Cell Type-dependent Regulation of the Hypoxia-responsive Plasminogen Activator Inhibitor-1 Gene by Upstream Stimulatory Factor-2*

Transcriptional regulation of the plasminogen activator inhibitor type-1 (PAI-1) gene is an important issue since PAI-1 plays a crucial role in various pathological conditions. The transcription factor USF-2 was shown to be a negative regulator for rat PAI-1 expression, and therefore it was the aim of this study to evaluate the role of USF-2 for human PAI-1 expression. We found in human hepatoma cells (HepG2) that USF-2 induced human PAI-1 expression via two classical E-boxes and the hypoxia-responsive element (HRE) within the promoter. Gel-shift analyses showed that E-box 4 and E-box 5 bound USFs, and although the HRE contributed to the USF-dependent effects, it did not bind them. By contrast, USF-2 inhibited PAI-1 promoter activity in primary rat hepatocytes suggesting that PAI-1 expression depends on either the promoter context or USF activity which might be cell type-specific. Cotransfection of human or rat PAI-1 promoter luciferase constructs with expression vectors for wild-type USF-2 or USF-2 mutants in human HepG2 and rat H4IIE cells as well as in primary rat hepatocytes revealed that the effects of USF on PAI-1 expression depend on the cell type rather than the promoter context and that the USF-specific region domain of USF accounts for the observed cell type-specific effects.


EXPERIMENTAL PROCEDURES
All biochemicals and enzymes were of analytical grade and were purchased from commercial suppliers.
Animals-Male Wistar rats (200 -260 g) were kept on a 12 h day/ night rhythm with free access to water and food. Rats were anesthetized with pentobarbital (60 mg/kg of body weight) prior to preparation of hepatocytes.
Cell Culture-Hepatocytes were isolated by collagenase perfusion. Cells (1 ϫ 10 6 per dish) were cultured in a normoxic atmosphere of 16% O 2 , 79% N 2 , and 5% CO 2 (by volume) in medium M199 containing 0.5 nM insulin, 100 nM dexamethasone as permissive hormones and 4% fetal calf serum for the initial 5 h of culture. Cells were then cultured in serum-free medium from 5 to 24 h at normoxia. Then, the medium was changed, and cells were cultured under normoxia or hypoxia (8% O 2 , 87% N 2 , 5% CO 2 (by volume)) for the next 24 h. HepG2 and H4IIE cells were cultured under normoxia in minimal essential medium supplemented with 10% FCS for 24 h. Then, medium was changed and cells were further cultured under normoxia or hypoxia.
RNA Preparation and Northern Analysis-Isolation of total RNA and Northern analysis were performed as described (15). Digoxigenin-labeled antisense RNAs served as hybridization probes; they were generated by in vitro transcription from pBS-PAI-1 using T3 RNA polymerase and pBS-␤-actin using T7 RNA polymerase and RNA labeling mixture containing 3.5 mM 11-digoxigenin-UTP, 6.5 mM UTP, 10 mM GTP, 10 mM CTP, 10 mM ATP. Hybridizations and detections were carried out essentially as described before (15). Blots were quantified with a videodensitometer (Biotech Fischer, Reiskirchen, Germany).
Western Blot Analysis-PAI-1 and USF-2 Western blot analysis was carried out as described (6). In brief, media or cell lysates were collected, and 100 g of protein were loaded on 10% SDS-polyacrylamide gels and after electrophoresis blotted onto nitrocellulose membranes. The primary mouse antibody against human PAI-1 (American Diagnostics, Pfungstadt, Germany), a primary rabbit antibody against USF-2 (Santa Cruz Biotechnology, Heidelberg, Germany), as well as primary mouse antibody against hemagglutinin-tag (Santa Cruz Biotechnology) was used in a 1:100, 1:200, and 1:500 dilutions, respectively. The secondary antibody was a goat anti-mouse IgG (Santa Cruz Biotechnology) or a goat anti-rabbit IgG (Santa Cruz Biotechnology) used in a dilution of 1:5000, respectively. The ECL Western blotting system (Amersham Biosciences, Freiburg, Germany) was used for detection.
