Glucocorticoid Induction of Epithelial Sodium Channel Expression in Lung and Renal Epithelia Occurs via trans-Activation of a Hormone Response Element in the 5′-Flanking Region of the Human Epithelial Sodium Channel α Subunit Gene*

In airway and renal epithelia, the glucocorticoid-mediated stimulation of amiloride-sensitive Na+ transport is associated with increased expression of the epithelial Na+ channel α subunit (αENaC). In H441 lung cells, 100 nm dexamethasone increases amiloride-sensitive short-circuit current (3.3 μA/cm2 to 7.5 μA/cm2), correlating with a 5-fold increase in αENaC mRNA expression that could be blocked by actinomycin D. To explore transcriptional regulation of αENaC, the human αENaC 5′-flanking region was cloned and tested in H441 cells. By deletion analysis, a ∼150-base pair region 5′ to the upstream promoter was identified that, when stimulated with 100 nm dexamethasone, increased luciferase expression 15-fold. This region, which contains two imperfect GREs, also functioned when coupled to a heterologous promoter. When individually tested, only the downstream GRE functioned in cis and bound GR in a gel mobility shift assay. In the M-1 collecting duct line Na+ transport, mαENaC expression and luciferase expression from αENaC genomic fragments were also increased by 100 nm dexamethasone. In a colonic cell line, HT29, trans-activation via a heterologously expressed glucocorticoid receptor restored glucocorticoid-stimulated αENaC gene transcription. We conclude that glucocorticoids stimulate αENaC expression in kidney and lung via activation of a hormone response element in the 5′-flanking region of hαENaC and this response, in part, is the likely basis for the up-regulation of Na+transport in these sites.

Transepithelial sodium transport in the collecting ducts of the kidney, in airway epithelia, and in sweat and salivary glands occurs principally via amiloride-sensitive epithelial sodium channels (ENaC) 1 (1). This channel is a hetero-multi-meric complex, composed of three distinct but homologous subunits termed ␣, ␤, and ␥ ENaC, that was first cloned from rat colon (2,3). Two of the major physiologic functions of this channel are to absorb lung liquid at birth and to regulate Na ϩ and K ϩ homeostasis. Targeted disruption of the ␣ENaC subunit causes fatal neonatal respiratory failure, whereas disruption of the ␤ or ␥ENaC subunit produces death from hyperkalemia (4 -6). Overactivity of the channel causes Na ϩ retention, hypokalemia, and hypertension, a phenotype that can be produced by activating mutations in one of the channel subunits, by excessive production of mineralocorticoids (MC) or by excessive stimulation by endogenous or exogenous glucocorticoids (GC) (7).
GC and MC are two important physiological regulators of amiloride-sensitive epithelial Na ϩ transport in target epithelia where their cognate receptors are expressed. Dexamethasone or aldosterone increases ␣ENaC mRNA expression in rat kidney cortex and in primary cultures of inner medullary collecting duct cells (IMCD) without any effect on ␤ and ␥ ENaC expression (8 -11). This effect on ␣ENaC correlates temporally with an increase in amiloride-sensitive Na ϩ transport in primary IMCD cultures (12). The effect of adrenal steroids on the abundance of ␣ENaC mRNA is not confined to the kidney. Dexamethasone, but not aldosterone, increases ␣ENaC expression in fetal and adult rat lungs in vivo, and in fetal lung explants and cultured airway epithelial cells in vitro (8,(13)(14)(15). The mechanism of the increase in steady state levels of ENaC mRNAs following adrenal steroid treatment is unknown but has been presumed to be at the level of ENaC gene transcription.
In this study, we identify the promoter and enhancer elements in the 5Ј-flanking region of the human ␣ENaC gene that regulate basal and GC-mediated induction of ␣ENaC gene expression. In addition we provide evidence that enhanced Na ϩ transport by two different cell lines (lung and kidney) is tightly linked to enhanced expression of ␣ENaC.
sham Pharmacia Biotech, and RU38486 was a generous gift from Roussel Uclaf (Romainville, France). Culture materials were from Life Technologies, Inc., and all radionucleotides were from NEN Life Science Products. Stock solutions of cycloheximide and actinomycin D were made in Me 2 SO and stocks of dexamethasone, spironolactone, and RU38486 made in ethanol.
