentiated cells) was further analysed in treated and non-treated cells using single-cell trajectory inference (Figure 7a-d). The estimated pseudotime (differentiation progress) of each cell revealed that the progression from holoclone-forming cells to meroclone-and paraclone-forming cell types was significantly faster in ROCKi-6D + cells than in Control-6D − (Figure 7e, p < 0.05). However, this acceleration disappeared once ROCKi treatment was withdrawn (ROCKi-6D + 6D − ), with the differentiation rate in this culture similar to that of the culture without ROCKi treatment (Control-6D − 6D − ) (Figure 7f, p < 0.05). These results indicated that short-term ROCKi treatment accelerates the differentiation of holoclone-forming cells into meroclone-forming cells, but this acceleration was reversible once treatment was stopped. In addition, the acceleration from holoclone-forming cells to meroclone-forming cells during this short-term treatment was transient and may not exhaust holoclone-forming cells as the corresponding cell numbers were not changed after treatment was withdrawn (Figure 6b).

ROCKi Treatment Rapidly Activated AKT1/2 and ERK1/2 Signalling
It has been reported that Y-27632 treatment could suppress dissociation-induced apoptosis in murine prostate stem cells and iPSCs by inhibiting the activation of PTEN and also by preventing the deactivation of AKT, which, in turn, promotes anti-apoptotic and pro-cell cycle pathways through the putative PI3K-AKT-mTOR signalling [15].

ROCKi Treatment Rapidly Activated AKT1/2 and ERK1/2 Signalling
It has been reported that Y-27632 treatment could suppress dissociation-induced apoptosis in murine prostate stem cells and iPSCs by inhibiting the activation of PTEN and also by preventing the deactivation of AKT, which, in turn, promotes anti-apoptotic and pro-cell cycle pathways through the putative PI3K-AKT-mTOR signalling [16]. Another study showed that U0126, an inhibitor of extracellular signal-regulated kinase 1/2 (ERK), could block Y-27632-induced cell proliferation in human hair follicular culture [19]. However, it is not clear whether ROCKi has these signalling influences in treated keratinocytes. Therefore, the activation of AKT1/2 and ERK1/2 was investigated. As ROCKi prevents dephosphorylation of myosin light chain (MLC2) by inhibiting the activity of MLC phosphatase [50], MLC2 phosphorylation in cells treated with ROCKi was assessed. Phosphorylated MLC2 was decreased within 15 min and at its lowest 30 min post-treatment (Figure 8a,b), indicating the impact of ROCKi in treated keratinocytes. Further, both AKT1 and AKT2 were activated with a significant increase in phosphorylated AKT1 (Ser473) and AKT2 (Ser474) within 15 min of treatment (Figure 8a). Finally, the phosphorylated states of c-RAF (Ser338), MEK1/2 (Ser217/221), and ERK1/2 (Thr202/204), which are involved in the ERK1/2 signalling pathway, were assessed. Cells treated with ROCKi significantly increased the phosphorylation of c-RAF and MEK1/2 in 15 min. However, the level of phosphorylation of ERK1/2 did not change (Figure 8a,b), suggesting a spatiotemporally regulated ERK signalling. Together, these results indicated that ROCKi stimulates primary human keratinocyte proliferation by 'dual' activation of AKT1/2 and ERK1/2 signalling pathways. Another study showed that U0126, an inhibitor of extracellular signal-regulated kinase 1/2 (ERK), could block Y-27632-induced cell proliferation in human hair follicular culture [18]. However, it is not clear whether ROCKi has these signalling influences in treated keratinocytes. Therefore, the activation of AKT1/2 and ERK1/2 was investigated. As ROCKi prevents dephosphorylation of myosin light chain (MLC2) by inhibiting the activity of MLC phosphatase [49], MLC2 phosphorylation in cells treated with ROCKi was assessed. Phosphorylated MLC2 was decreased within 15 min and at its lowest 30 min posttreatment (Figure 8a,b), indicating the impact of ROCKi in treated keratinocytes. Further, both AKT1 and AKT2 were activated with a significant increase in phosphorylated AKT1 (Ser473) and AKT2 (Ser474) within 15 min of treatment (Figure 8a,c). Finally, the phosphorylated states of c-RAF (Ser338), MEK1/2 (Ser217/221), and ERK1/2 (Thr202/204), which are involved in the ERK1/2 signalling pathway, were assessed. Cells treated with ROCKi significantly increased the phosphorylation of c-RAF and MEK1/2 in 15 min. However, the level of phosphorylation of ERK1/2 did not change (Figure 8a,d), suggesting a spatiotemporally regulated ERK signalling. Together, these results indicated that ROCKi stimulates primary human keratinocyte proliferation by 'dual' activation of AKT1/2 and ERK1/2 signalling pathways.

