Loss of keratin 14 expression from immortalized keratinocytes by promoter methylation

Abstract

Immortalized keratinocytes can offer a low‐cost experimental platform for human skin research, with increased cell yield compared to primary cultures. However, the usefulness of these surrogate cell models is highly dependent on their ability to retain the phenotypic attributes of the parent cells. Keratins K14 and K5 are the hallmarks of undifferentiated, mitotically active basal keratinocytes. We observed occasional progressive loss of K14 expression in growing keratinocyte cell lines, with persistent retention of K5 and an epithelial phenotype, and investigated possible reasons for this. We show that K14 repression occurs by DNA promoter methylation of KRT14 gene and is compounded by histone deacetylation and by the presence of EGF. In vivo, keratinocytes shut down K14 synthesis as they commit to terminal differentiation and move from the basal to spinous layer, but by laser‐capture microdissection of human epidermis we could detect no evidence of increased selective KRT14 methylation in this normal process. Loss of K14 expression suggests that epidermal identity of cultured keratinocytes can be compromised in certain tissue culture situations, possibly due to the immortalization method and persistent EGF supplementation.

1. INTRODUCTION

The seminal work by Rheinwald and Green marked the start of in vitro studies on keratinocytes, ¹ when for the first time, keratinocytes became amenable to a wide variety of experimental manipulations. Cultures of human keratinocytes typically experience a cell growth ‘crisis’ from which rare cells selectively emerge to establish spontaneously immortal lines. The low incidence of this event presents a major hurdle to establishing continuous cell models of disease from patient‐derived biopsies. A lack of ‘standardized’ cell lines also makes it more difficult to compare results across different studies. To address these problems, various immortalization methods have been established to bypass cellular senescence‐inducing events such as telomere shortening, DNA damage response and cell cycle checkpoints. These methods usually involve either the introduction of viral proteins, such as SV40 large T antigen, ² HPV E6 and E7, ³ adenovirus E1A and E1B ⁴ or, now more commonly, human telomerase reverse transcriptase. ⁵ Ideally, immortalized lines should possess extended proliferation capacity and retain a stable and similar, if not identical, genotype and phenotype to the parent cells. In practice, immortalized cells often show genetic aberrations such as TP53 mutations ⁶ and aneuploidy. ⁷ , ⁸ Yet phenotypically these cells retain many clear features of the parent line, such as contact inhibition, anchorage dependence, non‐tumorigenicity and continued expression of their major keratins K14 and K5, so that immortalized cells have sustained their position as mainstay tools for epithelial cell research.

Expression of keratins K14 and K5 is a robust phenotypic hallmark of basal keratinocytes, both in vivo and in vitro in tissue culture. However in the course of studies of the genetic skin disease epidermolysis bullosa simplex (EBS) using a number of model cell lines, we have observed stochastic and progressive loss of K14 protein, with retention of K5, in two keratinocyte cell lines generated earlier by HPV16 E6^E7 immortalization. ³ Following the Rheinwald and Green protocol, the culture medium used for these keratinocytes contained 10 ng/mL epidermal growth factor (EGF). ³ A few papers have reported reduction in K14 expression observed after viral transformation of keratinocytes, although causative factors are unknown. ⁹ , ¹⁰ , ¹¹ , ¹² Bowden and colleagues noted loss of K14 expression in cervical keratinocytes transformed with both HPV‐16 and v‐Ha‐ras, although HPV transfection alone only reduced K14 mRNA while K14 protein levels were maintained. ¹⁰ Ras is a downstream effector of the epidermal growth factor receptor (EGFR) and EGF is a common additive in keratinocyte tissue culture medium as it enhances growth and inhibits differentiation of cultured keratinocytes ¹³ , ¹⁴ , ¹⁵ upon binding to EGFR. In the epidermis, EGFR is expressed in the basal layer and plays a role in regulating epidermal homeostasis, ¹⁶ and this basal layer is where K14 expression is regulated. We show here that KRT14 repression is caused by DNA methylation of the KRT14 promoter, and that EGF in the culture media potentiates re‐methylation of KRT14 after DNA methyltransferase inhibitor withdrawal.

2. MATERIALS AND METHODS

2.1. Keratinocyte culture and drug treatment

The immortalized human keratinocyte cell lines NEB‐1 (from normal keratinocytes) and KEB‐7 (from an epidermolysis bullosa simplex patient) have been described previously. ³ N838 is a NEB‐1‐derived keratinocyte line in which keratin 14 expression has been knocked down using shRNA and which now only expresses simple epithelial keratins. These three feeder‐independent cell lines were maintained in RM+ medium (three parts Dulbecco's modified Eagle's medium [DMEM] to one part Ham's F12, plus hydrocortisone [0.4 μg/mL], transferrin [5 μg/mL], liothyronine [2 nM], adenine [1.8 mM], insulin [5 μg/mL], EGF [10 ng/mL] and 10% fetal bovine serum [FBS]). HaCaT keratinocytes ⁷ were maintained in DMEM (Hyclone) supplemented with 10% FBS. Experimental cell cultures were treated with 1 μM 5‐aza‐deoxycytidine (Sigma‐Aldrich) for 4 days, followed by a 3‐day recovery period in culture media without the inhibitor. The histone deacetylase inhibitor trichostatin A (TSA) was used at 0.3 μM concentration for two different treatment durations: long, 3 days (d2–d5) and short, and short overnight treatment (d4–d5).

