The Variant Hepatocyte Nuclear Factor 1 Activates the P1 Promoter of the Human α-Folate Receptor Gene in Ovarian Carcinoma¹

Epithelial ovarian cancer is the most fatal disease of the gynecological malignancies, and its low overall cure rate is partly because of late diagnosis, and partly because of a poor understanding of its pathogenesis and progression. Ovarian carcinoma arises from the OSE,³ the cells monolayer covering the ovaries. In the embryo, OSE cells together with the Mullerian and Wolffian ducts originate from the celomic epithelium of the gonadal ridge. At the first stages of tumorigenesis, OSE undergo phenotypic changes characteristic of Mullerian epithelium such as the expression of the specific epithelial markers E-cadherin and CA125 (1). In the past decade, we have focused on the role of αFR in ovarian tumorigenesis and progression because: (a) αFR is undetectable on OSE but becomes highly expressed in the initial step of transformation, being present in inclusion cysts;⁴ (b) αFR overexpression is common to the majority of ovarian tumors of different subtypes and expression increases in association with tumor progression (2, 3); and (c) RNA analysis indicates that αFR expression in normal kidney and placenta is high and at detectable levels in lung (4, 5). The αFR, a glycosylphosphatidylinositol-anchored membrane glycoprotein, is a member of a protein family that binds folates, and mediates the uptake of folate and antifolate drugs (6). This protein family shares biochemical and molecular properties but is encoded by independent genes that are expressed in a restricted, independent, and tissue-specific manner (6). Cell lines transfected to express αFR show a growth advantage (7), suggesting a regulatory role for αFR in cell proliferation. We showed that the overexpressed αFR distributes in low-density membrane microdomains in the absence of caveolin and that the receptor is physically associated with the src-family member p53–56 lyn and the Gαi-3 subunit of heterotrimeric G proteins (8). Interestingly, serial analysis of gene expression on a wide panel of ovarian tumors identified αFR as 1 of the 13 highly up-regulated genes, irrespective of tumor subtype (9).

In both cultured tumor cells and normal human tissues (10), the abundance of αFR transcripts is proportional to the receptor protein, suggesting a role for transcription regulation in modulating expression of the receptor. The αFR gene is composed of seven exons spanning ∼6.7 Kb (11, 12). The ORF is encoded by exons 4–7, whereas the reported 5′-UTR of the cDNA isoforms are encoded by exons 1–4. Sequences upstream from exons 1 and 4 (designated P1 and P4) that promote CAT transcription after transient transfection of KB and HeLa cells have been identified. Transcripts from the P4 promoter are the most abundant in KB cells and normal lung tissues, whereas P1 transcripts are the predominant mRNA species in normal kidney and cerebellum. The abundance of P1 and P4 transcripts in other normal human tissues is variable (4). On the basis of these results, we hypothesized that activation of the P1 and P4 promoters may be tissue-specific (12) analogous with other genes containing multiple promoters, e.g., the α- and β-retinoic acid receptors (13). The P4 promoter contains three SP1 sites, which together with the INR element, are essential for optimal promoter activity (12). By contrast, mechanism(s) regulating P1-driven transcription remains unknown. In this contest, we demonstrated an inverse relationship between αFR and caveolin-1 expression, reflecting the repressing effect of caveolin-1 on P1 promoter activity (14).

We have reported the cDNA sequence of three different αFR transcripts (clones #31, 4/6, and 51) isolated from an IGROV1 ovarian carcinoma cDNA expression library (15). Each cDNA shares a common ORF and 3′-UTR but contains a divergent 5′ terminal sequence. Comparison of the cDNA and genomic sequences (12), showed that clones #31 and #4/6 are transcribed from the P1 promoter but contain an alternatively spliced 66-bp fragment from exon 3. We found recently that post-transcription events such as RNA splicing regulate αFR expression in ovarian carcinoma (5). Indeed, splicing of the #4/6 transcript appeared to be regulated in a tissue-specific manner, and we proposed a P1 construct as a suitable tool for specific gene therapy against ovarian cancer. Furthermore, the inverse relationship between αFR expression and intracellular folate levels in KB nasopharyngeal epidermoid carcinoma cells (16) and SKOV3 ovarian carcinoma cells (8, 17) might also involve post-transcriptional regulation (18).

