Serum lipids, retinoic acid and phenol red differentially regulate expression of keratins K1, K10 and K2 in cultured keratinocytes

Abnormal keratinocyte differentiation is fundamental to pathologies such as skin cancer and mucosal inflammatory diseases. The ability to grow keratinocytes in vitro allows the study of differentiation however any translational value is limited if keratinocytes get altered by the culture method. Although serum lipids (SLPs) and phenol red (PR) are ubiquitous components of culture media their effect on differentiation is largely unknown. We show for the first time that PR and SLP themselves suppress expression of differentiation-specific keratins K1, K10 and K2 in normal human epidermal keratinocytes (NHEK) and two important cell lines, HaCaT and N/TERT-1. Removal of SLP increased expression of K1, K10 and K2 in 2D and 3D cultures, which was further enhanced in the absence of PR. The effect was reversed for K1 and K10 by adding all-trans retinoic acid (ATRA) but increased for K2 in the absence of PR. Furthermore, retinoid regulation of differentiation-specific keratins involves post-transcriptional mechanisms as we show KRT2 mRNA is stabilised whilst KRT1 and KRT10 mRNAs are destabilised in the presence of ATRA. Taken together, our results indicate that the presence of PR and SLP in cell culture media may significantly impact in vitro studies of keratinocyte differentiation.


Effect of SLP and PR on the expression of differentiation-specific keratins in 2D and 3D cultures.
We investigated the effect of SLP and PR on the expression of these keratins in NHEK cells growing in 2D cultures by immunocytochemistry. In monolayer cultures there was a clear difference in the K1, K10 and K2 expression when SLP and PR were removed from the medium but there was no difference in the expression of K14 ( Fig. 2A).
To investigate whether the presence of SLP and PR would affect K1, K10 and K2 expression in differentiating keratinocytes, we produced 3D cultures using NHEK and human dermal fibroblasts (HDF) under the 4 different conditions of SLP and PR (SLP + /PR + , SLP − /PR + , SLP + /PR − , SLP − /PR − ). Although the 3D cultures were grown for 10 days, the morphology of the suprabasal layers indicates that terminal differentiation was not complete. Nevertheless, keratinocyte differentiation was evident as indicated by expression of differentiation-specific markers. As expected, removal of SLP in the absence of PR induced keratinisation of the stratified epithelia. As shown in Fig. 2B, the lowest expression of K1 and K10 was observed in keratinocytes grown in SLP and PR containing medium (SLP + /PR + ) and the highest expression was observed when SLP were removed and the medium did not contain PR (SLP − /PR − ). This shows that keratinocytes became more differentiated in SLP − /PR − medium. This set of data is consistent with the western blotting data presented in Fig. 1D-G. The expression of K2 followed a pattern similar to K1 and K10 in the four different conditions except that the level of expression was low. In the absence of SLP and PR (SLP − /PR − ) expression of K2 was strongly induced in keratinising layers (Fig. 2B).  10 6 irradiated 3T3 fibroblasts in 2 ml growth medium per well of a 12 well plate for 3 days in four different RM + media conditions, PR containing RM + with or without SLP containing FCS (SLP + /PR + , SLP − /PR + ), no PR in RM + with or without SLP containing FCS (SLP + /PR − , SLP − /PR − ) in triplicates for each growth condition. Feeder cells were removed by squirting PBS/EDTA and keratinocyte lysates were collected and analysed using qPCR and WB. mRNA expression data are shown for different genes (A) KRT1, (B) KRT10 and (C) KRT2 as fold expression normalised to the expression of two housekeeping genes, POLR2A and YAP1 under different growth conditions. (D) NHEK (500,000 cells) were mixed with 1 million irradiated 3T3 fibroblasts in 4 ml growth medium per well of a 6 well plate for 3 days in the above four different RM + media conditions in triplicates for each growth condition. Feeders were removed as above and SDS cell lysate containing 20 µg total protein (measured by Lowry's method using BSA as standard) was analysed by western blotting for expression of K1, K2 and K10 under different growth conditions. GAPDH was used as a loading control. Relevant bands were cropped from different blots and grouped together. Original blots are shown in Supplementary Fig. S1. Quantification of protein bands was carried out by Image J and shown for (E) K1, (F) K10 and (G) K2. Statistical analyses: n = 3, Error bars = SEM, Student's t-test was performed to calculate p values using Microsoft Excel. One-way ANOVA (shown by a horizontal line over the graph) was used to determine the statistical significance in gene expression between different growth conditions. The p values are given by asterisks (*p < 0.05, **p < 0.01, ***p < 0.001 and ****p < 0.0001). SLP + /PR + was very low (Fig. 3B,D, Table 1). Absence of SLP with PR (SLP − /PR + ) did not make any difference in the level of KRT2 mRNA expression, however, it was increased significantly in the medium containing SLP + / PR − (Fig. 3C, Table 1). In N/TERT-1 cells, removal of SLP in the presence of PR (SLP − /PR + ) did not change KRT1 and KRT10 expression but decreased KRT2 expression. However, K1 and K10 expression was increased in SLP + / PR − compared with SLP + /PR + and SLP − /PR + (Fig. 4D). PR was able to suppress KRT1, KRT10 and KRT2 expression in the presence of SLP but both the mRNA and protein expression increased when PR was removed from the medium (compare SLP + /PR + with SLP + /PR − in Fig. 4A-D, Table 1). This expression pattern is different from the one obtained with HaCaT (Fig. 3). These data suggest that the method of immortalisation also determines how these genes respond to SLP and PR.

