Regulation of expression of the phospholipid hydroperoxide/sperm nucleus glutathione peroxidase gene. Tissue-specific expression pattern and identification of functional cis- and trans-regulatory elements.

A sperm nucleus glutathione peroxidase (snGPx), which is closely related to the phospholipid hydroperoxide glutathione peroxidase (phGPx), was recently discovered in late spermatids. Both GPx isoforms originate from a joint ph/snGPx gene, but their N-terminal peptides are encoded by alternative first exons. The expression of the two enzymes is differentially regulated in various cells, but little is known about the regulatory mechanisms. To explore the tissue-specific regulation of expression of the two isoenzymes, we first investigated their tissue distribution. Whereas phGPx is expressed at low levels in many organs, snGPx was only detected in testis, kidney, and in the human embryonic kidney cell line HEK293. Subcellular fractionation studies and immunoelectron microscopy revealed a cytosolic localization. To explore the mechanistic reasons for the differential expression pattern, we first tested the activity of the putative phGPx and snGPx promoters. The 5'-flanking region of the joint ph/snGPx gene exhibits strong promoter activity. In contrast, the putative snGPx promoter, which comprises 334 bp of intronic sequences, lacks major promoter activity. However, it strongly suppresses the activity of the ph/snGPx promoter. These data suggest negative regulatory elements in the first intron of the ph/snGPx gene, and DNase protection assays revealed the existence of several protein-binding sites. The corresponding trans-regulatory proteins (SP1, ERG1, GATA1, SREBP1, USF1, and CREBP1) were identified, and in vivo binding of EGR1 and SREBP1 was shown by chromatin immunoprecipitation. These data indicate for the first time somatic expression of the snGPx and provide evidence for the existence of intronic negative cis-regulatory elements in the ph/snGPx gene. Our failure to detect an alternative snGPx promoter suggests that transcription of the ph/snGPx gene may be regulated by a joint basic promoter. The decision, which GPx isoform is expressed in a given cell, appears to be made by alternative splicing of a joint primary transcript.

Selenium-containing glutathione peroxidases (GPx) 1 constitute a family of antioxidative enzymes that are capable of reducing organic and inorganic hydroperoxides to the corresponding hydroxy compounds utilizing glutathione or other hydrogen donors as reducing equivalents (1,2). Up to now, five selenium GPx subtypes have been identified (3,4), but there are only 4 distinct seleno GPx genes. The most recently discovered member (5) of this enzyme family is the sperm nucleus glutathione peroxidase (snGPx). It constitutes a 34-kDa selenoprotein that is expressed at a high level in late spermatids of various species and has been implicated in chromatin condensation during sperm development and in antioxidative protection of sperm DNA. It is localized in the spermatid nuclei, and its expression is strongly impaired under selenium deficiency (5,6). The snGPx shares a high degree of protein-chemical similarity (5, 7) with the previously described phospholipid hydroperoxide glutathione peroxidase (phGPx), because both GPx isoforms originate from a joint ph/snGPx gene (5). In fact, their C-terminal peptides (about 60% of the snGPx) are identical and they are encoded for by exons 2 to 7 of the joint ph/snGPx gene. In contrast, the N-terminal peptides of the two enzymes originate from alternative first exons (Scheme 1). The N-terminal peptide of snGPx 7 (15 kDa) contains an argininerich nuclear insertion sequence, which enables nuclear localization and is unique among selenoproteins (5). In contrast, the N-terminal peptide of the phGPx is much shorter (about 3 kDa) and lacks a nuclear import sequence. Because both GPx isoforms are encoded for by the same gene, the two enzymes may be considered splicing variants of a joint ph/snGPx pre-mRNA (3,4). However, it is not clear whether expression of the two proteins really involves the formation of a joint pre-mRNA or whether there are distinct transcription initiation sites and alternative promoters. Considering the given genetic composition (Scheme 1) it may be speculated that the 5Ј-flanking region of the ph/snGPx gene is of functional relevance only for phGPx transcription. In contrast, the immediate 5Ј-flanking region of exon E1b (first snGPx exon) that includes the 5Ј part of intron 1 (I1a in Scheme 1) and exon 1a (first phGPx exon) may contain regulatory elements for snGPx transcription. In fact, stringent in silico structural analysis of I1a indicated the presence of putative transcription factor binding sites. These data suggested the existence of an independent transcriptional regulation of phGPx and snGPx expression.
The various isoforms of seleno-GPx show a remarkable tissue-specific expression pattern (3). The phGPx is expressed at relatively low levels in most cells and tissues tested (8,9) but is much more prominent in testis (10). By contrast, snGPx mRNA has only been detected in testis and its expression appears to be restricted to late stages of spermatogenesis (5). Unfortunately, the regulatory mechanisms involved in high level expression of phGPx in germinative cells and repression of snGPx expression in most somatic tissues are unknown. Moreover, the molecular processes of transcriptional activation of snGPx expression during late sperm development remain to be investigated.
The lack of experimental data on the regulation of expression of the ph/snGPx gene and the remarkable tissue-specific expression pattern of the two GPx isoforms prompted us to study the regulatory mechanisms involved in ph/snGPx expression. Here we report for the first time that snGPx does not only occur in the nucleus of late spermatids but also in the cytosol of interstitial kidney cells. Moreover, we found that the regulation of expression of the joint ph/snGPx gene appears to be a complex process that requires cis-regulatory elements in the 5Ј-flanking region of the ph/snGPx gene and the binding of negative trans-regulatory proteins, such as SREBP1 and EGR1, to intronic sequences (I1a). The lack of major promoter activity in the 5Ј-flanking region of the first snGPx exon argues against the existence of alternative phGPx/snGPx promoters but strongly suggests alternative splicing of a joint ph/snGPx pre-mRNA as mechanistic reason for differential expression of the two enzymes.