Here, we investigated the elements of the P1 promoter that regulate αFR gene transcription in ovarian carcinomas. We identified a DNA-binding site in the 5′-untranslated exon 1 (from nucleotide +27 to +33), which is responsible for maximum activity of the P1 core promoter in ovarian carcinoma cells. Finally, vHNF1 expressed by ovarian carcinoma cells was shown to specifically bind this element. Together, these results suggest a crucial role for a specific element within the αFR P1 promoter that is bound by the homeoprotein vHNF1 in the transcriptional regulation of the αFR gene in ovarian carcinoma.

Human ovarian carcinoma {IGROV1 (a gift from Dr. Jean Benard, Institut G. Roussy, Villejuif, France), OVCAR3, SKOV3, and SW626 (ATCC) [the origin of the latter from an ovarian carcinoma has been questioned (19) and is now described as originating from a colon carcinoma metastasis to ovary], and 413OVA and 3507OVA (20)}, breast carcinoma [SKBR3 and MCF-7 (ATCC)], epidermoid carcinoma of the vulva [A431 (ATCC)], nasopharyngeal carcinoma [KB (ATCC)], lung adenocarcinoma [CALU3 (ATCC)], and embryonic immortalized epithelial kidney [HEK (ATCC)] cell lines were maintained in RPMI 1640 (Life Technologies, Inc., Paisley, United Kingdom) supplemented with 5% fetal calf serum, 2 mm l-glutamine, and gentalyn (100 units/ml) at 37°C in a humidified atmosphere of 5% CO₂ in air.

OSE cells were scraped from the surface of normal ovaries obtained at surgery from benign or malignant gynecological diseases other than ovarian carcinoma. OSE cells were maintained in culture for three to five passages in 199-MCDB105 medium supplemented with 15% fetal calf serum and 25 mg/ml 1entamicin as described (21). Immunohistochemistry to detect coexpression of cytokeratin 8 and vimentin confirmed the origin of the OSE cells (21). PBMCs were isolated from buffy coat of healthy donors by Ficoll-Hypaque (Pharmacia, Uppsala, Sweden) gradient centrifugation. PBMCs were activated for 72 h with 1 μg/ml of phytohemagglutinin (Wellcome, Dartfort, United Kingdom) and expanded for 3–7 days with 100 units/ml of r-IL2 (Eurocetus, Amsterdam, the Netherlands).

Total RNA from surgical samples was obtained from 11 patients at different stages of disease. Details are reported in the legend of Fig. 10. All of the human materials were obtained with informed consent from patients.

Cell surface expression of αFR was evaluated by immunofluorescence analysis using purified murine anti-FR monoclonal antibody MOv19 (2) as described (7).

RPA was performed essentially as described (22). The 284-bp 5′ EcoRI-AvaI restriction fragments from cDNA clone #4/6, and the 367 bp 5′ EcoRI-HincII restriction fragment from KB1 cDNA clone (12) were inserted into pGEM-4Z for in vitro synthesis of riboprobes designated IG1 and KB1, respectively. The constructs were linearized, and the riboprobes were transcribed using SP6 and T7 RNA polymerases (see Fig. 1 A). Total RNA (5–20 μg) from cell lines was purified using the total RNA Purification kit (Quiagen). KB cell and wheat germ RNAs were run in parallel as positive and negative controls, respectively. Each sample was also hybridized with a β-actin probe to control for RNA loading. A sequencing reaction was run in adjacent lanes and served as size markers.

Restriction fragments containing the 5′ flanking sequences of the αFR gene were subcloned into the promoterless pCAT basic vector. To analyze the promoter sequences flanking exon 1, the hKB1–3PD construct (12), designated pCAT-1.3K, was digested with MstII at position −41 relative to the first nucleotide of exon 1, and with HindIII (MCS of pCATbasic vector), blunt-ended, and religated to form the pCAT-0.3K construct. To analyze the activity of the 5′ upstream region of exon 1, P1-deleted constructs were obtained by PCR using the oligonucleotides listed in Table 1. These oligonucleotides included the restriction sites HindIII at the 5′-end and Xba1 at the 3′-end for cloning of PCR products into the MCS of pCAT. The pCAT-0.25K, -0.2K, -0.15K, and -0.08K constructs lacked nucleotides −41 to +50, −41 to +108, −41 to +148, and −41 to +177, respectively. The pCAT-0.05K and -(Δ21)0.3K constructs were obtained by a two-step PCR. For pCAT-0.05K a fragment from −34 to +41 was amplified using the sense-P1-0.3K oligonucleotide and the antisense-P1-0.05K oligonucleotide, which anneals at the 3′-end to the 5′-end of intron I. The PCR fragment obtained and pCAT-0.3K were used as templates for the second PCR with oligonucleotide sense- and antisense-P1-0.3K. For pCAT-(Δ21)0.3K, two fragments from −34 to +21 and from +41 to +256, respectively, were amplified in the first step. The antisense and sense primers of the two portions contained a 21-nucleotide overlapping region [oligonucleotides sense-P1-0.3K and antisense-P1-(Δ21), and sense-P1-(Δ21) and antisense-P1-0.3K, respectively, in Table 1]. The two fragments were annealed and used as a template for amplification with sense and antisense-P1-0.3K oligonucleotides.

