Preferential Transcription of Rabbit Aldh1a1 in the Cornea: Implication of Hypoxia-Related Pathways

Abstract

Here we examine the molecular basis for the known preferential expression of rabbit aldehyde dehydrogenase class 1 (ALDH1A1) in the cornea. The rabbit Aldh1a1 promoter-firefly luciferase reporter transgene (−3519 to +43) was expressed preferentially in corneal cells in transfection tests and in transgenic mice, with an expression pattern resembling that of rabbit Aldh1a1. The 5′ flanking region of the rabbit Aldh1a1 gene resembled that in the human gene (60.2%) more closely than that in the mouse (46%) or rat (51.5%) genes. We detected three xenobiotic response elements (XREs) and one E-box consensus sequence in the rabbit Aldh1a1 upstream region; these elements are prevalent in other highly expressed corneal genes and can mediate stimulation by dioxin and repression by CoCl₂, which simulates hypoxia. The rabbit Aldh1a1 promoter was stimulated fourfold by dioxin in human hepatoma cells and repressed threefold by CoCl₂ treatment in rabbit corneal stromal and epithelial cells. Cotransfection, mutagenesis, and gel retardation experiments implicated the hypoxia-inducible factor 3α/aryl hydrocarbon nuclear translocator heterodimer for Aldh1a1 promoter activation via the XREs and stimulated by retinoic acid protein 13 for promoter repression via the E-box. These experiments suggest that XREs, E-boxes, and PAS domain/basic helix-loop-helix transcription factors (bHLH-PAS) contribute to preferential rabbit Aldh1a1 promoter activity in the cornea, implicating hypoxia-related pathways.

Most mammalian corneas accumulate (5 to 40% depending on species) aldehyde dehydrogenase 3 (ALDH3A1) (unified nomenclature for the ALDH gene family [http://www.uchsc.edu/sp/sp/alcdbase/alcdbase.html]) mostly, but not exclusively, in their epithelium (47). While the function of the abundant ALDH3A1 in mammalian corneal epithelial cells is not established, detoxification of lipid peroxide radicals induced by UV light (26), absorption of UV (1), and a suspected structural role (47) are among the suggested possibilities. The fact that elimination of mouse ALDH3A1 (50% of corneal water-soluble protein) by homologous recombination in mice fails to show a corneal phenotype has increased the mystery of its function (44). Deficiencies in ALDH have also been implicated in keratoconus, which affects the corneal epithelium (18).

Rabbits are exceptional in that they accumulate ALDH1A1 rather than ALDH3A1 in the cornea (29). ALDH1A1 comprises about 3% and ALDH3A1 comprises about 5% of the total soluble protein in human cornea (31). In contrast to the situation with ALDH3A1 in other mammals, rabbit ALDH1A1 is more abundant in the stromal cells of the cornea than in the epithelium. Wound healing experiments after freeze-induced injury suggested that the prevalence of ALDH1A1 may contribute to transparency of rabbit stromal cells (29), analogous to the role of crystallins in the lens. Indeed, η-crystallin (ALDH1A8) is a lens crystallin, comprising 25% of the water-soluble protein of the elephant shrew lens (20). A protein equally similar to ALDH1A1 and ALDH2, Ω-crystallin, accumulates to high levels in the lenses of cephalopods (ALDH1C1 and ALDH1C2) and scallops (ALDH1A9) (48, 70). Unlike η-crystallin of elephant shrews (20), Ω-crystallins are enzymatically inactive with common substrates (48, 70). This differs from the highly expressed corneal ALDH1A1 and ALDH3A1, which are enzymatically active (40, 47); it remains unknown to what extent their enzymatic activity is necessary for their corneal function.

The cornea is an avascular tissue that must tolerate a 75% drop in the oxygen partial pressure at the corneal surface when the eyelids are closed (34). This hypoxia causes rabbit and human corneas to undergo a 4% thickening (39). Hypoxia-inducible factors (HIFs) responsible for hypoxia-mediated changes in gene expression bind hypoxia response elements (HRE) (5′-TACGTG-3′) and are members of the bHLH-PAS family which also mediate responses to xenobiotics, light, and developmental signals (23). Interestingly, an alternate transcript of one of the HIFs, Hif-3α^IPAS, encodes a hypoxia repressor protein called inhibitory PAS domain protein (IPAS) that may have a role in maintaining corneal avascularity (37, 38). In addition, stimulated by retinoic acid protein 13 (STRA13) is another repressor protein that is induced by hypoxia but which acts through E-boxes (of sequence 5′-CACGTG-3′) rather than HREs (61). An upstream xenobiotic response element (XRE) (5′-TNGCGTG-3′) (23) associated with rat Aldh3a1 that appears necessary to drive expression of the chloramphenicol acetyltransferase reporter gene in transfected rat corneal epithelial cells (6) further implicates hypoxia or other environmental stresses in the high level of expression of corneal genes. In this connection, some of the lens and cornea enzyme crystallins are genes known to be induced by hypoxia or xenobiotics in other tissues. For example, α-enolase (τ-crystallin in duck lens), glyceraldehyde-3-P-dehydrogenase (π-crystallin in gecko lens), and lactate dehydrogenase (ɛ-crystallin in duck lens) (47; J. V. Jester, J. Houston, B. Adams, J. Huang, D. Matinyare, W. M. Petroll, and H. D. Cavanagh, Abstr. Assoc. Res. Vision Ophthalmol., abstr. 4219, 2003) are induced by hypoxia (59). In addition, the highly expressed corneal ALDH3A1 and lens ζ-crystallin [NAD(P)H:quinone oxidoreductase] are induced by xenobiotics (43, 71).

In the present study, we have investigated the regulation of the rabbit corneal Aldh1a1 promoter in transient transfection and in transgenic mouse experiments. Together, our results suggest that E-boxes and XREs have been used to recruit the rabbit Aldh1a1 gene for a high level of expression in the cornea via hypoxia-related pathways. In addition to their relevance to the mechanism of tissue-specific gene expression, our results also may be pertinent to hypoxia-related corneal problems in diabetes (79) and contact lens use (34).

MATERIALS AND METHODS

Northern blot analyses.

Rabbit tissues were obtained frozen from an abattoir (Pel Freez, Rogers, Ark.). RNA was isolated using the TriPure isolation reagent (Roche Molecular Biochemicals, Indianapolis, Ind.). Five micrograms of total RNA was resolved by electrophoresis on a 6.6% formaldehyde gel. The gel was transferred to a Hybond N+ (Amersham, Piscataway, N.J.) membrane and UV cross-linked using a Stratalinker (Stratagene, La Jolla, Calif.). Blots were hybridized overnight at 65°C with 5 × 10⁶ cpm of [α-³²P]dCTP-labeled probe/ml in hybridization buffer (1 mM EDTA-0.5 M NaHPO4 [pH 7.2]-7% sodium dodecyl sulfate). Filters were washed for 15 min in 0.1× SSC (0.15 M NaCl plus 0.015 M sodium citrate)-0.1% sodium dodecyl sulfate at 65°C and autoradiographed on XAR film (Kodak, Buffalo, N.Y.).