Cell Lines and Culture Conditions-Human embryonic kidney cells 293 (HEK293) were obtained from the German Collection of Microorganisms and Cell Cultures (Braunschweig, Germany). The cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% (v/v) fetal calf serum and antibiotics (100 units/ml penicillin and 100 g/ml streptomycin) at 37°C under 5% CO 2 .
Animal Experiments-Mice (C57BL/N6, 10 -15 weeks old) were sacrificed by diethyl ether inhalation and the major organs were prepared. After extensive washing with PBS to remove blood, the tissue was homogenized on ice in 5 volumes of TBS (25 mM Tris-HCl, 130 mM NaCl, 5 mM KCl, protease inhibitor mixture, pH 7.4) in a glass Dounce homogenizer. Cell debris was spun down at 20,000 ϫ g and aliquots of the supernatants were subjected to immunoblotting. For nuclear extracts HEK293 cells, mouse testis or kidney were lysed in 10 mM HEPES buffer, pH 7.9, containing 10 mM KCl, 1.5 mM MgCl 2 , 0.1 mM EDTA, 0.35 M sucrose, 0.5 mM DTT, and a proteinase inhibitor mixture (see "Materials"). After 10 min on ice the cells were homogenized in a glass Dounce homogenizer and recentrifuged at 7000 rpm. The nuclear pellet was resuspended in 20 mM Hepes buffer, pH 7.9, containing 10% glycerol, 420 mM NaCl, 1.5 mM MgCl 2 , 0.1 mM EDTA, 0.5 mM DTT, and proteinase inhibitor mixture and was incubated on ice for 30 min. The suspension was centrifuged at 10,000 ϫ g for 15 min at 4°C and the supernatant was stored at Ϫ80°C. The protein concentration in all samples was measured by colorimetric assay (Roti-Quant, Roth, Karlsruhe, Germany).
Reverse Transcriptase-Polymerase Chain Reaction-Total RNA was isolated from mouse tissues or HEK293 cells using the RNA MIDI or MINI Kits from Qiagen (Hilden, Germany) and 3 g were reversetranscribed at 37°C for 90 min in 45 l of Tris-HCl buffer, pH 8.2, containing 8 mM MgCl 2 , 30 mM KCl, 1 mM dithiothreitol, 100 g/ml bovine serum albumin, 30 units of RNase inhibitor, 0.166 mM dNTPs, 150 pmol of oligo(dT) primer, and 15 units of reverse transcriptase. To stop the reaction samples were heated to 95°C for 10 min. RT-PCR signals were normalized for the signals of the human or murine glyceraldehyde-3-phosphate dehydrogenase (GAPDH), which were amplified with the following primer combinations: human GAPDH, 5Ј-TCG-GAGTCAACGGATTTGGTCGTA-3Ј and 5Ј-ATGGACTGTGGTCATGA-GTCCTTC-3Ј; murine GAPDH, 5Ј-TCGGTGTGAACGGATTTGGCCG-TA-3Ј and 5Ј-ATGGACTGTGGTCATGAGCCCTTC-3Ј. The following primer combinations were used for amplification of the murine and human snGPx: 5Ј-TCGCCGGATGGAGCCCATTCCT-3Ј and 5Ј-ACGC-AGCCGTTCTTATCAATGAGAA-3Ј (murine); 5Ј-CCGGCGGAAGAAG-CCCTGTCC-3Ј and 5Ј-CGAATTTGACGTTGTAGCCCGCG-3Ј (human). For amplification, 1 l of the reverse transcription reaction was used. After initial denaturation for 4 min at 94°C, 30 cycles of PCR were performed. Each cycle consisted of a denaturing period (40 s at 94°C), an annealing phase (murine GAPDH: 60s at 69°C, human GAPDH: 60s at 66°C, snGPx: 60 s at 67°C), and an extension period (120 s at 72°C). After the last cycle, all samples were incubated for an additional 10 min at 72°C. PCR products were separated by 2% agarose gel electrophoresis, and the DNA bands were stained with ethidium bromide.
Two-dimensional Electrophoresis and Immunoblotting-The cytosolic proteins of kidney homogenates were mixed with 10 l of sample buffer (1.43 g/ml urea, 10 mg/ml dithiothreitol, 10% (w/v) Triton X-100, 107 l/ml Servalyt 3-10, 26 l/ml Servalyt 6 -8, 16 l/ml Servalyt 2-4) and then subjected to isoelectric focusing (first dimension) in acrylamide cylinders, in which a linear pH gradient between 2 and 10 was SCHEME 1. Structure of the ph/snGPx gene and putative transcription factor binding sites in intron I1a. The gray boxes represent the phGPx exons (E1 to E7) and the introns (I1 to I6) are indicated by the line between the exons. The start codons for the mitochondrial and cytosolic phGPx isoforms are shown as 5Ј-ATG and 3Ј-ATG, respectively. ATG(sn) represents the translational initiation site of the snGPx. Sec indicates the localization of the catalytic selenocystein and secis marks the selenocystein insertion sequence required to prevent premature termination of translation at the amber triplet encoding for the selenocystein. Analysis of the transcription factor binding sites was carried out with the MatInspector program.
