A Highly Active Homeobox Gene Promoter Regulated by Ets and Sp1 Family Members in Normal Granulosa Cells and Diverse Tumor Cell Types*

One mechanism by which normal cells become con-verted to tumor cells involves the aberrant transcriptional activation of genes that are normally silent. We characterize a promoter that normally exhibits highly tissue- and stage-specific expression but displays ubiquitous expression when cells become immortalized or malignant, regardless of their lineage or tissue origin. This promoter normally drives the expression of the Pem homeobox gene in specific cell types in ovary and placenta but is aberrantly expressed in lymphomas, neuroblastomas, retinoblastomas, carcinomas, and sar-comas. By deletion analysis we identified a region between nucleotides (cid:1) 80 and (cid:1) 104 that was absolutely critical for the expression from this distal Pem promoter ( Pem Pd ). Site-specific mutagenesis and transfection studies revealed that this region contains two consensus Ets sites and a single Sp1 site that were necessary for Pem Pd expression. Gel shift analysis showed that Ets and Sp1 family members bound to these sites. Transfection studies demonstrated that the Ets family members Elf1 and Gabp and the Sp1 family members Sp1 and Sp3 transactivated the Pem Pd. Surprisingly, we found that Sp3 was a more potent activator of the Pem Pd than was Sp1; this is unusual, because Sp3 is either a weak activator or a repressor of most other promoters. Activation by either Elf1 or Gabp required an intact Sp1 family member binding site, suggesting

The malignant and metastatic characteristics of cancer cells derive in part from the deregulation of genes whose normal role is to control the division, differentiation, and migration of embryonic cells during development (1). One important class of genes that regulates both developmental and tumorigenic events is the homeobox gene family (2,3). This gene family encodes a large group of transcription factors that each contain a 60-amino acid DNA-binding motif termed a homeodomain.
Although much is known about the function of homeodomain transcription factors, little is known about the regulation of the genes that encode them, particularly in response to signals in tumor cells. One homeobox gene of interest in this regard is Pem, which we originally cloned by subtraction hybridization from a T cell lymphoma clone (4,5). Pem is the founding member of the recently defined PEPP homeobox subfamily, a small group of homeobox genes on the X chromosome that are all normally expressed in reproductive tissues (6 -8). Pem is expressed in parietal and visceral endodermal cells, where it appears to regulate the development of cells that eventually form portions of the placenta (5,9,10). In neonatal and adult mice and rats, Pem is specifically expressed in ovary, testis, and epididymis (10 -12). In the ovary, Pem expression is restricted to granulosa cells of preovulatory follicles. In the testis, Pem is specifically expressed in Sertoli cells at stages VI-VIII of the seminiferous epithelial cycle. In striking contrast to its tissue-and stage-specific expression in normal tissues, Pem is aberrantly expressed in a wide range of tumor cell types regardless of their origin (5). It is not known whether the ubiquitous expression of Pem in tumors reflects a causal role for Pem in promoting tumor cell growth and metastasis or whether instead Pem expression is a consequence of immortalization and malignant conversion. In support of the former, it was shown recently that Pem physically interacts with the tumor suppressor protein menin (13) and is a potent tumor antigen (14).
We showed previously (15) that Pem transcripts are derived from two promoters that are independently regulated in a tissuespecific manner. The proximal promoter (Pem Pp) is expressed exclusively in male reproductive tissues, whereas the distal promoter (Pem Pd) is preferentially transcribed in the female reproductive tissues ovary and placenta (7,15). In the present investigation, we determined which promoter is responsible for the ubiquitous expression of Pem in tumor cells. We found that the Pem Pd was responsible for this aberrant expression. We then defined the cis elements upstream of the Pem Pd transcription start site that were necessary and sufficient for expression in both tumor cells and normal granulosa cells. We identified specific Ets and Sp1 transcription factor family members that bound to these cis elements and controlled the expression of Pem in a ras-dependent manner. Our data suggest that the Pem Pd is normally a tissue-specific promoter that is aberrantly expressed in diverse tumor cell types because it is activated by ubiquitously expressed transcription factors that are regulated in normal cells but constitutively activated in tumor cells.