2.2. RNA isolation, cDNA synthesis, reverse transcriptase and real‐time PCR

RNA was extracted from keratinocytes using RNeasy Mini Kit (Qiagen) according to the manufacturer's protocol. Reverse transcription was carried out on 1000 ng total RNA using the Transcriptor First Strand cDNA kit (Roche). For RT‐PCR, KRT14 transcripts were amplified by Qiagen Taq DNA Polymerase Kit using 1.2 ng template and highly specific intron‐spanning primers (Table S1). For real‐time PCR, 1 μL of cDNA was used in the SYBR green master mix (Roche). Amplification, normalization and relative quantification of transcripts were carried out using a LightCycler 480 (Roche).

2.3. Immunoprecipitation and immunoblotting

Immunoprecipitation of keratin 5 (K5) interacting proteins in HaCaT, NEB‐1 and N838 was carried out using a protein A IP kit (Sigma‐Aldrich) in accordance with the manufacturer's instructions. Two microgram of anti‐K5 antibody (clone XM26) was used for each reaction. For details of all antibodies used in this study see Table S2.

2.4. Immunostaining method

For indirect immunofluorescence, cells were cultured on glass coverslips and fixed in cold methanol‐acetone (1:1) for 7 min on ice and washed twice with phosphate buffered saline (PBS). Primary antibodies used, as for immunoblotting, are listed in Table S2. Cells were incubated with primary antibody for 1 h at room temperature, rinsed thrice in PBS, incubated with secondary antibody for another hour, washed again and mounted on slides with Hydromount (National Diagnostics).

2.5. Bisulfite sequencing and statistical analysis

Genomic DNA was extracted from cell lines by QIAamp DNA Mini Kit (Qiagen) and 100 ng was used for bisulfite conversion by EpiTect Bisulfite Kit (Qiagen) according to the manufacturer's protocol. The final 140 μL solution was aliquoted into five PCR tubes and the conversion protocol was performed using a Biometra Thermal Cycler (Analytik Jena). The bisulfite‐converted DNA was cleaned up and used for assessing the KRT14 DNA methylation pattern. PCR products were amplified with primers devoid of any CpG dinucleotides and specific to converted promoter and enhancer sequences separately (Table S1). PCR products were separated on 1.2% agarose gels and purified using QIAquick Gel Extraction Kit (Qiagen), then cloned into the pCR2.1‐TOPO vector (Thermo Fisher Scientific). Colony PCR was performed on 20–30 white colonies per sample using vector specific primers, followed by enzymatic clean‐up to remove the residual primers and dNTPs via exonuclease I (NEB) and shrimp alkaline phosphatase (NEB) digestion. The cloned promoter and enhancer regions were subjected to Sanger sequencing. The resulting DNA methylation profile was analysed using a web‐based platform called QUantification tools for Methylation Analysis, QUMA ¹⁷ by aligning the target genomic sequences with the raw bisulfite sequences. The platform provides quality control, including sequence quality checks, conversion rate and statistical analysis. The difference at each CpG site between the two sample sets was analysed statistically with Fisher's exact test, whereas the entire sets of CpG sites were compared with the Mann–Whitney U‐test.

2.6. Immunohistochemistry and RNA in‐situ hybridisation

For human tissue analysis, formalin‐fixed tissue blocks of healthy human skin were obtained from surgical waste with full consent and local institutional ethical review board approval. Deparaffinized skin tissue sections were bleached with 10% hydrogen peroxide (Merck) in PBS for 2 h at 60°C to remove melanin. Antigen retrieval was carried out under pressure at 121°C with citrate buffer, pH 6 (Dako) followed by 1 h blocking in 10% goat serum (Dako). Sections were incubated overnight with primary antibodies (Table S2) at 4°C in a humidified chamber. Secondary antibody (EnVision™, Dako) incubation was carried out at room temperature for 30 min, and peroxidase activity was detected using diaminobenzidine‐tetrahydrochloride (DAB) substrate (Dako). RNA in situ hybridisation on sections was carried out using the RNAscope® assay (ACDBio) according to the manufacturer's protocol. Mounted sections were imaged on the Zeiss AxioImager.Z2 upright Wide Field fluorescence microscope controlled by Zen blue. Bright Field images were acquired by Axio Cam HRc using a 20x/0.75 Plan‐Apochromat Air lens (DIC).

2.7. Laser‐capture microdissection

Formalin‐fixed tissue blocks of healthy human skin were obtained from surgical waste with full consent and local institutional ethical review board approval. Normal skin tissue sections with 5 μm thickness were mounted onto nuclease free polyethylene naphthalate (PEN) membrane slides (Zeiss) and dried overnight at 40°C. They were then dewaxed and stained in 50% haematoxylin for visualizing skin structure to guide the laser‐capture microdissection. After staining, sections were dehydrated in ethanol, brought through xylene and completely dried before microdissection. Basal and suprabasal keratinocytes were collected separately and catapulted into the AdhesiveCaps of microfuge tubes (Carl Zeiss) using PALM MicroLaser Systems (Carl Zeiss). DNA was isolated from both keratinocyte populations using Arcturus Picopure DNA Extraction Kit (Applied Biosystems) and used for bisulfite sequencing.

2.8. Transcription factor binding site analysis

The list of putative transcription factor binding sites was obtained by conducting a search of the KRT14 promoter sequence on the PROMO website ¹⁸ with a dissimilarity percentage < 15% and shortlisting the transcription factors that had recognition sequences within the 14 CpG sites of the KRT14 promoter.