adjusted. SDS-PAGE (second dimension) was subsequently carried out in a 12.5% polyacrylamide gel. After SDS-PAGE, the two-dimensional chromatograms were either stained with 0.2% (w/v) Coomassie Brilliant Blue R-250 or were transferred to a nitrocellulose membrane NC45 (Serva, Heidelberg, Germany) by a semidry blotting procedure. To stain the immunoblots the nitrocellulose membrane was first incubated for 1 h with a 1:4000 dilution of a monoclonal antibody raised against the Sec46Cys mutant of the human cytosolic phGPx. This antibody prepared in our laboratory strongly cross-reacts with the rat and murine phGPx but does not distinguish between phGPx and snGPx. After extensive washing with PBS, the membrane was incubated with a peroxidase-labeled goat anti-mouse IgG antibody (Sigma) and immunoreactive protein bands were visualized with a 3,3Ј-diaminobenzidine/H 2 O 2 mixture.
Functional Promoter Assays-The pBlueTOPO reporter vector (Invitrogen, Groningen, The Netherlands) containing the ␤-galctosidase as reporter gene was used for analysis of promoter activity of subcloned promoter fragments. For the assays, putative promoter fragments of phGPx and snGPx were amplified by PCR and subcloned into the pBlue-TOPO reporter vector following the user's manual. PCR was carried out with a Biometra TRIO Thermoblock 2.51BB (Biometra, Göttingen, Germany) using purified plasmid DNA (Qiagen, Hilden, Germany) of the subcloned ph/snGPx gene as template and the following primer combinations were used (Fig. 5): 5Ј-flanking region of the ph/snGPx gene (construct C1, 325 bp), 5Ј-CGCGTCCCTATCACT-GGGGCATG-3Ј and 5Ј-TGGTGCCTGCCAGACCAGGCG-3Ј; immediate 5Ј-flanking region of the first snGPx exon (construct C2, 334 bp), 5Ј-TGGGCTACTGGGAACTTGGAGGA-3Ј and 5Ј-ATGCCCGCCGGTCT-GTGCGTC-3Ј; extended 5Ј-flanking region of the first snGPx exon (construct C3, 662 bp) 5Ј-CGCGTCCCTATCACTGGGGCATG-3Ј (same as for construct C1) and 5Ј-ATGCCCGCCGGTCTGTGCGTC-3Ј. For transient transfection HEK293 cells were resuspended at a density of 0.5 ϫ 10 6 cells/5 ml of medium and plated in 60-mm culture dishes. After a culturing period of 24 h cells were transfected with a calcium phosphate transfection kit (Invitrogen). For transient transfection we used 10 g of the ␤-galactosidase promoter construct and 5 g of an internal reference plasmid containing the luciferase cDNA (pGL3, Promega, Mannheim, Germany). After 72 h, cells were harvested and both ␤-galactosidase and luciferase activities were assayed. The ␤-galactosidase activities were normalized for transfection efficiency using the data of the luciferase assay.
DNase I Protection Assay-A genomic fragment containing intron I1 of the ph/snGPx gene was subcloned into the NcoI and PstI site of the Litmus29 vector (Promega). After excision with XbaI and BamHI the probe was radioactively labeled using T4 polynucleotide kinase. The labeled probe was then treated with PstI or NcoI to achieve selective labeling of either the sense or antisense strand. Aliquots of the labeled probe containing 10,000 -20,000 cpm of 32 P-labeled DNA were added to reaction mixtures consisting of a 40 mM Hepes buffer, pH 7.9, containing 100 mM KCl, 12.5 mM MgCl 2 , 1 mM EDTA, 20% glycerol, 1 mM DTT, 1 g of poly(dI-dC), and 10 -20 g of nuclear proteins in a total volume of 50 l. The mixture was kept on ice for 30 min before addition of 50 l of 5 mM CaCl 2 , 10 mM MgCl 2 , and 2-5 units of DNase I depending on the input of nuclear proteins. Digestion was carried out at 4°C for 5 min and terminated by adding 100 l of a solution containing 200 mM NaCl, 20 mM EDTA, and 1% SDS. DNA was phenol extracted and precipitated with ethanol. The digestion products were analyzed on 6% acrylamide sequencing gels containing 8 M urea. The protected sequences were identified by comparison to the migration of Maxam-Gilbert sequencing products. Gels were dried and exposed to x-ray films overnight at Ϫ80°C.
Electrophoretic Mobility Shift Assay (EMSA) and Supershifts-EMSA was carried out with the DIG Gel Shift Kit (Roche Molecular Diagnostics) following the users manual. In the first step singlestranded complementary oligonucleotides containing the binding sequences for transcription factors were annealed and end-labeled with digoxigenin. The labeled probes (48 fmol of double-stranded oligonucleotides) were then incubated for 30 min at 4°C with 10 g of nuclear extract proteins in 40 mM Hepes buffer, pH 7.9, containing 100 mM KCl, 12.5 mM MgCl 2 , 1 mM EDTA, 20% glycerol, 1 mM DTT, 1-2 g of poly(dI-dC), 0.1-0.2 g of poly(L) lysine (total assay volume of 15 l). Then the mixtures were subjected to electrophoresis on a 6% polyacrylamide gel with 0.5-fold TBE running buffer. The digoxigenin-oligonucleotide/protein complexes were transferred to a Hybond-N blotting membrane (Amersham Biosciences) and the shift bands were visualized following the users manual. For competition studies unlabeled probe competitor or consensus oligonucleotides (Santa Cruz Biotechnology, Santa Cruz, CA) were added at a 166-fold excess over the digoxigenin-labeled probe. For supershift studies, polyclonal antibodies from Santa Cruz Biotechnology were used. When either competitor or antibodies were used, they were preincubated with the nuclear extracts at 4°C for 10 or 60 min before the addition of the digoxigenin-labeled probe DNA.