Tissue Culture and RNA Isolation-The human lung epithelial cell line, H441, and the human colon carcinoma cell line, HT29, were cultured as described previously (16). The mouse renal cortical collecting duct (CCD) cell line, M-1, was grown in Dulbecco's modified Eagle's medium:F12 with 10% fetal calf serum (17). To examine the effects of dexamethasone on gene expression, cell cultures were switched to serum-free media and then exposed to various concentrations of dexamethasone or its vehicle for 24 h. Cycloheximide (10 M), actinomycin D (1 M), and the steroid receptor blockers, spironolactone (10 M) and RU38486 (10 M), were used in some experiments and compared with control cultures in the presence of vehicle alone. RNA was isolated from these cells as described previously (16).
Na ϩ Transport Measurements-H441 and M-1 cells were seeded on 12-mm Millicell PCF filters (Millipore, Bedford, MA), which had been pretreated with human placental collagen. H441 cells were grown for 6 days in RPMI medium with 6% serum and 100 nM dexamethasone and the medium changed daily. A day prior to electrical measurements, the cells were placed in steroid-free media and the following day placed in serum-free RPMI with or without 100 nM dexamethasone. The Millicell PCF filters were then transferred to a specially designed chamber (Jim's Instruments, Iowa City, IA) to measure transepithelial voltage, resistance (R T ) and short circuit current (Isc) at 37°C (18). M-1 cells were grown for 3 days in Dulbecco's modified Eagle's medium:F12 supplemented with 5 g/ml insulin, 5 g/ml transferrin, 5 nM triiodothyronine, 50 nM hydrocortisone, 10 nM sodium selenite, 50 g/ml gentamicin, 10 mg/ml bovine serum albumin, and 5 nM dexamethasone. The filters were then grown for 1 day without albumin and steroids and then with or without 100 nM dexamethasone and 10 M spironolactone for another 24 h.
To measure steady state levels of mouse ␣ENaC mRNA, a 451-bp fragment corresponding to sequence from the 3Ј portions of the coding region was amplified and cloned into pCRII and linearized at an internal BsmFI site to protect a 132-nt product. As a control a mouse ␤-actin template (pTR1-␤-actin, Ambion) was used to protect a 99-nt fragment. Antisense biotinylated cRNA probes were synthesized using the Brightstar Psoralen-Biotin kit (Ambion). Sample RNAs were co-hybridized with both riboprobes, digested with RNase A and T1, resolved by PAGE, transferred to nylon membranes, and developed by the Brightstar Biodetect system (Ambion) (8).
To quantitate mRNA expression, the autoradiograms were scanned with a PDI scanning densitometer and the density of individual bands measured using Quantity One software (Huntington Station, NY). Each ENaC band was normalized for the density of 18 S or ␤-actin band, and the data from three to four experiments pooled and analyzed by Student's t test and by analysis of variance.
Cloning of the 5Ј-Flanking Region of h␣ENaC-Sequence information from the 5Ј end of the SCNN1A locus was provided to us by the Chromosome 12 Mapping and Sequencing Center at Albert Einstein University. The transcription start sites for h␣ENaC have been mapped and previously reported (16). To clone the ϳ1500 bp of 5Ј-flanking region upstream of exon 1A, primer ␣25 (5Ј-ACCCAGCACCCAGAGA-GCAGACGAA) and primer ␣23 (5Ј-TCAGGCCCTGCAGAGAAGAGAG-AAGAGGTC) were used to amplify a DNA fragment from human genomic DNA. This fragment, which includes 55 nt of the 5Ј UTR of ␣ENaC-1, was directionally subcloned into pGL3basic (Promega, Madison, WI) upstream of the firefly (Photinus pyralis) luciferase coding region. The primers ␣18 (5Ј-GAGGGGGTGGCGAGGAATCA) and ␣21 (5Ј-CTCGAGCTGTGTCCTGATTC) were used to amplify ϳ700 bp of sequence upstream of ␣ENaC-2, digested with Bsu36I and then subcloned into pGL3basic. This genomic DNA fragment begins downstream of the transcription start site of ␣ENaC-1 and includes 682 nt 5Ј to the transcription start site of ␣ENaC-2 and 100 nt of its 5Ј UTR. Deletion variants of these constructs were created by using internal restriction enzyme sites or by re-amplification using internal primers.