3. RESULTS

3.1. Phenotypic drift in NEB‐1 keratinocytes with loss of keratin K14

We observed a phenotypic drift in immortalized human keratinocytes resulting in a progressive reduction in K14 protein expression with increasing passage in culture. This was seen in the HPV‐immortalized keratinocyte line NEB‐1, ³ which was being used as a normal control for skin disease studies. Prior to immortalization and for around 10 passages thereafter, NEB‐1 cells showed no keratin abnormalities (see e.g. ¹⁹ ) and, like the widely‐used HaCaT cells, expressed K14 and K5 strongly and uniformly. Progressive loss of K14 was thereafter observed stochastically at stages between passages 15 and 60 (NEB‐1 ^M ) with complete loss of K14 before 80 passages (NEB‐1^H), seen by immunoblotting and immunocytochemistry (Figure 1). In contrast, HaCaT, a spontaneously immortalized keratinocyte line, ⁷ retained uniform robust K14 expression at high passage numbers (>p80, HaCaT^H) (Figure 1A, Figure S1). No reduction in K5 staining was seen in any of the cell lines. We therefore undertook experiments to try and identify the causes and drivers of this unexpected phenomenon.

FIGURE 1.

Phenotypic drift in an HPV‐immortalized human keratinocyte line NEB‐1. (A) Immunofluorescence of NEB‐1 immortalized keratinocytes shows progressive reduction of keratin K14 (green) expression after prolonged serial passaging, whilst robust expression of the partner K5 (red) is maintained. HaCaT, a spontaneously immortalized epidermal keratinocyte line, retains abundant expression of both keratins at high passage (HaCaT^H). Nuclei counterstained with DAPI. Cells labelled according to extent of passaging: NEB‐1^L = low passage number (<30 passages), NEB‐1^M = medium (30–60 passages), and NEB‐1^H and HaCaT^H: High (> 80 passages). Scale bar = 60 μm (approx.). (B) Immunoprecipitation with K5 antibody and subsequent immunoblotting with other keratin‐specific antibodies (indicated below panels) demonstrate interaction of K5 with K17, K18 and K19. No K13 was detected. The blue box marks K14, confirming the K5–K14 interaction in HaCaT^H (positive control), while N838 (A NEB‐1 derivative carrying a K14‐targeting shRNA and not expressing other keratins) serves as the negative control. Letters H and L here refer to the heavy and light immunoglobulin chains of the K5 monoclonal antibody (clone XM26). Antibodies used are listed in Table S2.

Keratin 14 (K14) and its co‐assembling partner keratin 5 (K5) are the most robust and invariable biomarker proteins of epithelial keratinocytes, generating meshworks of cytoplasmic filaments that reinforce barrier epithelia and connect neighbouring cells by linking into desmosome junctions. Loss of expression of a keratin frequently results in a corresponding reduction of the partner keratin, as the keratin monomer is degraded in the absence of a binding partner. ²⁰ However, we observed that keratin 5 expression was still abundant in high passage NEB‐1 K14‐negative cells (NEB‐1^H, see Figure 1A legend), suggesting K5 is being stabilized by an alternative type I keratin. Epidermal keratinocytes in monolayer culture can variably express other type I keratins of barrier epithelia such as K17, K19, K15, even also sometimes K18 (a simple epithelial keratin expressed in monolayer internal epithelia and in fetal, but not adult skin ²¹ ). To confirm the interaction, NEB‐1^H was compared with other keratinocyte lines, either known K14‐positive HaCaT cells or control K14‐negative keratinocytes (such as N838, a NEB‐1 derivative in which K14 was silenced with shRNA). Immunoprecipitation with a mouse monoclonal antibody to K5 (clone XM26) pulled down a strong K14 band in HaCaT cells, but not in NEB‐1^H or N838 cell lines, confirming the expected interaction of K5 with K14 (Figure 1B). K5 also pulled down type I keratins K19 and K18, and to a lesser extent K17, but not K13. The interaction of K5 with other type I keratins was also observed in the positive control HaCaT line, which had strong K14 expression. Reverse immunoprecipitation with K18 and K19 antibodies confirmed the interaction with K5 (Figure S1). Whilst heterodimerization between type I and type II keratins is usually highly specific for the correct partner keratins, these associations are dynamic and weaker opportunistic type I–type II associations can occur, usually at a lower level. ²²

High passage NEB‐1 cells were also studied for epithelial‐mesenchymal protein signatures using shRNA knockdown lines as negative controls. Despite the lack of K14, NEB‐1^H retained both the cobblestone morphology and a protein signature typical of epithelia, as characterized by positive staining for E‐cadherin but no detectable N‐cadherin or vimentin (Figure S2).

3.2. Synergistic reactivation of KRT14 expression by DNA demethylation and histone acetylation

Attempts to identify sequence variants that might account for K14 loss in NEB‐1^H did not reveal any sequence abnormalities in the regulatory and coding regions of the KRT14 gene by Sanger sequencing (Table S3). Since we only found previously described benign allelic variants, ²³ , ²⁴ we explored the possibility of epigenetic silencing at this locus. Treatment of cells with the DNA methyltransferase inhibitor 5‐aza‐deoxycytidine (5‐aza‐dC, or ‘Aza’ in figures) highlighted the potential role of an epigenetic mechanism in KRT14 gene silencing. Demethylation with 5‐aza‐dC resulted in re‐expression of K14, both at the transcript and protein levels (Figure 2A,B). Used alone, a broad‐spectrum histone deacetylase (HDAC) inhibitor trichostatin A (TSA) did not reactivate K14 expression (Figure 2B). However, TSA treatment in combination with 5‐aza‐dC augmented K14 induction beyond that of 5‐aza‐dC alone (Figure 2B), suggesting that both DNA methylation and histone deacetylation contribute to K14 silencing. This implies that removal of DNA methylation is necessary before histone acetylation can augment induction of K14.

