Distinct Promoters Determine Alternative Transcription of gpx-4 into Phospholipid-Hydroperoxide Glutathione Peroxidase Variants*

A nuclear variant of phospholipid-hydroperoxide glutathione peroxidase (PHGPx, GPx-4) was considered to be derived from alternative pre-mRNA splicing in testis and to regulate sperm maturation. The genomic sequence of rat gpx-4 was established and investigated in respect to expression into the cytosolic, mitochondrial, and nuclear forms of PHGPx. In silico analysis suggested the presence of two distinct promoter regions, the upstream one leading to transcripts translating into cPHGPx or mPHGPx and the downstream one yielding nPHGPx. The promoter activity of both regions was verified by luciferase-based reporter constructs in A7r5 and H9c2 cells. The data reveal that the formation of nPHGPx is due to alternative transcription and not to alternative splicing. Transcripts encoding nPHGPx were most abundant in testis although not restricted to this organ. This observation points to a general role of the nuclear PHGPx variant in regulating cell division.

mouse, although vital, is more sensitive to redox-cycling herbicides and other pro-oxidant challenges (4 -6). Furthermore, the observation that Se-deficient or -supplemented cGPxϪ/Ϫ mice exhibit a similar survival curve supports the concept of cGPx as unique in protecting against oxidative stress (7). PHGPx has a broader substrate specificity, which also covers complex hydroperoxides such as phospholipid and cholesterol hydroperoxides (8). It was therefore considered to synergistically act with vitamin E in protecting membranes from lipid peroxidation (9). More recently, however, PHGPx has been implicated in regulating the synthesis of lipid mediators by silencing 5-and 15-lipoxygenase (10 -12), in suppressing hydroperoxideinduced apoptosis (13), in dampening cytokine-induced gene activation (14), in peroxynitrite scavenging (15), and in spermatogenesis (16).
Several observations indeed point to a specific function of PHGPx in testis: (i) PHGPx activity in this tissue is higher than in any somatic tissue (17); (ii) expression of the corresponding gene, gpx-4, correlates with the abundance of maturating germ cells (18); (iii) expression level is under indirect control of gonadotropins (18); (iv) in mature spermatozoa it is found as an oxidatively inactivated protein that represents a major component of the mitochondrial capsule (16 -19); (v) chimeric animals hemizygous for gpx-4 show mosaic testicular athrophy and abnormal sperm morphology (20) (vi) rescued enzymatic activity in clinical sperm samples correlates with fertility-related parameters (21); and (vii) promoter mutations in gpx-4 were detected in infertile subjects (22). More recently a selenoprotein band of 34 kDa, detected in testis (23), has been identified as a PHGPx variant with an N-terminal extension that comprises a nuclear targeting signal (24). This sperm nuclei, PHGPx ("sn GPx" in Ref. 24), is proposed to be involved in chromatin condensation and thereby to determine the differentiation of spermatogenic cells into spermatozoa (24).
How the nuclear PHGPx (nPHGPx) is formed is still debated. The PHGPx gene is composed of eight exons that, depending on species, are spread over 3 to 4 kb of DNA. It is transcribed into mRNAs of different lengths (25)(26)(27), which lead to three different isoforms differing in their N-terminal extensions: a cytosolic one (cPHGPx), a mitochondrial one (mPHGPx), and nPHGPx. cPHGPx and mPHGPx are derived from the same exon (designated 1A; see Fig. 1) by use of different translation starts, whereas the N-terminal extension of nPHGPx results from translation of an alternative exon (designated 1B; see Fig.  1). This phenomenon was first explained as resulting from alternative pre-mRNA splicing by Pfeifer et al. (24). The concept of splicing was also followed through by others (28). In the meantime, Moreno et al. (26) provided strong evidence for the existence of a second promoter that is located in the first intron and creates multiple transcription starts, including the alternative exon 1B, which is spliced to exons 2-7 and translated into nPHGPx. Attempts failed, however, to prove the activity of this alternative promoter region in reporter gene-transfected somatic cells (26,27). This was interpreted by Borchert et al. (27) to support the original idea of alternative splicing of a transcript derived from the "joint basic promoter" upstream of exon 1A, whereas Moreno et al. (26), argue that the supposed alternative promoter is not recognized in somatic cells because it likely depends on transcription factors regulating nPHGPx expression specifically in germ cells.
