Identification and Characterization of a Novel Tissue-specific Transcriptional Activating Element in the 5′-Flanking Region of theCYP2A3 Gene Predominantly Expressed in Rat Olfactory Mucosa*

CYP2A3 is expressed preferentially in rat olfactory mucosa and is believed to play important roles in maintaining cellular homeostasis in the chemosensory tissue. DNase I footprinting analysis revealed a single protected region in the proximal promoter of the CYP2A3 gene with nuclear extracts from olfactory mucosa, but not from liver, lung, kidney, or brain. The core sequence of the binding site, named the nasal predominant transcriptional activating (NPTA) element, is similar to that of nuclear factor 1, but it interacted with unique proteins detected only in the olfactory mucosa in electrophoretic mobility shift assays or on Southwestern blots. The NPTA element is conserved in ratCYP2A3, mouse Cyp2a5, and humanCYP2A6 genes and was found to be essential for transcriptional activity of the CYP2A3 promoter in in vitro transcription assays. NPTA-binding proteins were detectable at day 1 and were much more abundant at day 8 than at day 60 after birth. Furthermore, their levels decreased dramatically during chemically induced degeneration of the olfactory epithelium, paralleling the disappearance of CYP2A3 protein, and rebounded to higher than pretreatment levels during recovery. Thus, we have identified a novel transcriptional activation element potentially responsible for the olfactory mucosa-predominant expression of theCYP2A3 gene in rats and orthologous genes in mice and humans.

Numerous genes are uniquely or selectively expressed in the olfactory mucosa, such as the odorant receptors (1) and other components of the odorant signal transduction cascade (cf. Ref. 2), OMP 1 (3), the odorant-binding proteins (4), olfactomedin (5), UDP-glucuronosyl transferase (6), and two P450 isoforms, CYP2G1 (7,8) and CYP2A3 (9,10). Although not all of these genes are well characterized with respect to biological functions, their unique tissue-and cell-specific expression suggests functional importance in the olfactory epithelium, a tissue that is highly specialized for molecular recognition. However, except for a few recent studies (2,(11)(12)(13), little is known about the mechanisms regulating the tissue and cell type-specific expression of these genes.
P450 represents a superfamily of genes involved in the biotransformation and disposition of numerous endogenous and exogenous compounds in the body (14). Many different P450 isoforms are expressed at high levels in the olfactory mucosa in mammals (10,(15)(16)(17) and are believed to play important roles in maintaining cellular homeostasis in a tissue that is constantly exposed to high levels of air-borne foreign chemicals as well as endogenous substances (18).
CYP2A3 is a major P450 isoform in rat olfactory mucosa and, except for trace levels in the lung, has not been detected in any other tissues examined (9,10). Orthologous CYP2A isoforms are also expressed predominantly in the nasal mucosa in other species including mice, rabbits, and cows (10, 19 -21) and the human ortholog, CYP2A6, has also been detected in the olfactory mucosa (10,22). The onset of expression of the nasal CYP2As is detectable before birth in rabbits, rodents, and humans 2 and was found to increase rapidly after birth in rabbits (23). Immunohistochemical studies of rat and mouse olfactory epithelium indicated that CYP2A expression is restricted to the sustentacular and Bowman's gland cells of the chemosensory mucosa but is not detectable in the neuronal cells (24 -26). Studies with purified or heterologously expressed nasal CYP2As as well as microsomal studies indicated that, collectively, these isoforms are active in the biotransformation of odorants, such as coumarin and butanol, and endogenous compounds, such as testosterone, as well as many nasal toxicants (cf. Refs. 16 and 27).
CYP2A expression during olfactory neurodegeneration and subsequent regeneration has also been studied. An early report indicated that unilateral naris closure in mice results in loss of olfactory neurons in the open side and a reduction in CYP2A and CYP2G1 immunoreactivity in the nonneuronal cells (26). Loss of olfactory epithelial P450 expression was also found in bulbectomized mice. Similar observations were made in a methyl bromide-induced olfactory degeneration model in rats (28). More recent studies indicated that reduction of P450 expression also occurred during olfactory neurodegeneration induced by treatment of rodents with dichlobenil or coumarin (29,30).
The genomic structure, including the 5Ј-proximal regulatory region, of rat CYP2A3 (31), mouse Cyp2a5 (32), and human CYP2A6 genes (33) has been determined. To understand the mechanisms that dictate tissue-selective expression of the nasal CYP2A genes, we have examined the 5Ј-flanking region of the CYP2A3 gene in the present study using in vitro methods with nuclear extracts from rat and mouse tissues. Initial DNase I footprinting analysis revealed a single protected region in the proximal promoter sequence with nuclear extracts from olfactory mucosa, but not from liver, lung, kidney, or brain. The tissue-specific binding was confirmed by gel shift analysis, which revealed multiple bands. The core sequence of the binding site, named the NPTA (nasal predominant transcriptional activating) element, was determined and found to be essential for transcriptional activity of the CYP2A3 promoter in vitro. The developmental expression and olfactory epithelial degeneration-associated alterations in NPTA-binding proteins were also examined by gel shift analysis, and the nature of the NPTA-binding proteins was probed using supershift assays and Southwestern blots. Our results strongly indicate the presence of a novel transcriptional activation element that may be responsible for the olfactory mucosapredominant expression of the CYP2A3 gene in rats and orthologous genes in mice and humans.