Cell Transfection and Luciferase Assay-Freshly isolated rat hepatocytes (about 1 ϫ 10 6 cells per 60-mm dish), 4 ϫ 10 5 HepG2, and 4 ϫ 10 5 H4IIE cells per 60-mm dish were transfected as described (16). In brief, 2 g of the appropriate PAI-1 promoter Firefly luciferase (Luc) constructs were transfected together with 500 ng of USF-2a, U2⌬USR, U2⌬E5, and ⌬TDU2 expression vectors or in the controls with 500 ng of an empty vector. After 5 h the medium was changed and the cells were cultured under normoxia for 18 h. Then, medium was changed again and the cells were further cultured for 24 h under normoxia or hypoxia.
Preparation of Nuclear Extracts and Electrophoretic Mobility Shift Assay (EMSA)-Nuclear extracts were prepared by modification of a standard protocol essentially as described (17,18). The sequence of the human PAI-1 oligonucleotides used for EMSA are 5Ј-TCTTACA-CACGTACACACA-3Ј (Ϫ199/Ϫ181); 5Ј-ACAATCACGTGGCTG-GCT-3Ј (Ϫ571/Ϫ552), and 5Ј-AGTCTGGACACGTGGGGA-3Ј (Ϫ689/Ϫ670). Equal amounts of complementary oligonucleotides were annealed and labeled with [␥-32 P]ATP (Amersham Biosciences) using the 5Ј-end labeling kit (Amersham Biosciences). They were purified with the nucleotide removal kit (Qiagen, Hilden, Germany). Binding reactions were carried out in a total volume of 20 l containing 50 mM KCl, 1 mM MgCl 2 , 1 mM EDTA, 5% glycerol, 7 g of nuclear extract, 1 g of poly(dI-dC) and 5 mM dithiothreitol. After preincubation for 10 min on ice, 1 l of the labeled probe (10 4 cpm) was added, and the incubation was continued for an additional 10 min at room temperature. For supershift analysis 0,4 g of the USF-1, USF-2, and an ATF-1/CREB crossreactive antibody (Santa Cruz Biotechnology) were added to the EMSA reaction, which was then incubated for 45 min on ice. The electrophoresis was then performed with a 5% non-denaturing polyacrylamide gel in TBE buffer (89 mM Tris, 89 mM boric acid, 5 mM EDTA) at 250 V. After electrophoresis, the gels were dried and exposed to a phosphorimaging screen.

Induction of Human PAI-1 mRNA and Protein Expression by USF-2 under Normoxia and Hypoxia-In
HepG2 cells transfected with the empty control vector hypoxia enhanced PAI-1 mRNA by about 3.5-fold ( Fig. 1). Transfection of the USF-2 expression vector induced human PAI-1 mRNA by about 6.5-fold under normoxia and hypoxia (Fig. 1). The hypoxia-and USF-mediated increase of human PAI-1 mRNA was followed by an increase of human PAI-1 protein levels. Hypoxia enhanced PAI-1 protein levels by about 2-fold in accordance with data from a previous study (5). Overexpression of USF-2 enhanced PAI-1 protein levels by about 4-fold under normoxia and hypoxia (Fig. 1).