Chromatin Immunoprecipitation-10 7 HEK293 cells were washed twice with PBS and the cellular proteins were cross-linked to DNA by adding formaldehyde to a final concentration of 1% for 15 min at room temperature. Cells were subsequently washed, removed from the culture dishes by scraping, and resuspended in 1 ml of PBS. Cells were lysed in 1.5 ml of lysis buffer (10 mM Hepes buffer, pH 7.9, containing 10 mM KCl, 1.5 mM MgCl 2 , 0.1 mM EDTA, 0.5 mM DTT) and the proteinase inhibitor mixture. After 10 min on ice the cells were homogenized in a glass Dounce homogenizer and centrifuged at 7,000 rpm. The nuclear pellet was resuspended in 20 mM Hepes buffer, pH 7.9, containing 10% glycerol, 420 mM NaCl, 1.5 mM MgCl 2 , 0.1 mM EDTA, 0.5 mM DTT, and proteinase inhibitor mixture and was sonicated on ice to fragment the genomic DNA. The mixture was then centrifuged to remove cell debris and aliquots were stored at Ϫ80°C. For chomatin immunoprecipitation aliquots were diluted in RIPA buffer (50 mM Tris-HCl buffer, pH 7.5, containing 150 mM NaCl, 0.1% sodium desoxycholate, 1% Triton X-100, 0.25 mM EDTA, and the proteinase inhibitor mixture) and samples were incubated overnight at 4°C with 2.5 l of a rabbit polyclonal anti-Egr1 antibody (588) or with a rabbit anti-SREBP1 antibody (H-160) obtained from Santa Cruz Biotechnology. Immune complexes were precipitated by adding 50 l of protein Aagarose (Santa Cruz Biotechnology) for 3 h with low speed spinning. Precipitates were washed twice with 1 ml of RIPA buffer, twice with 1 ml of PBS, and DNA-protein complexes were eluted from the agarose by a 15-min incubation with 250 l of elution buffer (1% SDS, 100 mM NaHCO 3 ). Elution was repeated, the eluates were pooled, NaCl was added to reach a final concentration of 0.3 M, and the samples were incubated at 65°C for 4 h to reverse formaldehyde-induced cross-linking. The proteins were then digested with 2 l of proteinase K (20 mg/ml) in 40 mM Tris-HCl buffer, pH 7.5, containing 10 mM EDTA. The samples were incubated for 1 h at 60°C and the DNA was extracted with phenol/chloroform followed by ethanol precipitation. The DNA pellet was resuspended in 50 l of sterile H 2 O and 10 l of DNA solution was used as a template for PCR. The following primer combinations were designed to amplify the precipitated binding regions of the transcription factors (Fig. 11) and 35 PCR cycles were run: P1 (5Ј-CCATG-GTGGGCTACTGGGAACTT-3Ј), P1Ј (5Ј-TCCCGCGCCGAGGCCTAGC-C-3Ј), P2 (5Ј-AGGCCTCGGCGCGGGAGGTC-3Ј), P2Ј (5Ј-ATGCCCGC-CGGTCTGTGCGTC-3Ј). The PCR products were separated on 2% agarose gel and visualized by ethidium bromide staining.
Immunohistochemistry-For immunohistochemistry the animals (two rats and two mice) were perfused with ice-cold 4% (w/v) paraformaldehyde (dissolved in PBS). Kidneys were prepared, cryopreserved in 20% sucrose for 24 h at 4°C, and frozen in the gaseous phase of liquid nitrogen. Cryostat sections of 10 -20 m thickness were prepared and quenched in 50 mM NH 4 Cl/PBS for 30 min. The sections were blocked with 10% goat serum, 0.1% saponin in PBS for 1 h at room temperature and exposed to the primary monoclonal anti-PH-GPx antibody (1:200 diluted) overnight at 4°C. After washing in 0.1% saponin/PBS, sections were incubated overnight at ϩ4°C with a biotinylated anti-mouse IgG antiserum (Vector Laboratories, Burlingame, CA; diluted 1:250). For staining, the sections were exposed to avidin-biotin peroxidase complex reagent (Vector Laboratories) for 2 h at room temperature, and the immunoreaction was visualized with 3,3Ј-diaminobenzidine as a chromogen (incubation for 8 min at room temperature with 0.07% 3,3Јdiaminobenzidine and 0.001% hydrogen dissolved in 0.1 M PBS). Prior to inspection the sections were counterstained with hematoxylin eosin. Control staining was performed appropriately using a diluted preimmune serum.
Immunoelectron Microscopy-For ultrastructural studies, ph/snGPximmunostained sections were osmicated (1% OsO 4 in 6.84% sucrose dissolved in PBS) for 5 min, dehydrated in graded ethanol, and flat embedded in Epon medium (Merck, Darmstadt, Germany) between silane-coated slides. Ultrathin sections were cut on a ultratome, mounted on single slot grids coated with a Formvar film, and stained with lead citrate and uranyl acetate. A Zeiss EM 900 electron microscope (Zeiss, Jena, Germany) was used for examination.