Site-directed mutagenesis on the pCAT-0.3K construct was performed by two-step PCR: in the first step, two separate fragments were obtained using sense-P1-0.3K and each of the mutated antisense-oligonucleotides, and each of mutated sense-oligonucleotide and antisense-P1-0.3K, respectively. Table 2 lists the sense sequences of the mutated oligonucleotides. The resulting 91-bp and 193-bp fragments were annealed and used as a template for amplification with sense- and antisense-P1-0.3K.

Sequences of each insert were verified by sequencing through the insert-vector junctions.

Cells were transfected using positively charged liposomes (kindly provided by Dr. Silvia Arpicco, University of Turin, Italy) essentially as described (5). CAT activity was normalized to luciferase activity to correct for differences in transfection efficiency.

NEs were prepared, and GSA was performed essentially as described (11). Protein concentration was determined by the BCA method. Double-strand oligonucleotide probes were prepared by end-labeling 500 ng of DNA with 5′ polynucleotide kinase (10 units; New England Biolab, Beverly, MA) and 30 μCi of [α-³²P]ATP. Excess radioactive nucleotide was removed using a Microspin S200 HR column (Pharmacia). In all of the experiments specific DNA-protein complexes were competed with 10–100-fold molar excess of cold oligonucleotide. For supershift experiments, 1.5 μg of antibody targeted to HNF1, vHNF1, or G3α (all from Santa Cruz Biotechnology, Santa Cruz, CA) were incubated with the extracts for 2 h before to addition of the radiolabeled probe. Samples were resolved on a 6% nondenaturing gel.

RNA was extracted from 5 × 10⁶ cells using the Rneasy Total RNA kit (Qiagen, Hilden, Germany). For cell lines, 2 μg were reverse-transcribed using the Moloney murine leukemia virus reverse transcriptase and oligodeoxythymidylic acid primers according to the manufacturer’s instructions (GeneAmp; Perkin-Elmer). PCR amplification of 2 μl of the cDNA strand generated was carried out in a total volume of 20 μl. Samples were amplified at 95°C for 3 min, followed by 30 cycles at 95°C for 1 min, 55°C for 1 min, 72°C for 2 min, and completed with 1 cycle at 72°C for 10 min in an automated DNA Thermal Cycler (MJ Research, Inc., Watertown, MA). The sequence of both the sense and antisense primers were: (a) 5′-ATGGTTTCTAAACTGAGCCAGCTG-3′ and 5′-ACCTGTTTGTGGGAACGTAGGACC-3′, respectively, for HNF1α 2) 5′-ATGGTGTCCAAGCTCACGTCGCTC-3′ and 5′-CTCAGAGCAGGCATCATCGGACTG-3′, respectively, for vHNF1. For amplification of vHNF1 from surgical specimens, the sequence of both the sense and antisense primers were 5′-ACCCCTATGAAGACCCAGAAG-3′ and 5′-CTCAGAGCAGGCATCATCGGACTG-3′. Primers designed to amplify the ORF of αFR and β-actin have been reported elsewhere (16).

Five × 10⁶ cells were lysed with 1 ml of radioimmunoprecipitation assay buffer [50 mm Tris-HCl (pH 7.4), 150 mm NaCl, 1% NP40, 0.5% Sodium Deoxycholate, and 0.1% SDS]. SDS-PAGE and transfer on the membrane were performed essentially as described (14). For αFR, the immunoreaction was performed using the monoclonal antibody MOv19 and for vHNF1 the goat antiserum cited for the Supershift experiments.