Western blot analyses.

Total water-soluble protein was extracted from tissues in 50 mM Tris-HCl (pH 7.5) and 3 mM dithiothreitol with 1 tablet of Complete protease inhibitor (Roche). The insoluble fraction was removed by centrifugation at 16,000 × g for 10 min. Five micrograms of water-soluble proteins were subjected to electrophoresis in a 10% polyacrylamide bis-Tris NuPAGE gel (Invitrogen, Carlsbad, Calif.). Proteins from sodium dodecyl sulfate-polyacrylamide gel electrophoresis were transferred to nitrocellulose membranes, and blots were stained with SYPRO ruby stain (Molecular Probes, Eugene, Ore.) to ensure equal loading of protein samples. ALDH1A1 immunoblotting was performed using rabbit anti-human ALDH1A1 antibody. For immunoblots of nuclear extracts, 10 μg of rabbit stromal cell nuclear extract was used with goat anti-human HIF-1α antibody (sc-12542), goat anti-human aryl hydrocarbon nuclear translocator 1 (ARNT1) antibody (sc-8076), goat anti-human HIF-2α antibody (sc-8712), goat anti-mouse HIF-3α antibody (sc-8718), and goat anti-mouse AhR antibody (sc-8088) (Santa Cruz Biotechnology, Santa Cruz, Calif.).

Isolation of rabbit Aldh1a1 cDNA and genomic clones.

A rabbit corneal epithelial cell cDNA library constructed in lambda phage (Stratagene) was screened with radiolabeled cDNA fragments of rabbit Aldh1a1 generated by reverse transcription-PCR (RT-PCR) using RNA from rabbit corneas. A 0.56-kb amplicon was generated by RT-PCR of rabbit corneal RNA using the following primers: 5′-TCATAAACAATGAATGGCATGA-3′ (corresponding to nucleotides 124 to 145 of the human ALDH1A1 cDNA, accession no. K03000) and 5′-TAAAGATGCCACGTGGAGAG-3′ (corresponding to nucleotides 677 to 658 of the human ALDH1A1 cDNA). 5′ rapid amplification of cDNA ends (RACE) was performed on rabbit corneal RNA to identify the rabbit Aldh1a1 transcriptional start site. The 350-bp RACE product was radiolabeled and used to isolate genomic library clones from a rabbit genomic library (Clontech, Palo Alto, Calif.). Genomic library clones containing the rabbit Aldh1a1 gene were isolated and sequenced using primers derived from the 5′ RACE product of the cDNA. Upstream sequence −5154 to +43 was obtained and analyzed using MatInspector Professional (50). Rabbit genomic sequence was also compared to genomic sequence available from the human (http://www.ncbi.nlm.nih.gov/genome/guide/human/), mouse (http://www.ensembl.org/Mus_musculus/), and rat (http://www.ncbi.nlm.nih.gov/genome/guide/rat/index.html) genome databases.

Aldh1a1 reporter constructs.

Two regions (−1054 to +43 and −3519 to +43) including the 5′ flanking region of rabbit Aldh1a1 were amplified by PCR by using Pfu polymerase (Stratagene) with the following primers: 5′-TCCCCCCGGGATGCCTGAGGACGATTTTCC-3′ and either 5′-GGGGTACCAGGGGAGCTGGCAATTTTCACT-3′ (for the −1054/+43 amplicon) or 5′-CGACGCGTCGACAGAAAGTAGCCCACAAGCAT-3′ (for the −3519/+43 amplicon). KpnI/XmaI (−1054/+43) or MluI/XmaI (−3519/+43) restriction sites included in the primers were used to subclone the amplicons into pGL3-Basic (Promega, Madison, Wis.), containing the firefly luciferase reporter gene.

Cell culture and transfection analyses.

Primary cultures of rabbit corneal epithelial and stromal cells were isolated as described previously (28). Rabbit corneal stromal cells were maintained in minimal essential medium supplemented with 10% fetal bovine serum and with 1× penicillin, 1× streptomycin, 1% glutamine, and 1× Nystatin (Invitrogen). Primary cultures of rabbit corneal epithelial cells were grown in Medium 500 with 1× corneal epithelial growth supplement (Cascade Biologics, Portland, Ore.) with 1× penicillin, 1× streptomycin, 1% glutamine, and 1× Nystatin. The SIRC cell line (45), the human hepatoma Hep3B and HepG2 cell lines (32), the mouse hepatoma Hepa1c1c7 cell line (4), and the N/N1003A cell line (51) were grown as described previously. Conditions simulating hypoxia were introduced to cultured corneal cells by treatment with either 75 or 150 μM CoCl₂ (19). Dioxin treatment consisted of adding 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) to a concentration of 50 nM (19). Transfection experiments were performed using Fugene Reagent (Roche Molecular Biochemicals). One microgram of plasmid DNA was used in the transfection along with 100 ng of pRL-SV40 Renilla luciferase plasmid used as a control (Promega). Luciferase activity was measured using the Dual-Luciferase reporter assay system (Promega). pcDNA/IPAS was constructed as described elsewhere (37). pcDNA/HIF1-α was constructed from pEBB/HIF1-α (Novus Biologicals, Littleton, Colo.) using the pcDNA3.1 Directional TOPO expression kit (Invitrogen).

RT-PCRs.

Total RNA (0.5 μg) from primary cultures of rabbit corneal stromal cells was reverse transcribed into cDNA using random hexamers and Moloney murine leukemia virus (MMLV) reverse transcriptase (50 pmol; Invitrogen). One microliter of each reaction was amplified with human VEGF (accession no. X62568) (5′-CTACCTCCACCATGCCAAGT-3′ and 5′-GTCACATCTGCAAGTACGTTCG-3′), mouse Rpb1 (accession no. U37500) (5′-GCCATGCAGAAGTCTGGCCGTCCCCTCAAG-3′ and 5′-CTTATAGCCAGTCTGCAGATGAAGGTCAC-3′), or rabbit Aldh1a1 (5′-AGGGTTGAACATTGTCCCTGGT-3′ and 5′-AGTAGCCTTTATTCCCCCATGG-3′). PCR products were separated by electrophoresis and analyzed with a ChemiImager 4000 (Alpha Innotech, San Leandro, Calif.).

Transgenic mice and luciferase assays.

