The core promoter of human thioredoxin reductase 1: cloning, transcriptional activity, and Oct-1, Sp1, and Sp3 binding reveal a housekeeping-type promoter for the AU-rich element-regulated gene.

The selenoprotein thioredoxin reductase 1 (TrxR1) carries many vital antioxidant and redox regulatory functions. Its mRNA levels are known to be post-transcriptionally modulated via AUUUA motifs (AU-rich elements (AREs)), but the promoter yet remains unknown. Here we have cloned and determined the sequence of a 0.8-kilobase pair human genomic fragment containing the proximal promoter for TrxR1, which has transcriptional activity in several different cell types. The core promoter (-115 to +167) had an increased GC content and lacked TATA or CCAAT boxes. It contained a POU motif binding the Oct-1 transcription factor and two sites binding Sp1 and Sp3, which were identified with electrophoretic mobility shift assays using crude nuclear extracts of A549 cells. The TrxR1 promoter fulfills the typical criteria of a housekeeping gene. To our knowledge this is the first housekeeping-type promoter characterized for a gene with post-transcriptional regulation via ARE motifs generally possessed by transiently expressed proto-oncogenes, nuclear transcription factors, or cytokines and influencing mRNA stability in response to diverse exogenous factors. Expression of TrxR1 as an ARE-regulated housekeeping gene agrees with a role for the enzyme to maintain a balance between intracellular signaling via reactive oxygen species and protection of cells from excessive oxidative damage.

The mammalian thioredoxin system consists of thioredoxin (Trx), 1 thioredoxin reductase (TrxR), and NADPH. Trx is reduced by TrxR and participates in many different types of reactions (1) including synthesis of deoxyribonucleotides (2), redox control of transcription factors (3), reduction of peroxides (4), and regulation of apoptosis (5,6). Extracellularly it has immunoregulatory co-cytokine (7) and chemokine (8) activities. It is of importance to note that the redox status of Trx is essential for most, if not all, of its many vital functions. Consequently perturbations of the TrxR activity are implicated in a number of cell proliferative or immunological diseases, and the enzyme is increasingly being recognized as an important pharmacological target in a number of medical conditions (9). Knowledge of the molecular mechanisms guiding the TrxR1 expression therefore also has medical relevance.
Mammalian TrxR is a selenocysteine-containing oxidoreductase flavoprotein with a remarkably broad substrate specificity catalyzed by the unique carboxyl-terminal active site structure involving a conserved Gly-Cys-Sec-Gly (where Sec is selenocysteine) tetrapeptide with a redox active selenenylsulfide/selenolthiol motif (10 -14). The enzyme reduces not only protein disulfides such as that in oxidized Trx but also low molecular weight disulfide compounds, e.g. dithionitrobenzoic acid (15) and lipoic acid (16), low molecular weight nondisulfide compounds such as selenite (17) or alloxan (18), or even peroxides directly (19). Recently two TrxR isoenzymes have been identified, one mitochondrial (20 -22) and one present mainly in testis (20), both having the same overall domain structure as TrxR1. The 3Ј-UTR of all TrxR isoenzymes contain a SECIS (selenocysteine incorporation sequence) element necessary for co-translational selenocysteine incorporation (12,(23)(24)(25)(26). In addition, the 3Ј-UTR of TrxR1 contain AREs, which in untreated cells lead to a rapid TrxR1 mRNA turnover (25,27). AREs are typically found in cytokine, proto-oncogene, transcription factor, and other transiently expressed mRNAs (28), although the family of genes with possible AREs found using general data base searches is more diverse (29). Post-transcriptional regulation via AREs enables quick expression responses to various stimuli by a specific block in mRNA degradation through ARE-interacting proteins responding to intracellular signaling (28). It is interesting that TrxR1 contains functional AREs (25,27) because this enzyme is not transiently expressed only under specific growth conditions but is widely expressed in many diverse tissues (21,30,31,33) and cells (34). Further insights into the functional roles the enzyme plays on a cellular level can be gained from a knowledge of the mechanisms regulating its expression. However, no reports exist regarding the structure or function of the promoter guiding TrxR1 transcription. Here we report the identification and characterization of the proximal promoter for human TrxR1 found to display the typical characteristics of a promoter for a housekeeping gene. To our knowledge, this is the first example of an ARE-regulated gene with transcription guided from a housekeeping-type promoter. We propose that this functional organization may be explained by the intimate link between intracellular signaling via reactive oxygen species (32,35) and protection against oxidative damage, indicating a vital role played by TrxR1 for maintaining a functional balance in this regulation, and we propose the molecular mechanisms for how this can be achieved.