To test the glucocorticoid-responsive enhancer region in reverse orientation and with a heterologous promoter, an AvrII site was introduced at Ϫ289 using primer ␣27 (5Ј-CCTAGGCTCCGGGTCTGTGTCCA). A native Avr II site at Ϫ142 allowed the recovery of a 146-nt fragment, which was cloned in reverse orientation upstream of a P1 construct at Ϫ142 and a P2 construct at ϩ322. The glucocorticoid-responsive DNA fragment was also cloned upstream of a 459-nt h␥ENaC promoter (19) coupled to luciferase. To test individual glucocorticoid response elements (GRE) in transfection assays, single-stranded oligonucleotides that correspond to the putative GREs were synthesized and then annealed together. The flanking SacI and AvrII overhangs were used to clone these fragments into the native AvrII site upstream of ␣ENaC-1 (see sequence below).

SEQUENCES 1-3
Transfection and Functional Analysis of the 5Ј-Flanking DNA Clones-Subconfluent H441 cells grown in 12-well plates were used for transfection. 1 g of the firefly luciferase construct or the parent plasmid pGL3basic and 1-2 g of a control plasmid pSV␤-gal, where the Escherichia coli lacZ gene is cloned downstream of the SV40 promoter (Promega) was combined with Lipofectin (Life Technologies, Inc.) and added to each well. For transfection of M-1 cells, monolayers from two T75 flasks were trypsinized, and then resuspended in 2 ml of a solution containing 120 mM KCl, 150 M CaCl 2 , 10 mM K 2 HPO 4 , 10 mM KH 2 PO 4 , 2 mM EGTA 5 mM MgCl 2 , 100 mM ATP, 250 mM glutathione, and 25 mM Hepes, pH 7.6. A 500-l aliquot of cells was combined with 20 g of a luciferase construct and 40 g of pSV␤-gal in a 0.4-cm cuvette and then electroporated (Electroporator II, Invitrogen, Carlsbad, CA) at 330 V, 1000 microfarads. and 500 ohms. Each aliquot of cells was then diluted with complete medium and plated into four to six wells of a 12-well plate. 24 -48 h following transfection, cells were placed in serum-free medium and dexamethasone or vehicle were added where appropriate; another 24 h later, cell lysates were prepared for measurement of reporter gene activity. For HT29 cells, LipofectAMINE Plus (Life Technologies, Inc.) was used as the transfecting reagent. As HT29 cells contained a high level of endogenous ␤-galactosidase activity, we used pRL-SV40 (Promega), where the sea pansy (Renilla reniformis) luciferase gene is cloned downstream of the SV40 promoter, as the internal control plasmid. In some experiments, 1 g of the plasmid p6RGR, where the rat glucocorticoid receptor cDNA is under the control of the Rous sarcoma virus promoter (gift from D. Pierce and K. Yamamoto) was co-transfected with the luciferase vectors.
For preparation of cell lysates, the cells were washed in phosphatebuffered saline and then scraped into Lysis buffer (Luciferase assay kit, Promega). An aliquot of cell lysate was added to Luciferase Assay reagent and activity measured in a Monolight 2010 luminometer (Analytical Luminescence Laboratories, Ann Arbor, MI) for 5 s. For measurement of ␤-galactosidase activity, an equal aliquot of cell lysate was incubated with the substrate Galacton-plus (Tropix, Bedford, MA) for

FIG. 1. Short circuit current (Isc) measurements in H441 cells.
A, an increasing current is seen starting at day 2 after seeding. Cells were grown continuously in 100 nM dexamethasone. B, the current is almost entirely blocked by 10 M amiloride (Amilϩ) applied to the apical side of the membrane. C, cells grown for 24 h in steroid-free complete media and then placed in serum-free steroid-free media and treated with 100 nM dexamethasone (dex) or vehicle (ctrl). A robust dexamethasone-inducible Isc is evident at 24 h. *, p Ͻ 0.001 compared with control; #, p Ͻ 0.001 compared with 0 h. 30 to 60 min and then activity measured in a luminometer for 5 s. Lysates from HT29 cells were made with Passive Lysis Buffer (Promega) and firefly luciferase activity, reflecting transcriptional strength of cloned ENaC gene fragments, and sea pansy luciferase activity was sequentially measured in the same sample using the Dual Luciferase Reporter assay kit (Promega).