KpnI-SalI fragments (3.1 kb or 5.5 kb) containing either the −1054/+43 or the −3519/+43 rabbit Aldh1a1 promoter fused to the firefly luciferase gene were isolated from their respective plasmids, purified from an agarose gel using the Geneclean kit (Bio101, Carlsbad, Calif.), and injected into the pronucleus of a single-celled mouse embryo (FVB/N strain) obtained from superovulated FVB/N females. Injected embryos were transferred into CD1 females. Tissues from adult 8-week-old transgenic mice were homogenized in lysis buffer included in the luciferase assay system (Promega). Samples were centrifuged at 10,000 × g in a microcentrifuge, and firefly luciferase activity of the supernatant fraction was measured using the luciferase assay system (Promega) and the TR717 microplate luminometer (Applied Biosystems, Foster City, Calif.). The concentration of soluble protein was determined using the Bio-Rad protein assay (Bio-Rad Laboratories, Hercules, Calif.).

Preparation of rabbit stromal cell nuclear extracts.

The nuclear extracts were prepared from cultured rabbit stromal cells. Cells were trypsinized and washed with phosphate-buffered saline. All subsequent steps were performed on ice and in a cold room. The procedure was scaled down from that described by others (10). Homogenization was performed by 10 strokes of a Dounce homogenizer. Extracts were frozen in an ethanol-dry ice bath and stored at −80°C until use.

EMSA.

Electrophoretic mobility shift assay (EMSAs) were performed as described by others (10). Equimolar amounts of complementary oligonucleotides containing XRE2 (5′-TTGGGATTCCAGATTTCACGCATACATATA-3′ and 5′-TCGATATATGTATGCGTGAAATCTGGAATC-3′, corresponding to −3290 to −3261 upstream of rabbit Aldh1a1) or E-box (5′-CTTCTTCCAGGTTTCCCACGTGGGTGCAGG-3′ and 5′-TGCCCCTGCACCCACGTGGGAAACCTGGA-3′, corresponding to −3029 to −3000 upstream of rabbit Aldh1a1) were annealed. The annealed oligonucleotides were labeled with [α³²P]dCTP by filling in 5′ overhangs using the Klenow fragment (Invitrogen) and were purified on G-25 columns (Amersham Biosciences). Binding reactions were carried out in 20-μl volumes containing 1 μl of labeled oligonucleotide probe, 10 μg of nuclear extract, 50 mM NaCl and KCl, 20 mM HEPES (pH 7.9), 1 mM MgCl₂, 4% Ficoll, and 0.5 mM dithiothreitol. Binding reactions were carried out for 40 min on ice, after which 10 μl was loaded on a 6% nondenaturing polyacrylamide gel (Invitrogen) in 0.5× Tris-borate-EDTA buffer. Complexes were competed with 10-fold excess wild-type or mutated oligonucleotides. The mutated oligonucleotides for XRE2 and the E-box contained a single point mutation in the core binding sequences, XRE2 (5′-TTGGGATTCCAGATTTCCCGCATACATATA-3′ and 5′-TCGATATATGTATGCGGGAAATCTGGAATC-3′) and E-box (5′-CTTCTTCCAGGTTTCCCAGGTGGGTGCAGG-3′ and 5′-TGCCCCTGCACCCAGGTGGGAAACCTGGA-3′). For supershift experiments, antibodies were incubated with the nuclear extract for 30 min on ice before the addition of radioactively labeled oligonucleotides.

Nucleotide sequence accession number. The sequence of the 5′ end of rabbit Aldhlal was submitted to GenBank database under accession no. AY508694. The sequence of the −5154 to +43 region of the rabbit Aldhlal gene was submitted under accession no. AY508695.

RESULTS

Isolation and characterization of cDNA clones of rabbit Aldh1a1.

We amplified a rabbit Aldh1a1 cDNA fragment by RT-PCR, using rabbit lung poly(A)⁺ mRNA and rabbit corneal RNA with primers derived from human ALDH1A1 cDNA. This fragment was used to screen a rabbit corneal epithelial cDNA library. The rabbit Aldh1a1 probe hybridized to about 1% of total clones. Full-length cDNA clones were generated by 5′ RACE using rabbit corneal RNA. Our rabbit Aldh1a1 cDNA sequence agrees with that of others (40) (accession no. AY038801) published during the course of the present investigation; in addition, the present cDNA includes the 5′ end obtained by RACE. The 5′ RACE cDNA products were identical for lung and corneal mRNAs, indicating the same transcription start sites of the gene in both tissues. This is consistent with the same 1.9-kb Aldh1a1 transcript in all rabbit tissues tested by Northern analysis (Fig. 1A). Furthermore, hybridization of rabbit Aldh1a1 cDNA fragments to genomic Southern blots indicated a single gene locus for Aldh1a1 in the rabbit (data not shown).

FIG. 1.

Preferential expression of ALDH1A1 in rabbit corneal tissues. (A) Northern blot of 5 μg of total RNA extracted from rabbit tissues. RNA was probed with the amplicon of the 5′ RACE of rabbit Aldh1a1 and with a mouse 28s rRNA probe. (B) Immunoblots of 10 μg of total soluble proteins from rabbit tissues with rabbit anti-human ALDH1A1 antibody. The protein blot was stained with the SYPRO Ruby protein blot stain and visualized by UV illumination (not shown). Epi, corneal epithelium; Endo, corneal endothelium.

Characterization of ALDH1A1 expression in rabbit tissues.

Northern blot analysis of rabbit tissues revealed the preferential expression of Aldh1a1 in the cornea (Fig. 1A). After adjustment for amounts loaded, densitometry of Northern blots indicated that the rabbit cornea has at least 6.7 times more Aldh1a1 mRNA than the lung, and there were only trace amounts of Aldh1a1 mRNA in the liver, kidney, and testis. Western blot analyses of the different cell layers of the rabbit cornea showed that the stromal cells had 1.5 times more ALDH1A1 protein than the endothelial layer (Fig. 1B) and 5.6 times more ALDH1A1 than the epithelium, where ALDH1A1 was at a level comparable to that in the lung. Densitometry of SYPRO ruby-stained protein polyacrylamide gel electrophoresis blots (data not shown) of rabbit corneal extracts revealed ALDH1A1 to be 3.3, 16.1, and 10.9% of the total water-soluble protein of the epithelium, stroma, and endothelium, respectively. Western blotting analysis also revealed that the iris, sclera, and retina have higher levels of ALDH1A1 protein than the corneal epithelium, although less than the corneal stroma and endothelium (Fig. 1B). A cross-reacting 29-kDa-molecular-weight band was detected in the lens extract, although the identity of this band is not known. Thus, the rabbit Aldh1a1 gene is widely expressed, but specialized for the cornea, particularly within the stroma and endothelium.

FIG. 5.

