Previously we showed that spatial and developmental modulation of ARNT (HIF1β) expression in mouse epidermis is essential for maintenance of keratinocyte differentiation, proper formation of the barrier and normal desquamation. Here, using lentiviral suppression or induction of ARNT in TERT-immortalized (N-TERT) and HaCaT cells we assessed the nature and mechanisms of ARNT involvement in control of differentiation in human epidermal keratinocytes. ARNT depletion did not affect the levels of basal keratins K5 and K14, but significantly induced expression of several key differentiation markers (an effect abolished by EGF supplementation). Furthermore, ARNT deficiency resulted in the downregulation of amphiregulin (AREG) – the most highly expressed EGFR ligand in human keratinocytes – whereas upregulation of ARNT showed the opposite. In ARNT-deficient monolayer cultures and 3D epidermal equivalents, the downregulation of AREG was concurrent with a decline of EGFR and ERK1/2 phosphorylation. TSA, a potent suppressor of HDAC activity, abolished the effects of ARNT deficiency, implying a role for HDACs in ARNT-dependent modulation of the AREG–EGFR pathway and downstream epidermal genes. Total HDAC activity was significantly increased in ARNT-depleted cells and decreased with ARNT overexpression. ARNT-dependent shifts in HDAC activity were specifically attributed to significant changes in the levels of HDAC1, HDAC2 and HDAC3 proteins (but not mRNA) in both monolayer and 3D cultures. Collectively, our results suggest that ARNT controls AREG expression and the downstream EGFR–ERK pathway in keratinocytes, at least in part, by modulating HDAC activity. This novel regulatory pathway targeting advanced stages of epidermal differentiation might have important implications for skin pathology such as psoriasis, atopic dermatitis and cancer.
The aryl hydrocarbon receptor nuclear translocator (ARNT or HIF1β) is a basic helix-loop-helix protein that belongs to a Period–ARNT–single-minded (PAS) protein family. PAS proteins act as heterodimeric transcription factors activated by environmental stress with ARNT being a master dimerization partner in this protein family. Most characterised are the heterodimeric complexes of ARNT with aryl hydrocarbon receptor (AhR) and hypoxia-inducible factors (HIF-α proteins), formed upon AhR or HIF-α activation by halogenated aromatic hydrocarbons or hypoxia, respectively (Wenger and Gassmann, 1997; Beischlag et al., 2008). ARNT is also essential for normal development and its knockout in mice leads to early embryonic lethality due to abnormal angiogenesis, cardial dysfunction, defective haematopoiesis and cranial malformations (Kozak et al., 1997; Maltepe et al., 1997). Possible roles for ARNT acting as a homodimer independent of HIF1α and AhR have also been suggested (Sogawa et al., 1995; Swanson and Yang, 1999).
Given the wide-ranging role of ARNT in adaptation to environmental stress (Gu et al., 2000), its activity is likely to be implicated in control of adaptive responses in normal skin and in pathogenesis of certain environmentally mediated skin disorders. Recently, we and others showed an essential role for epidermal ARNT in mouse skin development, keratinocyte differentiation, barrier function and dermal vascularisation (Takagi et al., 2003; Geng et al., 2006; Wondimu et al., 2012). In contrast to the existing view that ARNT is ubiquitously present in mammalian tissues, we found that its expression in mouse and human skin is subject to significant developmental and spatial changes. In the epidermis, ARNT expression gradually declines through the spinous layer and is totally excluded from the granular layer suggesting a negative regulatory link between ARNT activity and advanced steps of keratinocyte terminal differentiation (Geng et al., 2006). We also showed that alterations in the structure and function of the epidermis in K14-ARNT KO mice are associated with elevated expression of a number of differentiation markers mapped to the epidermal differentiation complex on mouse chromosome 3. Abnormal epidermal differentiation in these mice was also accompanied by deregulation of certain ligands for the epidermal growth factor receptor (EGFR) (Geng et al., 2006).
EGFR belongs to a family of ErbB tyrosine kinase receptors. Upon activation by one of its ligands such as amphiregulin (AREG), beta-cellulin (BTC), epidermal growth factor (EGF), epiregulin (EREG), epithelial mitogen (EPGN), heparin-binding EGF-like growth factor (HBEGF), and transforming growth factor alpha (TGFα) EGFR forms homo- or a heterodimer (with another ErbB family member) leading to its autophosphorylation (Yarden and Schlessinger, 1987). This triggers many downstream signalling cascades including the protein kinase C and Ras-activated ERK1/2 MAP kinase pathways (Roberts and Der, 2007; Getsios et al., 2009), with the later considered as a hallmark of EGFR activity (Schlessinger, 2000). EGFR-mediated signal transduction results in effects on cell proliferation, migration and differentiation in various tissues and organs including the skin (Yarden and Sliwkowski, 2001; Pastore et al., 2008). In the epidermis, dividing keratinocytes of the basal layer are typified by elevated activity of EGFR (Wang et al., 2006), which maintains their sustained proliferation (Hansen et al., 2000; Jost et al., 2000). In turn, the inhibition of EGFR blocks proliferation and induces terminal differentiation of epidermal keratinocytes (Peus et al., 1997; Getsios et al., 2009) associated with induction of differentiation markers such as K1 and K10 (Poumay and Pittelkow, 1995). Furthermore, EGFR signalling is essential for normal development of the epidermis and its appendages (Luetteke et al., 1994; Miettinen et al., 1995; Sibilia and Wagner, 1995; Hansen et al., 1997). The activation of EGFR is also linked to pathogenesis of certain skin diseases such as psoriasis (Nanney et al., 1986) and squamous cell carcinoma (Uribe and Gonzalez, 2011) while disruption of the EGFR pathway inhibits development of papillomas and carcinomas from immortalized epidermal keratinocytes (Woodworth et al., 2000). In wound healing, EGFR plays a positive role by promoting keratinocyte proliferation and migration (Repertinger et al., 2004).