EXPERIMENTAL PROCEDURES
Animals-All animals were obtained from breeding stocks maintained at the Wadsworth Center. Tissues from male and female Wistar rats at ages of between 1 and 60 days and male C57BL/6 mice at the age of about 2 months were used for preparation of nuclear extracts. To induce degeneration of the olfactory epithelium, 60-day-old male rats (200 -250 g) were treated with coumarin (Sigma; 50 mg/kg body weight, injected once, intraperitoneally, in 0.1 ml of dimethyl sulfoxide) as recently described (30). Animals were sacrificed at 0, 3, or 48 h or 7 days following the injection, and tissues from four rats per group were combined for preparation of nuclear extracts.
Preparation of Nuclear Extracts-The protocol of Dignam et al. (34), modified by Kudrycki et al. (11), was followed for the preparation of crude nuclear extracts from various tissues of rats and mice. Tissues from at least four rats or eight mice were pooled for each preparation. Nuclear extracts prepared using this protocol are referred to as "crude nuclear extracts" to distinguish from the more enriched "nuclear protein extracts." Transcriptionally active nuclear protein extracts of rat olfactory mucosa were used for in vitro transcription assays and were prepared essentially according to the method of Gorski et al. (35), modified by Hattori et al. (36) using sucrose density gradient and ammonium sulfate fractionation. All preparations were stored in small aliquots at Ϫ85°C.
Electrophoretic Mobility Shift Assays-Assays were performed essentially as described elsewhere (37) with minor modifications. Binding reaction mixtures, which were preincubated at room temperature for 10 min, contained 15-20 g of crude nuclear extracts in 10 mM Tris-HCl buffer, pH 7.9, 50 mM NaCl, 1 mM dithiothreitol, 1 mM EDTA, 8% glycerol, and 3-5 g of sonicated salmon sperm DNA (Stratagene) or 5-10 g of poly(dI-dC) (Amersham Pharmacia Biotech) and, for competition experiments, unlabeled competitor oligonucleotide (at 50 -200fold excess). Labeled probes (about 1 ng and 10,000 cpm) were added to the mixtures to a final volume of 20 l and incubated for another 20 min at room temperature. The DNA-protein complexes and unbound probes were separated by electrophoresis through 6% polyacrylamide gels and detected by autoradiography. For supershift experiments, up to 7 l of a control rabbit serum or a polyclonal rabbit anti-CTF1 serum (38) was incubated with the nuclear extracts for 1 h at room temperature before the addition of the labeled probe (2A3pm1).
DNase I Footprinting Analyses-DNase I footprinting analyses were performed with the Core Footprinting System from Promega, according to the manufacturer's instructions. DNA-protein binding reactions were performed as described above for the gel shift assays, except that the reaction mixtures contained 40 -60 g of crude nuclear extracts and labeled DNA fragments (about 2 ng and 40,000 cpm) in a final volume of 50 l. The binding reaction mixtures were incubated on ice for 20 min before mixing with 50 l of the DNase reaction buffer (10 mM MgCl 2 and 5 mM CaCl 2 ) and incubated at room temperature for an additional minute. DNase I (0.1-1 unit) was then added, and the DNase digestion was allowed to proceed for 60 -90 s at room temperature. The reactions were terminated by the addition of 100 l of a DNase stop solution (200 mM NaCl, 30 mM EDTA, 1% SDS, and 100 g/ml yeast RNA). DNA was recovered from the reaction mixture by phenol/chloroform extraction and ethanol precipitation, and the washed pellets were air-dried and resuspended in 5 l of a gel loading buffer (95% (v/v) deionized formamide, 50 mM Tris borate (pH 8.3), 1 mM EDTA, and 0.1% (w/v) each of xylene cyanol and bromphenol blue). The samples were denatured by heating at 90°C for 3 min and analyzed by electrophoresis through 6% polyacrylamide, 7 M urea DNA sequencing gels. DNA sequence ladders were produced by the method of Maxam and Gilbert (39).
Southwestern Blotting Assays-Southwestern blotting was performed essentially as described by Abe and Kufe (40), with minor modification. Crude nuclear extracts were submitted to SDS-polyacrylamide gel (10%) electrophoresis using the Laemmli system (41). The resolved proteins were transferred electrophoretically to nitrocellulose membranes (42) and renatured by incubating with 100 ml of 5% (w/v) nonfat dry milk in 10 mM Tris-HCl (pH 7.9) at room temperature for 3 h. To detect specific DNA-binding proteins, the sheets were first incubated in about 5 ml of a binding buffer (10 mM Tris-HCl (pH 7.9), 50 mM NaCl, 10 mM MgCl 2, 0.1 mM EDTA, 2 mM dithiothreitol, and 15 g of poly(dI-dC) or 5 g of salmon sperm DNA per ml) and, for competition experiments only, unlabeled oligonucleotide probes (at a 200-fold excess) at room temperature for 30 min to block nonspecific binding. A 32 P-labeled oligonucleotide probe was then added at 20,000 cpm/ml (approximately 2 ng/ml), and the incubation was continued for another 2 h. The blots were washed twice (5 min each) with a wash buffer (same as the binding buffer, but containing 100 mM NaCl), and the DNA-protein complexes were visualized by autoradiography. The sizes of the DNA-binding proteins were estimated using prestained molecular weight markers from Bio-Rad (low range).