Rabbit Aldh1a1 promoter activity in transfected cells treated with dioxin (TCDD) or CoCl₂. (A) The −3519/+43 promoter fragment of the rabbit Aldh1a1 gene driving expression of the firefly luciferase reporter gene in transiently transfected cells that were left untreated (white bars), treated with 50 nM TCDD (light grey bars), treated with 75 μM CoCl₂ (medium grey bars), treated with 150 μM CoCl₂ (dark grey bars), or treated with both 150 μM CoCl₂ and 50 nM TCDD (black bar). The difference in activity of the promoter when treated with 150 μM CoCl₂ is statistically significant (P < 0.05). Results are the averages for 10 transfection experiments. (B) RT-PCR amplification of amplicons of VEGF, RNA polymerase II (Rpb1), and Aldh1a1 in rabbit cornea stromal cells with (lane 2) or without (lane 1) 150 μM CoCl₂. Lane 3 is the control PCR with no template added. (C) Transfection of rabbit corneal stromal (black bars) and epithelial (grey bars) cells and of the HepG2 cell line (white bars) with the −3519/+43 rabbit Aldh1a1 promoter construct containing point mutations in XRE1, XRE2, or the E-box. The mutated sites in the different constructs are crossed out. Results are the average of 6 transfection experiments. The reduction in activity when all three sites are mutated is statistically significant (P < 0.05).

Isolation of the 5′ upstream region of the rabbit Aldh1a1 gene.

In order to analyze the 5′ regulatory region of the rabbit Aldh1a1 gene, we isolated genomic library clones using a 425-bp probe generated from the corneal 5′ RACE product. Three lambda clones were isolated and characterized by Southern blotting. Starting with a primer for exon 1 of rabbit Aldh1a1, the nucleotide sequence for 5,154 bp of the 5′ upstream region of the Aldh1a1 gene was obtained. Truncated fragments of the promoter were subcloned into a firefly luciferase reporter vector for functional analysis.

The nucleotide sequence of the upstream region of the rabbit Aldh1a1 gene was compared with that of its human, mouse, and rat orthologs available from the human, mouse, and rat genome databases. Overall, the −5154 to −1 sequence of the rabbit Aldh1a1 gene was 60.2% identical to the −5026 to −1 sequence of the human ALDH1A1 gene, and several segments (−469 to −1, −796 to −567, −1034 to −917, −1906 to −1689, and −2078 to −1997) were more than 80% identical to the corresponding regions in the human promoter (Fig. 2). In contrast, the same region of the rabbit Aldh1a1 gene was only 46% identical to the −4991 to −1 sequence of the mouse Aldh1a1 gene. Moreover, the human ALDH1A1 promoter was only 49% identical to that of the mouse, and the −5160 to −1 sequence of the rat Aldh1a1 gene was 51.5 and 54.1% identical to those of the rabbit and human, respectively. The two rodent genomic sequences were 71.2% identical. These data show that the rabbit Aldh1a1 promoter sequence is more similar to that of the human promoter than to those of the mouse or rat promoters.

FIG. 2.

Isolation of the 5′ upstream region of rabbit Aldh1a1. Genomic nucleotide sequencing was performed upstream of the fragment identified by hybridization with the 5′ cDNA RACE probe. Percent identity plots of the −5154 to +43 region of the rabbit Aldh1a1 gene compared with the Aldh1a1 gene sequences from human (−5026 to −1), mouse (−4991 to −1), and rat (−5160 to −1) (top). Analysis was performed using Advanced PipMaker (58). The bar represents the −5154 to +43 region of the rabbit Aldh1a1 gene with putative transcription factor binding sites identified with MatInspector Professional (50) (bottom). In bold are XRE1, XRE2, XRE3, and the E-box.

There are many consensus transcription factor binding sites in the upstream region of the rabbit Aldh1a1 gene (Fig. 2, lower panel). In addition, many of the cis elements found in the human ALDH1A1 gene, such as retinoic acid response-like elements and binding sites for HNF-5, OCT-1, PEA3, GATA-1, and NF-IL-6 (75), are conserved in the rabbit promoter sequence. Of particular interest are three putative XREs (−714 to −709, −3274 to −3268, −5098 to −5093) in the rabbit Aldh1a1 genomic fragment, which are not found in the upstream region of the human ALDH1A1 gene. These XREs are of interest because an XRE has been implicated in corneal expression of the rat Aldh3a1 gene (6). The XREs of mouse Aldh3a1 are responsible for its dioxin-inducible expression (43). An E-box (of the sequence 5′-CACGTG-3′) (−3013 to −3008; Fig. 2) that could putatively bind the same or similar transcription factors which bind XREs (2, 27) is also present in the rabbit Aldh1a1 promoter fragment. We identified several XREs (5′-TNGCGTG-3′) and E-boxes (5′-CACGTG-3′) using MatInspector Professional (50) in the gene regulatory regions of other highly expressed corneal genes, including mouse TKT (57), mouse keratocan (35), mouse keratin 12 (36), and human alpha-enolase (80). Members of the bHLH-PAS family of transcription factors can bind these XREs and E-boxes (23). The IPAS protein, a bHLH-PAS repressor, has recently been proposed to be involved in maintaining corneal avascularity (37). Consequently, we examined the functional significance of the XRE and E-box consensus sequences in the rabbit Aldh1a1 promoter fragment in transient transfection and site-specific mutagenesis experiments.

Rabbit and human Aldh1a1 promoter activity in transfected cells.

Initially, wild-type constructs of the rabbit Aldh1a1 promoter driving expression of the firefly luciferase gene were tested in transient transfection experiments. Two promoter constructs (−1054/+43 and −3519/+43) were active in primary cultures of rabbit corneal stromal and epithelial cells (Fig. 3A). On average, the Aldh1a1 promoter was twice as active in the transfected stromal cells as in the transfected epithelial cells, consistent with the preferential expression of the Aldh1a1 gene in stromal cells in vivo. The transfected hepatoma cell lines HepG2, Hep3B, and Hep1c1c7 gave fourfold-lower levels of Aldh1a1 promoter activity, and transfected rabbit N/N1003A lens cells gave 10-fold-lower levels of promoter activity than transfected corneal stromal cells. Of note, the human −3835/+43 ALDH1A1 promoter showed eightfold-lower activity in the transfected primary cultures of rabbit stromal cells than did the rabbit promoter (Fig. 3B), consistent with the in vivo difference in the relative corneal expression of these genes in their respective species. In the transfected rabbit cornea epithelial cells, the human construct had three times less activity than the rabbit promoter. However, the human ALDH1A1 promoter construct had activity similar to that of the rabbit Aldh1a1 promoter construct in a rabbit lung cell line and in the hepatoma cell lines.

FIG. 3.