EXPERIMENTAL PROCEDURES
Cells, Chemicals, and Biochemicals-The SL12.1, SL12.3, SL12.4, N4TG1, M12, Ras/RAT-1, PS-1, S194/5, and 10T1/2 cell lines were * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ To whom correspondence should be addressed: Dept. of Immunology, Box 180, University of Texas M. D. Anderson Cancer Center, 1515 maintained in tissue culture plates containing Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum and 50 mg of both penicillin and streptomycin per ml. Drosophila melanogaster SL2 Schneider cells were kindly provided by Dr. Mitzi Kuroda (Baylor College of Medicine, Houston, TX) and were maintained in tissue culture flasks containing Schneider's D. melanogaster medium (Invitrogen) supplemented with 10% fetal bovine serum and 50 mg of both penicillin and streptomycin per ml. All cell culture reagents were obtained from Invitrogen. Antibodies for Sp1 (Sp1 SC-420), Sp3 (Sp3 D-20), and Elf1 (C-20) were obtained from Santa Cruz Biotechnology. Rabbit polyclonal antibodies, directed against recombinant GABP␣ and GABP␤1 (16), were gifts from Dr. Thomas Brown (Pfizer, Groton, CT).
Granulosa Cell Culture-Immature female rats (55-60 g, 23 days old) were injected with 1.5 mg of 17-␤-estradiol per day and then kept for 3 days before granulosa cells were harvested. The rats were treated in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals. All protocols were approved by the Institutional Committee on Animal Care of M. D. Anderson Cancer Center. Ovarian granulosa cells were isolated as described previously (17). Briefly, the ovaries were punctured multiple times with a 22-gauge needle to isolate granulosa cells. The granulosa cells were pooled and treated with 20 g of trypsin per ml for 1 min, and then 300 g of soybean trypsin inhibitor per ml and 160 g of DNase I per ml were added to remove necrotic cells. The cells were washed twice and cultured in Dulbecco's modified Eagle's medium:F-12 medium at 37°C in 95% air and 5% CO 2 for 16 h.
Transient Transfection Assays-For transfections, DNA concentrations were independently determined by using a fluorometer and analytical gel electrophoresis. Before transfection, all mammalian cell lines were replated in 6-well plates at a density of ϳ10 6 cells per well in 0.8 ml of serum-free medium. Plasmid DNA was suspended in 100 l of serum-free medium, mixed with 100 l of serum-free medium premixed with 15 l of LipofectAMINE (Invitrogen), and then incubated for 40 min at room temperature. The DNA-lipid complex was then added to the wells and incubated at 37°C in a CO 2 incubator for 12-14 h. The cells were then replenished with fresh medium containing serum and incubated for 48 h. Stably transfected SL12.4 and 10T1/2 stable cell lines were generated by providing 2 mg/ml G418 every other day until only antibiotic-resistant cells remained alive (2-3 weeks).
D. melanogaster Schneider cells and primary ovarian granulosa cells were transfected by using the calcium phosphate method (18). Briefly, plasmid DNA diluted in 2ϫ HBS buffer (280 mM NaCl, 1.5 mM Na 2 HPO 4 , and 50 mM Hepes (pH 7.2)) was precipitated by dropwise addition of 250 mM CaCl 2 . After incubation for 30 min, 300 l of the DNA precipitate was added dropwise to each well, and then the cells were incubated for 6 -8 h at 26°C (for the Schneider cells) or at 37°C (for the granulosa cells). The incubation medium was then carefully aspirated and replaced with fresh media, and then the Schneider and granulosa cells were incubated for 48 and 6 h, respectively.
Cell lysates for luciferase assay were prepared by centrifuging the cells at 1000 rpm (750 relative centrifugal force) for 2 min, and then 1ϫ lysis buffer (Promega) was added to the cell pellet. Luciferase activity was measured by the dual luciferase reporter assay system (Promega) according to the manufacturer's instructions. For all transient transfection experiments, Renilla luciferase activity from the Pem Pd was normalized relative to firefly luciferase activity expressed from a thymidine kinase promoter-driven construct (pGL2TK). Expression from plasmids containing Pem sequences was compared with the empty control vectors pRL-Null, pPac, or ⌬c-fos.