In Vitro Transcription Assays-In vitro transcription was performed essentially as described elsewhere (43,44) with minor modifications. Reaction mixtures contained 60 g of nuclear protein extract, ϳ1 g of a plasmid DNA template (in H 2 O), 1ϫ reaction buffer (10 mM HEPES-KOH (pH 7.6), 25 mM KCl, 6 mM MgCl 2, 3% glycerol, 0.6 mM each ATP, CTP, GTP, and UTP), and 1 l (40 units) of RNasin (Promega) in a total volume of 24 l. The reactions were conducted at 30°C for 60 min and were terminated by adding 400 l of a stop solution (0.5% SDS, 0.3 M sodium acetate (pH 5.0), 50 g/ml yeast tRNA, and 100 g/ml proteinase K) and incubating at 37°C for 20 min. The reaction mixtures were extracted with phenol/chloroform, and the RNA products were precipitated with ethanol. Pelleted RNAs were resuspended in 35 l of a hybridization buffer (42 mM PIPES, 420 mM NaCl, 53% formamide, and 1.1 mM EDTA) and hybridized with 5 l of a 32 P-labeled antisense oligonucleotide probe (20 ng/l) for detecting specific transcripts using an S1 nuclease protection assay. The mixture was added to 350 l of S1 reaction solution (150 mM sodium acetate (pH 8.0), 1 M NaCl, and 15 mM ZnSO 4 ) containing 300 units of S1 nuclease (Promega), and digestion was allowed to proceed at 30°C for 30 min, followed by the addition of 70 l of S1 stop buffer (1.8 M Tris-HCl (pH 8.0), 20 mM EDTA, and 150 g/ml yeast tRNA). The RNA-DNA heteroduplexes and DNA size markers (see below) were denatured; resolved on 10% polyacrylamide, 7 M urea gels; and detected by autoradiography.
Plasmids used as templates in in vitro transcription assays were constructed in pCR-Script SK AMP vector (Stratagene) using DNA fragments derived from a CYP2A3 genomic clone (31). Plasmid p2A3-1900 was constructed by inserting a SacI (Ϫ1900)/EcoRI (ϩ582) fragment of the CYP2A3 gene into the unique SacI/EcoRI sites in the pCR-Script vector. Plasmid p2A3-254 was constructed by inserting the BamHI (Ϫ254)/BglI (ϩ58) fragment of the CYP2A3 gene into the SrfI site of the pCR-Script vector following treatment with Pfu DNA polymerase (Stratagene) to generate blunt ends. Plasmid p2A3-254/M was made by site-directed mutation of p2A3-254 to convert a TG dinucleotide (Ϫ127 and Ϫ126) to a GT, using the Transformer site-directed mutagenesis kit from CLONTECH. The mutagenesis experiment was performed according to the protocol supplied by the manufacturer, using two mismatched oligonucleotide primers: one containing the intended sites of mutation in CYP2A3 proximal promoter (5Ј-CCTCCTT-GAGTGTGTGCTATGTCCCAAACTAGG-3Ј) and the other containing a mutation in the unique EcoRI site in the vector (5Ј-TTGATATCGAA-GTCCTGCAGCC-3Ј). A control plasmid, pOMP-280, was constructed by inserting a Ϫ280 to ϩ105 fragment derived from the rat OMP gene into the SrfI site of the pCR-Script vector. This genomic fragment was obtained by PCR using rat genomic DNA as template and primers designed according to published sequences (45): 5Ј-CCTGTGGAGTAT-TCTGTTCTTCAC-3Ј (upstream) and 5Ј-AGAACCAGCGGCATATCCA-GCT-3Ј (downstream). The structures of all promoter constructs were verified by sequencing.
The oligonucleotide probe used for detecting CYP2A3 transcripts in S1 nuclease protection assays is complementary to positions ϩ1 to ϩ52 of the CYP2A3 gene, and the S1 probe for OMP transcripts is complementary to ϩ1 to ϩ41 of the OMP gene. The sizes of the protected fragments detected in S1 nuclease assays were estimated from the positions of 32 P-labeled oligonucleotide fragments analyzed on the same gels. The 30-and 46-nucleotide oligonucleotides were chemically syn-thesized, while the 100-nucleotide fragment was gel-purified from the 100-base pair DNA size marker from Life Technologies, Inc.