We describe here reporter gene experiments revealing the activity of the homologous alternative promoter of rat gpx-4 even in somatic cell lines. We further detected nPHGPx-encoding transcripts not only in testis but also in somatic cells and tissues, which opens up the perspective of a potential role for nPHGPx in the regulation of cell division beyond germ line differentiation.
Cell Culture-The rat thoracic aorta smooth muscle cell line A7r5 and the rat myoblast H9c2 cell line were grown in Dulbecco's modified Eagle's medium as described (29,30) but supplemented with 50 nM sodium selenite.
Determination of the 5Ј-ends of gpx-4 Transcripts-The 5Ј-ends of PHGPx mRNAs were obtained by RACE cDNA amplification (SMART RACE cDNA amplification kit, Clontech) starting from total RNA of liver, heart, testis, and H9c2 cells. RNA was prepared using the RNeasy Mini-kit (Qiagen, Hilden, Germany) and retrotranscribed into 5Ј-RACE-ready cDNA samples of 1 g each. The first 5Ј-RACE PCRs were performed with reverse primers targeting exon 1A (5Ј-caggccaggcacagccagag-3Ј) or exon 1B (5Ј-caggaactcgtggctgttgctggtct-3Ј) to gain the cytosolic and mitochondrial or nuclear transcripts, respectively. For the second amplification, nested reverse primers for exon 1A or 1B (5Јtgctggcttcaataagcggctcagac-3Ј or 5Ј-gttgaaaggctcgggaatgggct-3Ј, respectively) were used. The agarose gel-purified RACE products were TAcloned and sequenced.
Real-time PCR Measurements-Total RNA was again extracted using the RNeasy Mini-kit. Genomic contamination was ruled out as described above. Only samples with an A 260/280 nm ratio of 1.8 were accepted.
Reporter Gene Experiments-Fragments of gpx-4 were generated by PCR with genomic DNA as template and gpx-4-specific primers and cloned into the KpnI/BglII site of the promoterless pGL3-Basic luciferase reporter vector (Promega). The primers were designed to yield the fragments Ϫ547 to Ϫ30, Ϫ547 to ϩ422, Ϫ313 to ϩ422, Ϫ179 to ϩ422, ϩ39 to ϩ422, ϩ86 to ϩ422, and ϩ86 to ϩ649. The accuracy of the constructs was verified by sequencing.
Cells were seeded onto 24-well plates, grown to 70% confluence, and transfected with 0.2 g of the above constructs and appropriate control vectors (pGLBasic vector for negative control and pRL-TK vector expressing Renilla luciferase (Promega) for control of transfection efficiency) using ExGen 500 (MBI-Fermentas, St. Lech-Rot, Germany). Cells were cotransfected with 0.2 g of one of the above experimental constructs or 0.2 g of empty vector. After a 24-h expression period, cell lysates were assayed for luciferase activity by the Dual-luciferase Reporter Assay System (Promega) on a plate-reading luminometer (Fluoroskan Ascent FL, Labsystem, Helsinki, Finland). Firefly luciferase activity of the experimental constructs was normalized by the Renilla luciferase activity obtained with the control vector pRL-TK and expressed as amplification factor in respect to the basic luminescence observed with pGLBasic/pRL-TK-transfected cells.