The transcription start sites and the abundance of P1-derived transcripts in ovarian and nonovarian carcinoma cell lines were assessed by RPA of total RNA from these cells using specific cDNA probes. Fig. 1,A shows the structure of the two cDNA constructs used for riboprobe synthesis and the αFR gene sequence. The IG1 and KB1 cDNA constructs were cloned from IGROV1 and KB cells, respectively (4, 15). Each probe contains a unique 5′ terminus encoded by exons 1 and 3, and exon 1, respectively, and shares a sequence of 156 bp because of a common splice site and 144 bp of ORF encoded by exon 4. This 156-bp protected fragment (Fig. 1,B, arrows) was evident in each lane containing RNA from ovarian cancer (IGROV1, OVCAR3, and SKOV3) and αFR-expressing nonovarian carcinoma cells (KB and SW626 cells). The IG1 riboprobe protected three additional bands ranging in size from 255 to 270 bp that were nearly the size of the IG1 probe (Fig. 1,B). This hybridization pattern was comparable with that obtained previously with kidney, lung, and placenta RNAs (5). The KB1 riboprobe protected multiple fragments in the RNA from OVCAR3, IGROV1, and KB cells. These fragments were present in variable abundance and ranged in size from ∼160 bp to 350 bp (Fig. 1 B), consistent with the profile observed for RNA from KB cells (12), and from normal kidney, lung, brain, placenta, and salivary gland (4). Because fragments of ≥180 bp correspond to transcripts encoded by exon 1, these results are consistent with multiple transcription start sites downstream from the P1 promoter.

The similarity in the protection patterns among all of the αFR-expressing cells indicates that the αFR 5′-exon region is conserved, because RPA can detect even a single base mutation.

The low abundance of protected fragments from SKOV3 and SW626 cells, which are clearly detected in gels exposed for longer time (data not shown), is consistent with the low level expression of the αFR in these cells (data not shown). No protected fragments were detected in any of the nonovarian carcinoma cell lines (A431, CALU3, MCF7, and SKBr3) with these riboprobes, even using twice as much RNA (Fig. 1 B).

To avoid misinterpretation because of the different length and, in turn, different labeling efficiency among the probes, we compared the intensity of the 156-bp protected fragment with each of the longer protected fragments within the same probe. Although P1-derived transcripts were present in all of the αFR-expressing cell lines (ovarian carcinoma cells, KB, and SW626), the relative abundance of the IG1 transcript fragments in ovarian carcinoma cell lines was equal to or greater than that of the 156-bp band, suggesting that the IG1 cDNA corresponds to the predominant transcript expressed in these cells. KB cells also showed protection by the IG1 riboprobe, but, based on the size of the protected fragments, the most abundant transcripts were those homologous with the KB1 probe and derived from multiple start sites within exon 1.

To determine whether an element(s) responsible for subsets of transcription start sites might be identified in the P1 promoter, we transiently transfected αFR-expressing IGROV1 cells with the following P1-CAT constructs: (a) pCAT-1.3K, which contains the entire exon 1 plus a 1000-bp upstream region, shown previously to contain promoter activity on transient transfection in HeLa cells (12); (b) pCAT-0.3K, which contains the exon 1 sequence plus 40 bp of the upstream intron; or (c) pCAT-0.08K, containing only ∼80 bp of exon 1. As shown in Fig. 2, both pCAT 1.3K and pCAT-0.3K were able to drive CAT expression in IGROV1 cells, whereas pCAT-0.08K essentially lacked transcription activity. Note that pCAT-0.3K yielded ∼3-fold higher CAT activity as compared with pCAT-1.3K, suggesting that the region from −41 to +252 (numbering the beginning of exon 1 as +1) contains the P1 core promoter and that inhibitory elements might be present in the 1000-bp sequence upstream of exon 1.

These results are consistent with the RPA findings, suggesting the presence of cis-regulatory elements within the −41 to +252 region important for cell-specific expression of the human αFR in ovarian carcinoma cells.