FIGURE 2.

K14 is epigenetically silenced in NEB‐1^H. (A) Treatment with the DNA methyltransferase inhibitor, 5‐aza‐2′‐deoxycytidine induced KRT14 transcripts in NEB‐1^H cells, detected by reverse transcriptase PCR using intron‐spanning primers. (B) Exposure of cells to trichostatin‐A (TSA; a histone deacetylase inhibitor) did not affect K14 protein expression, yet co‐treatment with 5‐aza‐2′‐deoxycytidine resulted in synergistic upregulation of K14 protein. (C) Schematic representation of KRT14 gene showing position of CpG dinucleotides in the upstream regulatory sequences. The highly conserved enhancer region was selected based on Romano et al., ²⁹ and only contains 7 CpG dinucleotides. In silico search revealed a CpG island with 22 CpG dinucleotides stretching from the promoter sequence to the coding region within the first exon. CpG dinucleotides are illustrated with vertical lines. CpGs in promoter and untranslated regions are designated with superscripts P and UTR, respectively. Nucleotide positions are labelled according to NCBI Reference Sequence NG_008624.1. Labels: Aza—5‐aza‐2′‐deoxycytidine; TSA—trichostatin‐A; +^L—Long TSA treatment (for 3 days: D2‐d5); +^S—Short TSA treatment (overnight, d4–d5).

We then also observed K14 silencing in another HPV‐immortalized human keratinocyte cell line, KEB‐7, a disease model for EBS which was generated (like NEB‐1) by HPV 16 E6^E7 immortalization, but using keratinocytes from an EBS patient carrying a K14 R125P mutation. ³ Similar to NEB‐1, 5‐aza‐dC treatment induced K14 in KEB‐7 (Figure S3), providing further evidence that K14 can be epigenetically silenced by methylation in long‐term cultures.

3.3. Gene silencing is limited to the KRT14 locus

Since epigenetic modifications can elicit global or regional changes, we examined whether the silencing was limited to the KRT14 gene or if other type I keratin genes in its vicinity were also affected. The KRT14 gene lies within the cluster of type I intermediate filament genes located on chromosome 17q21.2. K18, the only type I keratin encoded on chromosome 12, was used as a control, as it was shown earlier that tissue‐specific repression of this keratin is also regulated by DNA methylation. ²⁵ Immunoblotting showed that both the KRT15 and KRT19 genes, located close to KRT14, were active. The K19 protein was detected even in K14‐negative NEB‐1, and K19 levels were unchanged in the presence of 5‐aza‐dC. Interestingly this contrasts with K15, whose protein levels were not increased after 5‐aza‐dC treatment (Figure S4). Keratin 9, which is encoded between KRT19 and KRT14, was not investigated, since this protein is only expressed in palmoplantar epidermis. Expression of KRT16 was not seen at high passage, but it was inducible by 5‐aza‐dC. However, K16 is a stress‐associated keratin not usually expressed in basal cells and its levels are variable in culture. KRT17 expression (another stress keratin gene) showed no significant change. Overall, DNA methylation‐induced gene silencing was observed significantly and consistently only in the KRT14 gene, but not in the other keratins analysed here.

3.4. Epigenetic repressive marks in the K14 regulatory region

To map DNA methylation in KRT14, an in silico search was done to find CpG islands using Sequence Manipulation Suite ²⁶ and USCS CpG search site. ²⁷ A 319 bp long region encompassing 23 CpG residues with 65.8% GC content and 0.65 CpG frequency was identified within the promoter and exon 1 (Figure 2C) fulfilling the requirements for a CpG island. ²⁸ Bisulfite sequencing of the promoter region showed that overall DNA methylation increased from 48% in K14‐positive NEB‐1^L to 74% in the K14‐negative NEB‐1^H based on two independent experiments (Figure S5A). While none of the clones were fully methylated in all positions in NEB‐1^L cells, 100% methylation was detected in a quarter of the clones derived from NEB‐1^H cells (Figure S5B). Although the KRT14 enhancer region does not contain any dense CpG clusters, we mapped the methylation status of seven dinucleotides from that region as well, since it contains essential binding sites for transcription factors. ²⁹ Similar to the promoter, the enhancer sequence derived from NEB‐1^H showed higher methylation than that from NEB‐1^L.

3.5. EGF increases KRT14 promoter re‐methylation

K14 protein levels could be maintained up to 8 days after 5‐aza‐dC withdrawal, but this induction was short‐lived, as K14 expression was again lost once the cells were re‐plated (Figure S6). This led us to suspect that a component of the culture media could be responsible for the silencing. NEB‐1 cells are routinely cultured in EGF, and this growth factor was previously reported to increase DNA methyltransferase 1 (DNMT1) activity in ovarian cancer cells. ³⁰ In comparison, HaCaT, a spontaneously immortalized keratinocyte line usually cultured without EGF, ⁷ exhibited robust K14 expression (Figure 1A).