USF-2 Activated Human PAI-1 Promoter Luc Gene Constructs in HepG2 Cells-Sequence analyses of the human PAI-1 promoter revealed five E-box-like sequences from which only four and five are classical E-boxes. Therefore, they were named E4 and E5. To investigate the involvement of the HRE as well as E4 and E5 in the USF-2-dependent induction of human PAI-1 expression, the Ϫ806 bp wild-type human PAI-1 promoter Luc construct (pGl3hPAI-806) and its derivatives mutated in the HRE as well as in the classical E-boxes were cotransfected with the USF-2 expression vector or an empty vector into HepG2 cells. When the wild-type PAI-1 promoter construct was cotransfected with the empty vector, hypoxia enhanced Luc activity by about 2-fold. Cotransfection of the wild-type human PAI-1 promoter Luc construct together with the USF-2 vector resulted in an about 20-fold increase of A, the PAI-1 mRNA and protein levels were measured by Northern and Western blots, respectively. The hPAI-1 mRNA and protein levels under normoxia (16% O 2 ) were set to 100%. Values were presented as means Ϯ S.E. of at least three independent experiments. Statistics: Student's t test for paired values; *, significant difference 16% O 2 versus 8% O 2 ; p Յ 0.05. B, representative Northern and Western blot. 20 g of total RNA isolated from cultured HepG2 cells were subjected to Northern blot analyses and hybridized with digoxigenin-labeled PAI-1 and ␤-actin antisense RNA probes (see "Experimental Procedures"). 100 g of protein from the HepG2 cell culture medium was subjected to Western blot analyses with the hPAI-1 antibody. 100 g of protein from HepG2 cells were analyzed by Western blot with antibodies against the hemagglutinin-tag. Autoradiographic signals were detected by chemiluminescence and quantified by videodensitometry.
Luc activity under both normoxia and hypoxia. The construct pGl3hPAI-HREm containing the mutated HRE responded neither to USF-2 nor to hypoxia. Mutation of E-boxes E4 and E5 in the constructs pGl3hPAI-M4 and pGl3hPAI-M5 reduced induction under hypoxia and significantly decreased the USF-2-mediated induction of Luc activity to about 4-and 3-fold, respectively, compared with the control (Fig.  2). Double mutation of E5 and E4 in pGl3hPAI-M45 also diminished induction of Luc activity by USF-2 to about 2-fold. Thus, these data indicate an involvement of the HRE as well as E4 and E5 in hypoxia-and USF-2-dependent PAI-1 expression.
Binding of USF to E-box Sequences in the Human PAI-1 Promoter-To confirm the conclusion from the transfection experiments that USF-2 interacts with HRE, E4, and E5 within the human PAI-1 promoter, binding of nuclear proteins to oligonucleotides spanning the HRE, E4, and E5 was examined by EMSA. Furthermore, to investigate the presence of USF in these complexes, antibodies against USF-1 and USF-2 were included in the binding reaction (Fig. 3).
When the labeled oligonucleotide spanning the HRE was incubated with HepG2 nuclear extracts, three major DNA-protein complexes were detected, but the mobility of these complexes was not affected by incubation with antibodies against USF-1 and USF-2. The oligonucleotides spanning E4 and E5 bound also three major DNA-protein complexes. Addition of either USF-1 or USF-2 antibody supershifted the intermediate DNA-protein complexes bound to E4 and E5, confirming that this complex contains both USF-1 and USF-2. However, since it is known that proteins of the ATF/CREB family bind constitutively to HREs (19) it was tested whether the complexes formed with the HRE oligonucleotide contain ATF proteins. Addition of the ATF/CREB antibody to the reaction mixture resulted in a supershift not only with the HRE but also with E4 and E5 showing that the major DNA-protein complex contains ATF/CREB proteins. Thus, these data indicate that USFs can interact only with E4 and E5 within the human PAI-1 promoter.