To define the P1 promoter region responsible for the cell-specific expression of the αFR gene, we subcloned a series of 5′- and 3′-nested deletion mutants of the P1 core promoter pCAT-0.3K construct into a CAT reporter plasmid and transient transfected IGROV1 ovarian carcinoma cells with these constructs. Note that the pCAT-0.3K construct contains two putative TATA boxes at +30 to +33 and +126 to +129, a CAAT box at +56 to +59, and clusters of known transcription factors binding sites, i.e., Sp1 and AP1 binding sites (Fig. 3,A). Deletion of −41 to +50 (pCAT-0.25K construct in Fig. 3,B) reduced the level of CAT expression by nearly 75% (Fig. 3,C). Additional truncation to +108 and +148 (pCAT-0.2K and pCAT-0.15K constructs of Fig. 3 B, respectively) maintained CAT activity at the same average level observed with pCAT-0.25K, suggesting that the sequence from −41 to +50 is essential for P1 functionality. To determine whether this region per se was able to drive CAT transcription, we constructed a deletion mutant lacking a 150-bp region downstream of the putative TATA box (pCAT-0.05K). This construct only yielded a promoter activity comparable with the other deleted constructs, thus demonstrating that the sequence spanning the 5′ end of exon 1 contains an element(s) that regulates P1 promoter activity in IGROV1 cells.

To identify the pattern of transcription factor elements contributing to αFR promoter activity in ovarian cancer cells, three different oligonucleotides (AB3/4, NP1/2, and NP3/4) within the sequence −41 to +50 were generated (Fig. 4,A), and GSAs were carried out using NE from IGROV1 and from the αFR-nonexpressing cell line A431 (Fig. 4,B). Each of the oligonucleotide probes formed several specific DNA-protein complexes with NEs from both cell lines. Interestingly, NP3/4 oligonucleotide formed a low-mobility DNA-protein complex only with NEs from IGROV1 cells, of which the specificity was confirmed by oligonucleotide competition assay. Thus, GSA analysis was also performed with oligonucleotide NP3/4 and NEs prepared from a panel of αFR-expressing ovarian (IGROV1, SKOV3, OVCAR3, 413OVA, and 3507OVA), and nonovarian (KB and HeLa) cell lines, in addition to αFR-nonexpressing (A431, SKBR3, and MCF-7) and normal (OSE3 and 10, PBMC, and HEK) cells (Fig. 5). All of the ovarian carcinoma cell lines showed prominent bandshift (Fig. 5, arrow), which was specific because it was competed by the cold oligonucleotide, and particularly abundant in IGROV1, SKOV3, and 3507OVA cells. αFR-expressing cell lines KB and HeLa also showed the same bandshift but at lower intensity. Note that the relevant complex was absent in the two OSE cell lines.

To evaluate the transcription specificity of the P1 core promoter and of the NP3/4 binding element, CAT transfection experiments were carried out using the same αFR-expressing ovarian carcinoma (IGROV1, SKOV3, OVCAR3, 3507OVA, and 413OVA) cells tested in the previous GSA, in addition to αFR-nonexpressing carcinoma (A431 and MCF-7) and normal (OSE) cells, as shown in the fluorescence-activated cell sorter histograms reported in Fig. 6,C. Analysis of transcription driven by pCAT-0.3K, pCAT-0.25K, which lacks the entire sequence from −41 to +50, and pCAT-(Δ21)0.3K, which lacks only the sequence from −21 to +41 of the NP3/4 binding element (Fig. 6) revealed pCAT-0.3K-driven transcription only in ovarian carcinoma cell lines, with some detectable P1 promoter activity also in 413OVA cells, although immunofluorescence analysis showed that only a small fraction of the cells expressed αFR (Fig. 6 B); pCAT-0.25K yielded a 60–80% decrease in promoter activity in all of the ovarian carcinoma cell lines. P1 activity of pCAT-(Δ21)0.3K additionally decreased 20% and 50% in SKOV3 and OVCAR3 cells, respectively, as compared with pCAT-0.25K. P1 promoter construct activity appeared to be proportional to αFR expression in IGROV1, OVCAR3, 3507OVA, and 413OVA. None of these P1 constructs became active in normal (OSE) cells or in αFR-nonexpressing carcinoma cells (A431 and MCF-7). These results strongly suggested that the NP3/4 sequence contains an element required for P1 promoter activation in ovarian carcinoma cells.