Gel Mobility Shift Assay-Oligonucleotides that correspond to the functional ␣ENaC GRE or to a nonspecific sequence were synthesized and annealed together (see sequences below).

SEQUENCES 4 AND 5
For gel mobility shift experiments, 50,000 cpm of end-labeled doublestranded oligonucleotides were incubated with recombinant human glucocorticoid receptor (Affinity Bioreagents Inc., Golden, CO) in a 20-l reaction mixture that contained 20 mM Hepes, pH 7.9, 60 mM KCl, 5 mM MgCl 2 , 2 mM dithiothreitol, 10 ng/ml poly(dI-dC), 5 g/l bovine serum albumin, and 10% glycerol. For competition experiments, a 50fold excess of cold oligonucleotides were used. All constituents except the labeled probe were preincubated at 4°C for 30 min and then incubated with the labeled probe at 22°C for 30 min. Samples were then resolved on a 3.5% nondenaturing polyacrylamide gel (acrylamide: bisacrylamide 20:1) in 0.5ϫ TBE buffer run at 150 V.

RESULTS
GC increases ␣ENaC mRNA expression in lung and kidney cortex and in cultured epithelial cells derived from these tissues (8 -11, 15). To address the mechanisms whereby GC increase ␣ENaC mRNA expression, we used a human lung cell line H441 and mouse CCD cell line M-1. We found that H441 cells, when grown on permeable supports in the presence of dexamethasone, develop as a tight epithelium and have electrogenic ion transport measured as a positive Isc from the apical to the basolateral compartment (Fig. 1A). This current was almost completely blocked by 10 M amiloride applied to the apical side (Fig. 1B), indicating that the bulk of the Isc was accounted for by amiloride-sensitive electrogenic Na ϩ transport pathways. To examine the effect of dexamethasone on Na ϩ transport, H441 cells were placed in steroid-free media for 24 h and then exposed to 100 nM dexamethasone or vehicle for another 24 h. Dexamethasone increased Isc from 3.3 A/cm 2 to 7.5 A/cm 2 (Fig. 1C), similar to results obtained from primary cultures of airway epithelial cells (15).
We have previously reported that heterogeneity in h␣ENaC transcripts arise from alternate transcription start sites and from splicing at the 5Ј end of h␣ENaC (16). Exon 1A begins at the upstream transcription start site and gives rise to ␣ENaC-1 while a second transcription start site 724 bp downstream in an alternate first exon (exon 1B) gives rise to ␣ENaC-2,3,4 (Fig. 2). 24 h of treatment with dexamethasone increased expression of both ␣ENaC transcripts in a dose-dependent manner with the earliest effect seen at 10 nM (Fig. 3A). The basal expression of exon 1A-intiated transcripts was greater than exon 1B-initiated transcripts, and this difference persisted through the dose-response curve. The effect of GC on ␣ENaC expression was blocked by RU38486, a type II (glucocorticoid receptor; GR) antagonist, but not by spironolactone, a type I (mineralocorticoid receptor; MR) antagonist, confirming that these effects required GR binding (Fig. 3B). The effect of GC on ␣ENaC expression was blocked by co-administration of actinomycin D, an inhibitor of transcription, suggesting that GC stimulates ␣ENaC expression by increasing gene transcription rather than by affecting mRNA stability (Fig. 3C). Cycloheximide, a general inhibitor of translation, did not prevent GC-induction of h␣ENaC mRNA expression and may have augmented expression of the h␣ENaC-1 transcript (Fig. 3D).
To examine the mechanism of transcriptional regulation of the ␣ENaC gene, we cloned 5Ј-flanking sequences upstream of both transcription start sites and evaluated these by transient transfection assays. We focused our attention on identification of the putative promoters as well as the GC-regulated elements of the ␣ENaC gene. Constructs that included either transcription start site and portions of their proximal 5Ј-flanking sequence were able to stimulate luciferase expression in H441 cells (Fig. 4A). The ability of constructs containing sequence 5Ј to each transcription start site (e.g. Ϫ142 ϩ 44 and ϩ322 ϩ814) to increase luciferase gene transcription suggest that separate promoters direct expression of ␣ENaC-1 and ␣ENaC-2. Two of the five tested constructs robustly stimulated luciferase activity when treated with 100 nM dexamethasone (Fig. 4A). Analysis of the sequence common to these constructs (Ϫ487 and Ϫ142) revealed that it contained one or more GREs. The magnitude of the dexamethasone-induced luciferase gene transcription is similar to the magnitude of ␣ENaC mRNA stimulation in H441 cells (Fig. 3A), suggesting that transcriptional activation of this region is sufficient to account for the effect of dexamethasone on steady state ␣ENaC mRNA levels. A preliminary analysis of the nucleotide sequence in this region between Ϫ289 and Ϫ142 showed two imperfect GREs, TGTc-CannnTGTcCT (Up-GRE) and AGAACAnnnTGTcCT (Dn-GRE), which are candidate cis-elements to mediate this effect.
