Retinoic acid receptor expression in human skin keratinocytes and dermal fibroblasts in vitro

Retinoic acid has marked effects on skin and epithelia, and in excess inhibits normal keratinocyte differentiation and may induce mucous metaplasia; conversely, retinoic acid deficiency results in hyperkeratosis or squamous metaplasia (Sengel, 1976). Many studies have been directed at elucidating the effects of retinoic acid on keratinocytes, the major epidermal cell type, and it is clear that retinoic acid directly influences proliferation and alters the differentiation pathway of these cells in vitro (Green and Watt, 1982; Marcelo and Tomich, 1983; Redfern and Todd, 1988). Although there is little doubt that retinoic acid has direct effects on epidermis, some of the observed effects of retinoic acid on epithelial tissues and their derivatives in vivo may, in fact, be mediated by underlying mesenchymal or stromal cells (Tickle et al. 1989; Covant and Hardy, 1990; Hardy et al. 1990).

Specific cellular binding proteins (cellular retinoic acid binding protein or CRAJBP) and nuclear receptors for retinoic acid have been identified (Ong and Chytil, 1978; Daly and Redfern, 1987) and cloned (Petkovich et al. 1987; Giguere et al. 1987; Brand et al. 1988; Zelent et al. 1989; Krust et al. 1989); these presumably mediate the diverse effects of retinoic acid on different tissues. Retinoic acid receptors (RARs) are ligand-activated transcriptional regulators closely related in structure to steroid and thyroid hormone receptors (review: Redfern, 1992). Three main classes of RAR have been described and, for RAR-β and RAR-γ at least, alternative splicing and transcription from different promoters generates transcripts coding for RAR proteins differing at their amino-terminal ends (Zelent et al. 1991; Lehmann et al. 1991). RARs and the cytosolic retinoid binding proteins, CRABP and cellular retinol binding protein (CRBP), are expressed in precise spatiotemporal patterns during embryological development (Dollé et al. 1990; Ruberte et al. 1991). This developmentally regulated expression of specific retinoic acid receptors, coupled with the dramatic disruptions of normal pattern formation caused by excess retinoic acid (Tickle et al. 1982; Eichele, 1989), argues that retinoic acid functions as a regulatory molecule involved in the control of morphogenesis. Cell and tissue specificity in the biological effects of retinoic acid will be determined, at least in part, by the patterns of expression of different retinoic acid receptors.

With respect to skin, although retinoic acid can, in vitro at least, apparently determine whether keratinocytes differentiate to a squamous or mucus-secreting cell phenotype (Sengel, 1976), its role in the normal development and maintenance of epidermis in vivo is uncertain. Since interactions between epithelial and mesenchymal elements are important in both epidermal differentiation and the development of epidermal structures (Sengel, 1976), it is important to consider dermis as a potential mediator of the effects of retinoic acid on epidermis in vivo. Cultured cells can be relatively readily derived from normal human skin tissue and represent potentially useful models for studying the roles of retinoic acid in skin biology. We have addressed the question of whether or not retinoic acid is likely to have different developmental roles in different skin cell types by studying the patterns of expression of RARs in human keratinocytes and fibroblasts cultured in vitro.

Kératinocyte cultures were established from foreskin, retro auricular skin or foetal skin (abdominal), grown under serum-free conditions in medium MCDB 153 (Sharpe et al. 1989) and used after three to six passages. To induce stratification, the calcium ion concentration was increased from 0.07 mM to 1.2 mM, and the cells were used for experiments after 4 days at this higher Ca²⁺ concentration. Dermal fibroblast cultures were established by explant outgrowth from small blocks (˜8 mm³) of dermal tissue placed on a scratched surface of plastic tissue-culture dishes. Cells that grew from the explant were subsequently passaged and cultured in Dulbecco’s minimal essential medium (DMEM) containing 10% foetal bovine serum (FBS). The origin of each fibroblast primary cell line is given in Table 1.