Discussion
Epidermal sheet gene therapies require in vitro-cultured primary keratinocytes containing genetically modified KSCs to produce enough proliferated and differentiated keratinocytes for epidermal sheet formation. Ex vivo epidermal sheet gene therapy has been applied for the treatment of rare genetic skin diseases. The long-term survival of genetically modified epidermal sheet therapy relies on the presence of genetically modified KSCs in in vitro cultured epidermal sheets. However, the stemness of KSCs can be affected by in vitro culturing environment. Freshly isolated KSCs could undergo dissociation-induced cell death (anoikis), causing low yield of KSCs or holoclone-forming cells. This could prolong in vitro cell cultivation to obtain sufficient keratinocytes for epidermal sheet formation. The prolonged epidermal sheet culture could eventually exhaust the stemness of KSCs and hinder the clinical application of epidermal sheet-based therapies. We thus investigated whether the use of ROCKi in initial primary culture could prevent these problems whilst preserving the holoclone-forming cell population and maintaining KSCs' self-renewal properties and the ability of proliferation and terminal differentiation.
Previous studies of ROCKi treatment in keratinocytes isolated from human neonatal foreskin observed increased keratinocyte proliferation at the peak of day 6 with a 50-fold increase in colony forming efficiency [23]. A study has shown that long-term exposure (passage 122) of cultured keratinocytes to ROCKi increased proliferation without inducing any neoplastic transformation [51]. However, there were reports that more than a hundredday treatment of ROCKi in cultured cells could induce faster senescence after withdrawal of ROCKi treatment. In addition, ROCKi-induced expressions of IL-1β and IL-8 in oral keratinocytes treated with long-term ROCKi were not reversible [52]. These studies suggested that ROCKi treatment is beneficial for effective expansion of keratinocytes, but long-term use of ROCKi in culture may exhaust stem cells, leading to premature senescence and loss of stemness. In our study, a short-term (six-day) ROCKi treatment was used at the beginning of primary keratinocyte culture to achieve increased cell survival and efficiently expand KSCs in vitro to avoid the potential senescence. Our results showed that the influence of ROCKi on keratinocytes following a short-term treatment was transient and reversible.
We observed a 15-fold increase in CFE during ROCKi treatment. This was lower than a previously reported 50-fold increase [23], possibly due to differences in tissue origin (the skin from donors with ear reconstruction surgery in this study versus foreskin in Terunuma et al.), culture conditions, and donor-related variability. Despite these differences, we confirmed by related protein markers and single-cell transcriptomic analysis that the accelerated proliferation of primary keratinocytes following six-day ROCKi treatment was due to some holoclone-forming cells transforming to a large number of mero-and paraclone-forming cells. Such an initial boost in keratinocyte growth can save at least 1-2 weeks in the generation of epidermal sheet cultures.