Regulation of PAI-1 Promoter Luciferase Gene Constructs by Wildtype and Mutant USF-2 in Different Cell Lines-Since our results from the transfection studies and the results of our previous study with the rat PAI-1 promoter (6) implicate that the regulation by USF depends on the promoter context, different domains of USF, or the cell type, we inves-tigated the different regulation of human and rat PAI-1 promoter Luc constructs together with plasmids expressing various USF-2 mutants in human (HepG2), rat (H4IIE) hepatoma cells, and primary rat hepatocytes. The USF mutants included the protein U2⌬USR (⌬AA 208 -230) lacking USF-specific region (USR), the protein U2⌬E5, which does not FIGURE 3. Binding of USF to the human PAI-1 promoter region. A, the USF consensus sequence and oligonucleotides spanning the HRE, E-box 4, and E-box 5 used as probes are shown; bases matching the consensus sequence are underlined. B, nuclear extracts used in these studies were prepared from confluent HepG2 cells cultured under normoxia (16% O 2 ). The 32 P-labeled human PAI-1 HRE, E4, and E5 oligonucleotides were incubated with 7 g of nuclear extracts from HepG2 cells. In EMSAs with antibody, the nuclear extracts were preincubated on ice for 45 min with 0.4 g of USF-1, USF-2, or ATF-1/CREB antibodies before adding the labeled probes. The DNA-protein interaction complexes were separated by electrophoresis on 5% native polyacrylamide gels and visualized by phosphoimaging; S indicates supershifted USF complex, S * indicates supershifted ATF-1/CREB complex, and C indicates constitutive complex. In HepG2 cells cotransfected with either the human promoter construct pGl3hPAI-806 or the rat promoter construct pGl3rPAI-766 and USF-2, Luc activity was increased, and the hypoxia-dependent response of the human and the rat promoter was abolished (Fig. 4). Cotransfection of both promoter constructs pGl3hPAI-806 and pGl3rPAI-766 together with U2⌬USR no longer enhanced Luc activity in HepG2 cells (Fig. 4). When the human PAI-1 promoter construct was used together with expression vectors for U2⌬E5 or ⌬TDU2, Luc activity was decreased compared with USF-2-transfected HepG2 cells (Fig. 4). Likewise, in HepG2 cells cotransfected with the rat PAI-1 promoter construct and U2⌬E5 or ⌬TDU2 Luc activity was only slightly increased compared with the control (Fig. 4).
Similar to HepG2 cells, in H4IIE hepatoma cells cotransfection of the human promoter construct pGl3hPAI-806 or rat promoter construct pGl3rPAI-766 with USF-2 increased Luc activity, whereas cotransfection with U2⌬USR did not show the inducible effect (Fig. 5). After cotransfection of H4IIE cells with pGl3hPAI-806 and U2⌬E5 or ⌬TDU2 vectors, Luc activity did not change significantly compared with the normoxic controls. Furthermore, in H4IIE cells cotransfected with the rat PAI-1 promoter construct and the U2⌬E5 vector, Luc activity was increased (Fig. 5), whereas transfection of ⌬TDU2 together with pGl3rPAI-766 in H4IIE cells did not induce Luc activity (Fig. 5).
Next, we were interested to find out whether the USR or the part of the transactivation domain encoded by exon 5 is crucial for the observed effects of USF on PAI-1 expression in primary hepatocytes. In primary rat hepatocytes, USF-2 slightly repressed the human PAI-1 promoter Luc activity under normoxia and completely abolished hypoxiadependent up-regulation. Similarly, when primary rat hepatocytes were cotransfected with human pGl3hPAI-806 and U2⌬USR, Luc activity was still reduced under normoxia and hypoxia (Fig. 6). However, transfection with the U2⌬E5 vector again reduced Luc activity (Fig. 6). The rat PAI-1 promoter-dependent Luc activity was repressed by USF-2 in primary rat hepatocytes under normoxia and hypoxia (Fig. 6). In contrast, Luc activity was not repressed when primary rat hepatocytes were cotransfected with pGl3rPAI-766 and U2⌬USR. Together, these data support that the USF-dependent induction or repression of the human and rat PAI-1 promoter appears to be cell type-specific.

DISCUSSION
In this study, we have elucidated a complex role of the transcription factor USF-2 in the regulation of the human PAI-1 promoter in hepatoma cells and primary rat hepatocytes. We have demonstrated that in hepatoma cells USF-2 induced PAI-1 expression and that thereby the HRE and the classical E-boxes within the PAI-1 promoter played an essential role. Furthermore, USF-2 repressed PAI-1 expression in primary rat hepatocytes showing that the USF-2 effect on PAI-1 transcription is cell type-specific. In addition, we showed that mainly the USR domain within USF-2 appears to be essential for both the inducible and the repressive effect of USF-2 on PAI-1 expression.