EXPERIMENTAL PROCEDURES
Primer Extension Analysis-High quality mRNA was isolated from U937 cells with oligo(dT)-cellulose columns using the commercial QuickPrep Micro mRNA purification kit (Amersham Pharmacia Biotech). A TrxR1-specific antisense oligonucleotide (5Ј-TTA AAA TAT  CTT CCC TTC CCC GAG AC-3Ј) complementary to nt 170 -196 of the  previously published 5Ј-UTR (36) was [␥-32 P]ATP (3000 mCi/mmol, PerkinElmer Life Sciences) end-labeled with T4 polynucleotide kinase and mixed with 1 g of the purified U937 mRNA in 30 l of hybridization buffer (40 mM PIPES, pH 6.4, 1 mM EDTA, 0.4 M NaCl, 80% formamide). The mixture was heated at 85°C for 10 min, and then annealing was carried out at 56°C for 2 h. The annealed primer-RNA complex was precipitated with ethanol and subsequently used for primer extension. The extension reaction was performed with avian myeloblastosis virus reverse transcriptase for 30 min at 42°C according to the Primer Extension System protocol from Promega. A sample with no mRNA added was used as a negative control, and the 1.2-kb "kanamycin RNA" provided with the Primer Extension System (Promega) was used as a positive control together with the provided kanamycin oligonucleotide. The extension products were finally resolved by electrophoresis on an 8% denaturing gel containing 7 M urea and visualized with autoradiography using a PhosphorImager system.
Cloning of the Genomic Upstream Region-The PromoterFinder DNA walking kit (CLONTECH) was used with two nested TrxR1specific primers (5Ј-CAG GCC GTC CCC CGC GTG CTC CCA TC-3Ј and 5Ј-CTG GGC TCG CGG CTT TGT CTG GTT TC-3Ј) complementary to the 5Ј-UTR of the published cDNA sequence (36). Primary and secondary touch-down PCRs were performed (94 o C for 2 s, 72 o C for 3 min, 7 cycles; 94 o C for 2 s, 67 o C for 3 min, 32 cycles; 67 o C for 7 min) using the Expand Long Template PCR system (Roche Molecular Biochemicals). A single 0.83-kb PCR product was obtained, isolated, and cloned into a pGEM-T vector (Promega) with its sequence subsequently determined.
Luciferase Reporter Constructs-The reporter constructs were made using the promoterless firefly luciferase reporter vector pGL3-Basic from Promega. An MluI site-linked 5Ј-primer (5Ј-ATA CGC GTT GGA ATC TAA TAA GGA AGA G-3Ј) and a XhoI site-linked 3Ј-primer (5Ј-TAC TCG AGC TGG GCT CGC GGC TTT GTC T-3Ј) specific for the TrxR1 genomic sequence determined above were used to amplify the whole 827-bp cloned genomic fragment by standard PCR. The PCR product was cloned into a pGEM-T vector, and subsequently the 0.8-kb fragment was transferred in correct orientation into the MluI and XhoI sites in pGL3-Basic resulting in construct HA shown in Fig. 2. As a negative control, the genomic fragment was also cloned in the reverse direction ( Fig. 2, construct HB) using a XhoI site-linked 5Ј-primer (5Ј-ATC TCG AGT GGA ATC TAA TAA GGA AGA G-3Ј) and an MluI site-linked 3Ј-primer (5Ј-TAA CGC GTC TGG GCT CGC GGC TTT GTC T-3Ј). Subsequent deletion constructs were produced by digestion of the HA plasmid with the restriction enzymes MluI/AflII (Fig. 2, construct ⌬1), KpnI/NheI (Fig. 2, construct ⌬2), SstI (Fig. 2, construct ⌬3), or SstI/XhoI (Fig. 2, construct ⌬4) followed by self-ligation.
Determination of Reporter Gene Activity-Approximately 3 ϫ 10 4 cells/well in 24-well dishes were seeded 24 h before transfection using the TrxR1 promoter reporter constructs (0.5 g) or empty pGL3-Basic (0.5 g) with or without co-transfection with 0.1 g of pRL-TK or 0.4 ng of pRL-CMV plasmid. The transient transfections were performed using LipofectAMINE Plus Reagent (Life Technologies, Inc.) with 2 g of LipofectAMINE and 3 l of Plus reagent for 3 h according to the manufacturer's protocol. After 24 h the cells were washed with phosphate-buffered saline (Life Technologies, Inc.) and lysed in 100 l of Passive Lysis Buffer (Promega) at room temperature for 15 min. The firefly and Renilla luciferase activities in 20 l of the cell lysates were determined using the Dual-Luciferase Reporter Assay system (Promega) according to the manufacturer's instructions using a Turner Designs TD-20/20 luminometer.
Nuclear Extracts-Nuclear extracts were prepared from 1 ϫ 10 7 A549 cells grown in large culture plates that were scraped into phosphate-buffered saline and centrifuged for 10 min at 1800 ϫ g. The cells were resuspended in 600 l of hypotonic buffer (10 mM HEPES, pH 7.9), 1.5 mM MgCl 2 , 10 mM KCl, 0.2 mM phenylmethylsulfonyl fluoride, 0.5 mM dithiothreitol) and centrifuged at 1800 ϫ g for 5 min at 4°C. This pellet was resuspended in 400 l of hypotonic buffer and incubated for 10 min on ice whereupon the cells were homogenized with 10 strokes using a Dounce homogenizer with a B-type pestle. The homogenate was centrifuged for 15 min at 3300 ϫ g at 4°C, the supernatant was removed, and the nuclei in the pellet were resuspended in 180 l of low salt buffer (20 mM HEPES, pH 7.9, 25% glycerol, 1.5 mM MgCl 2 , 10 mM KCl, 0.2 mM EDTA, 0.2 mM phenylmethylsulfonyl fluoride, 0.5 mM dithiothreitol). KCl was added to a final concentration of 0.4 M while stirring gently with a pipette tip. Nuclei proteins were extracted for 30 min on ice with continuos gentle mixing with a pipette tip. Upon centrifugation for 30 min at 25 000 ϫ g at 4°C the supernatant with extracted proteins was frozen immediately at Ϫ80°C in aliquots, which were thawed gently on ice at time for use in shift assays.