RNA Isolation, Northern Blot, and Ribonuclease Protection Analysis (RPA)-Total cellular RNA was isolated as described previously by guanidinium isothiocyanate lysis and centrifugation over a 5.7 M CsCl cushion (4). The Pem cDNA probe used for Northern blot analysis was prepared as described previously (5). A cyclophilin cDNA probe (19) was used as a loading control.
RNA was analyzed by RPA, performed as described previously (11). The [ 32 P]UTP-labeled mouse Pem RPA probe was prepared by in vitro transcription, as described previously (Pem probe E) (15). The ␤-actin RPA probe was prepared in the same manner; it contained nt 135-169 of human ␤-actin exon 3 (GenBank TM accession number X00351). A set of RNA size markers generated from the century ladder template (Ambion, Inc.) was included in all gels.
EMSAs were performed by using a 32 P-labeled blunt-end doublestranded probe generated by annealing two complementary oligonucleotides (oligos) (MDA-473 and MDA-474). The probe (5 ϫ 10 4 cpm) was incubated for 30 min at room temperature in 20 l of binding buffer (10 mM Tris (pH 7.9), 50 mM NaCl, 1 mM dithiothreitol, 1 mM EDTA, 5% glycerol, and 1 g of poly(dI⅐dC)) containing 1 g of nuclear extract. For antibody supershift and blocking assays the reaction mixture was preincubated with monoclonal antibodies or polyclonal antisera (2 g) at room temperature for 20 min before addition of the 5Ј 32 P-labeled blunt-ended double-stranded oligonucleotide. The DNA-protein com-

Diverse Tumor Cells Express Pem
Transcripts from the Pem Pd-We showed previously (5) that despite its cell type-specific expression in normal cells, Pem is ubiquitously expressed in a variety of tumors cell lines regardless of their lineage. Here we determined whether Pem is expressed in tumor cells from its distal promoter (Pem Pd) or its proximal promoter (Pem Pp). We investigated this issue in six tumor cell lines derived from different tissues and cell lineages, all of which expressed Pem mRNA, as assessed by Northern blot analysis (Fig. 1A). To determine promoter usage, RPA was performed with a probe that distinguishes between Pem Pd-and Pem Pp-derived transcripts (15). As shown in Fig. 1B, all cell lines protected a band of ϳ80 nt that corresponded to the Pem Pd transcript. In contrast, none of the cell lines expressed the Pem Pp transcript, which is normally expressed in testes (Fig. 1B).
A 32-nt Element Is Necessary and Sufficient for Pem Pd Transcription-We next determined the sequences important for Pem Pd transcription. Promoter activity was assessed by measuring luciferase levels in transiently transfected SL12.4 T lymphoma cells. We first tested a construct containing 304 nt of sequence upstream of the exon 1-intron 1 junction. As shown in Fig. 2A, this Ϫ304 construct expressed high levels of luciferase activity (90 -110-fold more than empty vector), indicating that it contained sequences sufficient for high level Pem Pd transcription.
Deletion analysis was then performed to define functionally important cis elements in this 304-nt region. All deletion constructs tested had a common 3Ј end that included the transcription start sites defined previously between 22 and 40 nt upstream of the exon 1-intron 1 junction (15). Deletion of sequences from nt Ϫ304 to Ϫ104 did not appreciably reduce promoter activity ( Fig. 2A). In contrast, the Ϫ94 construct had much lower activity (about 10-fold less) than the Ϫ104 construct, indicating that sequences between nt Ϫ104 and Ϫ94 were critical for maximal Pem Pd transcription. Even less luciferase activity was detected from the Ϫ73 construct (60-fold less than the Ϫ104 construct). We therefore conclude that the region between nt Ϫ73 and Ϫ104 is indispensable for promoter activity.