To define the promoter regions P1 and P2, further deletions of the constructs Ϫ142ϩ44 and ϩ322ϩ814, respectively, were made and tested. In the case of the P1 promoter, a construct as short as Ϫ82 to ϩ44 was active, suggesting that the minimal promoter was contained within this region (Fig. 4B). In the case of the P2 promoter, a construct as short as ϩ476 to ϩ874 was active while the construct (ϩ596ϩ814) that included just ϳ70 nt of 5Ј-flanking sequence was not, suggesting that the minimal promoter included sequences 5Ј to this construct. (Fig. 4C).
To determine if the region between Ϫ289 and Ϫ143 in the 5Ј-flanking region of the h␣ENaC gene could transduce the glucocorticoid effect, a 146-bp sequence containing these elements were cloned in reverse orientation (Ϫ289Ϫ143INV) upstream of the P1 and P2 promoters and transfected into H441 cells. As expected for a classic "enhancer," these elements functioned in reverse orientation to confer glucocorticoid responsiveness to both promoters (Fig. 5, A and B). These experiments confirm that the defined region is sufficient to direct glucocor- Right panel, the genomic constructs were compared with the empty plasmid pGL3basic and the luciferase assay corrected for ␤-gal activity is shown. Each construct was tested with (black bar) and without (white bar) dexamethasone (100 nM for 24 h). *, p Ͻ 0.002 compared with pGL3basic; #, p Ͻ 0.02 compared with the absence of dexamethasone; ϩ, p Ͻ 0.002 compared with the absence of dexamethasone. (n ϭ 3-4 determinations). Panels B and C, mapping of promoters P1 and P2. B, the minimal sequence for transcriptional activity of P1 is contained within the Ϫ82ϩ44 fragment. *, p Ͻ 0.001 compared with pGL3basic; #, p Ͻ 0.001 compared with Ϫ99ϩ44 construct. C, the minimal sequence for transcriptional activity of P2 is contained within the ϩ476ϩ814 fragment. *, p Ͻ 0.002 compared with pGL3basic. ticoid-regulated gene transcription from P1 and P2. We also tested this ␣ENaC genomic fragment with the promoter for h␥ENaC. Our results show that the enhancer region of ␣ENaC confers glucocorticoid responsiveness to this heterologous promoter (Fig. 5C).
To further define transcriptional regulatory elements necessary for GC induction, a new set of 5Ј deletion constructs were made and tested in H441 cells (Fig. 6, A and B). To determine the role of each GRE, single copies of double-stranded oligonucleotides corresponding to each 15-nt element were cloned upstream of the h␣ENaC-1 promoter and tested in H441 cells (Fig. 6B). These results confirmed that the region containing the Dn-GRE (Ϫ248Ϫ142) was sufficient to confer GC enhancement. First, the Ϫ248ϩ44 construct, which excludes the Up-GRE, is capable of responding to dexamethasone (Fig. 6A); inclusion of the Up-GRE sequence (Ϫ289ϩ44) does not alter either basal or the GC-enhanced response (Fig. 6B). Second, the construct that contains only the Dn-GRE permits GC enhancement, but the construct containing only the Up-GRE does not (Fig. 6B). Finally, a 4-nt mutation that is predicted to disrupt GR binding abolished GC-mediated luciferase expression from the Dn-GRE (Fig. 6B). The region immediately 5Ј to the Dn-GRE also appeared to contain an inhibitory element as constitutive expression of the Ϫ248Ϫ142 construct was lower than the Ϫ142 ϩ44 construct (Fig. 6A).