The transition of epidermal keratinocytes through the defined stages of differentiation is associated with tightly regulated expressional shifts. Recently, evidence has begun to emerge demonstrating a key role of epigenetic mechanisms (e.g. histone modifications) in these regulations (Markova et al., 2007; Frye et al., 2007; Sen et al., 2008; Ezhkova et al., 2009; LeBoeuf et al., 2010; Fessing et al., 2011). It was previously reported that in mouse placenta, differentiation of trophoblast cells is controlled by ARNT through modulation of HDAC activity (Maltepe et al., 2005). In addition, class I HDAC1 and HDAC3, as well as class II HDAC4 and HDAC6, were shown to interact with HIF1α and enhance HIF1α/ARNT-mediated transactivation (Qian et al., 2006; Kim et al., 2007). Then, it was shown that altered HDAC1 activity plays a central role in Cyp1a1 expression driven by AhR – another conventional ARNT dimerization partner (Schnekenburger et al., 2007). Although these findings do not implicitly show direct ARNT–HDAC interaction, they suggest that HDACs are mediating downstream effects of ARNT activity.
Here, using N-TERT and HaCaT human epidermal keratinocyte cell lines with variable level of ARNT expression and 3D epidermal equivalent models developed using these cells, we show that ARNT deficiency positively regulates expression of keratinocyte differentiation markers through suppression of the EGFR ligand AREG and concomitant downregulation of EGFR activity and ERK1/2 phosphorylation. ARNT overexpression shows opposite effects. This pathway is controlled through modulation of HDAC activity, which appears to be the first step in this novel ARNT-dependent regulatory cascade targeting advanced stages of keratinocyte differentiation programme.
Modulation of ARNT activity in N-TERT and HaCaT keratinocytes
Using transduction with lentiviral particles created from Sigma Mission shRNA plasmids we developed two N-TERT keratinocyte cell lines with stable suppression of ARNT (N-KO1 and N-KO2; Fig. 1A–D). 3D equivalents developed using ARNT-depleted keratinocytes (N-KO) showed an evident decline in ARNT protein compared to control models (Fig. 1F).
Transduction of N-TERT keratinocytes with the lentiviral vector pCDH1-MCS1-EF1-copGFP (al Yacoub et al., 2007) containing the full ORF fragment of ARNT cDNA resulted in a significant increase of ARNT expression at both mRNA and protein levels (Fig. 1C,E). This N-OXP cell line showed a gradual loss of vector-retaining cells (up to 50% by day 44 of culture; supplementary material Fig. S2). Nevertheless, 3D epidermal equivalents developed using N-OXP keratinocytes showed at least twice the level of ARNT protein as compared to control models even at the time of 3D culture completion (day 42 post-transduction; Fig. 1E,G) thus confirming that gradual loss of ARNT overexpressing cells does not impede the use of this system as a valid model. Of note, ARNT transduction resulted not only in a general increase of ARNT protein level in 3D cultures but also in its ectopic appearance in the granular layer of ARNT-OXP epidermal equivalents (Fig. 1G).
In HaCaT keratinocytes, ARNT was effectively suppressed by ARNT shRNA (H-KO1 and H-KO2 cell lines; supplementary material Fig. S1A,B), or transiently upregulated with an ARNT/GFP-tagged expression vector (H-OXP1, H-OXP2; supplementary material Fig. S1C,D).
Depletion of ARNT promotes keratinocyte differentiation
While qPCR revealed no changes in the mRNA level of basal keratins K5 and K14, differentiation-associated keratins K1 and K10, transglutaminase 1 (TGM1) and involucrin showed a 2.5- to 3.5-fold increase in ARNT-depleted N-TERT (N-KO) cells. Transcription of late differentiation markers filaggrin and loricrin was upregulated in these cells by 30 and 70 fold respectively (Fig. 2A). Western blot (WB) analysis of N-KO keratinocytes demonstrated a corresponding increase in involucrin, TGM1 and K1/10 at the protein level (Fig. 2B, for quantification see supplementary material Fig. S3F). Filaggrin and loricrin proteins were not detected in N-KO cells cultured in monolayer, even though their mRNA levels were significantly induced by ARNT depletion. However, in 3D epidermal equivalents filaggrin protein was easily detected. WB results revealed a prominent increase of both filaggrin and K10 proteins in ARNT-depleted 3D cultures while in ARNT-overexpressing epidermal equivalents we found a substantial reduction of K10 and fully processed (40 kDa) filaggrin monomers (Fig. 2C, for quantification see supplementary material Fig. S3G). Importantly, supplementation of culture medium with EGF abolished the stimulatory effects of ARNT knockout on transcription of differentiation markers with the only exception being K10 (Fig. 2D).