To investigate the role of EGF on KRT14 re‐methylation, NEB‐1^H cells were seeded in the presence or absence of EGF and treated with 5‐aza‐dC for 24 h to induce K14 expression (Figure 3A and Figure S7). Cells were harvested for RNA and protein extraction, plus bisulfite sequencing from all four conditions. 5‐aza‐dC treatment led to a higher increase in K14 expression in the absence of EGF (sample NAz) than in its presence (sample EAz). Methylation levels of CpG sites located in the enhancer (7 sites) and promoter regions (14 sites) were examined (Figure 2C and Figure S8). Fine mapping of methylated CpGs in the KRT14 regulatory region showed that the DNA hypomethylating agent was more efficient in the absence of EGF. The lowest methylation levels were detected in enhancer (26%) and promoter (13%) regions from sample NAz (Figure 3B).

FIGURE 3.

KRT14 demethylation and remethylation occur in an EGF‐dependent manner. (A) Removal of EGF from keratinocyte culture medium enhances the effect of the demethylating agent 5‐aza‐dC on KRT14 gene resulting in higher protein expression (right) than in EGF‐containing control (left) as seen by K14 immunoblotting. (B) Methylation profiling of CpG sites in KRT14 enhancer and promoter regions revealed that EGF had no effect on the overall methylation levels in the untreated samples (EUn and NUn). However, EGF had significant impact on the demethylation by 5aza‐dC, indicated by higher fractions of methylated CpGs in EGF‐containing 5aza‐dC‐treated (EAz) samples (33% (enh) and 26% (prom)) compared to control (NAz) samples (26% (enh) and 13% (prom)). Bisulfite converted gDNA was PCR amplified and cloned into a TA vector, then 20–35 representative clones per sample were sequenced. Each row of circles represents the continuous sequence of a single amplicon (one subclone). The number of rows indicates the number of sequenced subclones. Methylation status of each CpG in the enhancer (7 CpGs) or the promoter (14 CpGs, see Figure S8) sequence is indicated as filled (methylated) or open (unmethylated) circles. Numbers at the bottom indicate overall methylation percentages in each condition. E: 10 ng/mL EGF; N: No EGF; Az: 1 μM 5‐aza‐dC; Un: Untreated with 5‐aza‐dC. (C) EGF supplementation resulted in K14 suppression in cells previously treated with 5‐aza‐dC. Cells expressing K14 after 5AzC treatment (EAz, NAz) from Figure 3A were replated and seeded in two aliquots. One aliquot of each pair was supplemented again with EGF (EAz>E and NAz>E), whereas the other was not (EAz>N and NAz>N). No further 5‐aza‐dC was added to any of the replated cells. K14 protein levels dropped in the presence of EGF in both cases when compared to the corresponding controls. (All panels in [C] are taken from the same blot.) Non‐adjacent tracks from a single blot remounted side by side for comparison. (D) EGF exposure boosts re‐methylation of KRT14 promoter after 5‐aza‐dC treatment. The promoter methylation levels are inversely correlated with K14 protein expression in the respective samples. No consistent effect was observed in the KRT14 enhancer region. Methylation profiling was performed by bisulphite sequencing (see Figure 3B).

The DNA methyltransferase inhibitor‐treated K14‐expressing cells from EAz and NAz samples were replated either with or without EGF, thus allowing a wash‐out of incorporated 5‐aza‐dC (Figure 3C). K14 expression was markedly reduced in the presence of EGF regardless of the sample of origin. While the methylation status of the enhancer region in the presence/absence of EGF showed no difference, re‐methylation of the CpG dinucleotides in the promoter region was significantly increased by the growth factor. The immunoblotting data was further supported by KRT14 promoter hypermethylation which was clearly increased in the presence of EGF (Figure 3D). Methylation of the KRT14 enhancer remained low regardless of EGF supplementation, suggesting that methylation of the KRT14 promoter region alone is sufficient for gene repression. Comparing the effect of EGF‐induced remethylation on specific residues, we found that CpG no.12—located in the KRT14 transcription start site—was the only common dinucleotide which was significantly methylated in samples analysed after removal of 5‐aza‐dC (Figure S9).

3.6. Transcription factor analysis of KRT14 promoter

As DNA methylation is known to affect the binding affinity of transcription factors, we searched for potential transcription factor motifs corresponding to the 14 CpG sites (numbered in Figure S8) within the KRT14 promoter using the PROMO transcription factor database. ¹⁸ Putative transcription factor motifs with the default dissimilarity index below 15% are listed according to their position in the KRT14 promoter (Table 1). Several of these factors (GCF, E2F‐1, AP‐2α, FOXP3 and XBP‐1) are already known to bind preferentially to hypomethylated DNA. ³¹ , ³² , ³³ Moreover, E2F‐1, AP‐2α and XBP‐1 expression are associated with increased proliferation. ³⁴ , ³⁵ , ³⁶ Interestingly, the motifs of two negative regulators of keratin expression (PEA3 and p53) were also identified. In particular p53, which binds preferentially to methylated CpG sites, ³⁷ forms a complex with Sp1 and represses the KRT14 gene. ³⁸ Thus, the change in KRT14 promoter methylation patterns may have resulted in a change in binding affinity to one or more of the above transcription factors.

TABLE 1.

List of putative transcription factor motifs corresponding to the CpG sites on the KRT14 promoter.

