Retinoic acid mediates post-transcriptional regulation of keratin 19 mrna levels

Stratified squamous epithelia exhibit variable, site-specific patterns of keratin expression (Doran et al., 1980; Moll et al., 1982). All stratified squamous epithelia express keratins K5 and K14 in the basal layer (Nelson and Sun, 1983). Nonkeratinizing stratified epithelia, which lack a stratum corneum, also express K19 in the basal layer (Ouhayoun et al., 1985; Ermich et al., 1988; Lindberg and Rheinwald, 1989). Retinoic acid (RA), a vitamin A metabolite, exerts profound effects on the level of expression of these keratins. High concentrations of RA suppress synthesis of K5 and K14 while enhancing K19 expression in cultured keratinocytes. These changes are associated with alterations in both mRNA and protein levels and rates of gene transcription (Doran et al., 1980; Fuchs and Green, 1981; Eckert and Green, 1984; Kim et al., 1984; Gilfix and Eckert, 1985; Stellmach and Fuchs, 1989; Tomic et al., 1990; Stellmach et al., 1991).

The effects of RA are believed to be mediated in the cytoplasm by the cellular retinoic acid binding proteins (CRABP) I and II (Sani and Hill, 1974; Bailey and Siu, 1987). Although the function of the CRABPs is uncertain, they may be involved in intracellular transport of RA from the cytoplasm to the nucleus (Takase et al., 1986; Barkai and Sherman, 1987), regulation of intracellular RA concentration (Maden et al., 1988), and RA metabolism (Boylan and Gudas, 1992; Posch et al., 1992). RA also exerts its effects through three structurally related ligand-dependent nuclear transcription factors, the retinoic acid receptors RARα, -β and -γ (Giguere et al., 1987; Petkovich et al., 1987; Brand et al., 1988; Krust et al., 1989; Zelent et al., 1989). The RARs are encoded by distinct genes belong to a larger family of regulatory proteins that includes the steroid and thyroid hormone receptors (Evans, 1988; Umesono et al., 1988; Graupner et al., 1989). Three other related proteins, RXRα, - β and - γ, are believed to coregulate target gene expression by dimerization with the RA, thyroid hormone and vitamin D receptors, thereby enhancing receptor binding to DNA response elements (Hamada et al., 1989; Mangelsdorf et al., 1990; Mangelsdorf et al., 1991; Yu et al., 1991; Kliewer et al., 1992; Mangelsdorf et al., 1992; Zhang et al., 1992). RA response elements have been identified in the promoter regions of several genes (Vasios et al., 1989; Schule et al., 1990; Sucov et al., 1990).

Previously, we reported that RARβ mRNA expression correlated with K19 message levels in normal cultured keratinocytes (Crowe et al., 1991). Further, the levels of these messages were directly dependent on extracellular RA concentration. However, this correlation was observed in only one of nine transformed keratinocyte lines examined by northern blot (Hu et al., 1991). We wished to determine the mechanism by which these changes in gene expression developed. We find that K19 and RARβ mRNA levels increase with time after RA exposure. Conversely, K5 mRNA expression decreases upon RA addition in a similar fashion. The observed changes in K5 and RARβ messages are primarily the result of RA-induced alterations in the basal transcription levels of these genes. However, the RA-mediated induction of K19 message is not the result of increased transcription but is in large part due to enhanced mRNA stability. We provide the initial evidence of an RA-dependent, post-transcriptional, cytoplasmic mechanism by which K19 mRNA expression is regulated in stratified squamous epithelia.