To determine if the identified Dn-GRE could bind GR in a mobility shift DNA binding assay, oligonucleotides corresponding to these sequences were synthesized, incubated with purified GR, and then analyzed by PAGE and autoradiography. To test specificity of interaction, the purified GR was preincubated with an excess of cold oligonucleotides corresponding to Dn-GRE, Up-GRE, or a nonspecific oligonucleotide prior to incubation with labeled oligonucleotide. Our results indicate that GR binds specifically to the functional Dn-GRE to retard its mobility (Fig. 7).
We next examined amiloride-sensitive Na ϩ transport and ␣ENaC gene expression in M-1 cells, a mouse CCD cell line. This cell line, established from an SV40 transgenic mouse, appears to have many of the characteristics of the native CCD including the presence of amiloride-sensitive Na ϩ transport (17). Dexamethasone treatment leads to marked stimulation of Na ϩ transport within 24 h (Fig. 8A). To determine the mechanism of this effect, we examined the expression of mouse ␣ENaC mRNA in these cells when treated with dexamethasone. We show a potent stimulation of ␣ENaC mRNA expression by 100 nM dexamethasone in these cells (Fig. 8B). To determine if the stimulation of m␣ENaC mRNA expression occurs at the level of gene transcription, an ␣ENaC-luciferase construct that included the GC-responsive region (Ϫ487 ϩ661) was transfected into M-1 cells and then the effect of 24 h of stimulation with 100 nM dexamethasone was studied. As in H441 cells, dexamethasone increased luciferase gene transcription via ␣ENaC genomic fragments about 5-10-fold (Fig. 8C). These experiments suggest that the GC-mediated induction of ␣ENaC gene transcription is similar in airway epithelia and in CCD.
To confirm that the GR is required for the GC effect on ␣ENaC gene transcription, we studied HT29 cells, a human colonic epithelial cell line. While these cells express the ␣ENaC mRNA constitutively (16), we had previously noted that there was no detectable increase in mRNA expression with GC. 2 When the ␣ENaC gene construct containing the GC-responsive regions (Ϫ487 ϩ661) was transfected into these cells, constitutive or basal activation was seen consistent with the presence of an active ␣ENaC promoter (Fig. 8D). When treated with 100 nM dexamethasone, no further increase was seen unless the GR was co-transfected with the luciferase construct (Fig. 8D). These results suggest that these cells lack GR or that the endogenous GR is not available for binding or trans-activation following stimulation with GC. Complementation with heterologously expressed receptor confirms that the GR is required for the GC effect on ␣ENaC gene transcription.

DISCUSSION
Corticosteroids are important physiological regulators of transepithelial sodium transport in the distal nephron, the distal colon, and the airway epithelia. With the cloning of the ENaC subunits, the effect of MC and GC on ENaC mRNA expression has been studied in a variety of epithelial cells and tissues from many species. The chronic effects of these hormones on Na ϩ transport coincide with an increase in mRNA expression of one or more subunits of the ENaC complex suggesting that synthesis of new ENaC channels may account for the increase in Na ϩ transport. However, the exact mechanism of the corticosteroid regulation of ENaC subunits is unknown.
We examined ␣ENaC expression in H441 cells, a human lung cell line where all ENaC subunits are expressed (␣, ␥ (Refs. 16 and 19); ␤ (see Footnote 2)). When allowed to grow as a polarized epithelium, these cells develop an amiloride-sensitive Isc that is enhanced by GC treatment (Fig. 1). We then used these cells to explore the effect of GC on ␣ENaC gene expression. It is important to note that this cell line expresses all four principal transcripts of h␣ENaC described to date and the transcription start sites are identical to that seen in human 2 C. P. Thomas, unpublished observation. kidney and lung tissue (16). We show that GC increased expression of both exon 1A-and 1B-initiated transcripts in a dose-dependent manner and that this effect was blocked by RU38486 and by actinomycin D. These results suggested that gene transcription following receptor binding is required for the GC effect on ␣ENaC expression in these cells.
To address the mechanisms of transcriptional regulation of the h␣ENaC gene and to identify its promoters, we cloned portions of the 5Ј-flanking region. Sequence analysis of the region upstream of each transcription start site did not reveal a canonical TATA box. TATA-less promoters tend to initiate transcription at multiple sites. Thus, it seemed possible that a single upstream promoter could initiate transcription at both ␣ENaC start sites. Our results show that genomic fragments directly upstream of each transcription start site were able to drive luciferase expression in H441 cells (Fig. 4A) consistent with the presence of dual promoters. The approximate limits of each promoter were mapped by deletional analysis and appear to include 82 nt of 5Ј-flanking region for ␣ENaC-1 and 191 nt of 5Ј-flanking region for ␣ENaC-2 (Fig. 4, B and C).