We have noted that ROCKi-treated keratinocytes were migrating non-cohesively with loose cell-cell contacts at the first three days of ROCKi treatment but became tightly packed at day 4. It has been reported that keratinocytes treated with ROCKi had fusiform-like phenotypes (spindle-shaped mesenchymal-like cells) [53] which was similar to cells with depolymerized F-actin [54]. Rho-associated kinase can activate LIMK-kinase 1 (LIMK-1) [55]. As LIMK-1 inactivates the actin depolymerization factor ADF/cofilin binding to F-actin to control actin dynamics, inhibition of Rho-associated kinase by ROCKi can therefore promote activation and binding of ADF/Cofilin to F-actin. This leads to F-actin depolymerization, i.e., loose cell-cell contacts or non-cohesive migration of cells, which we observed in the ROCKi-treated cells. We also detected mesenchymal-like cell morphology at the beginning of ROCKi treatment. Epithelial to mesenchymal transition (EMT) plays an essential role in wound healing. Keratinocytes can undergo EMT during wound healing and stay in this intermediate state (known as mesophase) to meet a high demand of cell proliferation for wound closure [56]. We speculate that isolation and culture of primary keratinocytes in vitro creates a situation like wound healing. ROCKi influences the wound healing-like situation by stimulating keratinocytes into EMT through the way of activating F-actin depolymerization and increased adhesion [54,57].
Mitochondrial mass/content increases in stem cells when the stem cells change from resting to activated states [58]. This is due to transition of glycolysis to oxidative phosphorylation (mitochondria-based ATP production), leading to increase in mitochondrial mass/content by the process known as mitochondrial biogenesis [59]. We observed an increased cell population with higher mitochondrial mass/content in ROCKi-treated cells, but we did not detect any changes at the transcriptional level related to mitochondrial biogenesis. Some studies show that, apart from the biogenesis, the inhibition of mitophagy in stem cells could also lead to increased mitochondrial mass [60]. This is because, in somatic cells, mitophagy selectively eliminates damaged mitochondria, whereas in stem cells, mitophagy may serve as a physiologic mechanism for the metabolic rewiring of differentiating stem cells by removing old mitochondria and replacing with new mitochondria [58,59]. We also observed that cells with higher mitochondrial mass and K14 expression were localised in the peripheries of the cell colonies. This indicated that the higher mitochondrial mass cells were at the status of proliferation in which they could relate to the KSCs differentiation to TA cells.
Transcriptomic analysis at single-cell resolution revealed gene expression changes between treated cells and non-treated cells during ROCKi treatment. We detected VIM and F3 differentially expressed in keratinocytes during treatment. Studies by others have also shown that both VIM and F3 are expressed in the skin, in particular during wound healing [47,61]. Although KRT7, KRT18 and KRT19 have been reported to be expressed in cultured keratinocytes, their exact biological role KRT7, KRT18 and KRT19 in cultured cells is not clear. As they have been considered carcinoma-related markers, the reduction in these genes may at least suggest that ROCKi-induced proliferation is not related to oncogenic transformation [43,44]. We noticed that there was no difference in the protein level of K14 in cells with and without ROCKi treatment. Several potential reasons may be behind this. For example, (i) K14 may not be a specific/unique marker for proliferation cells; (ii) early passaged primary keratinocytes (in our study) may be predominantly comprised of proliferation cells compared to differentiated cells. As abundant K14 were expressed in these cultures, minor change in K14 at protein level could not be detected by immunoblot. We used ANLN, AURKB, CCNA2, CKAP2L, FOXM1, HMGB2 and LMNB1 as holoclone-forming cell markers, KRT14, TP63, ITGA6, ITB1, BIRC5 as clonogenic markers, and SERPINB3, SFN, KRT10, TGM1, IVL, SPINK5 as differentiated cell markers [40] to identify the corresponding populations. We also analysed the rates at which cells differentiate through these cell types. Holoclone-forming cells in ROCKi-treated cell cultures produced mero-and paraclone-forming cells at a faster rate than non-treated cell cultures. Combined with the evidence of decreased holoclone-forming cell proportion and increased transcriptome changes in proliferation-related markers in ROCKitreated cells, our results suggest that holoclone-forming cells rapidly produced proliferating TA cells (meroclone cells), thereby increasing the cell numbers. This agrees with our observation of increased mitochondrial mass in ROCKi-treated cells. The observed decrease in the proportion of holoclone-forming cells during the ROCKi treatment was due to the increased proportion of proliferating TA cells while the absolute number of holoclone-forming cells were not significantly different between ROCKi-treated cells and control cells. Importantly, transcriptomic analysis further indicated that the proportion of the holoclone-forming cell population in ROCKi-treated cultures was not reduced after the withdrawal of a short-term ROCKi treatment, indicating that the holoclone-forming cells were not exhausted.