Activity Assay-The phospholipid hydroperoxidase activity was assayed measuring the decrease in absorbance at 340 nm with a coupled optical test (11). The assay mixture consisted of 0.1 M Tris-HCl, pH 7.6, containing 5 mM EDTA, 0.1% Triton X-100, 0.2 mM NADPH, 3 mM glutathione, and 1 unit of glutathione reductase in a total assay volume of 0.2 ml. The enzyme preparation (5-20 l) was preincubated at room temperature in the assay buffer for 10 min and then the reaction was started by addition of 10 l of a methanolic solution of high performance liquid chromatography-purified phospholipid hydroperoxide reaching a final substrate concentration of 50 M. The hydroperoxy phospholipids were prepared by incubating for 10 min soybean phospholipids in 0.2 M borate buffer, pH 9, containing 3 mM sodium desoxycholate and 1 mg/ml soybean lipooxygenase. After lipid extraction the conjugated dienes were prepared by reverse phase-high performance liquid chromatography (12) and the salt included in the solvent system was removed by solid-phase extraction. Final quantification of the peroxide preparation was carried out spectrophotometrically measuring the absorbance at 235 nm using a molar extinction coefficient of 25,000 (M cm) Ϫ1 .

RESULTS
Tissue-specific Expression of snGPx and Its Intracellular Localization-It has been previously reported that phGPx is ubiquitously expressed in small amounts in many cells and tissues (3,8,9) but at much higher levels in testis (10,13). In addition, Northern blot analysis suggested a high level expression of the snGPx mRNA in testis and its absence in liver, kidney, spleen, brain, and other organs (5). Within the testis the enzyme was mapped predominantly to the nuclei of late spermatids (5). Because spermatids are not suitable for mechanistic studies, the tissue-specific expression pattern of the snGPx mRNA was reinvestigated by semiquantitative RT-PCR. Here we found that the snGPx mRNA is present at a high level in testis and at lower levels in kidney (Fig. 1). In all other major organs (ileum, skin, heart, brain, liver, lung, stomach, spleen, muscle, and submandibular gland) we were unable to detect this messenger (not shown). In some experiments faint snGPx bands were observed in lung and stomach. In the snGPx-positive tissues (testis and kidney) we consistently detected an additional PCR product migrating with a higher molecular weight (Fig. 1). This PCR fragment was subcloned and sequenced, and the data indicated an incomplete splicing product that still contained intron I1b (Scheme 1). However, all other introns (I2-I6) had been spliced out completely. These data suggest problems of the splicing machinery during snGPx expression. It is likely that proteins binding to I1a of the joint ph/snGPx pre-mRNA may prevent proper splicing of the primary transcript.
For mechanistic studies on the regulation of gene expression, solid tissues have several limitations. Thus, we searched for a permanent cell line, in which the snGPx is expressed. For this purpose several human and murine cell lines were tested and expression of the snGPx mRNA was found in the human embryonic kidney cell line HEK293 (data not shown). Interestingly, this cell line also expresses phGPx as indicated by RT-PCR with phGPx-specific primer combinations.
To confirm the tissue-specific expression of snGPx on the protein level, immunoblots with a monoclonal anti-ph/snGPx antibody were carried out. For this purpose 20,000 ϫ g supernatants were prepared from murine tissues and equal protein amounts were applied to SDS-PAGE. From Fig. 2A it can be seen that immunoreactive proteins migrating in the expected 34-kDa region were present in the kidney lysis supernatant suggesting snGPx expression in this organ. In a separate experiment electrophoretic co-migration of the kidney snGPx with an snGPx standard prepared from human spermatids was observed. Moreover, an immunoreactive band of minor intensity was detected in the 20-kDa region that may be because of phGPx. By comparison, large amounts of this low molecular weight protein were found in the testis high speed supernatants. These data indicate that in testis cytosol the phGPx predominates and the snGPx appears to be absent.
The kidney snGPx exhibited a catalytic activity for reduction of hydroperoxy phosphatidylcholine of 0.057 units/mg of protein. This specific activity is similar to previously reported values (0.028 units/mg) (11). Our attempts to purify the native snGPx from murine kidney supernatants by conventional protein purification techniques were not successful. Running the supernatant over sequential Mono-Q FPLC and gel filtration resulted only in 8-fold enrichment of the immunoreactive material with substantial loss of enzymatic activity. In the most active FPLC fractions the specific activity could only be enriched by a factor of 2 (0.12 units/mg) compared with the lysis supernatant (see above).
It was previously shown that phGPx forms covalently linked polymers (10) that may migrate at a higher molecular weight region. Thus, the 34-kDa band observed in kidney cytosol may be because of a phGPx dimer instead of a snGPx monomer. To distinguish between these two protein species we carried out two-dimensional electrophoresis (Fig. 2B) and found that the immunoreactive material from the kidney lysate supernatant migrated with an apparent isoelectric point of about 7. However, a more alkaline isoelectric point has been reported for phGPx (7,10). It should be stressed that in two-dimensional electrophoresis, we did not observe major expression of the 20-kDa phGPx. Comparison of Coomassie-stained immunoblots revealed that snGPx is not a major protein in kidney cytosol. In the region where the immunoreactive material was detected, significant protein was not observed.
The relatively large amounts of snGPx detected in kidney cytosol were somewhat surprising because the enzyme contains a nuclear import sequence and thus, should be localized preferentially in the nuclei. However, subcellular fractionation studies combined with Western blot analysis indicated the predominant cytosolic localization of the snGPx in kidney and its lack in the nucleus (Fig. 2C).