For experiments with retinoic acid, dermal fibroblasts were used when approximately 70% confluent and the medium was replaced 12–24 hours before adding all trans retinoic acid (Sigma) in ethanol to final concentrations within the range 0.1 to 1000 nM. An equal volume (<5 μl per 10 ml medium) of ethanol was added to control cultures. Concentrations of retinoic acid stock solutions were estimated using an extinction coefficient of 36,500 at 343 nm. The adenosine 3′:5′-cyclic monophosphate (cAMP) analogues, N⁶,2′-O-dibutyryladenosine 3′:5′-cyclic monophosphate (dibutyryl-cAMP) and 8-bromoadenosine 3′:5′-cyclic monophosphate (8-bromo-cAMP) (Sigma), were dissolved in culture medium and added to the cell cultures to give a final concentration of 1 mM. The phosphodiesterase inhibitor, 3-isobutyl-l-methylxanthine (IBMX) (Aldrich), was dissolved in 1 M NaOH and used at a final concentration of 1 mM and the adenylate cyclase activator, forskolin (Sigma), was dissolved in ethanol and used at a final concentration of 10⁻⁵ M. The appropriate vehicle was added to control cultures for each experiment.

The effects of dibutyryl-cAMP on fibroblast proliferation were studied by seeding 50,000 cells into 35 mm diameter culture dishes; the cells were allowed to attach for two hours and dibutyryl-cAMP was added to a final concentration of 1 mM. Cells were refed with fresh medium containing 1 mM dibutyryl-cAMP each day, and after three days the cells were detached and counted using a haemocytometer. Control dishes set up in parallel were treated in the same way except that dibutyryl-cAMP was omitted from the medium.

Fibroblasts, seeded into 25 cm² tissue-culture flasks (0.5 × 10⁶ cells/flask) and used the following day, were treated for six hours with retinoic acid, forskolin or IB MX (as above). The cell monolayers were then washed twice with phosphate-buffered saline (Flow Laboratories, Dulbecco’s formula, without calcium and magnesium), extracted with 1 ml of 70% (v/v) ethanol in water and finally with 1 ml of 65% (v/v) ethanol. Ethanolic extracts were combined and a sample was lyophilized for assay with a cAMP radioimmunoassay kit (Amersham International, Amersham, UK).

Total cellular RNA was prepared by the guanidinium isothiocyanate/caesium chloride method (Chirgwin et al. 1979). RNA samples (15 μg per track for 0.5 cm wide slots and containing ethidium bromide as a visual check on RNA loading) were size-fractionated on 1.2% agarose/formalde-hyde gels and transferred by vacuum blotting with 1.8 M NaCl, 0.01 M EDTA, 0.1 M sodium phosphate, pH 7.4 (10×SSPE), to nylon membranes (Amersham). Membranes were hybridized at 42°C with ³²P-labelled probe using 50% formamide, 6xSSPE, 0.2% (w/v) Ficoll 400, 0.2% (w/v) polyvinylpyrrolidone, 0.2% bovine serum albumin (fraction V), 0.5% SDS, 5% dextran sulphate, 200 μg ml⁻¹ tRNA and 100 μg ml⁻¹ single-stranded carrier DNA as the prehybridization and hybridization buffer. After hybridization, membranes were washed 3–4 times in 0.1×SSPE, 0.1% SDS for >15 minutes each at 68°C and exposed to X-ray film with intensifying screens at — 70°C. For quantitative autoradiography, X-ray film was preflashed and the autoradiographs scanned using an LKB laser scanning densitometer or an image analysis system.

The RAR-α probe was a KpnI/SacI fragment (503 bp) from the 5′ end of the human RAR-αl cDNA (Petkovich et al. 1987). The human RAR-β probe consisted of the complete RAR-/S2 cDNA insert (1400 bp) of the plasmid pCOD20 (Brand et al. 1988). The human RAR-γ probe was the fulllength, 1500 bp γl cDNA insert (Krust et al. 1989). As a further check on RNA loading, the membranes for some experiments were reprobed with either a mouse 18 S ribosomal RNA cDNA probe (Edwards et al. 1985) or a rat β-actin probe consisting of a 1200 bp BglI fragment of the pRpA-1 cDNA clone isolated by P. Gunning. A mouse α- actin probe consisting of a 1150 bp PstI fragment of the cDNA clone pAM (Minty et al. 1981) was used as a test of fibroblast phenotype. Probes were labelled with [³²P]dCTP (Amersham International, 3000 Ci mmol⁻¹) to a specific activity of approximately 10⁹ disints min⁻¹μg⁻¹ (Feinberg and Vogelstein, 1983).