Electrophoretic Mobility Shift Assays-The electrophoretic mobility shift assay probes used (ordered from Life Technologies, Inc. Double-stranded oligonucleotides were generated by annealing equimolar complementary oligonucleotides in 50 mM Tris-HCl (pH 7.4), 1 mM EDTA, 100 mM NaCl, and 13 mM MgCl 2 with the following temperature program: 88°C for 2 min, 65°C for 10 min, 37°C for 10 min, 25°C for 5 min. The double-stranded oligonucleotides were endlabeled with [␥-32 P]ATP (3000 mCi/mmol, PerkinElmer Life Sciences) using T4 polynucleotide kinase, and the labeled probes were subsequently purified by Chroma-spin 30 (CLONTECH). For binding assays a mixture was prepared containing 32 P-labeled oligonucleotide (0.3 ng), 3 g of nuclear protein, 1 g of poly(dI⅐dC) (Amersham Pharmacia Biotech) adjusted to 20 l with binding buffer (20 mM HEPES, pH 7.9, 1.5 mM MgCl 2 , 0.2 mM EDTA, 1 mM dithiothreitol, 0.1 mM phenylmethylsulfonyl fluoride, 7.5% glycerol). Binding reactions were carried out for 15 min at 28°C whereupon a 15-l aliquot of each reaction was loaded onto a 4% nondenaturing polyacrylamide gel and run in 1ϫ Tris-glycine buffer at 120 V. Following electrophoresis, gels were dried and autoradiographed. In the competition assays 20 ng of competitor unlabeled oligonucleotides, with the nonspecific oligonucleotide being 5Ј-CTG ATG TCT TCA TCA TTC CTC AAA TTC TTG TAA GCT CTG C-3Ј, was added prior to the addition of the 32 P-labeled probe. For the supershift assays using polyclonal antibodies from Santa Cruz Biotechnology, Inc. 2 g of anti-Sp1 (catalogue no. sc-420x), anti-Sp3 (catalogue no. sc-644x), anti-Oct-1 (catalogue no. sc-232x), anti-Oct-4 (catalogue no. sc-9081x), or anti-NF1 (catalogue no. sc-870x) was added before or after addition of the 32 P-labeled probe as indicated in the figure legends.

Primer Extension and Cloning of the Human TrxR1
Promoter-We have previously found that murine cDNA species for TrxR1 exist in variants with different 5Ј-UTR domains, whereas human TrxR1 cDNA can only be found with one 5Ј-UTR domain sequence, albeit of different length in different cDNA clones (33). To localize the transcription initiation site of human TrxR1 and ascertain a lack of multiple transcription start sites, a 32 P-labeled primer complementary to the ϩ170 to ϩ196 region of the previously published TrxR1 cDNA (36) was extended using mRNA purified from U937 cells as a template. These cells express TrxR1 at high levels (37). 2 The primer extension gave a strong signal with a single 250-bp product (Fig. 1A). This revealed a transcription start site resulting in In A, primer extension analysis was performed with mRNA from U937 cells and a 32 P-labeled primer complementary to the ϩ170 to ϩ196 region of the previously published TrxR1 cDNA (36). This revealed a single extension product of 250 Ϯ 5 bp (arrow) demonstrating an ϳ54-bp longer 5Ј-UTR than the previously published sequence (36). The negative control was a sample with no mRNA added. The positive control was the pure "kanamycin control RNA" and specific primer provided with the Primer Extension System (Promega) generating an 87-bp product as expected. In B, the sequence of the subsequently cloned genomic fragment is shown with the region corresponding to the transcriptional initiation site indicated by the primer extension analysis shown with a shaded box. Aligned are also the three human TrxR1 cDNA sequences found in GenBank TM . These are identified in the figure as cDNA I (accession numbers AA173096 and BE247591), cDNA II (accession number AU077310), and cDNA III (accession number NM-003330). The latter, cDNA III, is the sequence upon which the primer extension shown in A was based with the exception an ϳ54-bp larger 5Ј-UTR than that of the previously published cDNA sequence (36). These 54 bp could in theory have encompassed a splice site, but considering the presence of a single 5Ј-UTR domain in U937 cells (Fig. 1A), we set out to clone the genomic region upstream of the known sequence with the aim to identify the TrxR1 promoter.