Preparation of Oligonucleotide and cDNA Probes-All oligonucleotides were synthesized by the Molecular Genetics Core of the Wadsworth Center. To prepare 32 P-labeled double-stranded oligonucleotide probes for gel shift assays, one strand of the synthetic oligonucleotide was labeled at the 5Ј-end with T4 polynucleotide kinase (Boehringer Mannheim) and [␥-32 P]ATP (6000 Ci/mmol, NEN Life Science Products) and then annealed to a 5-fold excess of unlabeled complementary strand. The resulting labeled double-stranded DNA probe was separated from unincorporated nucleotides using a Sephadex G-25 spin column (Amersham Pharmacia Biotech) and from the remaining singlestranded DNAs by gel purification with a 15% polyacrylamide, 7 M urea gel; the probe was recovered from gel using a Micropure separator from Amicon (Beverly, MA). Single-stranded DNA probes for S1 nuclease protection assays were labeled with 32 P and purified in the same manner. Unlabeled double-stranded oligonucleotides used as competitors were prepared by annealing equimolar amounts of the two complementary strands. 32 P-Labeled, double-stranded cDNA probes for DNase I footprinting experiments were prepared by PCR with one 5Ј-end-labeled primer and one unlabeled primer of the complementary strand and using the p2A3-254 plasmid (see above) as template. This ensures that the resulting cDNA probe is labeled on only one strand. The primers used for labeling the coding strand were 5Ј-32 P-CTGTCCTCTGTAATGCATAGTT-3Ј (Ϫ231 to Ϫ210) and M13 reverse primer in the multiple cloning region of the vector, and the primers for labeling the noncoding strand were 5Ј-32 P-GCAACTGTCTAATTTTA-3Ј (Ϫ67 to Ϫ83) and the KS primer complementary to vector sequence. Briefly, a 10-l reaction mixture containing 10 pmol of one PCR primer, 2 pmol of [␥-32 P]ATP, 1ϫ kinase buffer, and 8 units of T4 polynucleotide kinase was added to a 0.5-ml PCR reaction tube and incubated at 37°C for 60 min. The reaction was quenched by heating at 70°C for 10 min. The remaining components of the PCR reaction were then added in 40 l, containing 100 ng of template DNA, 20 pmol of the second PCR primer, and the other components in standard amounts. PCRs were performed in a Perkin-Elmer 9600 thermal cycler for 30 cycles of denaturing at 94°C for 15 s and annealing and extension at 68°C for 20 s, followed by a final extension at 72°C for 15 min. The labeled PCR product was gel-purified before use.
Other Methods and Materials-DNA sequence analyses were performed by the Molecular genetics Core of the Wadsworth Center using an automated sequencer from Applied Biosystems (model 373A, Foster City, CA). Plasmid DNAs were prepared using QIAprep Spin Plasmid kit, and purification of DNA fragments from agarose gels was conducted using QIAquick or QIAEX II agarose gel extraction kit (QIAGEN). Protein concentration in nuclear extracts was determined using BCA protein assay reagent (Pierce). Unlabeled DNAs were quantitated spectrally by absorbance at 260 nm. The concentrations of 32 P-labeled oligonucleotide probes were estimated based on the amounts of starting material in the labeling reactions, whereas the concentration of 32 Plabeled PCR products for footprinting experiments was estimated from the band intensity on ethidium bromide-stained agarose gels using known amounts of DNA as standards.

RESULTS
The proximal promoter of the CYP2A3 gene was initially analyzed by DNase I footprinting assay (Fig. 1). As shown in panel A, a single protected region was detected on the noncoding strand (Ϫ134 to Ϫ112) with nuclear extracts from rat olfactory mucosa, but not from lung, kidney, brain, and liver. Protein binding to this site was competed by an unlabeled oligonucleotide probe, 2A3pm1 (Ϫ140 to Ϫ108), that comprises the protected sequence. Identical results were also obtained with nuclear extracts from mice (panel B). With a probe labeled on the coding strand, a similar region (Ϫ129 to Ϫ108) was protected by olfactory mucosal nuclear extracts from either rats (panel C) or mice (data not shown).
The protected site, which is named the NPTA element, is about 80 bp upstream of a TATA box (Fig. 1D). A comparison of the NPTA element with those included in TFMATRIX transcription factor binding site profile data base (46) revealed that the binding site contained an NF-1-like sequence (47) (score of 87.3%) and an IK-2 (48) and a LYF-1 element (49) in reverse orientation. However, competition experiments with these ele-ments as well as with the binding sites for OLF-1 or upstream binding element (UBE) of the rat OMP gene (11) revealed that only NF-1 and UBE (which also contain an NF-1-like sequence) competed for binding to the NPTA element by olfactory epithelial nuclear proteins, and with the NF-1 and UBE sequences, only partial competition was observed ( Fig. 2 and Table I). Thus, the NPTA site is not likely to interact with OLF-1, IK-2, or LYF-1 factors, but it may bind NF-1 type transcription factors in vivo. However, the partial competition by the NF-1 and UBE sequences also suggests that the olfactory epithelial NPTA-binding proteins may be unique despite the apparent homology of the NPTA element with NF-1.
Multiple bands, representing putative NPTA factors, were detected in gel shift assays with nuclear extracts from olfactory mucosa (Fig. 2). To investigate whether the NPTA factors are members of the NF-1 family, a polyclonal anti-NF-1 antibody, named anti-CTF1 (38), which has been used in a number of studies to detect NF-1-related proteins (50 -54), was used in supershift assays in which olfactory epithelial nuclear extracts were exposed to the antibody prior to incubation with labeled 2A3pm1 probe. As shown in Fig. 2C, and in additional experiments not presented, the addition of anti-NF-1 led to the appearance of a weak supershift band as well as a strong well shift, which was accompanied by decreases in the intensity of the three higher bands, with the middle band (indicated by an arrowhead) barely detectable. However, the lowest, yet most prominent band was not affected. This indicates that proteins in the top three bands may be structurally related to NF-1, whereas the protein in the most predominant complex may be unique.