USF Modulates PAI-1 Expression via Binding to E-box Motifs-USF was originally identified as a transcription factor activating the adenovirus major late promoter (7). USFs mainly function through E-box core sequences, but their ability to bind non-canonical E-boxes (9, 20, 21) as well as pyridine-rich initiator (Inr) sites (22) has been reported. In our study, we showed that USF-2 activated the human PAI-1 promoter in HepG2 cells via the HRE, which contains a non-canonical E-box sequence and the classical E-boxes, E4 and E5 (Fig. 2). Mutations of E4, E5, or both decreased activation of the PAI-1 promoter by USF-2, while mutation of the HRE alone completely abolished it. In addition, in this study and in our previous one (6), we observed that the hypoxia-mediated response of the PAI-1 gene was abolished not only by mutation of the HRE but also after transfection of USF-2. Mutations of E4 and E5 also diminished the response of the PAI-1 promoter to hypoxia, and we have shown that they could be binding sites for HIF-1 under certain conditions (11). Thus, all these sites are critical for the action of USF and HIF-1 indicating that the competitive action between USF and HIF-1 as shown for the rat PAI-1 promoter (6) appears also to be likely within the human PAI-1 promoter. In all, these results proposed a model in which USF proteins may function as repressors of PAI-1 in certain cell types by binding to E-boxes thus preventing binding of proteins with strong transcriptional activity like HIF-1. A similar concept may hold true for other E-box binding proteins like Myc, which can displace HIF-1 from the p21cip1 promoter (23) or vice versa that HIF-1 can regulate the expression of some Myc target genes such as human telomerase reverse transcriptase (hTERT) and breast cancer anti-estrogen resistance 1 (BRCA1) (24,25).
Our present data suggested that the HRE is the major USF binding site in human PAI-1 promoter. Interestingly, electrophoretic mobilityshift analyses demonstrated that E4 (Ϫ566/Ϫ559) and E5 (Ϫ681/Ϫ674) bind USFs, and although the HRE (Ϫ194/Ϫ187) contributed to the USF-dependent regulation of the human PAI-1 gene, it did not bind FIGURE 5. Transcriptional activation of the human and rat PAI-1 promoter constructs by USF in the rat hepatoma cell line H4IIE. The human or rat PAI-1 promoter constructs (pGl3hPAI-806 or pGl3rPAI-766 Luc) were cotransfected either with an empty vector or with expression vectors encoding for U2⌬USR, U2⌬E5, and ⌬TDU2 into H4IIE cells. After transfection, cells were cultured in fresh culture medium for an additional 18 h under normoxia. Then they were cultured under normoxia (16% O 2 ) or hypoxia (8% O 2 ) for the next 24 h. The luciferase activity was estimated as fold induction compared with the Luc activity, measured in the respective controls. Values represent means Ϯ S.E. of four independent experiments, each performed in duplicate. Statistics: Student's t test for paired values; *, significant difference 16% O 2 versus 8% O 2 ; **, significant difference controls at 16% O 2 or 8% O 2 versus 16% O 2 or 8% O 2 with different constructs. HLH, helixloop-helix leucine zipper. FIGURE 6. Inhibition of the human and rat PAI-1 promoter constructs by USF in primary rat hepatocytes. The human PAI-1 promoter construct pGl3hPAI-806 was cotransfected either with an empty vector or with expression vectors encoding USF-2, U2⌬USR, U2⌬E5, and ⌬TDU2 into primary rat hepatocytes. The rat PAI-1 promoter construct pGl3rPAI-766 was cotransfected either with an empty vector or with expression vectors encoding USF-2 and ⌬USR. After transfection, cells were cultured in fresh culture medium for additional 18  USFs. Thus, these results again emphasize the role of the HRE, E4, and E5 in USF-2-dependent PAI-1 gene transcription and suggest a cooperative interaction among these elements within the promoter via a yet unknown cofactor. Similarly, a recent study proposed a cooperative model showing that E4 as a USF-1 binding site modulates the TGF-1␤dependent PAI-1 expression in human epidermal keratinocytes (26). Since we showed that a transcription factor from the ATF-1/CREB family binds constitutively to the HRE, it is tempting to speculate that CREB-binding protein is involved as a cofactor. It may then interact via the USR with USF proteins bound to E4 and E5 thus implicating that this complex cooperates with the general transcriptional machinery. This seems to be likely, since USF is related to the basal transcription factor TFII-I and both of them have been implicated in the recruitment of the general transcriptional complexes to TATA-less promoters and in stabilization of the general transcriptional machinery in TATA-boxcontaining promoters (22,27,28). The PAI-1 promoter contains a TATA box, and the USF action might be due to interaction with the basal machinery rather than with its stabilization, since luciferase assays showed that only mutation of the HRE is sufficient to abolish completely the USF effect. Furthermore, the cooperative mode is supported by a study showing that USF can act through both an E-box and a noncanonical E-box as both enhancer and initiator in the regulation of the vasopressin promoter (9). However, it is not known yet whether and to what extent USF, CREB-binding protein/p300, and the general transcriptional machinery interact, although preliminary evidence has been obtained (29). Thus, the details of this interaction need to be clarified in future studies.