3D epidermal equivalents developed using ARNT-depleted cells showed significant increase in the number of cornifying cell layers along with retention of nuclei (parakeratosis) as compared to control equivalents (Fig. 2F, H&E, arrows). Additionally, ARNT-deficient 3D culture models were characterized by a reduced number of granule-containing cells and a loss of granular layer contiguity (Fig. 2F, H&E).
Immunohistochemistry (IHC) on paraffin sections also revealed noticeable changes in the expression of several key differentiation markers in N-KO epidermal equivalents (Fig. 2F). In the control models, filaggrin was localized to the upper suprabasal layers corresponding to the granular layer of the normal epidermis. In N-KO models, filaggrin was increased in the uppermost layers (Fig. 2F, Filaggrin) and also became apparent in some immediate suprabasal cells suggesting premature induction of its translation or processing (supplementary material Fig. S2E, arrows). K10 was excluded from the basal layer of the control epidermal equivalents and showed low positivity in the uppermost layers being most prominent in the spinous cells overlaying the basal layer (ADD Fig. 2F, Keratin 10, control, arrows). In N-KO 3D models, the overall level of K10 was markedly increased (confirmed by WB, Fig. 2C) in all suprabasal layers including the uppermost ‘cornifying’ layers which were only weakly positive in controls (Fig. 2F, Keratin 10, N-KO2).
Finally, involucrin expression in control cultures was confined to the cytoplasm of suprabasal keratinocytes with most prominent positivity in the granular layer (Fig. 2F, Involucrin, arrows). ARNT-deficient 3D models showed significant increase of involucrin positivity, which appeared even in the nuclei of basal and spinous keratinocytes.
ARNT-overexpressing 3D cultures revealed no apparent differences in morphology as compared to control epidermal equivalents. At the same time, they appeared to be slightly thinner than control models (supplementary material Fig. S2C,D), while ARNT-deficient 3D epidermal equivalents were usually thicker compared to controls (Fig. 2F). This effect seems to be related to a negative correlation between the level of ARNT and the size of keratinocytes rather than to variations in the number of cell layers in corresponding 3D cultures.
ARNT deficiency has a negative effect upon keratinocyte growth
As shown by direct cell counting (Neubauer hemocytometer), in EGF-depleted medium, N-KO cells have a markedly reduced growth rates as compared to the control. This effect of ARNT deficiency was abolished when the medium was supplemented with EGF (Fig. 2E). The negative effect of ARNT deficiency upon keratinocyte growth was validated by WST-1 cell proliferation assay, which showed approximately a 25% decline of readings (irrespective of Ca2+ concentration) in N-KO monolayer cultures (supplementary material Fig. S3A).
Depletion of ARNT results in downregulation of amphiregulin and suppression of EGFR activity in human keratinocytes
In line with recent findings (Stoll et al., 2010a), our data showed that AREG is the most highly transcribed EGFR ligand in normal (control) N-TERT keratinocytes (about 70% of GAPDH expression level). TGFα message was detected at the level of about 10% of GAPDH while other EGFR ligands showed no or negligible expression in control cells (Fig. 3A). Comparing the level of AREG in N-KO cells with that of the controls revealed a significant downregulation of AREG mRNA – more than 50% in N-KO2 (Fig. 3B, shaded bars). Genes coding for other EGFR ligands showed insignificant changes (supplementary material Fig. S3C). The high dependence of AREG transcription on the level of ARNT was further confirmed by a prominent increase of AREG message in ARNT-overexpressing HaCaT keratinocytes (Fig. 3H).
Significantly less AREG protein was secreted into the culture medium by N-KO cells than by control keratinocytes (Fig. 3C). Correspondingly, in ARNT-depleted epidermal equivalents, AREG positivity (IHC) was fainter at the cell membrane and eradicated from the nuclei as compared to controls where strong nuclear and cell membrane positivity was observed (Fig. 3E, AREG). Concomitantly with downregulation of AREG in ARNT-depleted models, we found a prominent increase of AREG positivity in ARNT-overexpressing 3D epidermal equivalents (Fig. 3D).
EGFR phosphorylation at Tyr1173 (pEGFR) was significantly reduced in the N-KO cells cultured in monolayer (Fig. 3C). This was also confirmed by immunofluorescence: blind counting of pEGFR-positive foci performed in duplicate revealed that the number of foci per cell in N-KO keratinocytes was reduced by two fold as compared to the controls. This result was not apparent when the culture medium was supplemented with EGF (Fig. 3F,G). In the control 3D epidermal equivalents, IHC showed prominent pEGFR-positive signal at the plasma membranes of the spinous cells and in the cytoplasm of the uppermost (‘granular’) cell layers (Fig. 3E, pEGFR, arrows), whereas in N-KO 3D cultures pEGFR positivity was not observed. While ARNT deficiency significantly affected EGFR phosphorylation, it had no effect on EGFR transcription (supplementary material Fig. S3B).