Cornea-preferential activity of the rabbit Aldh1a1 promoter in transfected cells. (A) Promoter fragments of rabbit Aldh1a1 driving expression of the firefly luciferase reporter gene. Transfection into primary corneal stromal and epithelial cells, a rabbit corneal epithelial cell line (SIRC), a mouse hepatoma (Hepa1c1c7) cell line, human hepatoma (Hep3B and HepG2) cell lines, and a rabbit lens (N/N1003A) cell line. Therelative luciferase activity is given as the ratio of the luciferase activity generated by the promoter construct to that resulting from the empty luciferase vector. Each result is the average from 10 transfection experiments. The rabbit Aldh1a1 promoter fragments showed higher levels of luciferase activity in the rabbit corneal cells than in the other cell lines. (B) The −3519 to +43 region of rabbit Aldh1a1 (black bars) driving expression of the firefly luciferase reporter gene compared with the −3835 to +43 region of human ALDH1A1 (white bars) in transiently transfected cells. Each result is the average from eight transfection experiments.

Rabbit Aldh1a1 promoter activity in transgenic mice.

In order to test corneal preference, the −3519/+43 rabbit Aldh1a1 promoter fragment fused to the firefly luciferase gene was used to make transgenic mice (Fig. 4). This transgene was expressed in the cornea as well as in a number of other ocular and nonocular tissues (Fig. 4). In the nonocular tissues, luciferase activity was most prevalent in a sample of skeletal muscle (gracius leg muscle), followed by heart and lung (minimal). In the eye, there was preferential promoter activity in the cornea, but the transgene was expressed considerably in the iris and the sclera as well. Except for unexpectedly high activity in muscle and low activity in the lung, the pattern of activity of the −3519/+43 promoter in transgenic mice was similar to that of rabbit Aldh1a1 gene expression observed by Northern blot analysis (Fig. 1A). In order to better characterize the localization of luciferase expression in the corneas of transgenic mice with the −3519/+43 Aldh1a1 promoter/luciferase reporter transgene, we performed in situ hybridization using the firefly luciferase gene as a probe (data not shown). The highest expression of luciferase RNA detected by in situ hybridization was in the stromal cells; lower levels were observed in the epithelium. Detection of luciferase protein by immunohistochemistry using antiluciferase antibody was unsuccessful. Transgenic mice were also made using the −1054/+43 luciferase construct (data not shown). This truncated promoter fragment contains only XRE1. In three lines, the −1054/+43 promoter fragment directed a low level of luciferase expression to the cornea and about 12-times-greater luciferase expression to the retina of the transgenic mice (data not shown). The −1054/+43 fragment showed negligible activity in any other tissue examined.

FIG. 4.

Activity of rabbit Aldh1a1 promoter fragments driving firefly luciferase in transgenic mice. The −3519/+43 promoter fragment of the rabbit Aldh1a1 gene was used, driving expression of the luciferase reporter gene in 8-week-old transgenic mice. Average luciferase activities in relative units per microgram of soluble protein ± standard deviation for four lines of transgenic mice. Gracius leg muscle was isolated for analysis.

Response of the rabbit Aldh1a1 promoter to TCDD and CoCl₂ (simulated hypoxia).

Since XREs are known to activate promoters in response to dioxin (43) and to repress promoters when exposed to hypoxia (52), we tested rabbit Aldh1a1 promoter activity in transfected cells treated with TCDD (dioxin) or CoCl₂, which mimics hypoxia (17) (Fig. 5A). CoCl₂ has been shown to increase the protein stability of HIF-1α and HIF-2α under normoxic conditions (76), which suggests a mechanism for the mimicking of hypoxic effects. HIF-1α is usually induced in cells at oxygen levels less than 6% (59). The −3519/+43 promoter construct contains two XREs and the E-box but does not include XRE3, which is upstream of the −3519/+43 region (Fig. 2). In the HepG2 human hepatoma cell line, the −3519/+43 promoter construct was stimulated threefold by the addition of 50 nM TCDD. The stimulation was not observed when transfected rabbit stromal cells were treated with TCDD. An approximate threefold repression of rabbit Aldh1a1 promoter activity occurred when 150 μM CoCl₂ was added to both the transfected HepG2 and rabbit corneal cell cultures. This repression was not observed when 75 μM CoCl₂ was used. The addition of 50 nM TCDD to rabbit stromal cells cultured in 150 μM CoCl₂ prevented the repression caused by 150 μM CoCl₂ alone. In addition, the levels of endogenous expression of Aldh1a1 decreased 2.8-fold relative to RNA polymerase II (Rpb1) in response to simulated hypoxia (150 μM CoCl₂) in cultured rabbit corneal stromal cells (Fig. 5B). By contrast, levels of endogenous Vegf increased twofold relative to Rpb1, consistent with hypoxic stimulation of Vegf observed by others (59). Although less pronounced, the simulated hypoxia-mediated decrease in Aldh1a1 gene expression also occurred in the cultured epithelial cells (data not shown).

Site-specific mutagenesis of the rabbit Aldh1a1 promoter.

The activity of the −3519/+43 rabbit Aldh1a1 promoter in transfected rabbit corneal stromal cells was reduced three- to fourfold by simultaneous site-directed mutagenesis of the E-box and XRE1 and XRE2 (Fig. 5C). XRE3 is not present in the −3519/+43 rabbit Aldh1a1 promoter fragment and thus was not tested. Mutagenesis of one or two of the three sites inhibited expression to a lesser degree than mutagenesis of all three sites. Simultaneous site-directed mutagenesis of the E-box and XRE1 and XRE2 reduced the activity of the rabbit Aldh1a1 promoter 1-fold in rabbit corneal epithelial cells and 0.5-fold in the HepG2 cell line. Thus, the three mutations reduced activity of the promoter in all three cell types, but the reduction was greatest in the rabbit corneal stromal cells.

The repression of the promoter by 150 μM CoCl₂ was blocked by mutation of the E-box element and was less severe when only the two XREs were simultaneously mutated (data not shown). Moreover, simultaneous mutagenesis of the two XREs and the E-box eliminated the responsiveness of the promoter to 50 nM TCDD. Thus, these mutagenesis experiments demonstrated the importance of XRE1 and XRE2 and the E-box for the activity of the rabbit Aldh1a1 promoter in corneal stromal cells and for the responses to simulated hypoxia and dioxin.

Expression analyses of ARNT1, AhR, HIF-1α, HIF-2α, and HIF-3α in the cornea.