CpG position	Transcription factor class	Transcription factor	Dissimilarity (%)
1	Tryptophan cluster	c‐Ets‐2	2.14
2 and 3	TEA domain	ETF	13.12
2 and 3	Undefined	GCF	2.14
3	Forkhead/winged helix	E2F‐1	11.89
3	Nuclear receptors with C4 zinc fingers	RAR‐beta	10.7
4	Nuclear receptors with C4 zinc fingers	RXR‐alpha	1.7
5	Basic helix–loop–helix factors (bHLH)	AhR	14.41
5	Basic helix–loop–helix factors (bHLH)	AhR:Arnt	10.82
6	Undefined	ENKTF‐1	8.2
6	Forkhead/winged helix	E2F‐1	0.99
7	Tryptophan cluster	PEA3	9.94
7	Basic helix‐span‐helix (bHSH)	AP‐2alphaA	2.55
7	Forkhead/winged helix	E2F‐1	10.63
8	Undefined	ENKTF‐1	5.69
8	Forkhead/winged helix	FOXP3	9.51
8	SMAD/NF‐1 DNA‐binding domain	NF‐1	6.95
8	Forkhead/winged helix	E2F	6.81
9	Paired box	Pax‐5	1.54
9	p53 domain	p53	3.38
9	Nuclear receptors with C4 zinc fingers	RXR‐alpha	2.32
9	Forkhead/winged helix	E2F‐1	11.39
9	C2H2 zinc finger	Sp1	1.23
10	TATA‐binding proteins	TBP	1.87
11	Tryptophan cluster	PEA3	13.05
11	Basic leucine zipper (bZIP)	XBP‐1	13.65
12	Nuclear receptors with C4 zinc fingers	RXR‐alpha	3.39

3.7. DNA methylation is probably not the driver of KRT14 transcriptional control in epidermal differentiation

Synthesis of keratins is tightly regulated during epidermal differentiation, with K14 typically restricted to the basal layer of keratinocytes at homeostasis. The EGF receptor is strongly expressed in basal cells. Therefore, we looked for evidence of a link between epigenetic silencing of K14 and onset of terminal differentiation. We compared protein expression (by immunohistochemistry), mRNA distribution (by in situ hybridization) and methylation (by bisulfite sequencing of DNA from laser‐captured microdissected samples) in tissue sections of human epidermis. Basal layer keratinocytes actively transcribe KRT14 and KRT5, and this switches to KRT1 and KRT10 as cells commit to differentiation and transit from the basal to spinous layers. By in situ hybridization we observed that K14 transcripts are frequently observed in spinous and upper spinous layers, whereas K5 mRNA is, as expected, tightly restricted to lowest layers (Figure 4), suggesting a tighter control operating on K5 (type II keratin) transcription than on K14 (a type I keratin). A similar pattern of regulation was seen at the spinous‐granular transition, beyond which KRT10 (a type I keratin) transcripts persisted but KRT1 (type II) transcripts were clearly shut down (Figure 4, inset). Unlike the transcripts, both keratin 1 and 10 protein persisted into the uppermost layers. Only very low levels of DNA methylation were observed in both the enhancer and promoter regions of KRT14 (Figure S10), making it unlikely that DNA methylation alone is responsible for downregulating K14 during epidermal differentiation.

FIGURE 4.

Tighter restriction of expression of type II than type I keratins in human epidermis. Position‐matched skin tissue after in situ hybridization (RNA, left column) and immunohistochemistry (Protein, right column) showing expression of K14 versus K5, and differentiation keratins K1 and K10, in serial sections of human epidermis. Transcripts of K14 (a type I keratin) are sporadically seen in upper spinous layers (black arrowheads) whereas mRNA for K5 (a type II keratin) is tightly restricted to lower layers. Similarly, K1 (another type II) mRNA transcripts are downregulated at the spinous‐granular transition while K10 (type I) transcripts persist to the boundary with the stratum corneum (inset). NC: Negative control. DapB (RNA), anti‐mouse secondary (protein). Scale bar = 40 μm.

4. DISCUSSION

The use of immortalized cells for disease research is essential. For rare diseases, biopsies are becoming less available as diagnostics move increasingly towards DNA sequencing. Thus, disease model cell lines and their controls are being used ever more extensively and for longer periods of study, in the hunt for better understanding and new therapeutics. It is therefore important to understand the pitfalls of using immortalized cells for disease models and controls.

Here we have described examples of apparently healthy immortalized keratinocytes in culture which have lost a major phenotypic characteristic of keratinocytes that is expression of keratin 14, whilst retaining other basic hallmarks of epithelia. We show that epigenetic silencing is probably the primary mechanism of repression, and augmentation of K14 expression by 5‐aza‐dC and trichostatin A (TSA) co‐treatment suggests a synergistic repression driven by both DNA methylation and histone deacetylation. This accords well with earlier studies showing that methyl CpG binding proteins bind to methylated DNA promoters and recruit histone deacetylases. ³⁹

Silencing appears to be specific to KRT14, as 5‐aza‐dC had no significant effect on other keratins located in the same chromosomal keratin cluster. Keratin genes have very high GC content, with many CpG islands as defined by Antequera and Bird. ²⁸ Epigenetic regulation has been reported for four type I keratins, K13, K18, K19 and K23, ²⁵ , ⁴⁰ , ⁴¹ , ⁴² but we observed that although many keratin CpG islands lie within the gene body, only certain keratin genes have CpG clusters within the regulatory region. These are KRT4, KRT7, KRT8, KRT14 and KRT18, making these the most likely targets for epigenetic silencing.