Induction of ERK1/2 phosphorylation is a classical downstream effect of the EGFR-Ras signalling in many cell types including epidermal keratinocytes (Roberts and Der, 2007; Getsios et al., 2009). As shown by WB, ARNT depletion did not affect the total level of ERK proteins but significantly reduced ERK1/2 phosphorylation (Fig. 3I). Decrease of pERK1/2 in N-KO cells was also confirmed by IHC in the epidermal equivalents (Fig. 3E). In the control 3D cultures, pERK1/2 was localised to the uppermost layers of flattening cells (same pattern as for pEGFR, Fig. 3E, arrows) and to the nuclei of some basal keratinocytes (Fig. 3E, arrowheads). In N-KO equivalents, pERK1/2 positivity was lost or significantly reduced. No apparent difference in the level of non-phosphorylated ERK1/2 proteins between control and N-KO models was observed (Fig. 3E, ERK1/2).
Effects of ARNT depletion upon expression of downstream target genes are modulated through HDAC activity
Suppression of HDAC activity using trichostatin-A (TSA), a potent and specific inhibitor of class I and II HDACs, eradicated the negative effects of ARNT deficiency on the level of AREG transcription (Fig. 3B). Furthermore, TSA treatment eliminated the stimulatory effects of ARNT-deficiency upon transcription of keratinocyte differentiation markers as evidenced from the qPCR analysis of TSA-treated normal and N-KO keratinocytes when results were compared to corresponding (control and ARNT-deficient) untreated cell cultures (Fig. 4A). At the same time, comparison of all three manipulated cultures (normal TSA-treated, N-KO, TSA-treated N-KO) against untreated normal cells revealed that differentiation markers had different responses to TSA (Fig. 4B). Transcription of filaggrin, K10 and loricrin in N-TERT cells with a normal level of ARNT was significantly reduced by TSA; TGM1 and K1 mRNA showed a moderate increase while basal keratins K5 and K14 were not responsive to TSA treatment (as shown by both methods of comparison). Combination of TSA treatment with ARNT deficiency resulted in suppression of the stimulating effects of ARNT knockdown for filaggrin, K10 and loricrin mRNA while K5 and K14 again showed no response. Despite both TSA treatment and ARNT deficiency have a positive effect upon TGM1 and K1 transcription, combination of these two factors resulted in abrogation of their individual stimulatory effects.
In epidermal keratinocytes, the level of HDAC proteins and the total HDAC activity are dependent on ARNT
An HDAC activity assay showed that in ARNT-depleted N-TERT cells total HDAC activity was increased by 15–20%. In ARNT-depleted HaCaT keratinocytes this increase reached 50% (Fig. 5A,B). Correspondingly, ARNT overexpression in HaCaTs resulted in a ∼50% decrease of total HDAC activity (Fig. 5B). To identify specific HDACs accountable for these changes we first identified HDACs with a significant level of transcription in control N-TERT keratinocytes (Fig. 5C). In these cells, class I HDAC1 and HDAC3 appeared to be the most prominently expressed histone deacetylases. Among the SIRT genes (coding for class III HDACs), only SIRT2 showed a noticeable (but still low) level of expression. In ARNT-depleted N-KO1 and N-KO2 cells cultured in monolayer, the mRNA level of all HDAC and SIRT genes remained unchanged as compared to the controls (supplementary material Fig. S3D,E). However, at the protein level, depletion of ARNT resulted in a significant increase of HDAC1, HDAC2 and HDAC3 (Fig. 5D). In ARNT-depleted HaCaT cells the increase of HDAC1 and HDAC3 protein levels was even more prominent (Fig. 5E) matching the changes in total HDAC activity (Fig. 5A,B). Overexpression of ARNT in both N-TERT and HaCaT keratinocytes cultured in monolayer led to a prominent drop in HDAC1, HDAC2 and HDAC3 protein levels (Fig. 5F,G).
Immunohistochemistry in ARNT-deficient 3D epidermal equivalents revealed an increase in HDAC1 protein in the keratinocyte nuclei as compared to controls (Fig. 6A). In agreement with previous reports (Longworth and Laimins, 2006), HDAC3 protein in control 3D models was seen not only in the nuclei but in the cytoplasm and at the plasma membrane as well. In control cultures, the relatively low (compared to HDAC1) nuclear HDAC3 positivity was mostly confined to the lower cell layers whilst nuclei in the upper layers were HDAC3-negative (Fig. 6A, HDAC3, arrowheads). In the N-KO 3D models, high cytoplasmic positivity of HDAC3 was seen in all cells including the uppermost layers, which were HDAC3 negative in controls (Fig. 6A, arrows). Nuclear HDAC3 staining increased in the lower layers and also appeared in suprabasal keratinocytes (Fig. 6A). Increase of HDAC1, HDAC2 and HDAC3 protein levels in ARNT-deficient 3D cultures was confirmed by WB (Fig. 6B). Concomitantly, in ARNT-overexpressing 3D epidermal equivalents, the level of HDAC1 was significantly reduced (Fig. 6C).