Inspection of this 32-nt region revealed two putative Ets protein-binding sites (E1 and E2 in Fig. 2B) and one Sp1 protein- A-C, the Pem gene constructs indicated were transiently transfected (1 g) into SL12.4 cells, and then luciferase (luc) activity was measured. The values shown are Renilla luciferase activity expressed from the Pem constructs, normalized against firefly luciferase activity from the pGL2TK internal control vector (ϮS.E.). A shows expression from constructs containing the deletions indicated; the numbers refer to the nts upstream of the exon 1-intron 1 junction (the 1st black box is exon 1). Shown are the average values (relative to luciferase expression from the Ϫ304 construct, which was arbitrarily given a value of 100) from four independent transfection experiments performed in triplicate. B shows the two putative Ets family protein-binding sites (E1 and E2) and the putative Sp1 family protein-binding site (S1) that when mutated (indicated by lowercase letters) abrogate Pem Pd transcription. Shown are the average values (relative to luciferase expression from the Ϫ112 construct, which was arbitrarily given a value of 100) from six independent transfection experiments performed in triplicate. C shows that a single copy of the Pem promoter element sequence containing the S1-, E1-, and E2-binding sites (Ϫ81 nt to Ϫ112 nt) suffices for transcription from a heterologous minimal promoter (⌬c-fos). Shown are the average values (relative to that from the empty vector (⌬c-fos), which was given an arbitrary value of 1) from five independent transfection experiments performed in triplicate.
binding site (S1 in Fig. 2B). To determine whether these putative binding sites were necessary for Pem gene transcription, we mutated them and tested for transcriptional activity. As shown in Fig. 2B, mutation of either the E1 or S1 site in the Ϫ112 Pem construct strongly decreased promoter activity. Mutation of both sites led to a further decrease in promoter activity, suggesting the importance of both the Sp1 and the E1 Ets site for transcription. Mutation of the other Ets site (E2) led to a less drastic but still significant decrease in promoter activity (Fig. 2B).
Next we determined whether the region containing the Etsand Sp1-binding sites is sufficient to drive transcription. As shown in Fig. 2C, a single copy of a sequence containing these three sites (nt Ϫ81 to Ϫ104) was sufficient to drive transcription from a heterologous minimal c-fos promoter (4 -6-fold over basal level). This confirmed the functional importance of these Ets and Sp1 family binding sites, and thus we elected to call this region the Pem Pd promoter element.
To determine the specificity of this Pem Pd promoter element, we transfected the minimal Ϫ112 Pem construct containing this element into the immature T cell lymphoma cell clones SL12.1 and SL12.3, both of which express little or no Pem transcripts from the endogenous Pem gene (Fig. 3A). We found that the SL12.1 cell clone expressed luciferase activity that was indistinguishable from background levels, and the SL12.3 cell clone expressed trace levels of luciferase activity that was much lower than that from the SL12.4 cell clone (Fig. 3B).
Identification of Proteins Interacting with the Pem Pd Element-Having shown that the Sp1 and Ets family binding sites upstream of the Pem Pd transcription start site are necessary for Pem Pd transcription, it was imperative to identify the proteins interacting with these sites. Incubation of a doublestranded 26-mer oligo probe containing this region with SL12.4 nuclear extracts and analysis by EMSA revealed at least four specific protein-DNA complexes (Fig. 4A, complexes 1-4). An unlabeled oligo with a mutation in the Sp1-binding site (oligo M1) competed for bands 3 and 4 but not bands 1 and 2, and thus bands 1 and 2 represented proteins binding to the Sp1 site. Likewise, an oligo with mutations in the Ets-binding site (M2) competed with bands 1 and 2 only, indicating that complexes 3 and 4 contained proteins binding to the Ets site. An oligo with both the Sp1 and Ets (E1) sites (M3) had no effect on complex formation, demonstrating that these two sites are responsible for the four complexes.
To identify specific proteins that bind to the Pem Pd element, we used antibodies against Sp1 and Ets family members. Monoclonal antibodies against Sp1 and Sp3 supershifted complexes 1 and 2, respectively (Fig. 4B), consistent with our mutant-oligo competition experiments that demonstrated that complexes 1 and 2 contain proteins binding to the Sp1 family site in the Pem Pd element (Fig. 4A). To identify the transcription factors that bound to the Ets-binding site, we used a battery of antibodies against different Ets family members. We found that polyclonal antibodies against Elf1 and Gabp (␣ and ␤) inhibited the generation of complexes 3 and 4, respectively (Fig. 4C). In contrast, addition of monoclonal or polyclonal antibodies directed against various other Ets family members, including Ets1, Ets2, Erg1, and Elk1, failed to modify the nucleoprotein complexes formed with the Pem Pd probe (data not shown). We conclude that the Ets family members Elf1 and Gabp and the Sp1 family members Sp1 and Sp3 bind to the Pem Pd promoter element.