The available ph/snGPx antibodies do not distinguish between the two isoforms and thus, it was not possible to discriminate between the two GPx isoforms using immunohistochemistry. Fortunately, the Western blots indicated predominant expression of the 35-kDa snGPx in the kidney and thus, immunohistochemistry of kidney cross-sections will mainly reflect the cellular and subcellular distribution of the snGPx. In murine kidney, ph/snGPx is mainly expressed in cortical and medullary interstitial cells (Fig. 3). In contrast, the glomeruli and the epithelial cells of the tubuli and collecting ducts were free of immunoreactive material. Similar results were also obtained with rat kidneys (data not shown). These data con-FIG. 1. Expression of snGPx mRNA in murine testis and kidney. Total RNA was prepared from various murine tissues (100 mg wet weight; ileum, skin, heart, brain, liver, lung, stomach, spleen, muscle, submandibular gland, testis and kidney) and RT-PCR was carried out as described under ''Experimental Procedures.'' As control water was used instead of the RNA preparations. The intensity of the GAPDH band, which was used as internal standard for PCR amplification, was comparable for all RNA preparations (data not shown). Analysis of the snGPx negative tissues is not shown. The lower 527-bp band indicates the snGPx mRNA and the upper 740-bp band an incomplete splice product (see text).
trast with previous reports suggesting the expression of the phGPx in renal epithelial cells (14). Because of the higher density of interstitial cells in the medulla the level of ph/snGPx expression is apparently higher in this renal compartment (panel A). In additional experiments we investigated the intracellular distribution of the snGPx in renal cortex and medulla by immunoelectron microscopy. From Fig. 4 the immunoprecipitate is seen to localize in the cytoplasm and below the cellular membrane of interstitial (panel A) and vascular endothelial cells (panel C). Here again, the tubular epithelium did not show any immunoreactivity (panel B).
Functional Promoter Assays-If differential expression of phGPx and snGPx in a given cell is transcriptionally regulated, distinct transcription initiation sites and different basic promoter regions for the two isoenzymes would be expected. To find out whether there is significant promoter activity in the 5Ј-flanking region of exon 1b (Scheme 1) functional promoter assays were carried out. For this purpose ␤-galactosidasebased reporter gene constructs were transfected into HEK293 cells and ␤-galactosidase activity was assayed (Fig. 5). For control studies the ␤-galactosidase reporter gene construct without inserted oligonucleotide was used (construct C0). Construct C2, which contained a 334-bp fragment (I1a) of the 5Ј-flanking region of E1b (putative minimal snGPx promoter), only exhibited minor promoter activity. When compared with the putative ph/snGPx promoter, which involves the 5Ј-flanking region of the ph/snGPx gene and the 5Ј-untranslated region of the corresponding mRNA (construct C1), its activity was lower by 2 orders of magnitude. In additional experiments the impact of I1a on the activity of the phGPx promoter was tested (construct C3). Surprisingly, I1a was found to suppress the activity of the phGPx promoter in HEK293 cells by more than 95%. These data indicate that I1a may contain negative regulatory elements for ph/snGPx expression and that the corresponding trans-acting inhibitor proteins may be present in HEK293 cells.
Identification of Protein Binding Sites in I1a-The results of the functional promoter assays suggests that negative regulatory proteins bind to I1a. To investigate DNA/protein interactions in this region we carried out DNase protection assays using nuclear extracts of mouse testis and HEK293 cells as sources for the trans-regulatory proteins. The data presented in Fig. 6 indicate: (i) the existence of at least five distinct protein-binding regions (FP1-FP5) in I1b and (ii) the presence of the corresponding trans-acting proteins in the nuclei of testis and HEK293 cells. The lack of DNase protection by albumin (data not shown) indicates the specificity of protein binding.
Next, the sequences of the different protein-binding sites were determined and examined for the presence of transcription factor binding motifs. As evident from Fig. 7A FP1 contains the consensus binding sequences for the stimulating proteins (SP1-like factors) and for the early growth response gene product/Wilms tumor suppressor gene product (EGR/WT). EMSA were subsequently carried out to test the presence of the corresponding trans-acting binding factors in testis and HEK293 cells. With testis extracts we observed a single high molecular weight shift band that was competed off by nonlabeled competitor (Fig. 7B). By contrast, two shift bands were detected when nuclear extracts of HEK293 cells were used. Addition of a consensus oligonucleotide for SP1 caused a disappearance of the upper SP1 shift band (Fig. 7C). In a similar experiment the strong upper SP1 shift band from HEK293 extracts also disappeared, whereas the lower band of minor intensity remained unaltered. The lower shift band, however, was eliminated with a consensus oligonucleotide for the EGR transcription factor family (Fig. 7B). Because HEK293 cells are embryonic kidney cells adult kidney cells were also examined for expression of the corresponding transcription factors. As shown in Fig. 7D we observed two shift bands when murine kidney nuclear extracts were added to the assays. Here again, the upper band disappeared completely when a SP1 consensus oligonucleotide was added, whereas the lower band was quenched off in the presence of EGR competitor. In a separate experiment an EGR1 antibody was added to the reaction mixture and in this sample the EGR/WT shift band was removed. However, we were not able to detect a clear supershift signal. Taken together, these data suggest that footprint 1 is caused by the binding of a member of the SP1 transcription factor family (15) and/or EGR1 (16) to appropriate binding sites in I1a. The differential quenching behavior of testis and HEK293 nuclear extracts suggest that the testis extracts may not contain sufficient amounts of EGR1.