With the high-stringency post-hybridization washing conditions used for these experiments there is no detectable cross-hybridization of the RAR probes. The human RAR-α and RAR-γcDNA probes were provided by Martin Petkovich and Pierre Chambon, Strasbourg, France, and the RAR-β probe by Anne Dejean, Paris, France.

Keratinocytes proliferate to form a monolayer of cells when cultured in media with a low (0.07 mM) Ca²⁺ concentration. At higher Ca²⁺ levels (>0.1 mM), the cells stratify and grow in tight colonies (Hennings et al. 1980; Boyce and Ham, 1983). To ask whether or not RAR expression varies in relation to these Ca²⁺-induced changes in vitro, RNA was extracted from keratinocytes cultured continuously in low-calcium medium, and from keratinocytes cultured for four days in high-calcium medium. Both RAR-α and RAR-γ were expressed in keratinocytes and there was no consistent difference in RAR expression between cells grown in high-calcium or low-calcium medium (Fig. 1), and no differences between keratinocytes obtained from different sites or ages of donor. The RAR-αprobe detected two transcripts, approximately 3.6 and 2.8 kb, as has been described in other tissues (Rees et al. 1989), and the shorter of the two transcripts was more abundant. RAR-γ transcripts were detectable as a band at approximately 3.2–3.3 kb. RAR-β mRNA was undetectable in all keratinocyte RNA samples analysed (Fig-1).

Since RAR-α and RAR-β are inducible in response to retinoic acid in some cell types (de Thé et al. 1989; Redfern et al. 1990; Clifford et al. 1990; Leroy et al. 1991; Kastner et al. 1990), the effect of retinoic acid on RAR expression in keratinocytes was investigated. However, for cells grown in either low-calcium or high-calcium medium, there were no significant changes in RAR expression in response to treatment with 10⁻⁷ M retinoic acid for up to 24 hours (Fig. 1).

We have reported previously that of two human dermal fibroblast lines isolated, a cell line (f10.11) isolated from the chest skin of a 60-year-old male expressed RAR-β at a high level whereas a breast-skin fibroblast line (line SC) did not (Rees and Redfern, 1989). In view of this potential heterogeneity of dermal fibroblasts, we studied RAR expression in fibroblast cultures established from biopsies taken from different body sites and from individuals ranging in age from 24 weeks gestation to 89 years (Table 1). RAR-α was expressed in all fibroblast cultures tested and at a 1.5- to 2-fold higher level than keratinocytes, relative to RAR-γ (Figs 2 and 3). RAR-γ was also expressed in all samples, and at a level comparable to keratinocytes (Figs 2 and 3). Although there was some variation in RAR-γ signal intensity between the cell lines, this correlated with slight variations in RAR-α signal intensity and RNA loading. RNA was extracted from only one sample of the f10.11 cell fine but for other fibroblast cell lines the RARexpression patterns were consistent between different batches of cells.

With the exception of the dermal fibroblast cell line f10.il, RAR-β transcripts were either undetectable or present at a low level in cultured fibroblasts. The f10.ll fibroblast cell line was unique with respect to the very high level of expression of RAR-β. These cells are no longer in existence and to investigate the possibility that the individual from which the f10.ll cells were derived had consistently high levels of RAR-β expression in his dermal fibroblasts, new dermal fibroblast cultures were established from the same body site of the original donor. However, in these subsequent cultures the expression of RAR-β was undetectable (Fig. 4).