Cell types Genes
Influence of ATRA and PR on KRT1, KRT10 and KRT2 mRNA expression in NHEK. Although SLP include fatty acids, retinoic acids and their metabolites, we argued that the effect of removing SLP from FCS was primarily due to the presence of retinoids in SLP. We therefore attempted to replicate the effect of SLP by adding back all-trans retinoic acid (ATRA) and evaluating the expression of KRT1, KRT10 and KRT2. NHEK were grown with γ-irradiated feeder fibroblasts in RM + with CS-FCS (SLP − ) either in the presence or absence of PR. ATRA was added at a final concentration of 1 µM, 2 µM or 3 µM in NHEK culture medium for 24 h after which the cells were lysed to estimate mRNA expression by qPCR. As shown in Fig. 5, the mRNA expression of both KRT1 and KRT10 was reduced significantly in the presence of ATRA whether PR was present or not. Presence of PR reduced the effect of ATRA on KRT1 but not KRT10 (Fig. 5A,B). For KRT2, the expression pattern was starkly different as in the presence of PR the KRT2 mRNA was not influenced by ATRA but showed significant increase in the absence of PR (Fig. 5C).
To further substantiate the effect of PR, we grew NHEK with feeder cells in SLP − /PR − RM + medium in the absence of PR. Cells were grown for 72 h with 1 µM of ATRA or 0.01 mg/ml PR (the concentration used in normal culture medium) added in the last 24 h before harvesting the cells for qPCR analyses. As shown in Fig. 6, the expression of KRT1 and KRT10 was reduced in ATRA as well as in PR treated cells (Fig. 6A,B). However, when PR and ATRA were added together the effect was not synergistic, instead PR reduced the effect of ATRA on KRT1 and KRT10 expression (Fig. 6A,B). Interestingly, the expression of KRT2 in presence of ATRA did not change (Fig. 6C). This shows that ATRA treatment suppresses steady state level of KRT1 and KRT10 mRNA but does not influence KRT2 mRNA level. www.nature.com/scientificreports www.nature.com/scientificreports/ Effect of β-Estradiol on the expression of KRT1, KRT10 and KRT2 mRNA in NHEK. Due to the reported structural similarity between PR and β-Estradiol (ED) 52,53 , we hypothesised that PR could be mimicking the effect of ED. We therefore investigated the effect of ED on KRT1, KRT10 and KRT2 expression. NHEK were grown with feeder cells in PR free RM + medium with CS-FCS (SLP − /PR − ) before adding ED at different concentrations for 24 h followed by cell lysis for qPCR analysis. There was no systematic change in KRT1, KRT10 and KRT2 expression, instead increasing only at certain concentrations of ED ( Supplementary Fig. S4). This suggests that the effect of PR observed in this study could not be due to the structural similarity between PR and ED (Figs. 1, 3, 4 and 6).