A PCR on a human genomic library using nested primers based on the 5Ј-UTR of the placental cDNA sequence (36) gave an 827-bp product, which was cloned and sequence determined as described under "Experimental Procedures." The sequence has been deposited in GenBank TM under accession number AF247671. Using data base searches with this sequence we then discovered two human TrxR1 expressed sequence tag clones deposited in GenBank TM that perfectly overlapped with the genomic sequence and that extended the 5Ј-UTR of the previously known sequence exactly to the region indicated by the primer extension (Fig. 1B). We thereby concluded that the transcription start site was localized. The genomic sequence determined hence contained the transcriptional start site, 167 bp of the 5Ј-UTR, and 660 bp of the upstream 5Ј flanking region (Fig. 1B). This genomic fragment should encompass the core promoter of human TrxR1, which is functional in at least U937 cells (Fig. 1A) and human acute lymphatic leukemia pre-B-cells or in neuroepithelium, which are tissue sources of expressed sequence tag clones shown in Fig. 1B as form cDNA I.
Interestingly the longest cDNA variants verifying the transcription start site as mapped with the primer extension analysis were spliced further downstream and lacked a stretch of 238 nt (Fig. 1B, cDNA I) that was instead found to be maintained in another cDNA variant (Fig. 1B, cDNA II). This second variant (cDNA II) continued 14 nt further upstream of the previously published placental sequence (36) (Fig. 1B, cDNA  III). Both the cDNA II and cDNA III variants, but not the cDNA I variant, contained an upstream ATG (Fig. 1B). In cDNA III this ATG is in frame with the open reading frame and has in a previous study also been shown to be functional (27) yielding a 60-kDa TrxR1 in addition to the most common 55-kDa form (10 -12). Regulation of this alternative splicing pattern should be analyzed further, but in this study we have concentrated on continuing the characterization of the structure and function of the core promoter.
At present the genomic sequence that we have cloned cannot be found in GenBank TM . However, while preparing this manuscript the private sequencing effort for the human genome was published (38), and material from the Celera data base was thereby also made public. Searching this data base (publication.celera.com) gave a near perfect match of our 0.8-kb genomic sequence with nt 1667-2494 of the 16,175-bp long chromosomal fragment with Celera identification number HTBL682EW. This fragment has not yet been assigned to a specific chromosome and contains large stretches of nondetermined nucleotides. However, searching the Celera data base with the 500 first nucleotides of the human placental TrxR1 cDNA (36) gave a near perfect match to the same chromosomal fragment continuing the sequence with nt 2402-2687 of HTBL682EW corresponding to nt 8 -293 of the TrxR1 cDNA and with nt 4272-4382 corresponding to nt 294 -404 of the original TrxR1 cDNA (36). This confirmed our genomic sequence and the exon-splicing pattern proposed in Fig. 1B and, moreover, identified the Celera chromosomal fragment TABLE I Transcriptional activity of the upstream genomic TrxR1 fragment in different cells Four different cell lines, HeLa (cervix carcinoma), A549 (lung carcinoma), HT-1080 (fibrosarcoma), and 293 (embryonic kidney) were used for transfection studies to ascertain transcriptional activity of the human upstream TrxR1 genomic fragment. A firefly luciferase reporter construct was cotransfected with a Renilla luciferase control vector pRL-TK. The firefly/Renilla ratio is given in the table (mean Ϯ S.D.; t ϭ number of separate transfection experiments; n ϭ number of lysates from separate wells analyzed in total). Controls without the TrxR1 genomic fragment upstream of the firefly luciferase gave a firefly/Renilla ratio of Ͻ0.05 as determined in A549 and HeLa cells (see Table II for details). The genomic upstream region therefore showed strong transcriptional activity in all cell types tested with the highest value obtained using A549 cells.   Fig. 2) to identify a core region required or sufficient for the basal promoter activity shown in Table I. HeLa and A549 cells were transfected with the different constructs with or without cotransfection with the Renilla vectors pRL-TK or pRL-CMV supplied as internal controls for transfection efficiency. Constructs ⌬1-⌬3 maintained a level of luciferase activity similar to that of the full HA construct, whereas ⌬4, lacking the Ϫ115 to ϩ167 region, had completely abolished transcriptional activity.
The absolute values of the firefly as well as Renilla luciferase activity measurements are given in the table (mean Ϯ S.D. using a total of six separate wells from one transfection experiment). All of the firefly luciferase values are combined and summarized in Fig. 2 that a remaining EcoRI site kept at the 5Ј-end of the original sequence (36) has been removed. Note that cDNA I lacks a stretch of 238 nt that is present in cDNA II and cDNA III, whereas cDNA II lacks a subsequent 119-bp fragment that is present in cDNA I and cDNA III. The junctions between the cDNA forms (black arrows) most likely derive from alternative splicing, although the sequence at the middle junction indicated in the figure does not conform to the general GT-AG consensus of splice sites. Interestingly, cDNA II and cDNA III, but not cDNA I, contain an ATG (underlined) upstream of a second ATG codon (also underlined) with both of these ATG codons having previously been shown to function as alternative translational start codons using a human cDNA clone with the splicing pattern here identified as cDNA III (27). The asterisk shows the ϩ1 position considered in this study as the transcriptional initiation site. The hollow arrows in the figure indicate the restriction sites used for construction of the different luciferase reporter constructs shown in Fig. 2.