All fibroblast cell lines used for these experiments were obtained by explant outgrowth and no attempt was made to derive clonal cell populations; the variable level of RAR-β expression between different cell lines may reflect heterogeneity of dermal fibroblasts in vivo. One of the biopsies used to establish a fibroblast culture (bx4) was obtained from skin adjacent to a basal cell carcinoma. Myofibroblasts, as defined by their expression of α-actin (Oda et al. 1988), can appear in skin as part of a stromal reaction to tumour, and to assess the contribution of myofibroblasts to culture heterogeneity, Northern blots were reprobed with an a-actin probe under conditions of low stringency, allowing the detection of both α- and β-actin. The fibroblast line bx4 expressed a-actin at the highest level (Fig. 2); RNA from this cell line also gave a relatively strong signal for RAR-β. Expression of α-actin was also detectable in RNA from bx7 cells, derived from skin adjacent to a mole, but the signal intensity was low and this sample gave no detectable signal for the RAR-β probe at the exposures used. If it is true that myofibroblasts express RAR-β at a higher level than fibroblasts, the proportion of myofibroblasts present in the bx7 sample may have been too low to allow detection of RAR-β. The f10.11 fibroblasts expressed a-actin only at a low level by comparison with the bx4 and bx7 cells (Fig. 2).

To investigate the effects of retinoic acid on RAR expression in dermal fibroblasts, the breast-skin cell line (sc) was treated with 10⁻⁷ M retinoic acid for 24 hours and this resulted in a marked induction of RAR-β mRNA, relative to ethanol-treated control cells (Figs 2 and 3). This response was studied in more detail using the fetal fibroblast cells and the foreskin-derived fibroblasts fs5. In both fibroblast lines, the retinoic acid-mediated induction of RAR-β was rapid, detectable within two hours of adding retinoic acid to the cells and gave a 6- to 30-fold induction of RAR-β within 16 hours (Fig. 5). Induction of RAR-β was detectable with a retinoic acid concentration of 10⁻¹⁰ M and a response 50% of maximal was produced with retinoic acid concentrations in the range 10⁻¹⁰ to 10⁻⁹ M (Fig. 6). There was a slight reduction in response at a high retinoic acid concentration of 10⁻⁷ M. No changes in the abundance of RAR-α or RAR-γ mRNA were observed in response to retinoic acid (Figs 2 and 3).

It has been reported for embryonal carcinoma cells that the induction of RAR-β in response to retinoic acid is inhibited or attenuated by cAMP analogues (Hu and Gudas, 1990; Martin et al. 1990). cAMP is an important element of intracellular signal transduction pathways, controlling, for example, the activities of particular transcription factors by regulating protein kinases (Karin, 1989), and therefore could be important in the regulation of RAR-β expression in fibroblasts. Dermal fibroblasts were exposed to retinoic acid in the presence or absence of 1 mM dibutyryl-cAMP but this did not inhibit the retinoic acid-mediated induction of RAR-β (Fig. 7). The fact that others have obtained biological effects in response to the treatment of fibroblasts with dibutyryl-cAMP (Lin et al. 1988; Yoneda et al. 1988) suggests that penetration of the analogue into the cells should not be a problem. To verify that dibutyryl-cAMP had some biological effects on human dermal fibroblasts, the cells were grown for three days in the presence or absence of 1 mM dibutyryl-cAMP. Compared to the control, untreated cells, dibutyryl-cAMP significantly inhibited proliferation: mean cell numbers were 186,000 (n=18) and 101,000 (n=12) per 35 mm diameter dish for the control and dibutyryl-cAMP-treated cells, respectively (Student’s <28=8.765, P<0.001).