We next tested, by immunoblotting, for the presence of transcription factors that are known to bind XREs and E-boxes, namely HIF-1α (41), HIF-2α (69), HIF-3α (22), ARNT1 (64, 65), and AhR (64) in the nuclear extracts of rabbit cornea stromal cells (Fig. 6). The expression levels of STRA13 and IPAS could not be measured by immunoblotting due to the lack of specific antibodies. Furthermore, the HIF-3α antibody does not cross-react with IPAS, since it is raised against a peptide of HIF-3α that is not present in IPAS (38). The expression levels of HIF-1α, HIF-2α, and HIF-3α were not detectable by Western blotting in rabbit cornea stromal cells grown under normoxic conditions. Results of subsequent EMSAs, however, suggested that small amounts of HIF-3α were present at 0 and 150 μM concentrations of CoCl₂ (see below) despite its inability to be detected by Western immunoblotting. As expected, treatment of rabbit stromal cells with CoCl₂ resulted in upregulation of HIF-1α, HIF-2α, and HIF-3α. HIF-3α was induced by 75 μM CoCl₂ but curiously was not induced by 150 μM CoCl₂ (see Fig. 6, middle panel). Possibly, IPAS rather than HIF-3α was present at 150 μM CoCl₂, as suggested by the induction of HIF-3α^IPAS transcript by severe hypoxia (38). By contrast, HIF-1α was strongly induced by 150 μM CoCl₂. The difference in inducibility of HIF-1α and HIF-3α may be related to differences in inducibility of HIF-1α and HIF-2α observed by others (73), although HIF-2α was induced at trace levels by 150 μM CoCl₂ (Fig. 6, right panel). The molecular mass of rabbit HIF-3α was 74 kDa, close to the 73 kDa predicted for the mouse protein (22). Rabbit HIF-1α had a molecular mass of 110 kDa, which is close to that of the human protein (120 kDa) (72).

FIG. 6.

Detection of HIF-1α, HIF-2α, and HIF-3α in nuclear extracts of rabbit cornea stromal cells. Immunoblots of nuclear extracts from primary cell cultures of rabbit cornea stromal cells using goat polyclonal anti-human antibody for HIF-1α, HIF-2α, or HIF-3α protein. Equal loading was confirmed by staining with the SYPRO Ruby protein blot stain and visualizing with UV illumination (not shown).

The levels of ARNT1 and AhR detectable by immunoblotting were low in the nuclear extracts and were unchanged by treatment with CoCl₂ (data not shown). Possibly, the poor detection of ARNT1 and AhR was due to impaired antibody species specificity. ARNT1 has been detected by others in the corneas of rats, although AhR was not (6). Although our immunoblotting gave variable quantitative results, amplicons of Arnt1, AhR, Hif-1α, Hif-3α, Stra13, and Hif-3α^IPAS were obtained by RT-PCR and sequenced from rabbit corneal RNA, indicating expression of these genes in the cornea (data not shown).

Cotransfection of rabbit Aldh1a1 promoter constructs with ARNT1, AhR, IPAS, HIF-1α, HIF-2α, HIF-3α, and STRA13 expression plasmids.

In functional tests for XRE and E-box activity, the rabbit Aldh1a1 −3519/+43 promoter construct was cotransfected with expression plasmids pARNT/CMV4 and pAhR/CMV4 into various cells. pARNT/CMV4 stimulated rabbit Aldh1a1 promoter activity an average of fourfold in primary cultures of transfected rabbit corneal stromal cells (Fig. 7A) and sixfold in the transfected HepG2 cell line (data not shown). pAhR/CMV4 stimulated rabbit Aldh1a1 promoter activity an average of 2.5-fold in the cotransfected HepG2 cell line (data not shown) but did not stimulate the promoter in the cotransfected stromal cells (Fig. 7A). Rabbit Aldh1a1 promoter activity was increased an average of 12-fold in cotransfection tests with pARNT/CMV4 and 5-fold with pAhR/CMV4 in HepG2 cells treated with TCDD (data not shown), consistent with the dioxin inducibility of this promoter. Stimulation by either pARNT/CMV4 or pAhR/CMV4 was not observed with use of the rabbit Aldh1a1 promoter construct with point mutations introduced simultaneously in the E-box, XRE1, and XRE2 binding sites (data not shown).

FIG. 7.

Cotransfection of rabbit Aldh1a1 promoter constructs with AHR, ARNT1, HIF-3α, HIF-2α, IPAS, HIF-1α, and STRA13 expression plasmids. (A) Cotransfection of the rabbit cornea stromal cells with the −3519/+43 rabbit Aldh1a1 promoter construct with 50 ng of pCMV/AhR, 50 ng of pCMV/ARNT1, or 50 ng of pCMV/HIF-3α. Results are the averages from eight transfection experiments. The increases in activity with cotransfection with pCMV/ARNT1 (P < 0.05) or pCMV/HIF-3α (P < 0.02) are statistically significant. (B) Cotransfection of rabbit cornea stromal cells with the −3519/+43 rabbit Aldh1a1 promoter construct with 50 ng of phEP-1 (HIF2-α expression plasmid), 50 ng of pcIPAS, 50 ng of pcHIF-1α, or 50 ng of pcSTRA13. Results are the averages for eight transfection experiments. The decrease in activity with cotransfection with pcHIF-1α (P < 0.05) or pcSTRA13 (P < 0.02) are statistically significant. (C) Cotransfection of rabbit cornea stromal cells with 50 ng of pcSTRA13 and the −3519/+43 rabbit Aldh1a1 promoter construct containing point mutations in XRE1, XRE2, or the E-box. The mutated sites in the different constructs are crossed out. Results are the averages from six transfection experiments.

Further evidence for involvement of XRE1, XRE2, and the E-box in Aldh1a1 promoter activity was obtained by cotransfection with the known hypoxic response transcription factors, HIF-1α (59), EPAS1 (also called HIF-2α) (69), HIF-3α (22), IPAS (37), and STRA13 (also called DEC1) (42). Rabbit Aldh1a1 −3519/+43 promoter activity decreased onefold in rabbit cornea stromal cells cotransfected with pcDNA/HIF1-α but was not affected by cotransfection with phEP-1, the HIF-2α expression plasmid (Fig. 7B). This repression by HIF-1α is consistent with 150 μM CoCl₂ causing an increase in HIF-1α (Fig. 6) and a repression of rabbit Aldh1a1 expression (Fig. 5A). The IPAS expression construct (pcDNA/IPAS) also inhibited the rabbit Aldh1a1 promoter in cotransfection experiments (Fig. 7B). Note that the HIF-3α expression construct stimulated rabbit Aldh1a1 promoter activity an average of 7.2-fold (Fig. 7A), although the inhibitory IPAS is an alternatively spliced variant of HIF-3α. It is also noteworthy that HIF-3α is induced by 75 μM CoCl₂ (Fig. 6), a concentration which does not repress the Aldh1a1 promoter (Fig. 5A), consistent with a stimulatory effect of HIF-3α on the Aldh1a1 promoter. Cotransfection with pcDNA3-STRA13 caused an average threefold repression of the rabbit promoter (Fig. 7B). Simultaneous point mutations introduced in the E-box, XRE1, and XRE2 binding sites, which caused a threefold decrease in luciferase expression in transfected rabbit stromal cells (Fig. 5C), eliminated the decrease in Aldh1a1 promoter activity observed by cotransfection with pcDNA/HIF1-α (data not shown). Moreover, HIF-3α did not activate the mutated Aldh1a1 promoter (data not shown). Mutation of the E-box, however, relieved the repression caused by cotransfection with pcDNA3-STRA13 (Fig. 7C).