Transactivation of the Pem Pd by Sp1, Sp3, Elf1, and Gabp-To determine the functional importance of Sp1 and Ets family members in Pem transcription, we performed a series of transient transfection experiments with the D. melanogaster Schneider SL12 cell line, which lacks Sp1 family members (20,21). As shown in Fig. 5A, transfection of increasing amounts of Sp1 expression plasmid (25-100 ng) resulted in a concentrationdependent increase (5-20-fold) in promoter activity. Similarly, Sp3 also transactivated the Pem Pd in a dose-dependent manner. In fact, Sp3 was a more potent activator (20 -800-fold) than Sp1. Sp1 and Sp3 together exerted an additive effect in transactivating the Pem Pd (Fig. 5A). Elf1 and Gabp also stimulated luciferase activity in a concentration-dependent manner (Fig. 5B). Transfection of Elf1 and Gabp simultaneously resulted in an additive increase in promoter activity ( Fig 5B). Transfection of all four transcription factor plasmids (Elf1, Gabp, Sp1, and Sp3) also resulted in additive stimulation of promoter activity.
To determine whether the three DNA-binding sites for these different transcription factors was necessary for transactivation, we performed cotransfection experiments with mutant Pem Pd constructs. In these experiments, we used the Sp3 expression plasmid rather than the Sp1 expression plasmid, as Sp3 had more potent transactivation activity than Sp1. Results from cotransfection of Sp3 expression plasmid with a Pem Pd construct containing a mutation at the Sp1 site (Ϫ112/S1) demonstrated that transactivation by Sp3 required its cognate binding site (Fig. 6). Surprisingly, Sp3 transactivation also required the Ets-binding sites, as mutation of either the E1 or E2 sites prevented the induction of Pem Pd transcription in response to Sp3 (Fig. 6). Thus it appeared that one or more Schneider cell factors must bind to the Ets-binding sites for Sp3-mediated transactivation to occur. Similarly, Elf1 and Gabp transactivation required not only the Ets-binding sites but also the adjacent Sp1 site (Fig. 6). This implies that Schneider cells express an endogenous Sp1 site-binding factor (although not an Sp1 family member) that participates with mammalian Ets factors to drive Pem Pd transcription. These results suggest that Ets and Sp1 family members functionally cooperate to activate Pem Pd transcription (see "Discussion").
To assess the functional importance of these transcription factors for Pem transcription in mammalian cells, we transiently transfected plasmids encoding DN mutants into the SL12.4 T cell clone. As shown in Fig. 7A, transfection of DN-Gabp (␣ and ␤) expression plasmids caused a dose-dependent repression of Pem Pd promoter activity (Fig. 7A). A DN-Sp3 expression plasmid also strongly inhibited Pem Pd transcriptional activity in a dose-dependent manner (Fig. 7B). Because the encoded DN-Sp3 protein lacks a transcription activation domain but has a DNA-binding domain, it probably inhibits the activity of all Sp1 family members, as they all have the same DNA binding specificity (21). Thus, the inhibition by this DN protein shows that one or more Sp1 family members are required for Pem transcription in T cells. We conclude that the Ets factor Gabp and at least one Sp1 family member are critical for Pem transcription in SL12.4 cells.  (Fig. 8A). The neuroblastoma cell line N4TG1 and the prostate cancer cell line PS-1 had a high level, which was from 150-to 200-fold above that of the empty control vector. The ras-transformed fibroblast cell line RAT-1 had promoter activity 30-fold above that of the control vector. The B-myeloma cell lines S194/5 and M12 had comparatively low promoter activity that were 5-6-fold above that of the empty control vector.
We found that four of the five tumor cell lines required all three transcription factor-binding sites (S1, E1, and E2) in the Pem Pd element for optimal transcription (Fig. 8A). The one exception was the M12 cell line, which appeared to only require the S1 and E2 sites, because it did not exhibit a statistically significant (p Ͼ 0.05) decrease in luciferase activity when the E1 site was mutated. Transfection experiments with the DN-Gabp and DN-Sp3 plasmids showed that the Ets factor Gabp and one or more Sp1 family members were necessary for maximal Pem Pd transcription in all the tumor cell lines (Fig. 8B). We conclude that Pem Pd transcription in a wide range of tumor cell types depends on Ets and Sp1 family members.