The derivation and properties of the cell lines used in this study have been described previously (Crowe et al., 1991; Hu et al., 1991). Cells were cultured in Dulbecco’s modified Eagle’s (DME)/F12 medium (3:1 v/v) plus 5% fetal bovine serum, 0.4 μg/ml hydrocortisone, 5 μg/ml insulin, 24 μg/ml adenine, 10⁻¹⁰ M cholera toxin, and 10 ng/ml epidermal growth factor. RA concentration of this media was 10⁻⁹ M (DeLeenheer et al., 1982). Cells were cocultivated with lethally irradiated 3T3 fibroblast feeder cells (Rheinwald and Green, 1975). Some cultures received all trans-RA (Sigma, St. Louis, MO) at a final concentration of 10⁻⁶ M for up to 48 hours prior to nucleus and RNA isolation. In the message stability experiments, cultures were treated with 10⁻⁶ M RA for 12 hours followed by actinomycin D (2 μg/ml) or with actinomycin D alone for 2 to 6 hours prior to RNA extraction. Control cultures received vehicle only.

Intact nuclei and cytoplasmic RNA were isolated as previously described (Groudine et al., 1981; Greenberg and Ziff, 1984). Cultured keratinocytes were collected by trypsinization and centrifugation at 400 g for 5 minutes. The cell pellet was washed twice in 5 ml ice cold phosphate buffered saline. Nuclei were isolated from 2×10⁷ cells by lysis in 0.01 M Tris-HCl (pH 7.4), 0.01 M NaCl, 3 mM MgCl₂, 0.5% NP-40. The nuclei were washed in this buffer and suspended in 0.1 ml of 40% glycerol, 50 mM Tris-HCl (pH 8.3), 5 mM MgCl₂, 0.1 mM EDTA. The nuclei were stored at −70°C prior to use.

Vanadyl ribonucleoside complexes (Gibco BRL, Gaithersburg, MD) were added to the cell lysis supernatant at a final concentration of 10 mM followed by centrifugation at 10,000 g for 10 minutes at 4°C. The supernatant was collected and digested with proteinase K (200 μg/ml) in 0.1 M Tris-HCl (pH 7.5), 0.22 M NaCl, 12.5 mM EDTA, 1% SDS for 30 minutes at 37°C. The solution was extracted with an equal volume of phenol/chloroform/isoamyl alcohol (25:24:1 by vol.) and precipitated with 2.5 volumes of absolute ethanol overnight at −20°C. The cytoplasmic RNA was washed in 70% ethanol and dissolved in 10 mM Tris-HCl (pH 7.4), 1 mM EDTA.

For northern blot analysis, RNA samples were heated to 65°C for 5 min in buffer containing a 7:2:1 (by vol.) mixture of formamide, 10× MOPS (0.2 M morpholinopropanesulfonic acid, 50 mM sodium acetate, 10 mM EDTA) and 37% formaldehyde. Thirty μg total RNA was loaded per lane onto denaturing 1% agarose gels containing 2.2 M formaldehyde. The samples were electrophoretically separated at 120 V using 1× MOPS running buffer and the RNA was capillary transferred to nitrocellulose membranes. Blots were UV-crosslinked using Stratalinker Model 1800 (Stratagene, La Jolla, CA), wetted briefly in 4× SSC (1× = 0.15 M NaCl, 0.015 M sodium citrate), and prehybridized in 50% formamide, 5× SSC, 50 mM NaH₂PO₄/Na₂HPO₄, 5 mM EDTA, 0.02% Ficoll, 0.02% polyvinylpyrrolidone, 100 μg/ml bovine serum albumin, and 100 μg/ml sheared calf thymus DNA at 42°C for 18 hours. For dot blot analysis, 20 μg of total RNA was incubated at 65°C for 15 minutes in 50% formamide, 7% formaldehyde, and 1 × SSC. Two volumes of 20× SSC were added to each sample, which was vacuum transferred to nitrocellulose membranes using a dot blot apparatus (Schleicher and Schuell, Keene, NH). Blots were UV-crosslinked and prehybridized as above. cDNA probes were radiolabeled with [α-³²P]dCTP (3000 Ci/mmol, DuPont, Boston, MA) using the random primed method (Feinberg and Vogelstein, 1984). Blots were hybridized with solutions containing 4× 10⁶ cpm/ml of radioactive probe for 18 hours at 42°C followed by three washes in 2× SSC/0.1% SDS for 45 minutes at 22°C and one wash with 0.2× SSC/0.1% SDS at 60°C for 30 minutes. Dot blots were quantified using a liquid scintillation counter.