RESULTS
Gene Organization-The overall organization of the rat PHGPx gene, gpx-4, which has been cloned here for the first time, appears to be essentially identical to that previously reported for mouse (31) and humans (32). It complies with the assumption that the coding region is constituted by 7 exons, the first one existing in two alternative forms that encode different proteins (Fig. 1). If exon 1A is translated starting from position ϩ1, mPHGPx is formed, which, upon cleavage of the targeting sequence during intake into the mitochondrion, becomes indistinguishable from the cPHGPx (33,34). Alternatively, translation can start at a downstream ATG just before the exon 1A/ intron 1 boundary to yield cPHGPx. If exon 1B is used with a translation start at ϩ 417, a different N terminus results, which, after 84 amino acids residues, transits into the typical PHGPx sequence due to translation of exons 2-7 with an un- A difference from the previously reported sequences of the N-terminal extension was observed. The sequence 12 QCAGIR-PYGP 21 (24) is replaced here by 12 CRQ 14 ; this diversity was detected independently by others 2 and thus is likely common. Otherwise, the total genomic sequence now established, Ϫ1497 to ϩ 2714 (not shown; GenBank TM accession no. AJ537598), is remarkably congruent with published cDNAs (25) and partial genomic sequences (24). Just the insertion of a gctccgg motif upstream of the translation start of mPHGPx, if compared with the mRNA reported by Puspha-Rekka et al. (25), deserves to be mentioned.
An in silico analysis of the entire rat gpx-4 sequence from Ϫ1497 to ϩ 2714 by the Genomatix PromoterInspector program (available at www.genomatix.de) suggested two distinct promoter areas (Fig. 1). The most upstream one corresponds in sequence and location to the promoter first identified by Huang et al. (35) in human GPX-4 and confirmed to be active by Maiorino et al. (22) and Borchert et al. (27). Being upstream of exon 1A, it may reasonably be presumed to regulate the products derived from this exon, i.e. cPHGPx and mPHGPx. The second promoter option is located in the first intron upstream of the alternative exon 1B. It is homologous to the mouse promoter postulated by Moreno et al. (26) to regulate the expression of nPHGPx, which depends on the transcription of exon 1B. This hypothetical promoter comprises 308 nucleotides that, according to MatInspector (36), contain 29 putative binding sites for transcription factors or regulatory proteins with a core similarity of 1 and an "optimized" matrix similarity. With identical criteria, a further 17 putative binding sites were identified in the region 340 nucleotides upstream of the putative promoter. In contrast to the corresponding region of the mouse gpx-4, no testis-specific regulatory elements were suggested in this area by the MatInspector program.
Typical splice sites that might yield an mRNA translating into nPHGPx, as suggested by Pfeifer et al. (24) and Borchert et al. (27), could not be detected irrespective of the assumed transcription start.
Proof of Second Promoter Activity-Left with no reasonable source to explain nPHGPx formation by alternative splicing and encouraged by the in silico prediction of the two promoters in gpx-4, we decided to reinvestigate the promoter activities with conventional luciferase-based reporter gene constructs and to test them in homologous, i.e. rat, mesenchymal cell lines.
Expectedly, the sequence Ϫ547 to Ϫ30 comprising the canonical upstream promoter only was active in both the embryonic rat heart myoblast line H9c2 and the smooth muscle cell line A7r5 (Fig. 2). If the whole stretch from Ϫ547 up to the translation start of exon 1B was lined to the reporter gene, activity declined to about 50% in both cell lines, suggesting that downstream elements bind inhibitory ligands. The removal of sequences upstream of the predicted promoter did not significantly affect the results. So far, the data is in line with the canonical promoter being localized in front of exon 1A. The constructs, starting at ϩ 39 and ϩ 86, however, which totally lack the canonical promoter, also showed significant activity in H9c2 cells, whereas in the A7r5 cells their activity was marginal. Furthermore, in both cell lines, increased activity was observed by including exon 1B almost completely into the construct lacking the canonical promoter (ϩ86 to ϩ649). These findings validate the prediction of the PromoterInspector program and, for the first time, unambiguously demonstrate that the hypothetical second promoter of gpx-4 is active in a cell type-dependent manner.