Ras Is Essential for Endogenous Pem Expression-Because the Ras signaling pathway is known to positively regulate Ets and Sp1 family members (see "Discussion"), we examined the role of Ras in Pem transcription. We found that transient transfection of a DN-ras expression plasmid almost completely abrogated Pem Pd transcriptional activity from the Ϫ112 Pem construct in SL12.4 cells, indicating that indeed Ras does activate the Pem Pd (Fig. 9A). To determine whether Ras also positively regulates Pem Pd transcription from the endogenous Pem gene, we took two approaches. First, we stably transfected SL12.4 cells with the DN-ras plasmid and selected cell clones that expressed the plasmid. We found that DN-ras-positive cell clones expressed lower levels of endogenous Pem mRNA (up to ϳ3-foldless) than did untransfected cells (Fig. 9B) or stably transfected cells that expressed little or no DN-ras (data not shown). Second, to determine whether Ras expression was sufficient to activate endogenous Pem gene expression, we stably transfected an activated Ha-ras expression plasmid into a Pem-negative cell line, 10T1/2. We found that constitutively active Ha-ras strongly induced gene expression from the endogenous Pem gene in 10T1/2 cells (Fig. 9B).

Pem Transcription in Normal Granulosa Cells Requires the Pem Pd Regulatory Element and Is
Ras-dependent-Last, we determined whether non-malignant cells require the same regulatory element for Pem transcription as do malignant cells. We therefore investigated Pem Pd transcription in primary ovarian granulosa cells, as mice granulosa cells normally express Pem (22). As shown in Fig. 10, we found that primary granulosa cells expressed high levels of luciferase from the Pem Ϫ112 construct (almost 100-fold over that from the empty vector). Mutation of Ets-and Sp1-binding sites in the Ϫ112 construct drastically reduced luciferase expression, suggesting that, like tumor cells, granulosa cells require these binding sites for transcription. Cotransfection of the DN-ras construct dramatically reduced expression. We conclude that, like tumor cells, normal granulosa cells express the Pem Pd in a ras-dependent manner that involves Ets and Sp1 family members. DISCUSSION We have characterized a promoter transcriptionally active in ovarian granulosa cells, placental trophoblast cells, and tumor cells from a variety of lineages (see Figs. 1, 8, and 10 herein; see FIG. 6. Transactivation of the Pem Pd requires the Sp1/Sp3and Ets-binding sites. Schneider cells were cotransfected with 0.5 g of either the wild-type Ϫ112 construct (WT) or versions of it that had mutations in the Sp1 site (S1) or one of the two Ets sites (E1 and E2) and 50 ng of pPac-Sp1, pPac-Sp3, pPac-Elf1, or pPac-Gabp (␣ and ␤). Shown is the average luciferase activity, calculated as described for Ref. 15). We showed that this Pem Pd was transcriptionally activated in response to the cell type-specific Ets transcription factors Gabp and Elf1, as well as the ubiquitous zinc finger transcription factors Sp1 and Sp3 (Figs. 4 and 5). Three closely spaced binding sites for these factors between nt Ϫ80 and Ϫ104 were sufficient for Pem Pd transcription (Figs. 2 and 6). We found that Pem Pd transcription required the ras pathway (Fig.  9), which is known to activate both Ets and Sp1 family transcription factors and is constitutively activated in a large proportion of tumors of different origins (23)(24)(25)(26).
That Ets transcription factors are essential for Pem transcription in diverse tumor cell types is consistent with the fact that many Ets family members play a role in tumor cell growth, invasion, and metastasis. For example, the founding member of the Ets gene family, the v-ets retroviral gene, was originally identified as an oncogene based on its ability to transform cells (27,28). Since then, at least 10 of the ϳ30 Ets domain-containing genes have been shown to play a role in cancer, including ETS1, ETS2, FLI1, TEL, ERG, and PSE (29). Our finding that Gabp was required for Pem transcription in every tumor cell line tested (Figs. 7A and 8B) indicates that this Ets transcription factor is also important in activating gene transcription in tumor cells. However, whether Gabp is an oncogene that causes cancer remains to be determined.