HTBL682EW as containing the initial region of the human TrxR1 gene. The only other Celera chromosomal sequence that matched with the 500 first nucleotides of the human TrxR1 cDNA was the 787-bp long KRDC2R9Q fragment of which nt 127-223 correspond to nt 405-500 of the TrxR1 cDNA sequence. The KRDC2R9Q fragment also has yet to be assigned by Celera to a specific chromosome, but because TrxR1 previously was experimentally found to be located on chromosome 12 (12q23-q24.1) by fluorescence in situ hybridization analysis (31) this assignment should now be possible for the two fragments. Finally, the location of the TrxR1 gene at chromosome 12 was also verified by searching the Celera data base with nucleotides 500 -2000 of the TrxR1 cDNA of which nt 716 -1872 were covered by exons found in the KRCE3UY3 fragment having the Celera-assigned location of nucleotides 109607037 to 109682324 on chromosome 12. Transcriptional Activity of the Human TrxR1 Promoter-Computer-based analysis (MatInspector) of the 0.8-kb genomic sequence herein cloned by us revealed many possible transcription factor binding motifs including a CCAAT box at nt Ϫ127 to Ϫ123 and several GC-rich regions (total GC content 46%), but no consensus TATA box could be found. To ascertain transcriptional promoter activity, the cloned 0.8-kb genomic fragment was transferred to a pGL3-basic vector in correct orientation upstream of a firefly luciferase reporter gene (construct HA). The reverse orientation (construct HB) served as a negative control. The HA construct, but not the HB construct, indeed showed strong promoter activity in the diverse cell types transfected, which included HeLa (cervix carcinoma), A549 (lung carcinoma), HT-1080 (fibrosarcoma), and 293 (embryonic kidney) cells (Table I). Deletion constructs (⌬1-⌬4) were then made to identify smaller regions required or sufficient for maintaining the basal promoter activity. In both HeLa and A549 cells constructs ⌬1-⌬3 gave a luciferase activity comparable with the full HA construct, whereas construct ⌬4, which lacked the Ϫ115 to ϩ167 region, had completely abolished transcriptional activity (Table II). The transfections were repeatedly analyzed with or without two different Renilla-coupled control vectors and gave consistent results as summarized in Fig. 2 where the reporter constructs are also schematically shown. The basal promoter activity clearly resided in the Ϫ115 to ϩ167 region (the ⌬3 construct) excluding the CCAAT box at Ϫ127 to Ϫ123. The ⌬3 construct had a total GC content of 54%, and a putative CpG island at Ϫ18 to ϩ90 was suggested by analysis with CPGPLOT (European Molecular Biology Laboratory, European Bioinformatics Institute) (39) using Ͼ0.6 as the threshold of observed/expected ratio of C plus G to CpG and Ͼ50% as the percentage of GC cut-off.
We next set out to analyze specific transcription factors that function in binding to the core promoter encompassed by the ⌬3 construct. For nuclear extract starting material we chose A549 cells because these cells had the highest basal activity (Table  II) and are known to express endogenous native TrxR1 at high levels (27).
Identification of Transcription Factors Binding to the Human TrxR1 Core Promoter-Electrophoretic mobility shift assays were performed with 10 overlapping oligonucleotides covering the whole ⌬3 construct (Fig. 3). Oligo-1 (Ϫ115 to Ϫ75) generated a single specific band shift as did oligo-9 (ϩ125 to ϩ167), however it was of a different size. Two specific complexes were seen with oligo-6. Oligo-4, oligo-7, and oligo-10 all demonstrated a similar pattern with three band shifts with oligo-4 also showing an additional fourth specific band. No shifted complexes could be demonstrated with the regions covered by oligo-2, oligo-3, oligo-5, or oligo-8. These results are shown in Fig. 3.
As an aid to determine the specific transcription factors producing the band shifts shown in Fig. 3, a detailed computer analysis was first performed using the MatInspector (40) and PromoterInspector (41) algorithms. This revealed, among other motifs, NF-1 consensus motifs at Ϫ85 to Ϫ82 and ϩ5 to ϩ8, a POU domain at Ϫ98 to Ϫ91, and Sp1/Sp3 motifs at Ϫ77 to Ϫ73, Ϫ30 to Ϫ21, and Ϫ17 to Ϫ7. These analyses thereby guided the choice of antibodies for use in supershift assays.