Detection of Exon 1B-derived Transcripts-As expected, the RACE technology applied to mRNA isolated from whole testis detected transcripts corresponding to both exons 1A and 1B (Table I). The 1B transcript extended 14 bases into the 5Ј-UTR, which agrees perfectly with the major transcription start region at Ϫ12 to Ϫ14 upstream of the AUG translation site of nPHGPx, worked out by Moreno et al. (26) for the mouse system. Surprisingly, however, an identical transcript could also be detected with mRNA isolated from heart and liver. This clearly indicates that the use of the alternate promoter is not restricted to germ line cells, and accordingly, nPHGPx is not a testis-specific protein.
In the H9c2 cells, which had been shown to use the alternate promoter efficiently, a similar and two additional RACE prod-2 R. Brigelius-Flohé, personal communication.

FIG. 2. Activity of supposed gpx-4 promoters by reporter gene analysis.
Fragments analyzed for reporter gene activity contained only the first promoter (Ϫ547 to Ϫ30); were extended to include the area of the second supposed promoter also (Ϫ547 to ϩ422, Ϫ313 to ϩ422, Ϫ179 to ϩ422); or contained only the area of the supposed second promoter (ϩ39 to ϩ422, ϩ86 to ϩ422, ϩ86 to ϩ649). Constructs are shown on the right, and colors are as described in the legend for Fig. 1. ucts were obtained, indicating additional transcription sites at Ϫ18, Ϫ73, and Ϫ108 relative to the nuclear translation start, which complies with the additional upstream transcription sites detected in the mouse system (26). Under our conditions, probing testis mRNA for the region upstream to exon 1A or 1B by reverse transcription PCR with suitable reverse primers and a series of forward primers nesting further 5Ј to the uppermost transcription start sites identified for each exon product by RACE, a corresponding series of the correct-sized bands was observed (Fig. 3). This suggests that, in both cases, transcription may be initiated even further upstream than suggested by RACE experiments.
In agreement with previous observations (25,33), some of the exon 1A-derived transcripts identified here (Table I)  Quantitative Transcript Analysis by Real-time PCR-Transcripts encoding the exon 1A-derived products cPHGPx and mPHGPx and the 1B-derived nPHGPx were detectable by real-time PCR in somatic cell lines and all tissues investigated, i.e. testis as well as brain, heart, liver, and kidney. In each case nPHGPx transcript level were substantially lower than those encoding the cytosolic and mitochondrial forms (Table II). In terms of percentage of total PHGPx transcripts, the nPHGPx transcript varied between 17% in testis and 0.4% for the A7r5 cell line that had shown low activity of the alternative promoter region. The data reveals that nPHGPx is most abundantly expressed in testis in terms of both, absolute and relative amounts, but transcription of gpx-4 into nPHGPx-encoding mRNA is not restricted to spermatogenic cells.

DISCUSSION
The controversy about the mechanism of nPHGPx transcription can now be settled with a high degree of accuracy. The original proposal (24,27,28), which assumes nPHGPx to result from alternate splicing, conflicts with the 5Ј-UTR sequences of nPHGPx-encoding transcripts ( Table I). None of them starts with a 5Ј-end corresponding to any known sequence upstream of exon 1A, and the genomic gpx-4 sequence upstream of the identified 5Ј-ends of the transcription starts do not contain any splice sites that could reasonably yield the sequences of the identified transcripts, as similarly argued by Moreno et al. (26) for the mouse system. Instead, multiple transcription start sites similar to those of the mouse system were identified here immediately and further upstream of the translation start site of exon 1B (Table I, Fig. 3), strongly supporting the concept that nPHGPx results from the use of an alternate promoter located in intron 1. The argument made in favor of the alternative splicing concept, that the intron region upstream of exon 1B does not display any promoter activity (27), is invalidated by the demonstration of such activity in H9c2 cells (Fig. 2). Furthermore, our data suggest the presence of downstream elements, increasing novel promoter activity, within exon 1B. The failure to detect the promoter