ARNT is generally thought to be a ubiquitously expressed protein. However, we have previously demonstrated strong developmental and differentiation-associated regulation of ARNT in mouse epidermis, which appeared to be essential for epidermal development, cornification, barrier formation, desquamation and epidermal–dermal communication (Geng et al., 2006; Wondimu et al., 2012). Here, we elucidate a novel ARNT-dependent mechanism that controls advanced stages of differentiation in human epidermal keratinocytes. Given the essential role of ARNT in the activity of HIF- and AhR-dependent pathways implicated in adaptive response to hypoxia, organic toxicants, ozone and UV (Wenger and Gassmann, 1997; Sogawa et al., 1995; Afaq et al., 2009; Fritsche et al., 2007) our studies also expose a potential mechanistic link between environmental stress and the process of keratinocyte terminal differentiation in normal and diseased human skin.
According to our results, suppression of endogenous ARNT significantly affects the process of differentiation in human keratinocytes. Altered expression of key differentiation markers in N-KO cells is concomitant with morphological changes in ARNT-depleted 3D epidermal equivalents i.e. increased number of cornifying layers, diminished granular layer, and parakeratosis – all signs of deregulated differentiation.
At the same time, the epidermal markers reveal different modes of response to variations in the level of ARNT in vitro: basal keratins K5 and K14 showed no sensitivity at all; suprabasal differentiation markers such as K1/10, TGM1 and involucrin were induced by ARNT deficiency at both mRNA (moderately) and protein levels; while filaggrin and loricrin – specific markers of the granular layer – were dramatically induced at mRNA level (approximately 10-fold that of other differentiation markers) but were not reliably detected at the protein level in control and N-KO monolayer cultures. We assume that monolayer culture may not provide the correct microenvironment for normal processing of these proteins – a situation analogous with β-casein expression in mammary epithelial cells cultured in monolayer or in 3D conditions (Li et al., 1987). In line with this assumption, in 3D epidermal equivalents (which represent a more adequate model of differentiating epidermis) filaggrin protein was easily detected (Fig. 2C). The changes between control and ARNT-depleted or -overexpressing samples were proportional for different filaggrin-positive bands – 40 kDa, 74 kDa and 120 kDa – presumably corresponding to filaggrin monomers, dimers and trimers (Dale et al., 1985; Sandilands et al., 2009). This fact, together with dramatic changes in filaggrin mRNA level (Fig. 2A) suggests that ARNT controls (negatively) expression of profilaggrin gene rather than proteolytic processing of filaggrin protein. Ectopic appearance of ARNT positivity in the granular layer of N-OXP 3D epidermal equivalents and concomitant drop in total filaggrin protein level in these models (Fig. 2C) further support this suggestion.
Overall, the induction of differentiation markers in ARNT-deficient keratinocytes and their suppression in ARNT-overexpressing 3D models (Fig. 2C) correspond to the absence of ARNT protein in the upper layers of control 3D epidermal equivalents (Fig. 1F; Control) and to the exclusion of this protein from the granular layer of normal mouse and human epidermis (Geng et al., 2006). In general, the epidermal phenotype in ARNT-deficient mice matches our findings in ARNT-KO 3D epidermal equivalents and monolayer cultures (profound acceleration of differentiation in both in vivo and in vitro systems). However, there are some discrepancies as well. In the epidermis of ARNT-KO mouse newborns we observed some reduction of loricrin and filaggrin (Geng et al., 2006) whereas ARNT-deficient 3D epidermal equivalents are characterized by profound induction of these proteins. We believe that reduction of loricrin and filaggrin – two specific granular layer proteins – in the ARNT-deficient mouse epidermis can be attributed to abrupt cornification and the virtual absence of a granular layer in these mice. Involucrin, which is normally expressed not only in the granular but also in the spinous layer, did not show a decline in ARNT-deficient mouse epidermis but rather is increased (Geng et al., 2006) as we observed in the human system (3D equivalents). The abrupt cornification characteristic of the ARNT-null mouse epidermis is not as prominent in 3D culture and therefore the granular layer, while abnormal, is present in this model. This allows us to perceive the effects of ARNT depletion upon the expression of loricrin and filaggrin. Thus, the appearing discrepancy in expression of loricrin and filaggrin between our previous in vivo and current in vitro results is likely attributed to the basic developmental and structural differences between intact epidermis (in vivo) and in vitro experimental 3D models.
Activation of the EGFR pathway by ligand binding is known to suppress differentiation and positively affect growth in keratinocytes (Monzon et al., 1996; Peus et al., 1997; Jost et al., 2000; Schneider et al., 2008). Here we found that effects of ARNT depletion upon keratinocyte differentiation (positive) and growth (negative) are abolished by supplementing cell culture medium with EGF thus suggesting that these effects are instigated by downregulation of certain EGFR ligand(s). The implication of the EGFR pathway in ARNT control over the expression of differentiation markers is further confirmed by decreased EGFR and ERK1/2 phosphorylation in N-KO cells and its restoration in EGF-supplemented medium. Decrease of EGFR phosphorylation in ARNT-depleted keratinocytes is also supported by significantly smaller number of pEGFR-positive cytoplasmic foci, which are presumably representing late endosomes. Previously it was shown that EGFR endocytosis is driven by its autophosphorylation (Waterman and Yarden, 2001). Therefore, a reduced level of EGFR internalization (Fig. 3F) corresponds with a decline of its phosphorylation (Fig. 3C,E) in ARNT-deficient cells.