Ets family members are also critical for the proper control of cellular proliferation and differentiation of normal cells (30). There is a wealth of knowledge about the role of Ets factors in some biological systems, but less is known regarding their role in the female reproductive tract, where Pem is expressed. In D. melanogaster, Elg, the fly orthologue of Gabp, is essential for normal egg development, as null mutations in Elg cause tinyegg syndrome (31). In mammals, Gabp positively regulates the expression of the FBP gene in placenta (32), and thus this Ets factor may be important for normal placental development and function. Part of the function of Gabp in placenta may be to allow Pem expression, as we showed previously (5, 15) that Pem is strongly expressed in this tissue, and we demonstrated here that Gabp positively regulated Pem transcription in D. melanogaster Schneider cells and several mammalian cell lines (Figs. 6 -8). Gabp may also positively regulate Pem expression in ovarian granulosa cells, as Pem is expressed in ovarian granulosa cells in vivo (10,22), and we found that its expression in these cells depends on the presence of an Ets factor-binding site (Fig. 10).
Our demonstration that Sp1 family members are essential for Pem transcription is not surprising given the fact that this set of transcription factors participates in the regulation of a wide variety of genes expressed in both normal and malignant cells (33)(34)(35)(36). Sp1-regulated genes include those expressed in granulosa and trophoblast cells, the type of cells that normally express Pem (32,37). Sp1 is at a wide range of levels in different tissues (35,38), which could partly explain the unique expression pattern of Pem in different tissues. For example, the high expression of Sp1 in normal trophoblasts (38) and some types of tumor cells (38,39) could, in part, explain the high expression of Pem in these cell types (5,10).
Surprisingly, we found that Sp1 was a less potent activator of Pem than was the related family member Sp3 (Fig. 5A). This response contrasts with most other genes, which are more strongly transcriptionally activated in response to Sp1 than to Sp3 (40 -42). In fact, rather than acting as an activator, Sp3 functions as a repressor for most gene promoters, including many that are activated by Sp1 (21,(43)(44)(45)(46)(47). In contrast, we found that Sp3 had an additive effect on Sp1-mediated transactivation of the Pem Pd. We speculate that Sp3 is a potent transcriptional activator of Pem because the Pem Pd regulatory element possesses only a single Sp1 family binding site (Fig. 2B). This notion is based on studies conducted on other gene promoters that demonstrated that a single Sp1 family binding site supports Sp3mediated transcriptional activation, whereas multiple Sp1 family binding sites support Sp3-mediated repression (48,49). This differential response has been suggested to result from the inability of a single Sp3 repressor domain to overcome the glutamine-rich activating region of Sp3 (48,50,51). Another factor that can dictate the transcriptional response to Sp3 is the type of cell in which it acts. For example, Sp3 activates the CD11d gene in myeloid cells but represses this same gene in B and T cells (21,(52)(53)(54). However, cell type cannot be the only determinant governing positive Pem Pd regulation by Sp3, as we found that the Pem Pd was induced by Sp3 in a wide range of cell types, including ones that engender negative regulation for other genes in response to Sp3 (21,36,37,55,56). Thus, we believe that intrinsic features of the Pem Pd, rather than cell type, dictates its strong induction in response to Sp3.
Our data support the notion that the Pem Pd is regulated by Ets and Sp1 family members acting in a cooperative manner. First, the close proximity of their binding sites in the Pem Pd regulatory element (Figs. 2B and 4A) suggests that Ets and Sp1 family members physically interact. Second, both Ets sites and the single Sp1 site were required for maximal Pem transcription in several different cell types, including normal granulosa cells and tumor cells from a variety of cell lineages (Figs. 6, 8, and 10). Third, any single mammalian Ets or Sp1 family member transactivated the Pem Pd less well than a combination of them (Fig. 5). An additive transcriptional response to Ets and Sp1 family members was observed in Schneider cells, which we believe is an underestimate, because our results suggested that there are endogenous D. melanogaster factors that bind to the Ets-and Sp1-binding sites and therefore elevate transcription levels in response to any single mammalian transcription factor (Fig. 6). That Ets and Sp1 family members cooperate to activate Pem Pd transcription is supported by several other studies showing that most Ets family members cooperate with other transcription factors, including Sp1 family members, to strengthen their transactivation potential (26). For example, the Ets factor Gabp has been shown to cooperate with Sp1 to activate at least three genes, including the FBP gene in placenta (32,57,58). Elf1 forms a complex with Sp1 or Sp3 to activate the transcription of the MRG1 and stem cell leukemia tal-1 genes (59, 60).