When the nuclear extracts of HEK293 cells and murine kidney were used as source of nuclear proteins an additional FIG. 2. Immunoblot analysis of snGPx expression. The 20,000 ϫ g supernatant was prepared from the 50% homogenates of solid murine tissues and aliquots were applied for analysis. After separation the proteins were transferred to a nitrocellulose membrane by semidry blotting and immunoreactive bands were stained using the monoclonal ph/snGPx antibody. A, SDS-PAGE: 50 g of lysate supernatant was applied to SDS-PAGE. B, two-dimensional electrophoresis: 50 g of kidney homogenate supernatant was used for isoelectric focusing (horizontal separation) and SDS-PAGE was carried out in the second dimension (vertical separation). Protein blotting and immunostaining was performed as described above. C, subcellular localization of snGPx in murine kidney. Nuclear extract and cytosol were prepared from murine kidney as described under ''Experimental Procedures'' and 50 g of total protein was applied to SDS-PAGE. Protein blotting and immunostaining were carried out as described above. very low shift band (Fig. 7, B and C) was observed and this band does not appear when testis nuclear extracts were used. Unfortunately, it has not yet been possible to identify the corresponding trans-acting protein.
FP2 (Fig. 8A) contains a GATA1 binding motif (17) and EMSA indicated a single shift band when the nuclear extracts from testis and HEK293 cells were used as source of the binding proteins (Fig. 8B). This band was competed off by addition of unlabeled competitor (Fig. 8B) and by a GATA consensus oligonucleotide (Fig. 8C). In FP3 (Fig. 9A) a SREBP1-binding site was detected (18) and EMSA experiments indicated single shift bands when the nuclear extracts of testis and 293 cells were incubated with the corresponding DNA probe (Fig. 9B). Addition of unlabeled competitor eliminated the shift bands. A similar result was obtained when a SREBP1 consensus oligonucleotide was added (Fig. 9C). FP4 and FP5 (Fig. 10) contain binding sites for USF1 (Fig. 10A) and CREB (Fig. 10B), respectively (19,20). After addition of the nuclear extracts of testis and HEK293 cells to the corresponding DNA probes we detected strong shift bands (Fig. 10, C and D). Addition of unlabeled competitor (data not shown) and consensus oligonucleotides caused a disappearance of these signals. Taken together, the EMSA experiments indicate that cis-regulatory elements for the binding SP1, EGR1, GATA1, SREBP1, USF1, and CREBP are present in the 5Ј-flanking region of E1b (Scheme 2) and that the corresponding trans-regulatory factors are expressed in murine testis and HEK293 embryonic kidney cells.
In Vivo Binding of ERG1 and SREBP1 to the 5Ј-Flanking Region of E1b (Putative snGPx Promoter)-The EMSA experiments indicated that the above mentioned transcription factors are present in the nuclei of testis and HEK293 cells and that they bind in vitro to the corresponding cis-regulatory sequences in the 5Ј-flanking region of E1b. To determine whether they also bind in vivo, chromatin immunoprecipitation was carried out. For this purpose DNA-binding proteins of HEK293 cells were covalently linked to genomic DNA by treatment of the cells with formaldehyde. The DNA-protein complexes were then sheared by sonication, and specific protein-DNA complexes were immunoprecipitated with antibodies against SREBP1 and ERG1. Covalent linkage was reversed and the precipitated double stranded DNA was amplified by PCR with binding site-specific primers. From Fig. 11B it can be seen that PCR signals were obtained when the DNA/protein adducts were immunoprecipitated with antibodies against SREBP1 and ERG1 indicating the in vivo binding of the transcription factors. In contrast, the corresponding controls, in which immunoprecipitation was performed without specific antibodies, did not show any PCR signal.  (magnification ϫ40). B, lack of immunoreactive material in glomeruli (magnification ϫ400). C and D, immunoreactive material in the renal medulla was also detected in interstitial cells (black arrows), whereas the tubular epithelium (red arrowheads) was immunonegative (magnification ϫ400 for C and ϫ1000 for D). Nonimmune controls of mouse kidney (E, glomerulum; F, medullary collecting ducts (magnification ϫ100).

DISCUSSION
GPxs constitute a family of functionally related enzymes and the phGPx and snGPx are two members of this family. Whereas phGPx is expressed at low levels in many cells and tissues (3,8), snGPx was only detected in late spermatids of several mammalian species (5). In these cells the enzyme was predominantly localized in the nucleus and has been impli-cated in chromatin structuring during sperm development and in antioxidative defense of the DNA (4 -6). In this paper we report for the first time that snGPx is also expressed in selected somatic cells (interstitial kidney cells). Immunohistochemical staining, immunoelectron microscopy, and subcellular fractionation studies indicate that the enzyme is located predominantly in the cytosol. This localization was rather surprising because snGPx contains a nuclear import peptide that has been proposed to target the protein to the nucleus. Currently, it is unclear why the enzyme is transported into the nucleus in late spermatids and why it remains in the cytosol in interstitial kidney cells. It is possible that nuclear import requires addi-  I1aϩE1a) were ligated into a ␤-galactosidase-based reporter gene and HEK293 cells were transfected with different constructs. To correct the ␤-galactosidase activity measured for transfection efficiency cotransfections with a luciferase containing control plasmid (pGL3) were carried out. ␤-Galactosidase and luciferase activity were measured as described under ''Experimental Procedures.'' A, length (bp) of the PCR fragments of the 5Ј-flanking regions used for reporter gene construction. B, ␤-galactosidase activity of the different promoter constructs corrected for luciferase transfection. As positive control, a 325-bp fragment of the 5Ј-flanking region of E1a (phGPx promoter) was used. As negative control (C0) the linearized vector was religated without any insert.

FIG. 6. Footprint analysis for intron I1a of the ph/snGPx gene.