In line with previous reports designating AREG as a principal EGFR ligand mediating autocrine growth in epidermal keratinocytes (Piepkorn et al., 1994; Nylander et al., 1998; Stoll et al., 2010a) we showed that AREG is the most abundant EGFR ligand in normal N-TERT cells. In N-KO keratinocytes AREG declined significantly at both mRNA and protein levels and its nuclear positivity was completely lost (Fig. 3B,C,E) whereas other conventional EGFR ligands showed no changes (supplementary material Fig. S3C). These findings suggest suppression of AREG as a likely mechanistic explanation for downregulation of EGFR pathway in ARNT-depleted cells.
Recently, it was shown that knockdown of AREG in human epidermal keratinocytes results in reduced ERK phosphorylation, inhibition of growth, and induction of K1/10, involucrin, loricrin and TGM1 (Stoll et al., 2010b) – exactly the same set of changes we observed in ARNT-depleted N-TERTs. This strongly supports our proposition that decreased EGFR and ERK phosphorylation and subsequent induction of differentiation markers in N-KO cells is mediated through downregulation of AREG, but not through modulation of EGFR expression. This view is supported by the lack of ARNT effects upon EGFR mRNA level (supplementary material Fig. S3B).
Loss of nuclear AREG in ARNT-depleted keratinocytes (Fig. 3E) is of particular interest here. It has been shown previously that the C-terminal transmembrane fragment of AREG can be internalized and accumulated in the nucleus to block keratinocyte differentiation (Stoll et al., 2010b). Similarly, internalization and nuclear accumulation of the C-terminal transmembrane fragment of another conventional EGFR ligand, HB-EGF, leads to silencing of transcription (Nanba et al., 2003; Toki et al., 2005). Taking these findings into account, our results identify ARNT as an upstream positive regulator of both membrane (EGFR activation) and nuclear functions of AREG. The fact that induction of K10 expression in N-KO cells is not amended by EGF supplementation suggests that, in contrast to other differentiation markers, transcription of K10 may be specifically regulated by nuclear C-terminal domain of AREG through the EGFR-independent mechanism mentioned above (Fig. 7).
It is increasingly apparent that HDAC activity is essential for epidermal development and differentiation (Markova et al., 2007; Ezhkova et al., 2009; LeBoeuf et al., 2010). ARNT controls differentiation of trophoblast cells in the mouse placenta through modulation of HDAC activity (Maltepe et al., 2005) and it is very likely that similar mechanisms may control epidermal differentiation. As shown by our results, TSA treatment abolishes effects of ARNT deficiency upon expression of AREG and differentiation markers thus strongly implicating HDAC activity in ARNT control over AREG–EGFR pathway and keratinocyte differentiation. However, we found that in normal N-TERT cells different epidermal genes respond to TSA treatment differently and, most interestingly, the mode of their response to TSA appeared to match the changes in their expression in response to ARNT depletion.
As evident from Table 1, the basal keratins K5 and K14 comprise a group of epidermal genes (Group 1), which, in contrast to differentiation markers, are not responsive to ARNT deficiency and also show no response to TSA. Previously, based on the studies of differentiation in mouse trophoblast it was suggested that HDAC inhibition phenocopies ARNT deficiency (Maltepe et al., 2005). In normal keratinocytes, we found this to be true for TGM1 and for K1 genes only (Group 2) which both are induced by TSA and upregulated by ARNT deficiency. Involucrin occupies an intermediate position between Group 1 and 2 genes as it is not responsive to TSA (as Group 1) but is positively regulated by ARNT depletion (as Group 2 genes). Filaggrin, K10 and loricrin genes (Group 3) are strongly induced by ARNT depletion and are significantly suppressed by TSA treatment in N-TERT keratinocytes. Moreover, TSA treatment of N-KO cells not only eradicates stimulatory effect of ARNT deficiency on Group 3 genes but significantly downregulates their expression as compared to control. Thus, for genes of Group 3, which are negatively regulated by ARNT, HDAC activity represents a key limiting factor. Also, there is an inverse correlation between the changes in expression of Group 3 genes and AREG in response to all kinds of manipulation (Table 1) further suggesting that filaggrin, K10 and loricrin genes are negatively regulated by ARNT through HDAC- and AREG- mediated signalling. Interestingly, HDAC1- and HDAC2-knockout mice show an absence of K10 and loricrin expression in the epidermis (LeBoeuf et al., 2010) thus supporting our conclusion that effects of ARNT upon AREG, EGFR and downstream differentiation markers are mediated by HDAC activity. It was previously reported that TSA alters expression of epidermal genes in a protein synthesis-dependent manner (Markova et al., 2007) implying existence of an intermediate protein regulator. Our data identify AREG as a putative regulatory link between HDAC activity and expression of filaggrin, loricrin and K10.