Using a panel of mutant or truncated binding site oligonucleotides in gel shift competition experiments, as summarized in Table I, the core sequence of the rat NPTA element has been tentatively identified as 5Ј-TGTTGGCTATGTCCCAAAC-3Ј, with the two underlined motifs and the distance between them being critical for binding activity. The banding patterns of olfactory epithelial NPTA-binding proteins in gel shift experiments were constant regardless of the different competitor used ( Fig. 2 and data not shown), which suggests that the core sequence used in the probe may not contain more than one independent binding site. Of interest, a second NF-1 like sequence (score of 87.6%) was found further upstream at Ϫ2053 to Ϫ2078 (in reverse orientation) of CYP2A3 promoter. However, this sequence has a CAA in place of the critical CCA motif and was not able to compete for protein binding to the NPTA element (Table I).
The core sequence of the NPTA element is highly conserved (5Ј-TG(C/T)TGGC(A/T)(A/G)TGTCCCAA(A/G)C-3Ј) among rat CYP2A3, mouse Cyp2a5, and human CYP2A6 genes, which are all expressed in the nasal mucosa. The sequences in CYP2A3 and Cyp2a5 promoters are virtually identical, with only one A/T substitution in a position shown to be unimportant for protein binding activity (Table I). The corresponding sequence in the human CYP2A6 gene is slightly more divergent, but the critical nucleotides are conserved. Gel shift analysis using the 2A3pm1 probe indicated that the NPTA sequences from mouse Cyp2a5 (2A5pm1) and human CYP2A6 (2A6pm1) genes were active in competing with the rat probe for specific binding to rat or mouse nuclear proteins (Fig. 2). However, the minimal NPTA binding sequence found in the human CYP2A6 gene may include two extra nucleotides at the 5Ј-end, possibly because of the small differences in internal sequences between the rat and the human genes (Table I).
To directly demonstrate the role of this novel element in transcriptional activation in the olfactory mucosa, in vitro transcription assays were performed using promoter constructs containing two different lengths of 5Ј-flanking sequence (Fig.  3A). Transcription from the truncated CYP2A3 genes was detected by S1 nuclease protection assay using a 52-mer antisense oligonucleotide probe corresponding to the 5Ј-end of the CYP2A3 mRNA. Primer extension experiments were also performed using an antisense primer corresponding to ϩ52 to ϩ31 of the CYP2A3 mRNA; the results (not shown) indicated that the CYP2A3 transcripts generated in vitro were initiated from the same site as were those detected in total RNA isolated from rat olfactory mucosa. Transcription from pOMP-280 plasmid, used as an internal control, was also determined in some experiments; the same promoter sequence from rat OMP gene was previously found to be sufficient to direct olfactory neuronspecific expression of a reporter gene in transgenic mice (11).
As shown in Fig. 3, B and C, CYP2A3 transcripts were abundant in reactions with either the Ϫ1900 to ϩ582 bp (p2A3-1900) or the Ϫ254 to ϩ58 bp (p2A3-254) promoter constructs but were barely detectable with the Ϫ254 to ϩ58 bp FIG. 1. DNase I footprinting analysis of CYP2A3 proximal promoter. A-C, a DNA probe (Ϫ231 to ϩ58) labeled with 32 P at the 5Ј-end on the noncoding strand was incubated with crude nuclear extracts from olfactory mucosa (Olf), lung, kidney, brain, and liver of 2-month-old rats (A) or mice (B). A second probe (Ϫ255 to Ϫ67) labeled with 32 P at the 5Ј-end on the coding strand was incubated with nuclear extracts only from rat olfactory mucosa (C). The amount of nuclear proteins used in each binding reaction was 60 g, except for the experiments in lane 2 of A and lane 3 of B, which contained 40 g. An unlabeled oligonucleotide probe containing the NPTA-binding site (2A3pm1; see D for sequence) was added in a 100-fold molar excess to the reactions in lane 4 or 5 (as indicated) to demonstrate binding specificity. Negative control reactions (Probe) were performed in the absence of nuclear extracts. The tissue sources of nuclear extracts are indicated at the top of each lane. DNA ladders (A ϩ G) were included for sequence identification. The hatched bar shows the region protected by NPTA-binding proteins, and numbers indicate nucleotide positions relative to the transcription start site of the CYP2A3 gene. D, sequences of the proximal promoter of the CYP2A3 gene (31) with the footprints of the NPTA-binding proteins on the coding (above) and noncoding (below) strands indicated. The entire sequence was cloned into the SrfI site of pCR-Script vector and used as a template for generating 32 P-labeled DNA probes by PCR, as described under "Experimental Procedures." The nucleotides are numbered to the left, with the transcriptional start site (arrow) as ϩ1 (31), and the sequence of the 2A3pm1 oligonucleotide competitor is underlined. The positions of a TATA box (double underlined) and the translation initiation codon for CYP2A3 protein (boldface type) are also shown. construct containing a mutant NPTA element (p2A3-254/M) corresponding to 2A3mut4 in Table I. Transcription from the OMP promoter construct appeared to be more active than from the CYP2A3 promoter constructs, and the OMP transcription rates were about the same in the three different reactions (panel C), confirming that the differences in CYP2A3 transcription rate were due to differences in the promoter structure. A densitometric analysis of the relative intensity of the CYP2A3 transcripts in different reactions indicated that the ratio of transcriptional activities from the three CYP2A3 constructs was about 35:17:1 (p2A3-1900:p2A3-254:p2A3-254/M). These results indicate that the NPTA element is important for transcriptional activation of the CYP2A3 gene. In addition, inclusion of sequences between Ϫ1900 and Ϫ254 apparently led to a higher level of transcription, suggesting the presence of additional upstream regulatory elements.