In addition to basal or constitutive activity, we identified GC-inducible transcriptional activity that mapped to a region upstream of ␣ENaC-1 and contained imperfect GREs. Deletional analysis demonstrated that the Dn-GRE but not the Up-GRE was functional (Fig. 6B). Comparison of the GREs showed that the Dn-GRE was an imperfect palindromic repeat with a 3-nt spacer that separates each hexameric half site and differs from the classic GRE (AGAACAnnnTGTTCT) by a sin-gle nt. The Up-GRE is an imperfect direct repeat with a 3-nt spacer that differs from the classic GRE by 3 nt. Steroid hormone receptors typically bind as homodimers to bipartite response elements that are palindromic and separated by 3 nt. However, several natural examples of direct repeats that serve as targets for GR binding are known including the mouse mouse mammary tumor virus promoter (20). Unlike palindromic repeats, direct repeats separated by only 3 nt have poor binding affinity for glucocorticoid and estrogen receptors but this binding improves when separated by 6 or 9 nt (21). Our results with transient transfection demonstrating that the palindromic repeat, Dn-GRE, is functional while the direct repeat, Up-GRE, is nonfunctional is in keeping with this paradigm.
A search of rodent sequences in GenBank using the Patscan program at http://www.mcs.anl.gov/compbio/PatScan/HTML/patscan. html revealed at least two genes whose GREs were functional and identical to the Dn-GRE in ␣ENaC (22,23). One of these, the rat phenylethanolamine N-methyltransferase (PNMT) gene, has a functional GRE, AGAACAgagTGTCCT, Ϫ513 nt to the transcription start site (22). In this system the GRE may not act entirely alone. Native PNMT expression is stimulated 40-fold with dexamethasone but only 5-fold with a reporter construct containing the GRE. In our experiments in H441 cells, the magnitude of dexamethasone induction of ␣ENaC mRNA expression matched that seen with the Ϫ248 ϩ 44 construct (Fig. 6B), suggesting that all the required cis-elements for glucocorticoid stimulation are present within this region. Clearly, sequences adjacent to the Dn-GRE (between Ϫ248 and Ϫ142) are required for the full glucocorticoid effect as the Dn-GRE alone increased gene transcription by just 3-fold compared with 15-fold with the Ϫ248ϩ44 construct (Fig. 6B).
We have demonstrated by transfection, gel mobility shift, and trans-activation assays that glucocorticoids acting through their cognate receptor stimulate ␣ENaC gene transcription via an imperfect GRE upstream of the h␣ENaC gene. GREs can bind MR in vitro, and so far no distinct mineralocorticoid response elements have been identified (24). Currently, the GREs are thought to be the natural targets for activated MR. In vivo, in aldosterone-responsive epithelia the effect of GC on ENaC expression is very similar to the effect of MC (1) and it is quite likely that the MC effect is also via the ␣ENaC GRE. It is important to note, however, that the molecular responses to MC and GC have some differences (12,25). Mechanisms proposed to explain these differences include the specific metabolism of endogenous GC to an inactive metabolite in aldosteroneresponsive epithelia and the combinatorial regulation by GR but not MR at composite response elements for steroid receptors and other trans-acting factors (26,27).
The effect of corticosteroids on ENaC gene expression in vivo is tissue-specific with distinct effects on different subunits in the kidney, colon, and lung. For example, dexamethasone and aldosterone increase ␣ENaC mRNA in the kidney, but ␤ and ␥ENaC mRNA in colon (8 -10, 28). Under certain culture conditions, we can induce ␣ENaC gene transcription in response to glucocorticoids in a colonic cell line, HT29 (Fig. 8D). The fact that the ␣ENaC gene is not induced in the colon in vivo even though steroid receptors are present implies that the effect of corticosteroids on these subunits is not a simple consequence of receptor binding and activation of a GRE. The factors that determine tissue-specific regulation may include steroid receptor cofactors such as SRC-1 and SMRT (29,30) but this remains to be elucidated.