To define the consensus sequence in the NP3/4 binding site responsible for P1 activation, we designed four mutated oligonucleotides for use in GSA with IGROV1 NE (Fig. 7). Mutated primers were generated by substituting AA bases with CC bases and TT bases with GG bases within the sequence +27 to +36 AATAATT (oligonucleotides mu1 to mu4 in Table 2; see also Fig. 8). In GSA analysis with IGROV1 NE radiolabeled mutated oligonucleotides mu1, mu3, and mu4 did not bind the relevant complex (Fig. 7,A), and when used cold at 10- and 100-fold excess, were not able to compete for binding with radiolabeled Np3/4 (Fig. 7,B). The oligonucleotide mu2 appeared to form a DNA-protein complex with the same mobility but less abundant than that obtained with the wild-type oligonucleotide (Fig. 7,A) and competed with NP3/4 binding at a concentration of 100-fold higher (Fig. 7 B). All of the mutated oligonucleotides competed slightly with the high-mobility DNA-protein complex, suggesting that the mutated oligonucleotides mainly affect the formation of the low-mobility NP3/4-protein complex in ovarian carcinoma cells.

To additionally examine the functional significance of the NP3/4-binding motif, the mutations described above were also introduced into the core promoter construct pCAT-0.3K and used to transfect IGROV1 cells (Fig. 8); all four of the mutated constructs showed 4–5-fold less activity than the wild-type construct. Moreover, P1 activity of the mutated constructs was comparable with that obtained in IGROV1 cells with the pCAT-0.25K construct (see Fig. 3). Together, these results demonstrate that the AATAATT motif within exon 1 is required for optimal P1 promoter activity.

Searching of the databases TRANSFAC and TESS indicated a possible TATA-like element and the consensus binding site for HNF1α and vHNF1 transcription factors within the region comprising NP3/4 sequence. To test this possibility, GSA was performed using as competitor a 10- and 100-fold molar excess of oligonucleotides (see Table 2) containing a TATA box element (oligonucleotides #1 and #2) and the consensus binding sites of HNF1 (oligonucleotide #3 and #4, which bind with high and low affinity, respectively; Ref. 23; Fig. 9,A). Neither TATA box oligonucleotides #1 or #2 showed competition at a concentration 100-fold or higher with the NP3/4 specific DNA-protein complex. On the contrary, both TATA box oligonucleotides competed the formation of the faster migrating complex, suggesting that at least in vitro this second complex may contain the general transcription factor TFIID. This result is in accord with that shown in Fig. 7, where the high-mobility complex is also formed with the mutated NP3/4 oligonucleotides. Both HNF1 oligonucleotides competed for the formation of the NP3/4 specific element. As expected, HNF1 high-affinity oligonucleotide #3 at 10-fold molar excess completely competed NP3/4 complex (low migrating), and HNF1 low-affinity oligonucleotide #4 partially inhibited the formation of all of the complexes at 100-fold molar excess.

Supershift experiments to verify the binding of HNF1 transcription factors to a P1 regulatory element using IGROV1 NE incubated with antibodies anti-HNF1α or vHNF1 followed by radiolabeled NP3/4 revealed supershifting only in anti-vHNF1-treated NE of the NP3/4-specific complex (low migrating) with the majority of the complex not entering into the gel.

RT-PCR performed with primers specific for HNF1 transcripts on OSE cells, a panel of ovarian carcinoma cell lines, and 2 αFR-nonexpressing carcinoma cells (A431 and MCF-7) demonstrated that vHNF1 is only expressed in ovarian carcinoma cells (Fig. 10,A). No amplification of HNF1α transcript was detected in any of the cell lines. The expression of vHNF1 was additionally confirmed by Western blot on total protein lysates prepared from the same cells lines (Fig. 10 B). The use of total cell lysates allowed us to analyze vHNF1 and αFR at the same time so that we could also compare the relative amount of the two proteins. A band at M_r ∼63,000 specifically reacted with the goat antiserum against vHNF1. Analysis of αFR revealed proportional increase at the protein level between expression of vHNF1 and αFR in the ovarian carcinoma cell lines tested (i.e., OVCAR3 and SKOV3 with high and low expression, respectively, of both αFR and vHNF1).

To validate these results, we also analyzed 9 representative RNAs extracted from tumors of patients with serous ovarian carcinoma expressing different amounts of αFR, together with 2 RNAs from tumors of patients with breast carcinoma. In all of the specimens from serous ovarian carcinoma, both vHNF1 and αFR mRNAs are varyingly expressed (Fig. 10,C), whereas both messengers resulted undetectable in specimens from breast cancer. In ovarian carcinoma specimens as well as in cell lines (Fig. 10 A), the presence of αFR transcripts was simultaneous to that of vHNF1, although there was no strict correlation between the levels of vHNF1 and αFR mRNA.