The nuclear runoff transcriptional assay has been described previously (Groudine et al., 1981; Greenberg and Ziff, 1984). Reactions contained nuclei in 20% glycerol, 2.5 mM DTT, 1 Mm MgCl₂, 70 mM KCl, 0.25 mM GTP and CTP, 0.5 mM ATP, and 100 μCi [α-³²P]UTP (3000 Ci/mmol, DuPont, Boston, MA). Nuclei were incubated for 30 minutes at 30°C. The reactions were terminated by addition of 40 μl of iodoacetate-treated DNase I (1 mg/ml) and incubated at 30°C for 5 minutes. Protein was digested by incubation with proteinase K (100 μg/ml) in 10 mM Tris-HCl (pH 7.4), 5 mM EDTA, 1% SDS at 42°C for 30 minutes, followed by extraction with phenol/chloroform/isoamyl alcohol. Yeast tRNA (100 μg/ml) was added to the aqueous phase, which was then precipitated at 0°C with an equal volume of 10% trichloroacetic acid (TCA), 60 mM Na₄P₂O₇. After 30 minutes, the precipitate was vacuum collected onto 25 mm glass microfiber filters (0.45 μm pore size) and washed three times with 10 ml 5% TCA, 30 mM Na₄P₂O₇. The filters were transferred to scintillation vials and incubated in 1.5 ml of buffer containing 20 mM HEPES (pH 7.5), 5 mM MgCl₂, 1 mM CaCl₂, 15 mM EDTA, 1% SDS at 65°C for 10 minutes. The filters were then incubated in 1.5 ml of 10 mM Tris-HCl (pH 7.5), 5 mM EDTA, 1% SDS for 10 minutes at 65°C. The two solutions were combined and extracted once with an equal volume of phenol/chloroform/isoamyl alcohol. The aqueous phase was precipitated overnight at −20°C with 2.5 volumes of absolute ethanol. The precipitate was recovered by centrifugation at 10,000 g for 30 minutes at 4°C and dissolved in 2 ml 10 mM N-tris(hydroxymethyl)methyl-2-aminoethanesulfonic acid (TES, pH 7.4), 10 mM EDTA, 0.2% SDS, 0.3 M NaCl at 6×10⁶ cpm/sample. Denatured K19, K5, RARβ, and β-actin cDNA probes crosslinked to nitrocellulose membranes were hybridized to the RNA solutions at 65°C for 36 hours. The filters were washed twice in 2× SSC at 65°C for 1 hour and quantified using a liquid scintillation counter. Values obtained were normalized to the βactin level for each sample following subtraction of average background counts.

The sequences of the cDNA probes used in this study have been reported previously (Eckert and Green, 1984; Hu and Gudas, 1990). The K19 probe was a 270 bp PstI fragment from the 3′ untranslated region of the K19 cDNA. The K5 probe contained 600 bp also from the 3′ untranslated region of the K5 cDNA. The RARβ probe was a 600 bp EcoRI fragment containing 3′ untranslated sequences and a partial coding sequence. The β-actin clone, obtained from Dr Steve Farmer, Boston University, was a 1500 bp PstI fragment consisting of 3′ untranslated and coding sequences (Bond and Farmer, 1983).

We wished to determine the magnitude of RA-induced changes in K19, K5 and RARβ mRNA expression (Fig. 1A). These genes have been reported to be RA-responsive (Fuchs and Green, 1981; Eckert and Green, 1984; Gilfix and Eckert, 1985; Brand et al., 1988). OKP7 and SCC25 cells were exposed to 10⁻⁶ M RA for 48 hours. K19 transcripts were easily detected and induced approximately 1-fold in strain OKP7 and 0.6-fold in SCC25 (Fig. 1B). Conversely, K5 message was significantly reduced by RA exposure (Fig. 1C). Levels of K5 mRNA decreased approximately 60% in both OKP7 and SCC25 cells after 48 hours of RA exposure. RARβ mRNA was also detected in these cells and was induced by RA. RARβ message levels were induced 3-fold in strain OKP7 and 4-fold in SCC25 cells (Fig. 1D).