Differential stability of KRT1, KRT10 and KRT2 mRNA in NHEK. As the steady state level of KRT2 mRNA was responding differently to ATRA compared with KRT1 and KRT10, we decided to investigate their relative decay rate by transiently blocking mRNA synthesis using actinomycin D (AD). Initial experiments suggested that 2 µg/ml AD was sufficient to inhibit transcription of KRT1, KRT10, KRT2 and c-MYC genes in NHEK. We used c-MYC as a control since its mRNA is reported to have a short half-life 54 . NHEK with feeder cells were grown in SLP − /PR − RM + for 24 h and treated with 2 µg/ml AD for up to 4 h and at different time intervals cells were lysed for qPCR analysis. The mRNA expression for KRT1, KRT10, KRT2 and c-MYC normalised for POLR2A and YAP1 reference genes is shown in Fig. 7A. Both KRT1 and KRT10 mRNA were very stable (half-life of more than 4 h) whereas KRT2 and c-MYC had a half-life of about 1 h and reached almost undetectable levels at 4 h (Fig. 7A). This showed that KRT2 mRNA in NHEK cells had lower half-life compared with KRT1 and KRT10 mRNA.

Effect of ATRA on KRT1, KRT10 and KRT2 mRNA stability in NHEK.
To study the effect of ATRA on the rate of KRT1, KRT10 and KRT2 mRNA decay, we grew NHEK with feeder cells in SLP − /PR − RM + for 24 h with or without 1 µM ATRA. AD at 2 µg/ml was used to inhibit transcription and samples were lysed at different time for qPCR analyses. Addition of AD to ATRA treated NHEK showed faster decay rate with half-life of HaCaT (500,000 cells) were grown in 4 ml growth medium per well of a 6 well plate for 3 days in the above four different RM + media conditions in triplicates for each growth condition. SDS cell lysate containing 20 µg total protein (measured by Lowry's method using BSA as standard) was analysed by western blotting for K1, and K10 under different growth conditions. GAPDH was used as a loading control. Relevant bands were cropped from different blots and grouped together. Original blots are shown in Supplementary Fig. S2. Statistical analyses: n = 3, Error bars = SEM, Student's t-test was performed to calculate p values using Microsoft Excel. One-way ANOVA (shown by a horizontal line over each graph) was used to determine the statistical significance in gene expression between different growth conditions. The p values are given by asterisks (ns = p > 0.05, *p < 0.05, **p < 0.01, ***p < 0.001 and ****p < 0.0001). (2020) 10:4829 | https://doi.org/10.1038/s41598-020-61640-9 www.nature.com/scientificreports www.nature.com/scientificreports/ 15 min for KRT1 and ~45 min for KRT10 mRNA compared with KRT2 which was significantly stabilised with increase in half-life from ~1 h to ~3.5 h. These observations suggest that ATRA treatment of NHEK accelerated the decay of KRT1 and KRT10 mRNA but had the opposite effect on KRT2 mRNA (Fig. 7B-D).

Discussion
Although FCS and PR are routinely found in medium used to culture epidermal keratinocytes, the effect of SLP, present in FCS, and PR themselves on keratinocyte differentiation have never been fully investigated. In the literature, researchers have identified the potential for influence and used de-lipidised serum 34,35 , CS-FCS 55,56 or serum free medium 39,57 supplemented with an effector (either retinoid or steroid) to investigate its specific effect on differentiation. In some studies, low serum concentration in the medium has been used to dilute out any effect of serum lipids 43,58 . Despite their ubiquitous presence in culture media no study has systematically compared keratin expression in medium containing FCS with CS-FCS or studied the effect of PR on the expression of differentiation-specific keratins. In this study, we have compared the expression of differentiation-specific keratin genes KRT1, KRT10 and KRT2 in FCS with CS-FCS with or without PR and showed how SLP and PR substantially influence differentiation-specific keratin expression.