Taken together, these results indicate that vHNF1 is expressed on ovarian carcinoma cells and can activate transcription of genes such as αFR.

In this study we functionally characterized the 5′-end region of exon 1 in the αFR gene to identify the regulatory transcriptional mechanism(s) in ovarian carcinoma. On the basis of RPA analysis indicating heterogeneity of the 5′-UTR products generated from the P1 promoter because of different start sites and alternative splicing of exon 3, we searched for the element(s) regulating such a complex transcript structure. The P1 promoter was active only in the panel of FR-expressing ovarian carcinoma cells tested, and a region within the 5′ untranslated exon 1 was found to be required for P1 promoter activity in cells. Within this region, a cis-regulatory element at nucleotide +27 to +33 formed a specific DNA-protein complex only with NEs from ovarian carcinoma cells and was responsible for P1 activation in the cells. Finally, we identified vHNF1 as one of the transcription factors bound to this cis-regulatory element. Thus, vHNF1 regulates tissue-specific transcription in ovarian carcinoma.

The relevance of the 5′-UTR in regulating the mouse orthologue of the human FR gene has been described (24), although the underlying mechanism remains undefined. It has been hypothesized that the 5′-UTR might regulate expression of a given gene in two different ways. First, the 5′-UTR might assume a stable stem and loop structure, and act as tissue-specific translational enhancer (25, 26). On this contest, a M_r 46,000 cytosolic factor was identified that binds to an 18-bp sequence in the 5′-UTR of the mRNA encoded by the c32 KB cell cDNA, and may regulate αFR translation (27). We also observed that exon 3 contains a consensus splice acceptor 66 bp upstream of its 3′ terminus as well as several AUG triplets that might differentially regulate αFR translation, and our recent evidence indicates that only ovarian carcinoma cells efficiently use the 3′ and 5′ consensus splice sites of exon 3 in the P1-derived pre-mRNA maturation process (5). Moreover, the presence of one or more AUG codons and/or short ORFs upstream of the main ORF can inhibit cap-dependent translation (28). The second 5′ UTR-mediated regulatory mechanism is at the transcriptional level; our RPA experiments indicate that the αFR gene is regulated mainly at transcription level in various cell lines, suggesting that the transcription activity of the 5′UTR is the most important regulator of αFR gene expression. Both organization and transcription of the αFR gene are complex (12). Transcripts characterized by novel nonhomologous 5′ termini encoding 5′-UTRs are generated by alternative splicing of upstream exons and by transcription from the P1 and P4 promoters (11). Transcripts from these promoters are expressed in a restricted and tissue-specific manner. The structure of the αFR transcripts expressed by ovarian cancer cells differed, with most P1 transcripts homologous to the cDNA clone #4/6 and including 66 bp from exon 3, which was alternatively spliced from exon 1 and 4 sequence. Furthermore, most P1 transcripts appeared to initiate from only two or three sites in ovarian cancer cells rather than from the multiple sites observed in KB cells and human tissues (12). Thus, the regulation of the P1 promoter and the processing of αFR mRNA in ovarian cancer cells differ from those in KB cells and normal human tissues (4).

Our RPA analysis suggested that the elements for optimal αFR gene transcription resided in exon 1. Indeed, the 5′-flanking region of the αFR gene between position −41 and +252 conferred strong transcriptional activity on the FR-expressing ovarian carcinoma cells, and the region between position −41 and +50 was necessary to retain this promoter activity in all of the ovarian carcinoma cells tested. Additional P1 promoter deletion analysis together with GSAs revealed a consensus sequence at +21 to +41 that determines tissue-specific promoter regulation, forming a specific DNA/protein complex with NE from ovarian carcinoma cells, but not with NE from normal ovary epithelium, other carcinoma cell lines, or PBMCs. Among the ovarian carcinoma cell lines tested, SKOV3 cells behave somewhat differently, with higher P1 promoter activity than in IGROV1 and OVCAR3 cells despite a lower FR expression (see Fig. 6). In SKOV3 cells, the telomerase reverse-transcriptase promoter is differently regulated, perhaps in turn leading to differences in transcription factor abundance and stoichiometry, because no correlation between genetic abnormalities and transcription factor activation was observed (29). SKOV3 was also the only ovarian cell line shown to be able to up-modulate αFR in folate-depleted medium (30). Furthermore, we did not observe differences in P1 promoter activity after transiently transfecting our ovarian carcinoma cells cultured in high folate versus folate-depleted medium with P1 promoter-CAT constructs (data not shown). Sadasivan et al. (31) have additionally demonstrated recently that in the αFR-expressing nonovarian KB cell line, when cultured in folate depleted medium, αFR expression is up-modulated because of an increase of both αFR transcription and mRNA half-life. However, our results suggest that the reduction of folates in the culture medium does not affect αFR gene transcriptional regulation in ovarian carcinoma cells. The contrasting results obtained for the SKOV3 cell line suggests that αFR transcription is more tightly regulated, but this still needs to be additionally explored.