Oligo-1 encompassed both an NF-1 site and the POU domain. We could not detect supershifts with anti-NF-1 antibodies using any of the 10 oligonucleotides, but the positive control did not give a shifted complex (not shown) indicating the lack of NF-1 in A549 cells. In contrast, antibodies against Oct-1 resulted in a specific supershift of oligo-1 highly resembling that seen with a commercial Oct-1-positive control oligonucleotide (Fig. 4). The other shifted complexes were unaffected by anti-Oct-1 antibodies (Fig. 4) in agreement with a lack of consensus POU domains in these oligonucleotides. This result showed that the predicted POU domain covered by oligo-1 was functional in binding the Oct-1 transcription factor present in the A549 nuclear extract. As a further control, we also used antibodies against Oct-4, another POU-binding transcription factor, which gave no result either with any of the TrxR1 promoter oligonucleotides or with the positive control oligonucleotide (not shown) in agreement with the notion that Oct-4 is only expressed in normal cells of the early stages of embryogenesis (42). In such embryonic cells, however, TrxR1 may possibly be one of the target genes for Oct-4.
Oligo-4 and oligo-10 contained GC-rich regions with predicted Sp1 sites, and these oligonucleotides also generated The correct orientation is that in construct HA with the reverse orientation of construct HB made as a negative control. The consecutive deletion constructs (⌬1-⌬4) were made using the restriction sites shown in Fig. 1B. In the lower half of the figure are the luciferase activities measured after transfections of either A549 or HeLa cells. These data are the combined mean Ϯ S.D. of all the firefly luciferase activity measurements given in detail in Table II. several band shifts (Fig. 3). To determine whether it was the Sp1 motifs that were functional, supershift assays with antibodies recognizing either Sp1 or Sp3 were subsequently performed. The anti-Sp1 antibodies indeed bound to the slowest migrating DNA-protein complexes given by oligo-4 and oligo-10 ( Fig. 5A), which is typical for functional Sp1 motifs (see the commercial Sp1-positive control in Fig. 5A and Ref. 43). Notably the largest band shift with oligo-7 was also affected, although oligo-7 contained no evident Sp1 site as predicted by the previous computer analyses; this shift was, however, eliminated and not supershifted (Fig. 5A). The two faster migrating shifts seen with oligo-4 and oligo-10, and also with oligo-7, were clearly supershifted using anti-Sp3 antibodies (Fig. 5B), which again is typical for functional Sp1/Sp3 sites (43). Oligo-10 was designed to encompass the GC-rich motif just prior to the transcription start, which otherwise was divided between oligo-3 and oligo-4 (Fig. 3). However, because oligo-3 failed to give a band shift (Fig. 3), the fraction of the GC-rich region covered by oligo-3 is not functional in binding Sp1, Sp3, or any other transcription factor in an A549 nuclear extract. Moreover, unlabeled oligo-10 competed out the three different band shifts with labeled oligo-4 and vice versa (not shown), and we conclude that Sp1 and Sp3 bind to the GC-rich region found in the section overlapping between oligo-4 and oligo-10, which also contains a tandem perfect consensus Sp1 motif. The different functional motifs of the human TrxR1 core promoter found in the present study to bind Oct1, Sp1, and Sp3 or to give band shifts of a yet uncharacterized nature are summarized in Fig. 6.

DISCUSSION
This is the first study of any mammalian thioredoxin reductase promoter; therefore no comparison of functional motifs conserved between species or between TrxR isoenzymes can be made. Nonetheless some conclusions can be drawn regarding the basal activity and structure of the promoter for human FIG. 3. Transcription factor binding sites. Ten overlapping oligonucleotides covering construct ⌬3 from Fig. 2 were designed as shown in the upper part of the figure. Electrophoretic mobility shift assays were subsequently performed with the resulting autoradiograms as shown. Lanes A are negative controls without the addition of nuclear extract. In lanes B-D proteins of crude nuclear extracts from A549 cells were included. In lanes C an excess of the homologous but unlabeled oligonucleotide was also added, whereas in lanes D an excess of an unlabeled unrelated oligonucleotide was added. All shift assays were performed with 0.3 ng of 32 P-labeled oligonucleotides and 1 g of poly(dI⅐dC). Nuclear protein was added at 3 g/sample, and in the competition assays (lanes C and D) 20 ng of unlabeled oligonucleotide was included. Specific shifts are indicated with asterisks.
TrxR1 described here. The cloned 0.8-kb genomic fragment had strong transcriptional activity in the different cell types analyzed. Because the deletion analysis revealed that the promoter activity was maintained within the Ϫ115 to ϩ167 region (the ⌬3 construct), we considered this region the core promoter and concentrated on its characterization.
The core promoter lacked classical TATA or CCAAT boxes but had an increased GC content and a probable CpG island involving the first part of the first exon. Using the shift assay approach four regions of the core promoter were found not to bind transcription factors present in crude nuclear extracts of A549 cells, whereas six specific fragments were identified as clearly binding nuclear proteins. We were able to identify transcription factors participating in the binding complexes for four of these six regions. The only regions producing shifts in the electrophoretic mobility shift assay analysis for which we have yet to identify the participating transcription factors are those covered by oligo-6 and oligo-9. In the other analyses of the transcription factor complexes, it was clear that Oct-1 bound oligo-1, whereas Sp1 bound oligo-4, oligo-7, and oligo-10. However, it should be noted that oligo-2 and oligo-3 also contained possible Sp1 sites as predicted by the computer analyses; these were found to be nonfunctional in this analysis. Moreover, the shift of oligo-7 was different from that of oligo-4 and oligo-10 in that Sp1 antibodies abolished the shifted band instead of producing a supershift, probably indicating that Sp1 indirectly bound other transcription factors binding oligo-7 and that this protein-protein interaction and thereby the complete DNA binding was abolished by the antibodies. Oligo-4 and oligo-10 bound Sp1 directly, and these two fragments also shared a strong consensus tandem Sp1 site. Finally we also showed that Sp3 bound the same oligonucleotides as those complexed by Sp1. The overall structure of the core promoter, its basal activity, and the profile of binding transcription factors have led to some conclusions regarding the nature of the regulation of transcription for human TrxR1.