We compared the effect of SLP and PR on the expression of differentiation-specific keratins in NHEK with two cell lines, HaCaT and N/TERT-1, widely used as substitutes for normal keratinocytes [48][49][50][51] . Removal of SLP and PR from the medium increased mRNA and protein expression for KRT1, KRT10 and KRT2 genes in NHEK ( Figs. 1 and 2). When primary keratinocytes are isolated from skin, K2 shows strong expression on western blotting 20 however it is rapidly downregulated and becomes undetectable when primary epidermal keratinocytes are cultured 19 , therefore, it was surprising to see K2 induction by removing SLP and PR from the medium. Although the induction of K2 was very low, nevertheless it was detectable (Fig. 1E). This is the first study where detectable quantity of K2 has been induced in 2D cultures of NHEK and it is possible that the quantity could be improved and YAP1 under different growth conditions. (D) N/TERT-1 (500,000 cells) were grown in 4 ml growth medium per well of a 6 well plate for 3 days in the above four different RM + media conditions in triplicates for each growth condition. SDS cell lysate containing 20 µg total protein (measured by Lowry's method using BSA as standard) was analysed by western blotting for K1 and K10 proteins under different growth conditions. GAPDH was used as a loading control. Relevant bands were cropped from different blots and grouped together. Original blots are shown in Supplementary Fig. S3. Statistical analyses: n = 3, Error bars = SEM, Student's t-test was performed to calculate p values using Microsoft Excel. One-way ANOVA (shown by a horizontal line over each graph) was used to determine the statistical significance in gene expression between different growth conditions. The p values are given by asterisks (*p < 0.05, **p < 0.01, and ***p < 0.001).
www.nature.com/scientificreports www.nature.com/scientificreports/ by further manipulation of culture conditions. Neither HaCaT nor N/TERT-1 expressed K2 under any experimental condition used in this study. Although HaCaT in 3D cultures do form stratified layers they do not form stratum corneum and do not express precursors of cornified envelops such as filaggrin, loricrin and involucrin as much as in NHEK 49,59 so it was not surprising they did not express K2. Expression patterns of KRT1 and KRT10 transcripts in HaCaT in the absence of SLP and PR were similar to NHEK (compare Fig. 1A,B with Fig. 3A,B). Interestingly, expression of KRT2 mRNA in HaCaT was significantly increased in the presence of SLP when there was no PR which was very different for NHEK (compare Figs. 1C with 3C). N/TERT-1, on the other hand, has been shown to have normal differentiation characteristics in monolayer and 3D organotypic cultures with proper stratum corneum formation similar to that observed with NHEK 50,60 . Surprisingly, removal of SLP did not induce either KRT1, KRT10 or KRT2 expression in N/TERT-1 cells. However, the transcripts of all the three keratins showed increased expression in the presence of SLP without PR which was also reflected in protein expression (Fig. 4A-D). This indicates that PR suppresses differentiation-specific expression of keratins in N/TERT-1 cells but only in the presence of SLP. This effect of SLP and PR was very different to that observed in NHEK and HaCaT cells and perhaps reflects the method used to immortalise N/TERT-1 cells.
FCS is an ill-defined component of most culture media, which contains a wide range of biological molecules including growth factors, hormones, vitamins, serum proteins, fatty acids, lipids including retinoids and minerals that are needed by the cells for their growth 61 . Removing the lipids from the serum by charcoal stripping would deplete the serum of retinoids (retinol + ATRA) and that could influence the keratin gene expression observed . Two-way ANOVA was used to measure the statistical significance between PR + and PRgroups for each keratin mRNA (KRT1, KRT10, and KRT2), p values are given by asterisks (ns = p > 0.05, *p < 0.05, **p < 0.01, ***p < 0.001 and ****p < 0.0001). Figure 6. PR supresses mRNA expression for KRT1 and KRT10 in NHEK. NHEK (200,000 cells) were mixed with 10 6 irradiated 3T3 fibroblasts in 2 ml of CS-FCS containing and PR free RM + medium per well of a 12 well plate in triplicates for each growth condition. PR was dissolved in this medium at a concentration similar to that used in normal culture medium (0.01 mg/ml). NHEK were cultured in this medium for 3 days either with or without 1 µM ATRA added in the last 24. The feeder was removed by squirting PBS/EDTA after which cell lysates were collected for qPCR gene expression analysis for (A) KRT1, (B) KRT10 and (C) KRT2. Control cells were treated with DMSO/EtOH (0.001%/0.01%). Data are shown as fold expression normalised to the expression of two housekeeping genes, POLR2A and YAP1. Statistical analyses: n = 3, Error bars =SEM, Student's t-test was performed to calculate p-values using Microsoft Exel. One-way ANOVA (shown by a horizontal line over each graph) was used to determine the statistical significance in gene expression between different growth conditions. The p values are shown by asterisks (ns = p > 0.05, *p < 0.05, **p < 0.01 and ****p < 0.0001). www.nature.com/scientificreports www.nature.com/scientificreports/ in this study. Although using 1 µM ATRA in CS-FCS containing medium suppressed expression of at least KRT1 and KRT10 transcripts, it is far higher than the physiological levels of retinoids found in FCS and therefore could not be the sole reason for the change in keratin gene expression. The level of retinoids (retinol + ATRA) in bovine serum reported in the literature is between 600-700 nM 62 and therefore our culture media containing 10% FCS should contain between 60-70 nM of retinoids. Removal of this quantity of retinoids is unlikely to affect keratin gene expression as much as shown in Figs. 1-4. This suggests that charcoal treatment is perhaps removing additional components from CS-FCS which might be keeping keratin expression suppressed. A recent study by Tu and co-workers used proteomic analysis to show that FCS contains a total of 143 proteins out of which 14 including insulin-like growth factor 2 (IGF-2), IGF binding protein (IGFBP)-2 and -6 were reduced by charcoal treatment 56 . Presence of these 14 proteins in FCS could be suppressing the expression of KRT1, KRT10 and KRT2. Reduced level of these proteins in CS-FCS by charcoal could be responsible for inducing the gene expression as shown in Figs. 1-4.