Mutations comprising +27 to +33 within the αFR 5′-region decreased P1 activity up to 6-fold. Database searches of the region from +27 to +33 revealed a putative TATA box and the consensus binding site for HNF1 transcription factors. The competition studies with mutated NP3/4 and TATA box-specific oligonucleotides showed that the TFIID basal transcription factor(s) can bind P1 promoter in vitro but not at the site responsible for P1 optimal activity. Instead, we found that transcription factor vHNF1 binds the region from +27 to +33.

α- and variant-HNF1 proteins were initially identified based on their interaction with a sequence essential for liver-specific transcription of several genes postulated to determine the hepatic phenotype (32). Additional characterization showed that HNF1 proteins are expressed in other tissues including kidney, intestine, stomach, and pancreas. Recently, in humans, mutations in vHNF1 and other HNF proteins in human pancreatic β cells and kidney have been associated with an early onset of type II diabetes and/or severe renal defects (32, 33). In all of the species analyzed thus far, the expression of vHNF1 precedes activation of the HNF1α gene during embryogenesis and appears to have a role in regulating the proper growth and differentiation of the primitive ectoderm in pregastrulating embryos (34). In a vHNF1^−/− mouse model, early embryonic lethality was observed because of an abnormal extraembryonic region and poorly organized ectoderm, as well as defective differentiation of the parietal and visceral endoderm (34). These abnormalities might be because of the lack of expression of another transcription factor enriched in liver, HNF4. In adult mice, vHNF1 is expressed in the kidney tubules, collecting ducts, uterus, and in the oviduct (Mullerian duct derivatives), and in the epididymis and seminal vesicles (Wolffian duct derivatives; Refs. 35, 36). Interestingly, αFR was found expressed in 14-week-old fetal kidney tubules and in adult oviduct, kidney tubules (37), placenta, and endometrium (38). Thus, both αFR and vHNF1 are expressed in tissues derived from the Mullerian duct, whereas we have reported here that they are not expressed in OSE cells, from which ovarian carcinomas are originated (1). Of importance, analysis of αFR and vHNF1 transcripts by RT-PCR on tumor cell lines and specimens revealed a concomitant presence of the two messengers only in ovarian carcinomas. In addition, we found a correlation between the amounts of αFR and vHNF1 proteins in ovarian carcinoma cell lines. All together these results support the notion that vHNF1 is necessary for αFR expression. We hypothesize that dysregulation followed by the transition from mesothelial phenotype to one more similar to Mullerian epithelium lead to expression of development-associated transcription factors such as vHNF1, which in turn activate genes such as αFR.

The gain or loss of gene function through dysregulated transcription may have important implications in tumor pathogenesis (through perturbations of normal cellular processes such as proliferation, differentiation, or programmed cell death) and in clinical considerations (diagnosis, prognosis, and/or therapy). Although the mere presence of αFR during these early alterations is not the causing agent of OSE transformation,⁵ our findings suggest that activation of αFR gene transcription during very early alterations of OSE cells takes place through vHNF1, rendering them more prone to malignant transformation.

Finally, therapeutic strategies can be contemplated that exploit the P1 promoter to drive expression of a cytotoxic gene in a tissue-specific manner and to additionally identify vHNF1 target genes with a role in the biology of ovarian carcinoma.

We thank Dr. Martino Introna for helpful suggestions. We thank Graziana Piantanida for technical assistance, Gloria Bosco for assistance in preparing the manuscript, and Mario Azzini for photography.