Acceleration of thrombomodulin gene transcription by retinoic acid: retinoic acid receptors and Sp1 regulate the promoter activity through interactions with two different sequences in the 5'-flanking region of human gene.

The interactions between retinoic acid- (RA)-dependent transcriptional regulatory sequences of the 5'-untranslated region of the thrombomodulin gene and nuclear RA-responsive proteins were studied using human pancreas BxPC-3 cells. Deletion mutants of pTM-CAT plasmid revealed the presence of distal and proximal RA-responsive regions containing direct repeat with 4 spaces (DR4) and three of four Sp1 sites, respectively. Cotransfection of a pTM-CAT plasmid with expression plasmids of RA receptors (RARalpha, RARbeta, and RARgamma) augmented the promoter activity under the condition of lower retinoid X receptor-alpha (RXRalpha) expression, whereas the activity was greatly diminished when RXRalpha was highly expressed. An electrophoretic mobility shift assay with cDNA containing the DR4 indicated that heterodimers of RAR and RXRalpha interacted with the DR4 site, although the interaction gradually disappeared with the increase in the ratio of RXRalpha/RAR. On the other hand, Sp1 protein interacted especially with the tandem Sp1 site corresponding to the first and second Sp1 sequences of the four Sp1 sites in the proximal RA-responsive region. The binding of Sp1 to Sp1 sites was independent of RAR-RXR heterodimer but increased with the increase in Sp1 concentration in the presence of unknown factor(s) of reticulocyte lysate. Upon treatment of the cells with RA, time-dependent increases in the ratio of RARbeta to RXRalpha and the phosphorylated form of Sp1 were observed. We concluded that two genomic DNA regions, the DR4 site (-1531 to -1516) and the first and second Sp1-binding sites (-145 to -121), were involved in the RA-dependent augmentation of thrombomodulin gene expression through increased interactions of the two regions with heterodimer of RAR-RXRalpha and nuclear Sp1, respectively.

Thrombomodulin (TM) 1 is an essential cofactor for activation of protein C by thrombin on vascular endothelial cells (1)(2)(3). TM expression of human endothelial cells is decreased by tumor necrosis factor-␣ (4 -6), interleukin-1 (6,7), endotoxin (8), and phorbol ester (5,6) and oxidized LDL (9, 10), but we have found that it is increased by all-trans-retinoic acid (t-RA) (11,12) and/or cAMP (11,13) in human umbilical vein endothelial (HUVE) cells, through the acceleration of transcriptional activity. Several studies on the regulatory region of human TM gene have been reported (14 -17), and Dittman et al. (17) showed the existence of an RA response element (RARE) in the 5Ј-flanking region of the gene. The direct effects of retinol and its derivatives (retinoids), such as t-RA and 9-cis-RA (9C-RA), on the expressions of many genes are apparently mediated by nuclear receptor proteins that are members of the steroid and thyroid hormone receptor (TR) superfamily of transcriptional regulators (18,19). Nuclear retinoid receptor dimers, of which retinoid X receptor (RXR) is a mandatory constituent, are required for effective activation of the t-RA and/or 9C-RA response pathways (20 -25). However, the direct repeats of RARE separated by 4 base sequences (DR4) at Ϫ1531 to Ϫ1516 from the transcription start site of TM may not be a candidate for nuclear retinoid receptors, because the DR4 sequence has been shown to be specific for a heterodimer of RXR and TR (25). It is also known that dimers of RA receptors such as homodimer of RXR␣ and heterodimers of RARs and RXR interact with repeated RAREs spaced by 1 base (DR1) and 5 bases (DR5), respectively (26 -28). Accordingly, it is unclear whether or not the RARE of the TM gene in human cells is functionally responsive to RARs or RXR, although Dittman et al. (17) recognized ligand dependence of RARE in the TM gene using Chinese hamster ovary cells.