The transcriptional activities of nuclear extracts from 8-and 60-day-old rats were compared using the p2A3-254 construct. As shown in Fig. 3D, a higher level of CYP2A3 transcripts was detected in reactions with nuclear extracts from the younger rats. Densitometric analysis of results from two independent experiments indicated that the differences in CYP2A3 transcript levels between the two age groups were about 3-fold.
The tissue-selective expression of NPTA-binding proteins was confirmed in gel shift assays with nuclear extracts from the olfactory mucosa, lung, kidney, testis, brain, and liver of both rats and mice (Fig. 4). Specific NPTA-binding proteins were detected in the olfactory mucosa, whereas no shifted bands were detected in lung, kidney, testis, and brain. A very weak signal was detected in the liver, but it had a lower mobility than the bands detected in the olfactory mucosa and thus may represent different binding protein(s). The integrity of the nuclear extracts from the different tissues was verified in experiments not presented using two different probes in gel shift assays, including the U-site in several olfactory signal transduction genes (2) and the UBE site in the OMP gene (11); the patterns of specific DNA-protein complexes detected in these tissues were similar to those reported in the earlier studies (2,11).
The developmental expression of the NPTA-binding proteins was also examined (Fig. 5). Protein binding to the rat NPTA element was detectable as early as postnatal day 1 and was much greater at day 8 than at day 30 or 60, although the relative abundance of the four different bands detected was fairly constant at different ages (Fig. 5A). In another series of experiments (Fig. 5B) with crude nuclear extracts prepared from nasal tissues of adult rats at different stages after intraperitoneal injection of coumarin, total protein binding was found to decrease dramatically during coumarin-induced degeneration of olfactory epithelium, paralleling the disappearance of CYP2A3 protein at 48 h as reported earlier (30), and to recover at higher than pretreatment levels at 7 days posttreatment, coinciding with the partial recovery of CYP2A3 expression at this time point (30). Interestingly, the two lower bands were preferentially diminished during degeneration, at 48 h post-treatment, while an additional band appeared, which had a migration rate lower than the four bands normally detected. At 7 days post-treatment, the top three bands were cated by arrows (A, B). When indicated, 2A3mut11 probe (see Table I) was added at a 100-fold excess as a negative control. For supershift experiments (C), 5 l of control serum or anti-NF-1 serum (as indicated) was incubated with nuclear extracts from rat olfactory mucosa for 1 h at room temperature before the addition of labeled 2A3pm1. The positions of the supershift and well shift bands are indicated by double arrows, and the band that nearly disappeared upon the addition of anti-NF-1 is indicated by an arrowhead.

FIG. 2. Characterization of NPTA-binding proteins in rat and mouse olfactory mucosa by electrophoretic mobility shift assays.
Gel-shift assays were performed with 32 P-labeled doublestranded 2A3pm1 oligonucleotide probe and crude nuclear extracts prepared from the olfactory mucosa of 2-month-old rats (A, C) or mice (B). Binding reaction mixtures contained 20 g of nuclear proteins, 1 ng of labeled 2A3pm1, and other components as described under "Experimental Procedures," and the reactions were carried out in the presence or absence of unlabeled double-stranded oligonucleotide competitors at a 100-fold molar excess. The sequences of the competitors are shown in Table I, including the authentic 2A3pm1 probe; conserved NPTA sites from mouse Cyp2a5 (2A5pm1) and human CYP2A6 promoters (2A6pm1), respectively; an authentic NF-1 binding site; and the UBE site and the proximal OLF-1 site from the rat OMP gene. Nuclear extracts were omitted in the lanes labeled Probe. The positions of DNA-protein complexes in rats and mice differ slightly and are indi-much more abundant than the two lower bands, which is different from the pattern seen in normal animals.