The findings of this study that USF-2 acted as an inducer of human PAI-1 expression appeared to be contrasted by our previous findings with the rat PAI-1 gene where USF-2 acted as inhibitor. First, we thought that these differences might be due to sequence variabilities within the promoter, since sequence analysis of the human and rat PAI-1 promoter revealed a number of differences. While complete conservation between the HRE in human and rat PAI-1 promoters was found, the two classical E-boxes, named E4 and E5, were found only in the human but not in the rat PAI-1 promoter. In addition, the USF-2 binding site in the rat promoter (6) is absent from the human PAI-1 promoter. However, our transfection experiments with wild-type human PAI-1 promoter pGl3hPAI-806 or rat PAI-1 promoter pGl3rPAI-766 Luc gene constructs together with a USF-2 expression vector in human (HepG2) and rat (H4IIE) cells as well as in primary rat hepatocytes showed that USF-2 induced both the human and the rat PAI-1 promoter constructs in either the human or the rat hepatoma cell lines (Fig. 4, 5). In addition, we found that USF-2 repressed both promoters to a different extend in primary rat hepatocytes (Fig. 6). Thus, these results show that the action of USF as activator or repressor of PAI-1 expression depends on the cell type rather than on differences between the promoters.
The USR Domain as a Critical Part within USF-2-Several functional domains of the USF-2 protein such as the USF specific region (USR) and especially the part of the transactivation domain that is encoded by exon 5 of the USF gene have been proposed to be important for USF activity (14,39). The USR has been shown to be necessary and sufficient for transcriptional activation by USF-2 of promoters containing both a TATA-box and an initiator element (Inr), whereas the exon 5 is required together with USR for transcriptional activation of promoters containing only the TATA-box but no Inr element (14,39).
In this work, we showed that in hepatoma cells the USR domain appears to contribute predominantly to the activity of USF-2 regulating the PAI-1 promoter, which contains E-box motifs, a TATA-box, but no Inr element. In addition, we obtained results showing that the USR is also required for PAI-1 gene repression by USF-2 in primary rat hepatocytes (Fig. 6). Furthermore, when the sequence corresponding to exon 5 was deleted from USF-2, no effects on the transcriptional activity of USF-2 were observed with one exception; enhanced transcriptional activity with the rat PAI-1 promoter in H4IIE cells (Fig. 5). This suggests that this part can inhibit the activity of USF-2 only in H4IIE cells, i.e. in a cell type-specific manner. Thus, our findings are in line with a study in which the transcriptional activity of USF proteins appeared also to be controlled by an unknown cofactor recognizing the USR domain (39), which might be either differently expressed or modified in a cell typespecific manner.
In summary, we found that the human PAI-1 promoter is regulated by USF via E4, E5, and the HRE. Thereby, the different regulation of the PAI-1 promoter occurred in a cell type-dependent manner where the USR domain of USF plays a crucial role implicating the interaction with a so far unknown cofactor.