That Sp1 and Ets family members and Ras are ubiquitously expressed in many tumor cell types may partly explain why Primary granulosa cells transfected with 1 g of either the wild-type Ϫ112 construct (WT) or versions of it that had mutations in the Sp1 (S1) or Ets (E1) sites. Also shown are granulosa cells cotransfected with the Ϫ112 construct (1 g) and DN-ras (2 g). Shown is the average luciferase activity, calculated as described for Fig. 3B, from two independent transfection experiments performed in triplicate.
Pem is ubiquitously expressed in tumor cells. In addition, we showed that more than one member of each family is capable of triggering Pem transcription, thereby further increasing the range of tumor cell types that could express Pem. For example, some cell types that express low levels of Sp1 express high levels of Sp3 (49), both of which we found activated the Pem Pd (Fig. 5). Another probable reason for Pem expression in a wide variety of tumor cells is that Ras, which we demonstrated is required for Pem transcription (Fig. 9), is constitutively activated in ϳ50% of human tumors as a result of mutation (23,61). Ras induces the phosphorylation and consequent activation of Ets family members (23), and thus Ras may drive Pem expression in tumor cells as result of its ability to stimulate the activity of Ets transcription factors. Consistent with this notion, Pem is widely expressed in T cell tumors, but it is not expressed in normal T cells (5), perhaps because the Ets factor Elf1, which is present in both normal and malignant T cells, is only constitutively activated when T cells become transformed (62). Ras proteins have also been reported to activate gene transcription via Sp1 family members (25,63), which could be another mechanism by which Pem is transcriptionally activated in response to the Ras pathway.
Why is Pem ubiquitously expressed in a wide variety of tumor cells but in a cell type-and stage-specific manner in normal cells? First, despite the ubiquitous expression of Sp1 and Sp3, they are known to display complex and intricate interactions with other transcription factors (35,38,58,64) that may confine their ability to activate the Pem Pd in normal cells and expand this ability in tumor cells. Second, we found that only some members of the Ets family can activate Pem transcription, and thus the normally restricted expression pattern of individual Ets transcription factors will confine Pem expression to certain cell types. Third, and perhaps most importantly, we speculate that even if a cell contains a particular Ets factor required for Pem transcription, in most cases the Ets factor will not be phosphorylated and hence will not be capable of activating Pem transcription. This follows from the fact that extracellular stimuli that activate Ras and downstream mitogen-activated protein kinase pathways are known to cause the phosphorylation of many Ets factors, an event that appears to be necessary for Ets factors to regulate the transcription of specific target genes (65,66). Consistent with this notion, Pem is weakly expressed in quiescent liver macrophages but is dramatically induced when these cells are treated with lipopolysaccharide, a known activator of Ras and mitogen-activated protein kinase pathways in macrophages (22,53). Quiescent macrophages constitutively express all the transcription factors known to activate Pem transcription (Elf1, Gabp, Sp1, and Sp3) (46,47,54,67,68), and thus the induction of Pem in these cells probably results from the activation of one or more of these factors by phosphorylation rather than an induction in their levels.
Our discovery that the Pem homeobox gene is regulated by a tissue-and stage-specific promoter that is aberrantly expressed in tumors raises several questions. First, what are the signaling pathways that normally trigger Pem transcription in response to Ets and Sp1 family factors? The Ets factor Gabp is known to be activated by the c-Jun N-terminal kinase and extracellular signal-regulated kinase mitogen-activated protein kinase pathways (69,70), and thus it will be of interest to determine whether these pathways signal Pem transcription in normal granulosa and trophoblast cells. Second, what receptorligand interactions trigger Pem transcription in normal cells? Virtually nothing is known about either the receptors or ligands that activate Ets transcription factors in the female reproductive cells that express Pem. Third, what is the precise mechanism by which tumor cells bypass the normal regulatory constraints of Pem to constitutively express this homeobox transcription factor? A full understanding of the regulatory networks that control the Pem homeobox gene promoter that we have defined in this report may be useful toward elucidating the transcriptional perturbations that occur when normal cells become malignant.