To assess the effect of RA on gene transcription, labeled nuclear RNA from OKP7 cells was isolated and hybridized to K19, K5, and RARβ cDNA probes. Levels of K19 nuclear transcripts decreased following RA exposure. This decrease was large and rapid (within 3 hours of RA addition) and remained relatively constant at one-half the level of the untreated control (Fig. 2A). This result was surprising since 10⁻⁶ M RA produces an increase in K19 mRNA levels in these cells (Fig. 1A). This would indicate that the observed RA-dependent increase in K19 message is not the result of increased K19 gene transcription. Similar results were obtained with SCC25 cells (data not shown).

K5 transcription also decreased significantly by 3 hours after RA addition (Fig. 2B). The level continued to decrease until 12 hours after RA addition and then remained relatively constant through 48 hours at one-half the level of the unexposed control group. This is consistent with the observed RA-dependent decrease in K5 message levels (Fig. 1B) and provides additional evidence for transcriptional downregulation of the K5 gene by RA.

Previous studies revealed that RARβ message levels increased following RA exposure in many cell types (Brand et al., 1988; Hu and Gudas, 1990; Crowe et al., 1991). To determine the mechanism by which this change takes place in cultured keratinocytes, nuclear runoff assays were performed. As shown in Fig. 2C, levels of RARβ nuclear transcripts increased by 50% within three hours of RA addition indicating that RA influences RARβ gene expression by transcriptional upregulation. This mechanism was consistent with previous studies in F9 cells (Martin et al., 1990).

Paradoxically, RA downregulated K19 gene transcription in OKP7 and SCC25 cells while cytoplasmic K19 mRNA levels increased. We therefore wished to determine if RA could enhance K19 message stability in cultured keratinocytes. Cells were exposed to 10⁻⁶ M RA for 0 or 12 hours followed by actinomycin D (2 μg/ml) for 0 to 6 hours. As shown in Fig. 3, K19 message decayed more rapidly in cells that were not exposed to RA. The calculated half-life of K19 mRNA in OKP7 cells not exposed to 10⁻⁶ M RA was 2 hours 48 minutes ± 17 minutes (± s.d.). In cells exposed to RA for 12 hours, K19 mRNA half life was 8 hours 11 minutes ± 32 minutes, reflecting a ∼2-fold increase in message stability. A similar decay pattern was observed in SCC25 cells (data not shown). These results indicate that the observed increase in cytoplasmic K19 message is in large part due to an RA-dependent mechanism that stabilizes K19 transcripts.

The most intriguing finding of this study is the apparent downregulation of K19 gene transcription and stabilization of cytoplasmic K19 transcripts by RA. Previous work has not directly investigated the effect of RA on K19 transcription (Stellmach and Fuchs, 1989). Since K19 mRNA levels are upregulated by RA, it would seem logical to assume the existence of an RA response element (RARE) in the K19 promoter region similar to that found in other genes (Vasios et al., 1989; de The et al., 1990). However, no consensus RARE has been reported in the K19 gene to date. These experiments show that induction of K19 message by RA is not primarily due to transcriptional increases but is in large part the result of an RA-mediated mechanism that stabilizes K19 transcripts. This result was also observed in the transformed cell line SCC25, which exhibits RA responsive K19 expression (data not shown), suggesting a similar mechanism. These findings suggest that RAdependent K19 gene regulation is more complex than previously believed.

Recently this same mechanism was proposed to explain RA-mediated induction of human prolactin gene expression (Gellersen et al., 1992). RA exposure produced a 2-fold increase in prolactin secretion and prolactin mRNA levels in the B lymphoblastoid line IM-9-P3. This increase in the steady-state levels of prolactin message was not due to enhanced transcription as determined by nuclear runoff experiments, but due to message stabilization. In RA treated cells, the half-life of prolactin mRNA increased from 9 hours to 22 hours. These studies point to a previously undefined role for RA in the regulation of gene expression by enhanced message stabilization.