Intracellular cAMP levels may be increased using phosphodiesterase inhibitors or activators of adenylate cyclase. Both IB MX, a phosphodiesterase inhibitor, and forskolin, an activator of adenylate cyclase, attenuated the retinoic acid-mediated induction of

RAR-β relative to untreated control cells (Fig. 8). However, the basal level of RAR-β expression in unstimulated cells was reduced in response to IBMX and forskolin (Fig. 8). Relative to this reduced level of RAR-β expression, RAR-β mRNA levels were increased by adding retinoic acid to IBMX- or forskolin-treated cells (Fig. 8). To see if the effects of IBMX and forskolin on RAR-β expression were specific to RAR-β, blots were reprobed for RAR-α. The expression of a gene unrelated to RARs was investigated by also probing the RNA samples from IBMX-treated cells with an interleukin-1β (IL1-β) probe. RAR-α transcript abundance increased approximately 2-fold in response to IBMX alone whereas IL1-β expression decreased (Fig. 8). Forskolin had no detectable effect on RAR-α expression. These results show that the effects of IBMX in reducing RAR-β expression were not specific to RAR-β and, since RAR-α mRNA abundance was increased in response to IBMX, were not due to a general reduction in cellular mRNA levels. In addition to the difference between IBMX and forskolin in their effects on RAR-α expression, there was a marked difference in their effects on intracellular cAMP concentrations: forskolin increased intracellular cAMP levels 7-fold after six hours incubation (Fig. 9) whereas there was no detectable change in intracellular cAMP in response to IBMX. While we cannot rule out the possibility that cAMP does affect RAR-β expression in forskolin-treated fibroblasts, the lack of effect of dibutyryl-cAMP and the effects of IBMX in the absence of measurable increases in intracellular cAMP suggest that other pathways mediate the changes in RAR-β and RAR-α expression in response to IBMX and forskolin.

It has been suggested that RAR-γ is the predominant RAR form expressed in skin (Zelent et al. 1989). However, the present results clearly show that RAR-α is also expressed in cultured keratinocytes and dermal fibroblasts, in agreement with previous studies on whole rodent and human skin (Rees and Redfern, 1989; Leroy et al. 1991). Overall, keratinocytes and dermal fibroblasts had similar patterns of RAR expression in vitro, expressing both RAR-α and RAR-γ but differing in the relative abundance of RAR-α mRNA. A recent report in which RAR-α was described as undetectable in human skin and cultured keratinocytes and fibroblasts (Elder et al. 1991) is likely to be an artefact due to the use of hybridization probes too short for the washing stringency employed.

Although one dermal fibroblast cell fine, f10.11, had a high level of expression of RAR-β, this appears to be an anomaly and was not a feature of other dermal fibroblasts derived from the same body site of the original donor, or of other dermal fibroblast lines analysed. Contamination of the original f10.11 primary culture with an established cell fine is unlikely as the pattern of RAR expression in these cells was different from that in other cell lines cultured in the same laboratory. Thus, the high RAR-β expression in f10.11 cells either represents an altered phenotype that has arisen during culture or other mesenchyme-derived cells within the dermis may have distinctive patterns of RAR expression. This raises the important question of whether RAR expression patterns of the remaining cell lines accurately reflect those of keratinocytes and fibroblasts in vivo. In whole human skin, expression of both RAR-αand RAR-γis detectable, whereas RAR-β is not, or only at a low level (Rees and Redfern, 1989; Leroy et al. 1991). Furthermore, recent in situ hybridization studies (Viallet et al. 1991) show that the expression of RAR-α and RAR-γ, but not RAR-β, is detectable in the dermis and epidermis of mouse skin. We would therefore argue that RAR-expression patterns in human dermal fibroblasts and keratinocytes remain at least qualitatively stable in vitro.

A major difference between keratinocytes and fibroblasts is their differential responsiveness to retinoic acid. The mechanism of RAR-β induction in fibroblasts has not been fully characterized but is likely to result from an RAR-mediated increase in the rate of RAR-β transcription, as has been shown for hepatocellular carcinoma cells (de Thé et al. 1989) and embryonal carcinoma cells (Martin et al. 1990). A retinoic acid response element (RARE) has been defined within the RAR-β promoter (de Thé et al. 1990; Sucov et al. 1990). Similar RAREs are present within the promoters of other retinoic acid-responsive genes (review: Redfern, 1992). The finding that RAR-β induction in fibroblasts occurred at nanomolar retinoic acid concentrations suggests that the response was indeed receptor-mediated: a response 50% of the maximum within a 0.1 to 1 nM range is indicative of a V_D comparable to values of 0.2–0.4 nM reported for RAR-αby Nervi et al. (1989) and Yang et al. (1991).