The expression of both KRT1 and KRT10 was suppressed when 1 µM ATRA was added in RM + containing CS-FCS (Fig. 5). However, presence of PR appears to reduce the effect of ATRA on KRT1, but not on KRT10 expression (compare PR − with PR + in Fig. 5A,B), which was very similar to the effect of SLP (compare SLP + / PR − and SLP + /PR + in Fig. 1A,B). The effect of ATRA on KRT2 mRNA was the most pronounced in the absence of PR increasing by more than 2 folds, which is consistent with DNA microarray data showing increase in KRT2 expression in retinoid exposed keratinocytes 39 . In the presence of PR however there was no increase in KRT2 expression (Fig. 5C). This effect of PR on keratin expression was novel, it was investigated further by adding back PR in PR-free medium. Both ATRA and PR on their own suppressed expression of KRT1 and KRT10 but the effect was not synergistic and when added together PR reduced the effect of ATRA (Fig. 6). Unlike KRT1 and KRT10, the KRT2 expression either increased (Fig. 5C) or did not change (Fig. 6C) by ATRA suggesting KRT2 behaved differently. Taken together, these results suggest that the effect of SLP on keratin expression could be primarily due to the bioactivity of retinoids present in FCS. It has been proposed that due to structural similarity between PR and oestrogen, PR mimics the activity of oestrogen and at concentrations used in tissue culture media, it can increase cell proliferation by 200% and progesterone receptor content of MCF-7 cells by 300% 52,53 . However, later studies have suggested that the oestrogenic activity of PR was primarily because of lipophilic impurities contaminating the commercial preparations of PR 63,64 . This was also shown by our data as PR was able to suppress KRT1, KRT10 and KRT2 expression however ED did not suppress keratin expression, instead stimulating the expression only at certain concentrations suggesting that PR works through a different mechanism from ED ( Fig. 6; Supplementary Fig. S4).
The mRNA abundance for a protein is determined by the balance between rates of gene transcription and mRNA half-life 65 . Our measurement of transient mRNA decay rates showed significant differences in KRT1, KRT10 and KRT2 mRNA in RM + medium containing CS-FCS. This is the first report where mRNA decay rates of differentiation-specific keratins have been investigated. Surprisingly, KRT2 mRNA had the fastest decay rate compared with KRT1 and KRT10 mRNA which did not show significant decay over 4 h (Fig. 7A). These results suggest that the rapid mRNA decay together with the transcriptional suppression could be factors which makes K2 protein mostly undetectable in 2D keratinocyte cultures 19 . The mRNA decay rates are regulated by cis-acting sequence determinants, which recruit trans-acting factors and largely control mRNA stability. One of the most characterised cis-elements regulating mRNA decay are AU-rich elements (ARE) located in the 3′ untranslated region (UTR) of 5-8% of human genes of short-lived mRNA which function as a signal for rapid degradation 66,67 . There are three classes of ARE, class I contains multiple, overlapping copies of AUUUA motifs dispersed over the 3′ UTR surrounded by U-rich regions; class II contains multiple, overlapping copies of AUUUA, and class III lacks AUUUA but contains predominantly U-rich sequences 68 . These ARE recruit AU-binding proteins which assemble the degradation machinery to mRNA for their destruction 68,69 . ARE mediated decay cannot be the mechanism for the regulation of KRT1 and KRT10 expression because of two reasons, first their mRNA were stable with no significant change over 4 h suggesting the half-life must be more than 4 h (Fig. 7A) and second, the 3′ UTR of none of them contained any class of ARE. The KRT2 mRNA was however less stable but as it also did not contain a typical ARE so should be degraded involving mechanisms other than ARE. In addition to 3′ ARE, other regions have been identified in the coding sequence of a number of genes including c-FOS 70 , c-MYC 71 and β-tubulin 65,72 , which determine stability of individual transcript and it is possible that KRT2 mRNA stability is also regulated by sequences in its coding region.