On the other hand, it is well known that the transcription factor Sp1 binds to GC boxes (GGGCGG or CCCGCC) and activates transcription of a subset of genes that contain those boxes in animal cells (29,30). Four Sp1-binding sites are located just upstream of the TATA box in the 5Ј region of the TM gene (14 -16). Darrow et al. (31) indicated that RA-induced expression of the tissue plasminogen activator gene during F9 teratocarcinoma cell differentiation might involve Sp1. The binding abilities and functional significance of Sp1 for the promoters of various genes, such as murine ornithine decarboxylase (32), tissue factor (33), interleukin-6 (34), and prolactin receptor (35) genes, have been reported. However, there is no direct evidence concerning interactions between Sp1, RA receptor proteins, and the four Sp1 sequences in the TM gene, especially in the case of HUVE cells treated with retinoids.
The results of transactivation and electrophoretic mobility shift assay (EMSA) indicate that up-regulation of TM gene expression by retinoid is associated with increases in the ratio of RARs to RXRs and in the amount of Sp1, followed by enhancement of the interactions between heterodimers of RARs-RXR␣ and the DR4 sequence and between Sp1 and the consensus binding sites of the TM gene. Furthermore, our results suggest that retinoid itself does not function directly as the accelerator of these bindings and that the bindings of Sp1 to the Sp1 sites is dependent on the phosphorylation of the protein.
Construction of Plasmids-The 5Ј-flanking DNA fragment of the TM gene was obtained by PCR from a genomic DNA template that was prepared from HUVE cells. The primers were designed based on the sequences reported by Shirai et al. (14) and Dittman et al. (17). The PCR products of 689-and 1319-bp fragments were ligated at the XhoI site to prepare a 1702-bp TM fragment (from Ϫ1730 to Ϫ29 with respect to the untranslated site, which corresponds to Ϫ1562 to ϩ140 with respect to the transcription starting site). The DNA fragments were subcloned into the pGEM-4Z vector, and the sequences were determined by the 373S DNA sequencer (Applied Biosystems Inc.). Expression plasmids, designated as pTM549-CAT and pTM1562-CAT, were constructed by inserting the TM cDNA (689 or 1702 bp) at the SalI site of a promoterless pMAM-CAT vector, which was prepared by the ligation of the CAT expression fragment derived from pMAM-neo-CAT vector and the neo gene-deleted pMAM-neo vector after deletion of the Rous sarcoma virus-long terminal repeat sequences. The TM gene was systematically digested with exonuclease III and mung bean nuclease after treatments with NheI and SphI. Deletion end sites were confirmed by DNA sequencing with the 5Ј-DNA primers derived from pMAM plasmid.
Preparation of Mutant Plasmids and Purification of Plasmid DNA-The deletion mutants of pTM-CAT plasmid were prepared by ligation of PCR products over the mutation sites. pTM346-CAT plasmid was digested with MluI and XhoI, and then the appropriate ligation fragment was introduced. To introduce a mutation at the DR4 site of the gene, pTM1562-CAT plasmid was digested with PstI and ligated after Klenow fragment treatment (the plasmid was designated as pTM1562-CAT d(Ϫ1524/Ϫ1521)). Amplified plasmid DNAs in JM109 cells were purified by CsCl density gradient centrifugation in a Hitachi RP120VT rotor.
Cell Culture and Transfection-BxPC-3 cells were cultured into 60-mm diameter noncoated dishes with RPMI 1640 medium containing 10% fetal calf serum (FCS), and the medium was replaced with MCDB medium without FCS at ϳ50% density. Plasmid DNAs treated with a cationic liposomes, TransFast, were prepared in polystyrene tubes and added to the cell cultures. They were allowed to stand for 2 h in a CO 2 incubator, and then the culture medium was added to the cells. Plasmid DNA used amounted to 10 and 5 g/dish for pTM-CATs and pSV-␤galactosidase, respectively, and various amounts for pCMX-RARs. 24 h after the addition of culture medium, t-RA (10 M in 0.1% dimethyl sulfoxide (Me 2 SO), final concentration) or 9C-RA (10 M in 0.1% Me 2 SO, final concentration) was added to the DNA-transfected cells for 24 h. The cells were recovered and sonicated, and the resulting supernatants were used for determinations of CAT and ␤-galactosidase activities. HUVE cells were cultured in Dulbecco's modified Eagle's medium containing 20% FCS without endothelial cell growth supplement and heparin, and secondary cultures on collagen-coated dishes were used for experiments (11,36).