Lack of classical TATA or CCAAT boxes, an increased GC content with functional Sp1 site(s) in the proximal promoter region, and a CpG island close to the transcriptional initiation site are features typical of housekeeping genes such as glycolytic enzymes, thymidylate synthase, adenine deaminase, and dihydrofolate reductase (44), the latter being studied as a model housekeeping promoter (45). Also Oct-1 is ubiquitously expressed and believed to govern the transcription of many housekeeping genes (46). TrxR1 mRNA is also well known to be expressed in many diverse lineages of cells and growth conditions as illustrated by the presence of TrxR1 mRNA detected in all human cells or tissues analyzed with general Northern blot screening methods (27,31) or by the fact that cDNA sequences for TrxR1 have been cloned from more than 30 different types of cells or tissues. 3 The transcriptional activity of the core promoter in the different cell types analyzed in the present study is hence in agreement, in principle, with the ubiquitous expression of the native gene. Based upon this expression pattern in combination with the structure and function of the core promoter as described herein, we thereby propose that TrxR1 should be considered to clearly belong to the class of human housekeeping genes. However, a significant difference from other housekeeping genes is found in the pattern of regulation of TrxR1 in response to cellular signaling.
TrxR1 is known to display a significant and fast (within hours) increase of protein as well as mRNA upon treatment of cells with different exogenous agents. This includes human epidermoid carcinoma A431 cells treated with epidermal growth factor, H 2 O 2 , or 1-chloro-2,4-dinitrobenzene (20) or thyrocytes treated with calcium ionophore (A23187) and phorbol 12-myristate 13-acetate (47). The latter agents induced TrxR also in human umbilical vein endothelial cells, although that induction was much less pronounced due to more than 10-fold higher basal TrxR levels in these cells compared with thyrocytes (48). In human bone marrow-derived stromal cells (KM-102) both phorbol 12-myristate 13-acetate in combination with A23187 or, alternatively, treatment with interleukin-1␤ or lipopolysaccharide significantly increased the TrxR1 mRNA levels within 4 h (27). In peripheral blood monocytes and myeloid leukemia cells (49) as well as osteoblasts (50) TrxR1 mRNA 3 As compiled from GenBank TM data in the NCBI UniGene cluster "Hs. 13046 TXNRD1" the human TrxR1 mRNA has been cloned as cDNA from at least the following cells or tissues: aorta, amnion fluid, B-cells, bladder, bone, bone marrow, brain, cervix, central nervous system, colon, ear, eye, foreskin, gall bladder, germ cell, head and neck, heart, kidney, lung, lymph, ovary, pancreas, placenta, prostate, skin, stomach, testis, thyroid, tonsil, uterus, and whole embryo. Additional data base mining revealed TrxR1 cDNA that was derived from at least liver, spleen, melanocytes, umbilical vein endothelium, retina, senescent fibroblasts, and T-cells.

FIG. 4. Identification of Oct1 binding.
Supershift assays were performed with the oligonucleotides that generated band shifts in Fig. 3. Polyclonal antibodies against Oct1 were added either before (b) or after (a) addition of the labeled oligonucleotide to the nuclear protein extract. See Fig. 3 (B lanes) for other experimental conditions. The band shifts obtained with the Oct1-positive control and oligo-1 (asterisks) were supershifted (arrows), whereas none of the other shifts were affected by these antibodies as seen in the figure. levels were shown to be increased above basal levels in a fast but transient manner by vitamin D 3 treatment. How is the increase of TrxR1 levels upon these diverse exogenous stimuli transmitted? It is clear that housekeeping-type promoters also may have responsive promoter elements and distant enhancers or silencers. It is interesting to note that Sp1 and Oct-1 are FIG. 5. Identification of Sp1 and Sp3 binding. To examine whether Sp1 or Sp3 participated in any of the band shifts seen in Fig. 3, supershift assays were performed with polyclonal antibodies against either Sp1 (A) or Sp3 (B) added before (b) or after (a) the labeled oligonucleotide. See Fig. 4 for experimental conditions. The slowest migrating band (asterisks in A) generated with the Sp1-positive control oligonucleotide as well as oligo-4 and oligo-10 was supershifted with Sp1 antibodies (arrows in A), whereas the two slower migrating bands (asterisks in B) were supershifted with antibodies recognizing Sp3 (arrows in B). Interestingly the slower migrating complex obtained with oligo-7 disappeared after using anti-Sp1 antibodies, but no supershifted complex could be detected (A, oligo-7). This result was repeated in two separate experiments, and even after overexposure of the autoradiography no supershifted oligo-7 complex was seen using anti-Sp1 antibodies despite the disappearance of the slowest migrating band. However, supershifts of the two other complexes with oligo-7 were seen using the antibodies against Sp3 (B, oligo-7). The sequence of oligo-7 contained no clear Sp1 or Sp3 binding motif as predicted with computer analysis (see "Results" and "Discussion").