activity of the intron region preceding exon 1B (26,27) is likely explained by its relatively low and variable response in somatic cells, which is reflected in the relative and absolute low abundance of pertinent transcripts in tissues other than testis (Table II) and/or by the use of heterologous systems for reporter gene experiments in the previous work (26,27). In short, therefore, translation of gpx-4 into nPHGPx instead of cPHGPx or mPHGPx is caused by the  4 and A-C), were probed with a series of forward primers annealing 5Ј to the uppermost transcription start as identified by RACE technology (see below). As template retrotranscribed, total mRNA from testis was used. Reverse primers were within exon 2 for mPHGPx and at the boundary of exons 1B and 2 for nPHGPx, respectively. Lanes are as follows: 1, forward primer annealing to the gpx-4 position Ϫ50 to Ϫ33; 2, Ϫ81 to Ϫ64; 3, Ϫ166 to Ϫ149; 4, Ϫ313 to Ϫ290; A, ϩ236 to ϩ253; B, ϩ174 to ϩ191; C, ϩ86 to ϩ103. L marks the molecular weight ladder. For either transcript, no signal was detected using primers annealing further 5Ј upstream than Ϫ313 to Ϫ290 and ϩ86 to ϩ103 (not shown). See "Experimental Procedures" for other details.   (20) and clinical sperm samples of infertile patients (21,37), are best explained by the role of PHGPx as a structural protein of the mitochondrial capsule in sperm. Although only cPHGPx were detected in the capsule material (16), both mPHGPx and cPHGPx may be considered to provide the raw material for the capsule formation, because these two forms can no longer be distinguished after processing. However the oligospermia consistently observed in transgenic animals (20), in severely selenium-deficient experimental animals (38), and in infertile males with low PHGPx content in sperm (21,37) points to a distinct role for the active enzyme in earlier stages of spermatogenesis, the PHGPx form involved being unclear. The appearance of nPHGPx in late spermiogenesis (24,26,28) coincides with DNA condensation. The ability of PHGPx to act on -SH groups of protamine as alternate substrates (24,39) prompted the idea that its nuclear variant is indispensable for chromatin condensation by cross-linking protamine thiols. This process is also considered to guarantee appropriate sperm function and arrest of proliferation, once the germ line cells have reached their final state of differentiation (24,26).
Under consideration for this postulated role of nPHGPx in sperm differentiation, its ubiquitous expression documented here may offer a key to understanding the essential nature of PHGPx in embryonic development. Surprisingly, all efforts to obtain homozygous PHGPx (Ϫ/Ϫ) mice have failed thus far. 3 This outcome of gene disruption experiments cannot be explained by any of the established roles of PHGPx. Knock-out of gpx-1 and gpx-2 yielded offspring with a mild phenotype. GPx-1 (Ϫ/Ϫ) mice developed normally and displayed no phenotype whatsoever as long as they were not oxidatively challenged. Appropriate controls demonstrated that GPx-1 is the GPx variant that dominates antioxidant defense (40), revealing that a disturbed hydroperoxide metabolism cannot account for prenatal lethality. Similarly, the gastrointestinal GPx-2 could be knocked out without impairing development. Even when crossbred with GPx-1 (Ϫ/Ϫ) mice, the resulting double knock-out mice were vital and only later developed an intestinal inflammatory syndrome (41). It is thus extremely unlikely that PHGPx, which is less abundant than any of its congeners, is of vital importance because of its antioxidant potential. Modulation of cytokine responses, as reported for gpx-4 overexpression, cannot likely be reconciled with embryogenesis either, because it has been similarly reported for GPx-1 (42), and needless to say, the role of PHGPx as a structural protein of spermatozoa is irrelevant to embryonic development. Rather a unique role of PHGPx in cellular differentiation would be compatible with the outcome of the reverse genetics approach. The use of protamine -SH groups as an alternate substrate of PHGPx, as demonstrated for testis (39,24), might be taken as a model for a more general principle of how the enzyme might regulate cell division, and the nuclear form would surely be predestined to fulfill such demands.