DNase protection assays (footprints) were carried out as described under ''Experimental Procedures.'' The areas of quenched signals indicate the protein-binding sites (FP1 to FP5). Unspecific binding was tested with albumin as a probe, but we did not observe any signal quenching (data not shown). The sequences of the footprints (Figs. 7-10) were determined with the G ϩ A ladder. tional factors, which may not be present at the required concentration in the kidney cells.
The kidney-specific function of the snGPx has not yet been investigated. Because the enzyme is mainly expressed in interstitial cells but not in the glomeruli, the tubulus epithelium, or the collecting ducts, it is unlikely to play a role in urine production. Alternatively, the renal snGPx may be involved in antioxidative defense as it has been suggested before for other GPx isoforms (21). It is well known that there is an active oxidative metabolism in renal cortex, which may be accompanied by the formation of reactive oxygen species (22,23).
Compared with other selenoperoxidases, which are encoded for by separate genes, phGPx and snGPx originate from a joint ph/snGPx gene (4,5). Nevertheless, both isoforms are differen- tially expressed in various cells and tissues, but virtually nothing is known about the regulatory mechanisms. Here we report that the minimal promoter of the phGPx (cytosolic isoform), which comprises the 5Ј-flanking region of the ph/snGPx gene and the 5Ј-untranslated region of its mRNA, exhibits major promoter activity. By contrast, the promoter activity of the 5Ј-flanking region of E1b (putative snGPx promoter) was almost negligible. These data are difficult to reconcile with the existence of alternative phGPx/snGPx promoters. They rather suggest that transcription of the ph/snGPx gene is driven by joint regulatory sequences located upstream of the 3Ј-ATG of the ph/snGPx gene. Interestingly, the joint ph/snGPx promoter can be strongly down-regulated by inhibitory proteins that bind to intronic regulatory sequences of I1a. Possible suppressor protein candidates include members of several transcription factor families (e.g. SP, EGR, GATA, SREBP, USF, and CREBP). Functional intronic cis-regulatory elements have been reported to be involved in the regulation of expression of several genes (24,25), but the detailed mechanisms of action are unknown.
RT-PCR of snGPx transcripts indicated the presence of incomplete splicing products that selectively contain I1a (Fig. 1). Interestingly, we have never observed splicing problems with introns 2-5 of the ph/snGPx gene. Although the mechanistic reasons for this incomplete splicing have not been investigated in this study, the identified transcription factors may be involved. It has been reported before that transcription factors that prefer to bind to double stranded DNA are also capable of binding to single stranded RNA (26,27). Thus, they may be of regulatory importance for the splicing process. Alternative exons are frequently flanked by suboptimal splicing sites, and an increasing body of experimental evidence has indicated that effective outsplicing of alternative exons requires the participation of splicing enhancers and/or silencers in addition to the constitutive spliceosomal machinery (28). The presence of such auxiliary proteins may help to define the exact splicing sites, and may also be important for maturation-dependent and tissue-specific alterations of the protein expression patterns. cis-Regulatory pre-mRNA sequences that bind the trans-acting splicing factors may reside in the alternative exon itself (29), in the flanking introns (30,31), and even in an adjacent exon (32). Inclusion of an alternative exon in the final mRNA is a multistep process that involves timely differentiated removal of the two flanking introns. In fact, in several alternatively spliced FIG. 10. Electrophoretic mobility shift assay of footprints FP4 and FP5. The labeled oligonucleotide dimers (GCG-GGCTCTGACCTGGTCACGCGCCCTG-CCCGGCC containing the consensus binding sequence for USF; see panel A and TGCCCGGCCTGGTGCGTCACAG-TCGGCGGCGCCTT for CREBP, see panel B) were incubated with the nuclear extracts of murine testis (10 g of protein) and HEK293 cells (10 g of protein) in the presence or absence of consensus oligonucleotide as described under ''Experimental Procedures.'' After 30 min of incubation at 4°C the entire binding samples were applied to PAGE. A and B, the footprint sequences were determined from The arrows indicate the positions of the primers used for amplification of the immunoprecipitated DNA. B, sheared DNA-protein complexes were precipitated with antibodies against SREBP1 and EGR1 and precipitated genomic DNA was amplified (see A). Input, sheared DNA prior to immunoprecipitation. pre-mRNA species it has been demonstrated that one flanking exon is removed before the other (33,34). Although the detailed splicing mechanism of the ph/snGPx pre-mRNA was not the focus of this study, we have identified functional cis-regulatory sequences in the 5Ј-flanking intron of alternative exon E1b, which may be related to the splicing process. Additional stringent in silico search for similar cis-regulatory elements in E1b also suggested the existence of SP1/WT1-binding sites and preliminary DNase protection assays suggested their functionality (data not shown). Similarly, the 3Ј-flanking intron of E1b (I1b) also contains binding sites for several transcription factors (GABP1, OCT1P1, and GATA3), but their functionality has not been tested. Taken together, these data suggest the presence of cis-regulatory elements in the alternative exon E1b and in the flanking introns I1a and I1b. Such a constellation is frequently found in genes containing alternative exons, and the cis-regulatory sequences may serve as binding regions for splicing enhancers and/or silencers.
In the light of these findings the functional cis-regulatory elements identified in I1a and the corresponding trans-regulatory proteins may be of dual functional importance: (i) they may act as negative regulators of transcriptional initiation of the ph/snGPx gene (Fig. 5) and (ii) they may exhibit auxiliary function for alternative splicing of the joint ph/snGPx mRNA. Work is in progress in our laboratory to provide more detailed information on either process.