CAT and ␤-Galactosidase Assays and Measurement of TM Antigen Level-For measurement of CAT activity, 1-deoxy[dichloroacetyl-1- 14 C]chloramphenicol was used as the substrate, and the 3Ј-acetylated form of chloramphenicol was detected by autoradiography following thin layer chromatography on a Silica Gel 60 plate using a solvent of chloroform/methanol (19:1, v/v). ␤-Galactosidase activity was measured according to Promega's protocol, and the activity was used as an internal control to normalize the transfection efficiency of individual pTM-CAT plasmids. Total TM antigen in HUVE and BxPC-3 cells was measured after solubilization of cells with 50 mM Tris-HCl, pH 7.5, containing 0.15 M NaCl, 0.5% (w/v) Triton X-100, and 1 mM benzamidine hydrochloride for 30 min at 4°C (12).
Northern Blot and Nuclear Run-on Analysis-Total RNA was isolated from HUVE and BxPC-3 cells (100-mm diameter dishes) with an ISOGEN kit according to the recommended protocol. Northern blot hybridization was performed as described previously (11). For RARs and RXR␣, RNAs were absorbed on a filter using a slot blotter after the confirmation of each specific transcript on 1% denatured agarose gel electrophoresis. A 2057-bp TM cDNA fragment (corresponding Ϫ188 to ϩ1869, where ϩ1 is the first translated base, subcloned into pUC 119 at HindIII and BamHI sites), which was cloned from a placenta cDNA library, was used as a probe. Hybridization probes for RAR␣, RAR␤, RAR␥, and RXR␣ were obtained by treatment of pCMX plasmids with appropriate endonucleases, whereas Sp1 (DNASIS accession number J03133) probe was prepared by PCR using specific primer and cDNA of HUVE cells as a template. The identification of probes was performed by means of endonuclease treatments. The lengths of the probes used were 823 (RAR␣), 882 (RAR␤), 804 (RAR␥), 1074 (RXR␣), and 732 bp (Sp1). Probes were labeled with [␣-32 P]dCTP by using the random primer extension labeling system. Nuclear run-on study was performed using [␣-32 P]dUTP-labeled RNA of nuclear fraction of cells treated or not treated with 10 M t-RA for 6 h. The labeled RNA probe was purified by DNase I and proteinase K treatments, then absorbed on a nitrocellulose filter, and further purified with trichloroacetic acid washing and additional DNase I treatment. The probe eluted by SDS treatment was recovered by ethanol precipitation. Linearized pCMX and pCMX-TM plasmids, of which the latter was constructed by insertion of the 2.6-kb TM cDNA fragment into HindIII and BamHI sites of pCMX, were absorbed on a Hybond-N filter using a slot-blotter and fixed by a transilluminator. The filter was incubated with the labeled RNA probe for 36 h at 42°C in the presence of 40% formamide and 10% dextran sulfate. The intensities of the signals developed by exposure of filters on x-ray films were determined using a Fuji BAS 1500 imaging analyzer (Fuji Photo Film Co., Tokyo, Japan) and expressed relative to the signal of the control band.