FIG. 6. Summary of functional motifs in the core promoter of human TrxR1. This figure summarizes the binding of transcription factors to the core promoter region of human TrxR1 as found using A549 nuclear extracts. Oct1 binds to a consensus POU domain at Ϫ98 to Ϫ90 and Sp1/Sp3 bind to a GC-rich region immediately upstream of the transcription start site (arrow), which at Ϫ17 to Ϫ7 forms a strong tandem consensus Sp1/Sp3 motif as shown in the figure. Sp3 also bound to a region corresponding to the 5Ј-UTR (oligo-7), although no consensus motif could be identified and Sp1 seemed to indirectly bind to that same region. The areas where transcription factors were found to bind but which have not yet been identified are indicated by dashed circles. known to be able to physically interact (51), an interaction that cooperatively stimulates expression in the case of the small nuclear U2 RNA gene (52). Such Sp1/Oct-1 interaction may be one mechanism involved in executing the basal activity of the TrxR promoter. Binding of both Sp1 and Sp3 to the same GC boxes, as found in the present study, is known, and because Sp3 acts either as a repressor or an activator depending on the cellular context it has been suggested that variations in the cellular ratio between Sp1 and Sp3 may be another level of regulation. This mechanism of regulation has been reviewed elsewhere (43). Also, both Oct-1 and Sp1 are, like many other transcription factors, known to be sensitive to their redox status (53,54). The redox regulation of Oct-1 or Sp1 is, however, highly intricate, complex, and difficult to conclusively assess. For example, when transcription factors (including Sp1) are oxidized they often lose DNA binding activity (53,54). However, in vivo induction of O-linked glycosylation of Sp1 upon cellular production of reactive oxygen species was found to increase its transcriptional activation (55). On the other hand, oxidative damage to the Sp1-binding DNA motif in a promoter by formation of a single 8-oxodeoxyguanosine residue may disrupt the Sp1-dependent activation (56). The promoter-driven regulation of TrxR1 transcription should certainly be studied further in diverse cellular contexts, also including distant regulatory elements. Yet an additional and alternative model for TrxR1 regulation can also be proposed considering the unique functional organization of a housekeeping-type promoter in combination with ARE-mediated post-transcriptional regulation.
The presence of 3Ј-UTR AREs may enable a quick stabilization of mRNA and can thereby up-regulate protein levels in fast response to various signals (28). Upon many different exogenous stimuli, ROS are also produced as common mediators for intracellular signaling (35,56). One regulatory protein that is rapidly up-regulated upon formation of ROS is the p38 mitogen-activated protein kinase (57). The stress-activated p38 mitogen-activated protein kinase in turn up-regulates mitogenactivated protein kinase-activated protein kinase 2 (MK2) (58). Interestingly MK2 was recently shown to induce stabilization of ARE-containing mRNAs thereby executing their stabilization under intracellular formation of ROS (59 -61). TrxR1 contains several functional AREs (25,27) and is also known to be up-regulated by many exogenous agents (see above), which in turn are known to mediate intracellular ROS formation as a common denominator (32,35,56). In addition in cells, the TrxR1 protein has been reported to be rapidly inactivated by ROS targeting the selenocysteine residue (20). This chain of events makes it possible to propose an interesting model for TrxR1 regulation and function. With a strong constitutive transcription (suggested here) combined with ARE-regulated mRNA turnover and generally a short mRNA half-life in nonstimulated cells (25), TrxR1 thereby has the inherent capacity for a fast response to an increase of intracellular ROS in their role of stabilizing the mRNA via MK2 and the ARE motifs; this would occur concomitant with a momentary inactivation of the enzyme (20). However, once more TrxR1 has been synthesized rapidly as a result of the stabilized mRNA, the antioxidant properties of the newly produced enzyme would then be able to carry the cells back to a correct basal balance of the intracellular redox status while having allowed a transient burst of ROS to be a necessary component for the many diverse systems of intracellular signaling. This proposed model for TrxR1 regulation needs to be confirmed and is now the basis for our further studies of the cellular function and regulation of TrxR1.
In conclusion, we have cloned the proximal promoter for human TrxR1, and based upon its structure, basal activity, and the ubiquitous expression we have proposed that TrxR1 fulfills all the criteria of a housekeeping gene. However, a housekeeping gene with ARE-regulated mRNA degradation is to our knowledge a novel concept. This functional organization could possibly be explained by a role for TrxR1 in continuous protection of cells against oxidative damage while being in tune with and allowing bursts of reactive oxygen species generated at events of intracellular signaling cascades.