Several exogenous factors could affect the cells' susceptibility to DNA methylation. Long‐term EGF treatment is reported to increase DNA methyltransferase activity through activation of the Ras and ERK/MAPK pathway. ³⁰ , ⁴³ In the epidermis, DNMT1 is, like EGFR, expressed in basal cells and maintains methylation patterns in dividing keratinocytes using hemi‐methylated DNA as a substrate, keeping terminal differentiation genes silenced ⁴⁴ so cells can still divide. When we tested the effect of EGF on K14 expression after withdrawal of 5‐aza‐dC treatment, increased re‐methylation was indeed observed when EGF was present (Figure 3).

Another possible factor is viral‐mediated immortalization, which has been widely used in the establishment of cell lines from patient biopsies. There is in fact increasing evidence linking K14 reduction to the presence of human papilloma virus ⁹ , ¹² , ⁴⁵ K14 was earlier reported to be down‐regulated in an HPV16‐immortalized keratinocyte line with v‐Ha‐ras‐induced constitutively active EGF signalling, although the mechanism of K14 repression was unknown. ¹⁰ Doorbar and colleagues ⁴⁶ showed that HPV16 inhibits keratinocyte differentiation through Notch‐1 inactivation by viral E6 protein, which helps the virus by leading to expansion of the proliferative keratinocyte compartment where the virus can multiply. Evolution may have favoured this capability in papilloma viruses and loss of K14 by whatever mechanism would certainly prevent the keratinocytes differentiating. The cells used in our study had both been immortalized by HPV16 E6^E7, and it is possible that some residual HPV material in the cells is linked to the observed silencing of K14.

HPV and EGF thus represent a potent combination. In serially passaged human keratinocytes supplemented with EGF, immortalization with high‐risk HPV E6^E7 was shown to be associated with sequential and progressive promoter methylation of a number of host genes. ⁴⁷ K14 loss has been associated with the presence of HPV both in vitro ⁹ , ¹² and in vivo, ⁴⁵ where K14 silencing is more frequently seen in high‐grade HPV lesions. EGFR is up‐regulated in 90% of late‐stage HPV‐infected cervical cancers ⁴⁸ and is associated experimentally with increased pathogenesis. ⁴⁹ Long‐term culture of keratinocytes immortalized with HPV16 exhibited extensive features of epithelial‐mesenchymal transition, ⁵⁰ and EGFR signalling itself is also known to activate Slug expression upon keratinocyte activation in wound healing. ⁵¹ Keratinocyte activation involves keratin and desmosome remodelling for cell migration, and a reduction of K14 during wound healing may increase plasticity required for migration. K14 is the primary type I keratin interacting with desmosomes, and it was previously shown that K14‐null lines show increased migration in scratch‐wound closure assays. ⁵² Although classic EMT markers such as vimentin and fibronectin were not detected here in the K14 null NEB‐1^H cells (Figure S2), monolayer cultured cells are arguably in a constitutively activated state, and it is plausible that EGF‐mediated K14 silencing could be an early step towards de‐differentiation.

Viruses commonly manipulate the proliferative lifespan of the host cell, and this trait has been exploited to generate immortalized cell‐lines, including generation of cell culture models of human diseases. However, along with gaining immortality, these cells sometimes lose key characteristics of the parental cell.

In this paper we have described a partial mechanism for the observed downregulation of K14 expression upon long‐term culture, by DNA methylation of the K14 promoter, facilitated by HPV and augmented by EGF exposure. The most dramatic downregulation of K14 seen in vivo is the often very sharp cut‐off in the expression of K14 seen in the transition from basal cells to suprabasal cells in the epidermis. However, we have not yet found any evidence to link this to methylation‐driven epigenetic silencing, although further work could be done. The phenomenon of K14 loss is not common, and in the course of generating more than 20 immortal cell lines from human keratinocytes, we have only observed it in the two HPV16‐immortalized cell lines described here, KEB‐7 and NEB‐1, and not even in all strains of these. The speed at which K14 is lost when it happens, and the heterogeneity (possibly at the level of methylation) of affected cells at quite early passage, suggests a possible growth advantage of K14‐negative clones. Nevertheless although this phenomenon may be infrequent and the underlying mechanisms far from fully understood, the extensive use of immortalized lines in research emphasizes the importance of regular assessment of model cell lines in order to ensure phenotypic integrity. In particular, KEB‐7 and NEB‐1 should be used with caution and regularly checked for retention of K14 expression.

AUTHOR CONTRIBUTIONS

Conceptualization: I.S., E.B.L. Methodology: I.S., R.K., D.P.L., E.G.H. Validation: R.K., D.P.L. Formal analysis: I.S., R.K. Investigation: R.K., I.S., D.P.L., E.G.H. Data curation: R.K., I.S. Writing—original draft preparation: I.S., R.K. Writing—review and editing: I.S., R.K., E.B.L. Visualization: I.S., R.K., E.B.L. Supervision: I.S., E.B.L. Project administration: R.K., E.B.L. Funding acquisition and resources: E.B.L.

CONFLICT OF INTEREST STATEMENT

All authors declare no relevant conflicts of interest.

Supporting information

Figure S1. Reciprocal immunoprecipitation with K18 (antibody LDK18) and K19 (antibody LP2K) antibodies confirmed the interaction with K5 (XM26) in HaCaT^H and NEB‐1^H. Letters H and L refer to the antibody heavy and light chains respectively.