Thus, we have demonstrated that ARNT modulates keratinocyte differentiation through HDAC- and EGFR-dependent mechanisms. AREG appears to be the key intermediate player linking together these two pathways. There is evidence that PAS proteins such as AhR and HIF2α (which both require dimerization with ARNT for their activity) control the expression of AREG in urinal tract and cancer tissues (Choi et al., 2006; Stiehl et al., 2011). These findings are in line with the role of ARNT in control of AREG/EGFR pathway in the epidermis revealed by our studies.
Importantly, while HDAC activity is essential for ARNT effects upon AREG expression and keratinocyte differentiation, HDACs themselves appeared to be dependent on ARNT level: depletion of ARNT results in the prominent induction of total HDAC activity and significant increase in HDAC1, HDAC2 and HDAC3 protein levels. Concomitantly, upregulation of ARNT leads to decrease of HDAC1 level (Fig. 6B,C).
All these interactions are shown on a schematic drawing (Fig. 7) representing a putative model for ARNT control over epidermal differentiation. This model suggests that ARNT, being nuclear (and hence active) specifically in the lower layers of 3D epidermal equivalents (Fig. 1F), is implicated in the deterrence of differentiation in spinous keratinocytes. As keratinocytes approach granular layer, ARNT is downregulated leading to an increase in HDAC activity and suppression of AREG/EGFR pathway thus triggering the expression of specific granular layer proteins such as filaggrin and loricrin. This view also fits prevalent expression of ARNT in lower layers of mouse and human epidermis and its decline in the granular layer (Geng et al., 2006). Our model, while instructive, still remains a subject for further validation since the role of ARNT dimerization partners (HIF-α proteins and AhR) in control of the ARNT-AREG-EGFR pathway and mechanisms of ARNT control over activity of specific HDACs are not yet clear, nor is the role of AREG in control of K10 expression (Fig. 7, question marks). These issues will be addressed in our future studies.
Materials and Methods
Cell culture, modulation of ARNT expression and drug treatment
Human N-TERT keratinocytes immortalized with hTERT (a gift from J. Rheinwald, Harvard Medical School, Boston, MA) were maintained in GIBCO keratinocyte serum-free medium supplemented with BPE, Ca2+ and EGF (final concentrations 25 µg/ml, 0.4 mM, 0.2 ng/ml, respectively) (Rheinwald and Beckett, 1981). In experiments analysing the effect of EGF this supplement was omitted from the media. To stably deplete ARNT in N-TERTs we used transduction with lentiviral particles created from Sigma Mission shRNA plasmids (SHGLY-NM_001668 and non-targeting control plasmid SHC002; see supplementary material Table S1) as described previously (al Yacoub et al., 2007). Successfully transduced N-TERT keratinocytes were selected for and further maintained with 0.5 µg/µl puromycin at low Ca2+ (0.4 mM). Comparison of two ARNT-depleted N-TERT cell lines (N-KO) with the non-targeting shRNA control showed 60–80% decrease of ARNT mRNA (Fig. 1A,C) and 81% and 96% decrease of ARNT protein for N-KO1 and N-KO2, respectively (Fig. 1B,C). Immunocytochemistry demonstrated a high level of nuclear ARNT protein in control N-TERTs and its visible decline in the nuclei of N-KO cells (Fig. 1D).
For viral overexpression of ARNT in N-TERT cells, ARNT cDNA ORF fragment was cut out and purified from the OriGene expression vector (RG216724; OriGene Technologies, Rockville, MD) by restriction enzyme digest with BamHI and NotI (R0136S and R0189S; NEB, Hitchin, UK). Lentiviral transfer vector pCDH1-MCS1-EF1-copGFP (al Yacoub et al., 2007) was digested with BamHI and NotI and the obtained ARNT fragment was ligated in. The ligation reaction was transformed into Turbo Competent E. coli (C2984H; NEB). Resulting ampicillin-resistant colonies were mini-prepped and checked for the correct insertion by restriction digest and direct sequencing. These preparations (along with empty pCDH1-MCS1-EF1-copGFP vector for control purposes) were then used to create lentiviral particles further utilized for transduction of N-TERT keratinocytes (N-OXP) as described previously (al Yacoub et al., 2007). Efficiency of transduction in N-OXP keratinocytes was monitored by FACS analysis of GFP fluorescence (supplementary material Fig. S2) using a BD Biosciences (Oxford, UK) LSR Fortessa Flow Cytometer.
HaCaT keratinocytes (ECACC; Salisbury, UK) were transfected with ARNT shRNA expressing vectors (Sigma; Dorset, UK) using Lipofectamine 2000 and selected for stable transformants with puromycin. Three constructs targeting different regions of the ARNT mRNA were used (supplementary material Table S1). Cells transfected with non-targeting shRNA vector were used as a control. The use of constructs 1 and 2 resulted in ∼40 and 60% reduction of ARNT mRNA (supplementary material Fig. S1A) and ∼90% reduction of ARNT protein level whereas construct 3 gave only 20% reduction (supplementary material Fig. S1B). For further studies we selected cell lines obtained using constructs 1 and 2 only (H-KO 1 and 2). Transient over-expression of ARNT in HaCaT cells (H-OXP) using lipofectamine-mediated transfection with an ARNT/GFP-tagged expression vector (OriGene; Rockville, MD) resulted in ∼200-500× increase of ARNT mRNA level (supplementary material Fig. S1C) and about 1.5- to 3.5-fold increase of total ARNT protein (supplementary material Fig. S1D).