Preparations of Nuclear Extracts and in Vitro Translation
Receptors-BxPC-3 cells transfected with pCMX-hRARs and/or pCMX-hRXR␣ were cultured for 24 h in the presence or absence of t-RA (or 9C-RA). The cells were recovered with a cell scraper and centrifuged at 2,000 rpm for 5 min. The resulting pellet was homogenized in a Potter-Elvehjem type homogenizer and further centrifuged at 2,000 rpm for 5 min to obtain the nuclear fraction. This fraction was suspended in a buffer containing 50 mM Tris-Cl (pH 8.3), 20% glycerol, 5 mM MgCl 2 , 0.1 mM EDTA, 0.5 mM phenylmethylsulfonyl fluoride and stored at Ϫ80°C until used as a nuclear extract. RAR␣, RAR␤, and RXR␣ were also prepared in a rabbit reticulocyte lysate translation system using linearized pCMX-hRARs and pCMX-hRXR␣ plasmids as templates for RNA synthesis with T7 RNA polymerase according to the manufacturer's instruction. The amounts of translated proteins were estimated by Western blotting after SDS-polyacrylamide gel electrophoresis (SDS-PAGE).
EMSA and Supershift Assay-Double-stranded oligonucleotides (25mer each) containing DR4 or Sp1 sites were end-labeled with [␥-32 P]ATP by T4 polynucleotide kinase. The upper strand of oligonucleotides with consensus sequences underlined were 5Ј-CGTTTGGTCA-CTGCAGGTCAGTCCA-3Ј for DR4; 5Ј-CTGTGTTGCACGTGCAAGCT-CCGTA-3Ј for scramble DR4; 5Ј-CCTGTCGGCCCCGCCCGAGAACCT-C-3Ј for the third Sp1 site; 5Ј-ATCCCATGCGCGAGGGCGGGCGCA-A-3Ј for the second Sp1 site; 5Ј-GCGCAAGGGCGGCCAGAGAACCCA-G-3Ј for the first Sp1 site; and 5Ј-GCGAGGGCGGGCGCAAGGGCGG-CCA-3Ј for the tandem Sp1 sites containing second and first Sp1 sites. The reaction mixtures for the EMSA contained 0.5 ng of labeled probe (1 ϫ 10 5 cpm/ng), 1 g of poly(dI-dC), and 1 l of nuclear retinoid receptors or 1 unit of Sp1 or AP2 in a final volume of 6 l of 20 mM Hepes-KOH, pH 7.8, 50 mM KCl, 5 mM MgCl 2 , 1 mM ZnCl 2 , 0.5 mM DTT, 0.1 mM phenylmethylsulfonyl fluoride, 5 mM spermidine, 0.008% Nonidet P-40, and 15% glycerol. As sources of nuclear retinoid receptors, nuclear extracts of BxPC-3 cells transfected with pCMX-hRARs and/or pCMX-hRXR␣ plasmids or in vitro translation products were used. As a reference, the preparation obtained by using the same procedure without any plasmid was employed. Competition experiments were performed with a 100-fold excess of unlabeled probe. The mixture was incubated for 30 min at 25°C and subjected to 4% PAGE using a buffer consisting of 10 mM Tris-HCl, pH 7.9, 1.5 mM EDTA, and 5 mM sodium acetate. For the supershift assay, 0.5 g of the respective antibody was further added, and the mixture was incubated for an additional 15 min at 25°C. After electrophoresis, the gels were dried on filter paper and subjected to autoradiography.
Western Blot Analysis-Nuclear extracts were subjected to SDS-PAGE according to the method described previously (37). After SDS-PAGE, protein was transferred to a nitrocellulose membrane and incubated with polyclonal antibodies of RARs, RXR␣, or Sp1. Detections of RA receptors and Sp1 were carried out using biotinylated anti-rabbit IgG and horseradish peroxidase-conjugated streptavidin, and the resulting light emission developed by Renaissance Western blot Chemiluminescence Reagent was captured on Kodak X-Omat autoradiography film.