Figure S2. NEB‐1^H retains an epithelial gene signature despite K14 loss. Epithelial/mesenchymal gene signatures were studied in low and high passage NEB‐1 lines using shRNA knockdown line N838 as negative control as this line has completely abolished KRT14 transcripts. Non‐adjacent tracks from a single blot remounted side by side for comparison. For antibodies and their sources see Table S2.

Figure S3. Loss of K14 expression at high passage numbers (A) and its re‐induction by 5‐aza‐deoxycytidine (B) in another HPV immortalized keratinocyte line, KEB‐7. ³

Figure S4. Silencing primarily affects the KRT14 locus in the keratin cluster on chromosome 17q21.2, shown by immunoblotting. K18 (KRT18 gene on chromosome 12q13) was included as a positive control of DNA methylated silencing. Aza: 5‐aza‐2′‐deoxycytidine.

Figure S5. Overall promoter DNA methylation level increases in the K14‐negative NEB‐1^H cells. (A) Comparison of methylation levels of CpG sites in high and low passage NEB‐1 cells are shown from two independent bisulfite‐sequencing experiments. Location of CpGs in KRT14 promoter region is numbered according to Figure S8. Note, that first (#1) CpG residue was outside of the PCR target region in this experiment. (b) Detailed promoter methylation pattern of NEB‐1^L and NEB‐1^H clones from experiment #2; none of the clones were fully methylated in all positions in NEB‐1^L, while more than half of the clones derived from NEB‐1^H were methylated at 80%–100%. Bisulfite converted gDNA was PCR amplified and cloned into a TA vector, then 10–20 representative clones per sample were sequenced. Each row of circles represents the continuous sequence of a single amplicon (one subclone). The number of rows indicates the number of sequenced subclones. Methylated and unmethylated CpG residues are depicted with filled and open circles, respectively.

Figure S6. K14 expression was induced by DNA methyltransferase inhibitor in high passage NEB‐1 cells (NEB‐1^H), but it could not be sustained after drug withdrawal upon further cell passages. Low and high passage NEB‐1 cells were treated once with 5‐aza‐dC and then passaged without it. K14 protein expression was maintained in NEB‐1^L cells even at passage 8, but it was lost from NEB‐1^H after replating without 5aza‐dC.

Figure S7. EGF suppresses K14 expression at the transcript level; methyltransferase inhibitor 5‐aza‐dC is more efficient in the absence of EGF (sample NAz). Total RNA samples were reverse transcribed to cDNAs from all four culture conditions and real‐time PCR was carried out with 1 μL of cDNA in the SYBR green master mix (Roche). Amplification, normalization and relative quantification of transcripts were performed using a LightCycler 480 (Roche). Primers are listed in Table 1.

Figure S8. Location of CpGs in the KRT14 promoter. CpG content of promoter region highlighted with grey colour within KRT14 promoter and 5′ sequence. UTR sequence is in italics, TATA box and ATG translational start site in bold letters. CpGs within the CpG island marked with asterisks. The CpG sites in the promoter region whose methylation patterns were studied here are numbered.

Figure S9. Comparison of methylation levels at the 14 CpG sites in the KRT14 promoter showing effect of EGF on demethylation (A) and re‐methylation (B). Promoter methylation data from Figure 3 presented as pair‐wise comparisons. The mean methylation levels at each dinucleotide are illustrated by the small pie chart; black indicates the percent methylation of the 14 CpG sites from the bisulfite sequenced clones presented in Figure 3. Statistical analysis was performed by QUMA web‐based tools. ¹⁷ Fisher's exact test was used to calculate statistical significance of the difference between two bisulfite sequence groups at each CpG site and Mann–Whitney U‐test was used to evaluate that of the entire set of CpG sites. Sample labels as on Figure 3. Red asterisks indicate statistically significant shifts (p = <0.05). (A) Overall demethylation was significantly higher in the absence of EGF; (B) significant remethylation was seen in the presence of EGF regardless of previous treatment of the cells.

Figure S10. DNA methylation is probably not the driver of K14 repression during keratinocyte epidermal differentiation. Undifferentiated K14‐expressing basal monolayer (B) and differentiated K14 negative suprabasal (SB) keratinocytes were collected separately by laser capture microdissection (LCM) from haematoxylin and eosin stained human skin sections (obtained as surgical waste with full local ethical approval). DNA methylation mapping of the KRT14 regulatory (enhancer + promoter) region was performed on both keratinocyte populations. In situ, KRT14 methylation levels are very much lower in the suprabasal layers, suggesting that KRT14 methylation is unlikely to by itself drive downregulation of K14 expression at the start of epidermal differentiation.

Table S1. List of all primers used in this study.

Table S2. List of primary antibodies used in this study. Abbreviations of detection methods used: ICC, immunocytochemistry; IHC, immunohistochemistry; IP, immunoprecipitation; WB, immunoblotting.

Table S3. List of KRT14 allelic variants found by Sanger sequencing in NEB‐1 and primary keratinocytes.

The variants found in the NCBI SNP database are indicated above the columns by their SNP identifier ID codes.

Koh R, Szeverenyi I, Lunny DP, Eng GH, Lane EB. Loss of keratin 14 expression from immortalized keratinocytes by promoter methylation. Exp Dermatol. 2024;33:e15143. doi: 10.1111/exd.15143

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.