The steady-state levels of functional mRNAs are determined in part by their cytoplasmic decay rates (for review see Brawerman, 1987). This process plays an important role in controlling gene expression. The metabolic stability of individual mRNAs differs widely. Recent studies have provided some insight into the nature of mRNA decay and have identified structural features that determine its susceptibility to decay. The selectivity of mRNA decay can best be explained by the action of specific factors that recognize unique sites in mRNA. Endonucleolytic cuts triggered by these interactions result in rapid mRNA destruction. A specific sequence promoting mRNA decay (5′-AUUUA-3′) has been identified in the 3′ noncoding regions of a variety of mRNAs (Shaw and Kamen, 1986).

The manner in which this sequence promotes mRNA decay is unknown, although it may be recognized and cleaved by a specific endonuclease. Interestingly, this element is also present in the 3′ noncoding sequence of both the human K19 and prolactin mRNAs (Truong et al., 1984; Stasiak and Lane, 1987; Eckert, 1988). The role of this sequence in regulating K19 mRNA stability remains to be determined. The indication that RA-dependent K19 message stabilization takes place in the cytoplasm provokes speculation about the existence of RA-dependent factors that stabilize K19 mRNA. Inducible cytoplasmic factors that bind to the 3′ AUUUA motif of several genes have been isolated (Bohjanen et al., 1991; Gillis and Malter, 1991; Vakalopoulou et al., 1991). CRABPs are the predominant RA binding proteins in the cytoplasm, are induced by RA (Astrom et al., 1992), and their expression has been shown to affect that of other RA responsive genes (Boylan and Gudas, 1991). It is interesting to suggest a possible new role for CRABPs as modulators of K19 mRNA expression. CRABP, perhaps in conjunction with as yet uncharacterized factors, may interact with specific mRNA sequences or secondary structure, perhaps in the 3′ noncoding region, making the message less vulnerable to endonucleolytic attack. Studies focusing on the effect of altered CRABP levels on message stability and an examination of specific protein-RNA interactions would be necessary to prove this hypothesis. This mechanism may help explain our previous observation of RA unresponsive K19 expression in some squamous cell carcinoma lines (Hu et al., 1991). Experiments aimed at understanding K19 regulation in these cell lines may indicate whether a nuclear or cytoplasmic process accounts for the unresponsiveness.

These experiments also provide additional evidence for transcriptional downregulation of the K5 gene in cultured keratinocytes. Previous studies (Stellmach and Fuchs, 1989; Stellmach et al., 1991) using normal and transformed epidermal cells have found that decreased transcription of the K5 gene correlates with reduced K5 message levels by 6 hours following RA addition. This study found that the decrease in K5 nuclear transcripts begins even earlier, by 3 hours after RA addition. These findings imply the existence of a negative RA response element in the K5 promoter as other studies have suggested (Tomic et al., 1990). Negative RA response elements have been found in the flanking sequences of other genes (Lipkin et al., 1992). Whether these transcriptional effects are the result of direct receptor-mediated action or involve additional factors has not been determined.

Induction of K19 mRNA expression by RA is primarily the result of increased message stability rather than transcriptional upregulation. These results point to an RA-dependent post-transcriptional mechanism of K19 regulation located in the cytoplasm. This discovery suggests promising future experiments aimed at understanding this mechanism. Future work will examine K19 expression not only in cells where the gene is RA responsive but also in cells where it is not. Detailed studies examining the possible interaction of cytoplasmic factors and K19 mRNA may elucidate a novel mechanism for regulating expression of differentiation-specific keratin intermediate filaments in stratified squamous epithelia.

Thanks are due to Dr James G. Rheinwald for his support and comments on the manuscript, Dr Steve Farmer for providing the β-actin clone, and Ms Liesbeth Brown for expert cell culture assistance.