For suppression of total HDAC activity cells were grown in culture medium containing 100 nM TSA for 48 hrs prior to taking lysates for protein and RNA isolation. Development of 3D epidermal equivalents using N-KO or N-OXP cell lines resulted in formation of epidermis-like structures with significant depletion or increase of ARNT levels, respectively, as was confirmed by IHC (Fig. 1F,G) and WB (Fig. 1E).
3D cell culture
N-TERT keratinocytes (2×105) were resuspended in 400 μl PCT epidermal keratinocyte medium (CnT-57; CELLnTEC, Bern, Switzerland) and added to Millicell PCF 12 mm inserts with 0.4 um pore size (PIHP01250; Millipore, Hertfordshire, UK) placed in a 60 mm Petri dish. Medium was added so that levels were equal with that inside the insert. After formation of a confluent monolayer the medium was replaced inside and outside the insert with 3D medium (CnT-02-3DP1; CELLnTEC). The next day, the medium was aspirated from inside the inserts. Inserts were left for 14 days with medium changes every three days. Membranes and attached cells were then removed from the inserts, fixed in formalin overnight, processed through graded alcohols to xylene and embedded in paraffin blocks. 4-µm-thick sections were cut for H&E and IHC staining using a Leica RM2135 rotary microtome.
RNA isolation and real-time PCR
RNA was isolated from cells grown to 70–80% confluency using Tri Reagent (Sigma, Dorset, UK) according to the manufacturer's protocol. RNA was diluted 1∶10 and used in qPCR reactions with Invitrogen Superscript III Platinum SYBR Green One-Step qRT-PCR. Reactions were run on a Bio-Rad mini-opticon real-time PCR machine as previously described (Weir et al., 2011a). Primers used are shown in supplementary material Table S2.
Protein isolation and western blotting
Protein lysates were obtained from 70–80% confluent keratinocytes as reported previously (Geng et al., 2006) and processed for WB (Weir et al., 2011a) using appropriate antibodies (supplementary material Table S3).
Cells were cultured on 2 well permanox chamber slides (Lab-Tek; Scotts Valley, CA). For assessment of ARNT localisation cultures were fixed with acetone for 2 min at −20°C. For examination of EGFR phosphorylation, cells were fixed with 4% paraformaldehyde at RT, permeabilized with 0.1% Triton X-100 for 5 min and blocked for 30 min with 0.3% BSA in PBS blocking solution. After fixation and blocking, cells were incubated with primary antibodies (supplementary material Table S3) made up in either PBS or blocking solution. Washes were repeated followed by incubation with secondary antibodies made up in PBS. Cells were mounted using ProLong Gold antifade reagent with DAPI (Invitrogen) and examined using Zeis Axioskop2 microscope with fluorescence.
Deparaffinized sections of 3D epidermal equivalents were boiled in the microwave for 3×3 min in 10 mM sodium citrate (pH 6.0) for antigen unmasking, incubated with primary antibodies (supplementary material Table S3) at RT for 1 hr with blocking and processed further using the Vectastain Elite ABC kit (pk-6200; Vector Laboratories, Peterborough, UK). Samples were developed using the DAB peroxidase kit (pk-4100, Vector Laboratories) and counterstained with Haematoxylin. Samples treated with no primary antibodies were used as a negative control.
WST-1 assay and cell counting
The water-soluble tetrazolium salt (WST-1) assay, a standard method to assess cell viability and proliferation (K301-500; BioVision; distributed by Cambridge BioSciences, Cambridge, UK), was performed as described previously (Weir et al., 2011b). To induce keratinocyte differentiation in this experiment, ARNT-deficient N-TERT cells were exposed to high Ca2+ (1.4 mM) as compared to normal cell culture conditions (0.4 mM). For direct counting, cells were released with 0.05% trypsin and counted in triplicate using a Neubauer hemocytometer. Trypan Blue was used to exclude dead cells from the live cell count.
HDAC activity assay
HDAC activity was performed using a Fluorometric HDAC activity assay kit (ab1438; Abcam, Cambridge, UK). Equal amounts of cell lysate (10–50 µg) were used and the assay was carried out as per the manufacturer's description. Results were retrieved using a SpectraMax Gemini XS fluorescence plate reader (Molecular Devices, Berkshire, UK).
The authors are grateful to Celine Pourreyron (Skin Cancer Group, University of Dundee) for her valuable help with some of the techniques used in this study. We also acknowledge the Flow Cytometry Core Facility at the College of Medicine, Dentistry and Nursing at the University of Dundee for assistance with FACS analysis.
Our work was funded by Cancer Research UK [grant number C5314/A6695].
Supplementary material available online at http://jcs.biologists.org/lookup/suppl/doi:10.1242/jcs.095125/-/DC1
- Accepted March 22, 2012.
- © 2012. Published by The Company of Biologists Ltd