To determine the number and size of the proteins that directly interact with the NPTA element, Southwestern experiments were carried out with 2A3pm1 as a probe. As shown in Fig. 6, following protein renaturation, two NPTA-binding proteins were detected in olfactory epithelial nuclear extracts from both rats and mice. Consistent with results from gel shift assays, the proteins detected in rat and mouse olfactory epithelial nuclear extracts had slight differences in size. However, their molecular weights (all in the range of 40 -50 kDa) appear to be lower than those of affinity-purified human and mouse CTF/NF-1 binding proteins, which ranged from 52 to 66 kDa (55). No signal was detected with nuclear extracts from rat or mouse liver and detection of olfactory epithelial NPTA-binding proteins was blocked by excess unlabeled probes (Fig. 6, lanes e-h), indicating binding specificity. In additional control experiments, no signal was detected in olfactory mucosa when the DNA probe was replaced by labeled 2A3mut4 (see Table I) probe, which contains a mutated NPTA binding site (data not shown). These results indicate that the NPTA element can directly interact with at least two nuclear proteins from both rat and mouse olfactory mucosa. In other experiments not presented, these proteins were not detected by antibodies to NF-1 on immunoblots, confirming that the NPTA-binding proteins may be different from known NF-1 factors. DISCUSSION Unique difficulties exist with studying gene regulation in the olfactory mucosa, including the presence of multiple cell types and the lack of cell lines that maintain proper phenotype or gene expression. As a consequence, the mechanism of olfactory mucosa-specific gene expression is not well understood. To date, the mechanism of tissue-specific gene expression in the nonneuronal cells of the olfactory mucosa has not been examined. However, using in vitro methods, two recent studies have been successful in identifying a neuron-specific promoter motif (OLF-1) in several genes specifically expressed in olfactory neurons, including the OMP gene and five genes involved in the olfactory signal transduction cascade (2,11). In one of these studies, a 0.3-kb promoter fragment of the OMP gene containing one OLF-1 motif was sufficient for olfactory neuron-specific expression of a reporter gene in transgenic mice (11). The OLF-1-binding protein has been cloned from an olfactory mucosa cDNA library (12). Further studies indicated that, in addition to the olfactory neuron, the OLF-1 protein is also expressed in neuronal precursors of other sensory structures and is believed to play a role in neurogenesis (56). A more recent study showed that OLF-1 protein (which is renamed O/E-1) is present in several tissues in addition to olfactory neurons and developing B-cells; however, two other genes in the same family of transcription factors, O/E-2 and O/E-3, are expressed at a high level only in the olfactory tissue and may provide redun- I Characterization of NPTA-binding site in the CYP2A3 gene using various oligonucleotide competitors in gel shift assays Gel shift assays were performed with 32 P-labeled 2A3pm1 oligonucleotide probe and nuclear proteins from rat olfactory mucosa as described in the legend to Fig. 2. The various unlabeled competitors were added at a 100-fold excess. Identical results were also obtained with nuclear extracts from mouse olfactory mucosa. Ϫ, no competition; ϩ, partial competition; ϩϩϩ, strong competition.
b Numbers indicate positions relative to transcription start site, except for the CYP2A6v2 gene, for which the numbers indicate distance from the translation start codon.
c Underlined nucleotides highlight mutations in each probe; ▫, internal deletion. The distal NF-1-like site in CYP2A3 promoter (2A3pm2/NF-1) is in reverse orientation. dancy in the regulation of olfactory neuronal development (13).
The mechanism of tissue-specific and developmental regulation of xenobiotic-metabolizing P450s was not studied until recently, and most studies were focused on the liver. Multiple liver-enriched transcription factors, including HNF-1␣, HNF-3, HNF-4, and CCAAT/enhancer-binding protein ␤, and more ubiquitously expressed factors, such as Sp1, GABP ␣/␤, and NF2d9, were found to be involved in the regulation of the liver-specific expression of one or more P450 genes in the CYP2 family, such as CYP2E1 and several CYP2C and CYP2D genes (cf. Ref. 57). Moreover, unique regulatory circuits appear to be utilized for controlling each of the P450 genes. Transcriptional regulation of the CYP2A3 gene has not been studied previously, although the transcriptional start site has been determined (31) and a HNF-4-like sequence, named HepG2-specific P450 2C factor 1 motif, or HPF1, was found to be present at about 55 nucleotides upstream of the start site (58). The HPF1 site is present in the proximal promoter of many CYP2A, -2C, and -2D genes and appears to be important for hepatic expression of some, but not all, CYP2C genes (58,59). A more recent study indicated that this HNF-4-like motif may be involved in activating the hepatic transcription of the Cyp2a4 gene (60), which is highly homologous to the nasal CYP2As but is not expressed in the olfactory mucosa (10). It remains to be determined whether the HPF1/HNF-4 site is functional in CYP2A3 expression in the olfactory mucosa.
Using various in vitro methods, we have identified a conserved binding site potentially involved in transcriptional activation of rat CYP2A3, mouse Cyp2a5, and human CYP2A6 genes in the olfactory mucosa, named NPTA element. The NPTA element contains an NF-1-like binding site but interacts FIG. 4. Gel shift analysis of tissue-specific interaction of nuclear proteins with the NPTA-binding site. Gel shift assays were performed as described in the legend to Fig. 2, with 32 P-labeled 2A3pm1 probe and crude nuclear extracts (20 g in each lane) from various tissues of 2-month-old rats (A) and mice (B) as indicated. Nuclear extracts were omitted in the lanes labeled Probe. When indicated, OLF-1 probe (see Table I) was added at a 100-fold excess to olfactory epithelial binding reactions as a control for nonspecific binding (Olf ϩ OLF-1).