Although retinoids suppress transcription of most keratin genes by direct interaction between ligand bound receptors and retinoid responsive elements on keratin genes 44,45 , the steady-state mRNA for several keratins including KRT2, KRT4, KRT7, KRT13, KRT15 and KRT19 have been shown to increase 39,40,43,58,73 . In this study we show that steady state mRNA for KRT1 and KRT10 is reduced by ATRA whereas for KRT2 it is induced. One possible explanation could be that the transcription of KRT1/KRT10 and KRT2 are reciprocally regulated by ATRA. This is supported by the fact that normal expression of K1 and K2 is mutually exclusive at different body sites of the murine skin 74 . Another possible mechanism could be the ATRA induced stabilisation of KRT2 transcript as has been proposed for KRT19 for the increase in its steady-state mRNA 43 . In the presence of ATRA the KRT1 transcripts become most unstable (half-life reduced from more than 4 h to just 15 min), followed by KRT10 (half-life reduced from more than 4 h to 45 min) whilst KRT2 mRNA becomes more stable (half-life increased from 1 h to 3.5 h). It has previously been reported that exposure of cells to retinoic acid also stabilises transcripts for a number of genes including proteolipid protein 75 , prolactin 76 , KRT19 43 and calbindin_D28K 77 . At the same time destabilisation of transcripts for TNF-α 78 and fibroblast growth factor 79 by retinoic acid has also been reported. Although the exact mechanism of transcript stabilisation or destabilisation by retinoic acid is not clear, a number of proteins have been identified which bind to transcripts to regulate stability (reviewed in 80 ). One of the best characterised is HuR, also known as embryonic lethal abnormal visual-like protein 1 (ELAVL1), family of proteins which are ubiquitously expressed 81 . These proteins strongly bind specific transcripts and protect them from the degradation machinery. When cells are exposed to ATRA cellular retinoic acid binding proteins I and II (CRABPs) are activated in the cytoplasm. It is tempting to speculate that ATRA-CRABP complexes could dissociate the transcript-protecting proteins thereby inducing transcript degradation as has been observed for K1 and K10 (Fig. 7A-C). It is also conceivable that the ATRA-CRABP complex could attract other proteins to a specific transcript thereby increasing their stability as has been observed for KRT2 mRNA.
In summary, we have shown that CS-FCS induced expression of differentiation-specific keratins KRT1, KRT10 and KRT2 in NHEK. The HaCaT and N/TERT-1, the two cell lines frequently used as a substitute for NHEK, the effect of SLP and PR was significantly different from NHEK. The effect of SLP was mimicked by ATRA which suppressed the expression of KRT1 and KRT10 but increased KRT2 expression. PR on its own also suppressed expression of these keratin genes. We observed that the increase in KRT2 transcript was because of stabilisation of its transcript by ATRA whereas KRT1 and KRT10 mRNA were destabilised. This suggests that post-transcriptional mechanism plays an important role in retinoid mediated regulation of differentiation-specific keratins. Overall this study emphasises the need to better understand how in vitro culture conditions affect the differentiation mechanism of keratinocytes, which could otherwise result in misinterpretation of biomarkers for skin and mucosal pathologies characterised by dysregulated keratinocyte differentiation such as cancer, inflammatory diseases or wound healing.