EMSAs using the E-box and XREs from the rabbit Aldh1a1 promoter.

Labeled oligonucleotides containing the E-box or XRE2 binding sites formed DNA-protein complexes in EMSAs using nuclear extracts from rabbit stromal cells and from HepG2 cells (Fig. 8). The XRE2 oligonucleotide formed a large complex with HepG2 extracts that was the same size as that formed with an oligonucleotide containing the XRE from the human UGT1A1 gene (77) (data not shown) and that was more intense when the HepG2 cells were stimulated with dioxin before obtaining the nuclear extract (Fig. 8A). The XRE2 oligonucleotide formed a complex with nuclear extracts from rabbit corneal stromal cells only when they were treated with CoCl₂ (Fig. 8A). Neither untreated nor TCDD-treated rabbit corneal stromal cell extract resulted in the formation of a complex (Fig. 8A). The complex formed by XRE2 with nuclear extracts from rabbit corneal stromal cells treated with CoCl₂ was smaller than that formed with the HepG2 extracts. It is noteworthy that the molecular mass of rabbit HIF-3α (74 kDa) (Fig. 6) is smaller than that of rabbit HIF-1α (110 kDa) (Fig. 6), rabbit HIF-2α (also 110 kDa), and human AhR (106 kDa) (49). Protein complexes that bind XREs and HREs are heterodimers of AhR/ARNT and HIF/ARNT, respectively (23). Both human and rabbit ARNT have a molecular mass of 87 kDa (25, 67). Thus, we suspected that the smaller complex induced by CoCl₂ might contain HIF-3α rather than HIF-1α or HIF-2α. This possibility was supported by the fact that this complex was most intense when derived from cells treated with 75 μM CoCl₂ (data not shown), which is consistent with the induction of HIF-3α in rabbit corneal stromal cells (Fig. 6).

FIG. 8.

Gel retardation assays using the XRE and E-box elements of rabbit Aldh1a1. (A) Complexes formed on XRE2 using either HepG2 nuclear extracts or rabbit corneal stromal cell extracts. HepG2 extracts were prepared with or without TCDD, and rabbit stromal cell extracts were prepared with or without CoCl₂. Arrows show complexes formed. The complex formed by XRE2 is competed with oligonucleotides containing wild-type XRE1, XRE2, or XRE3. The complex is not disrupted by competition with a mutated XRE2 site or with the wild-type E-box. Complexes formed with XRE2 can be disrupted by incubation with a goat polyclonal antibody for human HIF-3α but not with a goat polyclonal antibody for human HIF-1α. (B) Complexes formed on the E-box element using rabbit corneal stromal cell extracts. Rabbit stromal cell extracts were prepared with or without TCDD and CoCl₂. The position of the complex formed by XRE2 and the rabbit stromal cell extract is noted with an asterisk. The protein complex formed by binding to the E-box is not disrupted by competition with oligonucleotides containing the wild-type XRE2 or an oligonucleotide containing a mutated E-box.

The XRE2 complex from the corneal stromal cells was competed by unlabeled oligonucleotides containing XRE1, XRE2, or XRE3 but not by a mutated XRE2 oligonucleotide or by the E-box (Fig. 8A). Indeed, the E-box oligonucleotide formed a faster-migrating complex than that formed by the XRE2 oligonucleotide using rabbit stromal cell extracts (Fig. 8B). The complex formed by the E-box was competed by itself but not by the mutated E-box oligonucleotide or by wild-type XRE2 (Fig. 8B). Unlike the complex formed by XRE2, the E-box complex was not induced by CoCl₂ or dioxin (Fig. 8B).

We attempted to identify the proteins bound to the E-box and XREs by performing super-shift assays using antibodies for HIF-1α, HIF-2α, HIF-3α, and ARNT1. The antibodies for HIF-1α (Fig. 8A, right panel) and HIF-2α (data not shown) did not affect the protein complexes formed with the XRE2 oligonucleotide. The HIF-3α antibody disrupted the complex formed by XRE2 using extracts from stromal cells stimulated with CoCl₂ (Fig. 8A). This disruption by HIF-3α antibody of an XRE-bound complex from CoCl₂-treated rabbit corneal stromal nuclear extract is similar to experiments by others showing the disruption of an XRE-bound complex from dioxin-treated HepG2 cells by ARNT and AhR antibodies (77). Because HIF-3α should form a heterodimer with an ARNT family member (23), we attempted the supershift experiment with the human ARNT1 antibody. This antibody failed to disrupt the complex (data not shown), although we cannot rule out a species specificity problem with the ARNT1 antibody. The smaller complex formed using the E-box oligonucleotide and rabbit stromal cell nuclear extracts was not affected by addition of any of the antibodies used (data not shown).

DISCUSSION

We confirm (29) here that the level of ALDH1A1 is about five times higher in the stromal cells than in the epithelial cells of the cornea, although the epithelial cells also have substantial amounts of the enzyme. It is not known whether ALDH1A1 in the rabbit cornea fulfills precisely the same function or functions as ALDH3A1 in other mammals. ALDH1A1 protein expression increases in rabbit corneal stromal cells up to 16 days after birth and is associated with a developmental loss in corneal haze (Jester et al., ARVO 2003). Our northern and Western blotting tests showed, as expected, that rabbit ALDH1A1 is expressed in many tissues throughout the body and is at least five times more prevalent in the corneal stromal cells than in the tissue with the next-highest prevalence, the lung.

In the present investigation, a single Aldh1a1 cDNA has been isolated from a library constructed from rabbit corneal epithelial tissue, consistent with our Southern blot hybridization of genomic DNA, suggesting that rabbit Aldh1a1 is encoded by a single-copy gene (see “Isolation and characterization of cDNA clones of rabbit Aldh1a1”). This agrees with the results of others (40). However, there are two ALDH1 genes in the elephant shrew, where one (η-crystallin) is expressed highly in the eye, and the other (ALDH1-nl) is expressed in other tissues (20). As in mice and humans, no evidence was obtained for the existence of a homologue of rat Aldh-pb in the rabbit (14). Rat Aldh-pb is expressed at a low level in the liver and is induced by phenobarbital (30). It thus appears that there is but one differentially regulated rabbit Aldh1a1 gene.

5′ RACE analysis of Aldh1a1 cDNAs indicated that the same promoter is utilized for high Aldh1a1 gene expression in the cornea and lower expression in the lung. We speculate from these data that the other tissue-specific differences in rabbit Aldh1a1 gene expression are directed from the same promoter; however, this remains to be proved, and the particular regulatory elements involved require further investigation. This differs from the case of chicken Aldh1a1, which employs a different transcriptional start site and undergoes alternative RNA splicing in the retina and the liver (16). While one case of alternative promoter use is known for the high expression of crystallins in the lens (ζ-crystallin in the lens and liver of the guinea pig), high expression of crystallin genes in the lens is generally directed by the same promoter that is used for lower expression in other tissues (13).