We also checked whether cells maintained the ability to terminally differentiate following short-term ROCKi treatment in the 3D culture system. Both non-treated and ROCKi-treated cells were able to develop the epidermal-like structure with multiple layers and expressions of proliferation and differentiation protein markers. Notably, the proliferation marker ∆-P63α was upregulated in the cultures generated using ROCKi-treated cells, but this did not affect the process of terminal differentiation as there were fully developed cornified layers in the 3D culture generated using ROCKi-treated cells. In agreement with others observation, the level of ∆-P63α was similar in the cultures generated using nontreated cells and ROCKi withdrawal cells. This further confirmed the transient effect of short-term ROCKi treatment.
ROCKi plays its biological function via AKT1/2 and ERK1/2 signalling that regulates cell survival, growth, proliferation, and differentiation. The temporal activation of these kinases required to support cellular proliferation and differentiation is cell-type-dependent. Sustained (more than 2 h) ERK activation promoted proliferation in fibroblasts [62], but in neuronal-derived PC12 cells, sustained ERK activation inhibited proliferation, whereas transient activation (15 min) promoted PC12 cell proliferation [62]. Another study showed that the activation of ERK by epidermal growth factor could peak at 10 min but returned to basal level within 30 min [63]. Similar results were observed in HaCat cells where tumour necrosis factor (TNF) induced cell proliferation via ERK activation which peaked at 15 min but returned to basal level within 30 min in the presence of EGF receptor inhibition [64]. We also observed that ROCKi transiently activated both ERK and AKT signalling with the peak activation at 15 min in primary keratinocytes, but returned to basal level within 30 min. It suggests that primary keratinocyte proliferation stimulated by ROCKi is also caused by temporal activation of ERK and AKT, i.e., transient activation (15 min), but not sustained activation of AKT and ERK signalling.
In conclusion, while short-term (6-day) ROCKi treatment improved holoclone-forming cell survival and growth, the influence on keratinocytes was transient, reversible, and did not affect the characteristics of holoclone-forming cells. Our results provide new insights about the effect of ROCKi on human skin epithelial cell proliferation and differentiation. This could open a potential avenue for clinical application by leading to application for regulatory approval and clinical testing which benefits patients with rare genetic skin diseases.
Supplementary Materials: The following supporting information can be downloaded at: https:// www.mdpi.com/article/10.3390/cells12030346/s1, Figure S1. Cell morphology and colony formation; Figure S2. Cyto-staining for mitochondrial mass and K14; Figure S3. Unsupervised UMAP plot shows the distribution of VIM in cell clusters; Figure S4. Differential abundance analysis using DAseq; Figure S5. Quality control (QC) thresholds; Figure S6. An elbow plot showing the standard deviation of the first 50 principal components. Table S1  Informed Consent Statement: Skin samples were obtained from healthy children (10-to 16-year-olds) undergoing plastic surgery for ear reconstruction, following informed consent.

Data Availability Statement:
The R markdown notebook used to generate our single-cell RNAseq results is available at github.com/george-hall-ucl/reversibility_of_rocki_paper_scrnaseq/blob/main/ data_analysis_for_paper.md. Single-cell RNAseq data have been deposited in the Gene Expression Omnibus database under accession code GSE207130.