Culture medium. RM + (Rheinwald-Green Modified, also called FAD) medium 82,83 was prepared as described previously 84 . Cells were cultured in four different formulations of RM + medium; with (SLP − ) or without (SLP + ) charcoal treatment and with (PR + ) or without (PR − ) PR. 3D Organotypic cultures. NHEK cells used in 3D organotypic (OT) cultures were cultured in RM + with appropriate formulation prior to setting the 3D system. To set up OT, plastic inserts (pore size 0.4 µm) were placed in 12-well plates. Rat tail collagen solution was prepared on ice at 4 mg/ml in DMEM either with or without PR and neutralised using 0.5 N NaOH. Primary dermal fibroblasts were used in collagen matrix to support the growth of keratinocytes. Next day, medium was aspirated from inside and outside of each insert and 1 ml of RM + (with or without PR) and (with normal FCS or CS-FCS) was added underneath the inserts. The cells were allowed to grow at an air-liquid interface for 10 days to induce stratification and the medium underneath the inserts was changed every day 85 . Cell treatment. NHEK were co-cultured in 6 or 12 well plates with γ-irradiated 3T3 murine fibroblasts as feeder to support keratinocyte growth in different RM + formulations for 3 days before removing the feeder and using keratinocytes for gene expression analysis by qPCR (cells grown in 12 well plates) and western blotting (cells grown in 6 well plates). For specific treatment, the co-cultures were seeded in 12 well plate, grown in RM + containing CS-FCS, next day the cells were treated with reagents, ATRA (R2625, Sigma) or ED (E8875, Sigma) for 24 h. ATRA was first dissolved in DMSO and further diluted in ethanol (EtOH) to the desired concentrations. The concentrations of DMSO/EtOH used as a vehicle control varied from 0.003% to 0.03% (v/v). ED was dissolved in DMSO and further diluted in culture medium to working dilutions. DMSO (0.001%) was used as a vehicle control. PR (P3532, Sigma, UK) was added in PR free RM + culture medium at 0.01 mg/ml. The feeders were removed by squirting the culture with PBS/EDTA and collecting cell lysates for expression analysis by qPCR. When PR was added, PR free medium was used to grow cells as control. Actinomycin D (AD, 11805-017, Gibco, UK) was diluted from a stock (2 mg/ml in water) to working dilutions. mRNA isolation and cDNA synthesis. Cells were seeded at desired density in 6-well or 12-well plates, washed with PBS and lysed after specific treatment or media incubation. Lysates were either stored at −80 °C or subjected to mRNA extraction. Dynabeads mRNA DIRECT kit was used to extract polyadenylated (polyA) mRNA and the total RNA was isolated using RNeasy kit (Qiagen) according to manufacturer's instructions. The concentration of mRNA was measured using a NanoDrop spectrophotometer. For reverse transcription qPCR-BIO cDNA Synthesis Kit (PCR Biosystem, UK) was used according to manufacturer's instructions. The reverse transcription reaction protocol was as follows: 42 °C for 30 min, 85 °C for 5 min and 4 °C for 5 min. The resulting cDNA was diluted using nuclease-free qPCR H 2 O and immediately used in qPCR gene expression analysis or stored at −20 °C for later use.
Real-time quantitative PCR (qPCR). Five micromolar forward (F) and reverse (R) primer mixes were made by mixing equal volumes of forward and reverse primer stocks (100 µM) with suitable amount of nuclease-free H 2 O. The qPCR primer sequences used for different genes in this study along with amplicon size in parenthesis were as follows. . For PCR amplification 384-well format plates were used in a total volume of 5 μl per well consisting of 2.5 µl qPCRBIO SyGreen Blue Mix Lo Rox (#PB20.11-50, PCRBIO Systems, UK), 0.5 µl 5 µM F/R primer mix and 2 µl cDNA template. The cDNA samples were loaded at the bottom of the wells, followed by a mixture of SYBR Green and F/R primer added at the top wall of each well. Plates were sealed, centrifuged and inserted into the LightCycler 480 machine (Blizard institute's core facility) for qPCR according to the recommended protocol. Relative quantification of mRNA expression was measured using LightCycler 480 software (release 1.5.0). For normalisation, POLR2A and YAP1 were used as reference genes 51 .
Immunostaining of cells grown in 2 and 3D cultures. NHEK cells were trypsinised, counted and grown under different growth conditions on the top of collagen coated glass coverslips with irradiated 3T3 feeder cells. After 24 h, the cells were fixed with a mixture of acetone and methanol (1:1) and immunostained for K1, K2, K10 and K14 using the method described previously 84 . Stained samples were imaged using either Leica Epi DM5000 microscope or Leica Epi DM4000 equipped with a DFC350 FX digital camera under 20×/0.5 NA and/ or 40×/0.75 NA objective lenses. The images were acquired by Metamorph imaging system and processed using ImageJ bundled with 64-bit Java 1.8.0_112 (freely available at https://imagej.nih.gov/ij/download.html).