The sequence analysis of the rabbit Aldh1a1 5′ flanking sequence revealed greater similarity to the human homolog than to rodent homologs. This agrees with recent data indicating a closer phylogenetic relationship of rabbits with humans than with mice (11, 21). The expression pattern of ALDHs between humans and rabbits is also consistent with a closer evolutionary link of humans with rabbits than rodents. Humans express much less corneal ALDH3A1 (5% rather than ∼50% of the water-soluble protein) but more corneal ALDH1A1 (3%) than do mice (trace amounts at best) (31). The fact that rabbit ALDH1A1 comprises 16.1% of the total water-soluble protein of the corneal stroma is also consistent with our present finding of greater activity of the rabbit Aldh1a1 promoter than of the human Aldh1a1 promoter in the transfected corneal stromal cells. The low activity of the rabbit Aldh1a1 promoter in transfected noncorneal cells and the preferential rabbit Aldh1a1 promoter activity in the cornea in transgenic mice, generally (but not exactly) resembling the in vivo expression pattern of the endogenous Aldh1a1 gene, give confidence that the promoter sequences investigated in the present study are at least partly responsible for the characteristically high level of corneal expression of the rabbit Aldh1a1 gene.

The present site-specific mutagenesis and transfection tests implicate at least two XREs and an E-box for preferential corneal promoter activity of the rabbit Aldh1a1 gene. Comparable XREs and E-box sequences are well represented in many other genes that are expressed highly in mammalian corneas. The XRE and E-box involvement in rabbit Aldh1a1 promoter activity is consistent with the idea that environmental induction plays a key role in preferential gene expression in the cornea. The bHLH-PAS transcription factors which bind XREs and E-boxes mediate several environmental responses to hypoxia, circadian entrainment, and xenobiotics (23). That both mouse Tkt (55, 57) and Aldh3a1 (12) gene expression in the corneal epithelium are elevated at eye opening 2 weeks after birth supports an environmental role in the promoter activity of these genes. Moreover, rearing the newborns in constant darkness through the eye-opening stages of development delays the increase in the expression of these genes in the cornea (12, 55). In addition, the loss of Aldh3a1 expression in cultured rat epithelial cells is prevented by constant illumination (5), and corneal mimecan gene expression is stimulated by UV light via an E-box in the promoter (68).

That rabbit Aldh1a1 promoter activity is stimulated by dioxin and decreased by 150 μM CoCl₂ is also consistent with an inductive contribution to preferential activity in the cornea. There is a parallel between dioxin and hypoxic regulation of mouse Aldh3a1 (5) and rabbit Aldh1a1 (present study): neither is simulated by dioxin in the corneal cells, where they are presumably already maximally active, but both are dioxin inducible in liver cells and both are down-regulated by CoCl₂-simulated hypoxia in liver and corneal cells. Responses to both dioxin and hypoxia involve heterodimers of PAS domain-containing basic helix-loop-helix (PAS-bHLH) transcription factors (8). We show here that ARNT1 stimulates rabbit Aldh1a1 promoter activity, although we were unable to identify ARNT1 as the heterodimerization partner of HIF-3α in the rabbit corneal stromal cells. The E-box in the rabbit Aldh1a1 promoter may also bind homodimers of ARNT (65), heterodimers of ARNT/HIF1-α (41), and heterodimers of CLOCK and ARNT-like protein 1 (BMAL1) (15) or other bHLH proteins, such as upstream stimulatory factors (USF1 and USF2) (54) and the MYC/MAX/MAD transcription factors (60). In the rabbit Cyp1a1 promoter, XREs activated by AhR/ARNT heterodimers are displaced by USF1 homodimers (66). Our mutagenesis tests implicate the E-box for rabbit Aldh1a1 promoter activity; however, further investigations are necessary to establish the trans factors responsible.

STRA13/DEC1/SHARP-2, a bHLH repressor protein, may be involved in repressing the rabbit Aldh1a1 promoter during simulated hypoxia (150 μM CoCl₂) by binding an E-box sequence (61). The STRA13 promoter is transcriptionally activated by ARNT/HIF1-α heterodimers in response to hypoxia (42). STRA13 is also induced by serum starvation (63) and by retinoic acid (7), suggesting a role in feedback regulation of rabbit Aldh1a1, which is a retinoic acid biosynthetic enzyme. Transforming growth factor beta also induces STRA13 (78) and synergizes with hypoxia to activate expression of VEGF (53), an angiogenic growth factor. Transforming growth factor beta is known to induce the differentiation of corneal stromal cells into repair myofibroblasts (28) during corneal wound healing, and rabbit ALDH1A1 is also down-regulated during this cellular differentiation (62). IPAS, encoded by a hypoxia-inducible alternate transcript of the HIF-3α gene, is still another PAS domain protein that suppresses corneal angiogenesis by hypoxia (38) and which inhibited the rabbit Aldh1a1 promoter in the present transfection experiments. The dioxin signaling pathway may play an interconnected role with hypoxia in the cornea inasmuch as AhR null mice have excess vascularization in the peripheral cornea (33).

Taken together, we propose that the hypoxia-related pathways diagrammed in Fig. 9 contribute to the expression of the Aldh1a1 gene in the rabbit cornea. Under conditions of moderate hypoxia, such as might be expected in the nonvascularized cornea (especially when the eyelids are closed), HIF-3α/ARNT may be responsible for maintaining a high level of Aldh1a1 expression. Similarly, AhR/ARNT may also contribute to high activity of the Aldh1a1 promoter (not shown in Fig. 9). This AhR-related stimulation of Aldh1a1 gene expression in the cornea may be utilized as a detoxification mechanism for xenobiotic irritation of the environmentally exposed cornea. It is noteworthy that HIF-1α, whose transcriptional activity is inhibited by HIF-3α (24), is induced by severe hypoxia and suppresses Aldh1a1 promoter activity via activation of STRA13 (Fig. 9). This suppression may be overcome, however, by production of IPAS, a repressor of HIF-1α, from an alternate RNA transcript of HIF-3α. Since XREs and E-boxes are also enriched in promoters of other genes that are expressed highly in mammalian corneas, we suggest that similar pathways may be used for preferential gene expression in the cornea. The functional importance of XREs and E-boxes binding various bHLH-PAS transcription factors for the control of gene expression in the cornea is reasonable in view of the expected variability of oxygen concentration in this nonvascularized organ. Finally, a stress-connected mode of preferential gene expression in the cornea provides one more argument supporting a link between the lens and its stress protein-related crystallins and the cornea and its abundant cytoplasmic proteins (3, 9, 46, 47, 56, 74).

FIG. 9.

Model of bHLH-PAS regulation of rabbit Aldh1a1.