FIG. 3. Role of the NPTA element in
in vitro transcription of the CYP2A3 gene with nuclear extracts from rat olfactory mucosa. A, CYP2A3 and OMP promoter constructs used for in vitro transcription. Exons are shown as a solid bar, and the antisense oligonucleotide probes used for S1 nuclease protection assays (S1 probe) are shown as open boxes. The approximate positions of the NPTA site in the CYP2A3 promoter and the proximal OLF-1 binding site in the OMP promoter are also indicated. B-D, in vitro transcription was performed as described under "Experimental Procedures" using nuclear protein extracts prepared from 2-monthold male rats (B-D) or 8-day-old rats (male or female, randomized) (D). The transcription reactions contained 60 g of nuclear protein and 1 g of a plasmid containing one of the CYP2A3 promoter constructs (B) or 1 g each of a CYP2A3 plasmid and the pOMP-280 plasmid (C, D). RNA transcripts were detected by an S1 nuclease protection assay using 32 Plabeled, single-stranded oligonucleotide probes of different sizes (41-mer for OMP and 52-mer for CYP2A3) and visualized by autoradiography. The positions of the protected fragments corresponding to in vitro transcripts from the CYP2A3 promoter (*) and the OMP promoter (**) and single-stranded DNA size makers are shown.
with unique proteins (NPTA factors) detected only in the olfactory mucosa. Interestingly, the olfactory epithelial NPTA-binding proteins were detected at much higher levels at 8 days than at 30 or 60 days after birth, and nuclear extracts from 8-day-old rats also supported higher transcriptional activity of the CYP2A3 promoter in vitro. This is in contrast to the higher levels of CYP2A3 protein in the older rats than in the 8-day-old rats (data not shown) and suggests that the NPTA-binding proteins may be mainly involved in tissue-restricted developmental onset of CYP2A3 gene and that additional regulatory mechanisms may be involved in supporting high levels of expression in adult olfactory epithelium. To this end, the level of OLF-1-binding protein was also found to be higher at 10 days than at 21 days after birth in a previous study although the OMP gene was expressed at higher levels in the older group (11).
NF-1 (also named CCAAT-binding transcription factor, or CTF) represents a family of sequence-specific DNA-binding proteins that recognize sequences containing TGG motif. Several classes of NF-1 proteins having a highly conserved DNAbinding domain and variable transcription activation domain have been isolated (61-63), which differ by unique organ-and cell type-specific expression and by their function in either transcriptional activation or repression (64,65). Because bind-ing of olfactory epithelial nuclear proteins to the NPTA element was partially competed by an authentic NF-1 element and because three of the four bands detected in rat olfactory mucosa in gel shift assays were supershifted by anti-NF1, it is likely that one or more of the NPTA-binding proteins may represent novel members of the NF-1 family that are restrictively expressed in the olfactory mucosa.
NF-1-binding sites are present in many viral and cellular promoters. Among P450 genes, an NF-1 site was identified in the distal promoter of the CYP2C23 gene, which was protected by kidney nuclear proteins (66). An NF-1 site has also been found in the phenobarbital-responsive enhancer module in mouse Cyp2b10 gene and is required for the induction response (67,68). Also of interest, an NF-1-like site (named UBE) was identified in the OMP gene, an olfactory neuron-specific gene, but it formed complexes with nuclear proteins from a number of tissues (11,69). However, sequences similar to the NPTA element or NF-1 site were not found in the proximal promoter region (about 1 kilobase pair available) of rat olfactory mucosaspecific CYP2G1 gene. 3 The nature and cellular origin of the olfactory epithelial NPTA-binding proteins remain to be determined. The detection of multiple bands on gel shift assay and Southwestern blots indicated that the NPTA element may be regulated by more than one protein. However, the origin of the multiple bands detected in gel shift assays is not yet clear. They could represent different binding proteins expressed in cells at different differentiation stages in the olfactory epithelium or originating from different cell types in the epithelium and the submucosa. Both possibilities are consistent with altered banding patterns during degeneration and subsequent regeneration of the olfactory epithelium. Alternatively, at least some of the bands may be derived from dimerization of the NPTA-binding proteins with different partners.
Thus, we have identified a conserved transcriptional activation element that may be regulated by novel transcription factors, possibly of the NF-1 family, and may be responsible for the olfactory mucosa-predominant expression and early developmental onset of the CYP2A3 gene in rats and orthologous genes in mice and humans. The identification of these putative transcription factors and in vivo studies on the role of the NPTA element in the tissue-specific and developmental expression of the CYP2A genes are currently under way. 3 J. Zhang and X. Ding, unpublished results.  a and e) or mouse olfactory mucosa (lanes b and f) or from rat (lanes c and g) or mouse liver (lanes d and h) were submitted to electrophoresis in a SDS-polyacrylamide (10%) gel and transferred to nitrocellulose sheets. The proteins were renatured and incubated with 32 P-labeled 2A3pm1 probe (20,000 cpm/ml), and the DNA-protein complexes were visualized by autoradiography. To demonstrate binding specificity, a 200-fold excess of unlabeled 2A3pm1 probe was added to the hybridization buffer and incubated with the nitrocellulose sheet prior to the addition of 32 P-labeled 2A3pm1 probe (lanes e-h). The approximate positions of selected fragments of prestained molecular weight markers are shown on the left.