The effects of cAMP and cAMP-elevating drugs on RAR-β expression are difficult to interpret. Dibutyryl-cAMP itself had no effect, either on the basal level of RAR-β expression or on RAR-β induction by retinoic acid. This implies that the effects of IBMX and forskolin in reducing the constitutive level of RAR-β expression may be mediated by a mechanism other than directly via intracellular cAMP. This conclusion is supported by the finding that IBMX, unlike forskolin, did not increase cAMP levels detectably after a six-hour exposure, yet both IBMX and forskolin decreased constitutive expression of RAR-β but did not inhibit RAR-β induction by retinoic acid. Furthermore, IBMX alone increased the abundance of RAR-α, an effect not observed with forskolin. Although a putative cAMP response element (CRE) has been described within the RAR-β promoter (Zelent et al. 1991), there is no evidence that this CRE is functional; in embryonal carcinoma cells, Martin et al. (1990) have shown that the attenuation of RAR-β expression in response to dibutyryl-cAMP is mediated at a post-transcriptional level.

The finding that fibroblasts and keratinocytes differ with respect to changes in RAR-gene expression in response to retinoic acid raises two important questions: (1) does RAR-β expression vary in fibroblasts in vivo during normal development as a result of regulation of intracellular retinoic acid concentration; (2) what are the developmental and phenotypic consequences of RAR-β expression in fibroblasts in vivo and in vitro?

Detailed in situ hybridization studies of RAR expression during mouse embryogenesis have shown that the expression of all three receptors is spatiotemporally regulated (Dollé et al. 1990; Ruberte et al. 1991). For example, RAR-β shows spatially restricted patterns of expression in the mesenchyme of limb buds and facial structures (Dollé et al. 1990). However, whether or not these spatiotemporal changes in RAR-β expression are brought about through differential regulation of intracellular retinoic acid concentrations is unknown.

Studies on F9 teratocarcinoma cells suggest that RAR-β could determine whether the cells differentiate into parietal or visceral endoderm in response to retinoic acid. In the presence of retinoic acid alone, aggregated F9 cells differentiate into visceral endoderm, whereas in the presence of retinoic acid and cAMP, conditions in which RAR-β expression is attenuated (Martin et al. 1990), F9 cells differentiate into parietal endoderm (Strickland et al. 1980; Darrow et al. 1990). Unlike F9 cells, fibroblasts do not show a clear differentiation response to retinoic acid. Retinoic acid stimulates fibroblast proliferation and the increased expression of extracellular matrix proteins, but this may vary with growth conditions in vitro (Varani et al. 1990). Otherwise, the longer-term phenotypic consequences of retinoic acid treatment of fibroblasts are unknown.

In vivo studies have recently shown that RAR-β expression is increased in the limb-bud mesenchyme and embryonic mouse-lip dermis in response to retinoic acid (Tickle et al. 1989; Viallet et al. 1991; Noji et al. 1991). In the embryonic mouse lip, retinoid treatment results in the development of vibrissae follicle buds into exocrine glands (Hardy, 1968). Such morphological changes appear to be mediated by the dermis (Hardy et al. 1990) and are associated with increased dermal RAR-β expression (Viallet et al. 1991). Since the induction of RAR-β by retinoic acid in dermal fibroblasts is a rapid event, RAR-β is likely to be involved in the transcriptional regulation of genes specifying the phenotype of associated epithelial cells. The characterization of long-term phenotypic changes in fibroblasts in response to retinoic acid is of fundamental importance for understanding the role of retinoic acid in dermal-epidermal interactions.

We thank Pierre Chambon, Martin Petkovich and Anne Dejean for gifts of RAR probes, J. L. Rees and colleagues for the skin biopsies, J. L. Rees for preparing RNA from the original fibroblast line, Vai Randall for her valuable comments on the manuscript and Jane Taylor for the cAMP measurements and help with the fibroblast growth experiments. This research was supported in part by grants to C.P.F.R. from the Wellcome Trust and the North of England Cancer Research Campaign.