RA-dependent Augmentation of TM Expression and Acceleration of the Transcriptional
Rate-In the present study, we used human pancreas BxPC-3 cells instead of HUVE cells, for the following reasons. 1) BxPC-3 cells show less damage than HUVE cells after treatment with cationic liposomes, one of the mildest transfection methods (38); 2) their characteristics of t-RA-dependent TM expression are similar to those of HUVE cells (see below); and 3) they can be easily maintained under homogeneous conditions. Fig. 1 shows the changes in antigen and mRNA levels of TM in BxPC-3 cells after treatment with t-RA. Total TM antigen in BxPC-3 cells was 14.1 Ϯ 1.1 ng/1 ϫ 10 5 cells (HUVE cells contained 17.3 Ϯ 1.4 ng/1 ϫ 10 5 cells) and increased to 2.8 times the control on treatment with 10 M t-RA for 24 h (HUVE cells showed an increase to 2.5 times the control on similar treatment) (Fig. 1A). The TM mRNA level of BxPC-3 cells was increased by t-RA treatment and reached 3.7 times the control at 10 M t-RA for 24 h (Fig. 1B). The time courses of the t-RA-dependent increase in TM antigen and in mRNA levels of BxPC-3 cells were also similar to those of HUVE cells (data not shown). These dose-and time-dependent increases in TM antigen and mRNA levels of BxPC-3 cells after t-RA treatment were in accordance with the result in HUVE cells reported previously (11). To assess the effect of t-RA on TM gene transcription rate, a nuclear run-on study was performed on nuclei prepared from both HUVE and BxPC-3 cells treated or not treated with 10 M t-RA for 6 h (Fig. 1C). Treatment with t-RA increased the TM transcription rate to 310 and 325% of the respective control level for HUVE and BxPC-3 cells (Fig. 1C). Furthermore, TNF-␣-dependent downregulation of the TM antigen level in BxPC-3 cells was counteracted by synchronized treatment with t-RA (data not shown), as seen in HUVE cells (12).
RA-dependent Regulatory Sequence of TM Gene-We examined the promoter activity of TM by transient expression assay using constructs of pTM-CAT plasmids. Although we could not clearly identify the transcription initiation site for human TM mRNA by S1 nuclease protection analysis, a faint signal at Ϫ168 base from the ATG codon of the translation site was consistent with one of the two sites reported by Yu et al. (15), and thus the first untranscribed base Ϫ1 was defined as the Ϫ169 base from the ATG codon in this paper. Fig. 2 shows the sequence-specific promoter activity of TM as determined by measuring the CAT activity of BxPC-3 cells transfected with pTM-CAT plasmids, which have different TM 5Ј-flanking sequences. Compared with the CAT activity of BxPC-3 cells transfected with pTM1562-CAT, the activities of cells transfected with the deletion plasmids decreased gradually when the deletions of TM 5Ј-flanking sequences were extended to Ϫ549. Although the CAT activity of cells transfected with pTM549-CAT was half that in the case of pTM1562-CAT, plasmids with more extensive deletions, such as pTM479-CAT and pTM404-CAT, showed some recovery in CAT activity. When pTM213-CAT was transfected, the CAT activity was markedly increased as compared with that of pTM244-CAT, and a further deleted plasmid (pTM119-CAT) decreased the activity to a level less than that of pTM244-CAT plasmid. The results suggest that negative-and positive-acting elements are present in the sequences around Ϫ244 to Ϫ214 and Ϫ213 to Ϫ120, respectively, of the TM gene. The effect of t-RA on the promoter activity of TM was measured (Fig. 3). The ratio of increase of CAT activity by t-RA was different for each plasmid used. The 4.1-fold increase of the activity by t-RA treatment in the case of pTM1562-CAT decreased to ϳ2.3-fold when pTM-CAT plasmids containing the 5Ј-site from Ϫ1491 to Ϫ213 of the TM sequence were transfected, and further deletion of the 5Ј-sequence abolished the t-RA-dependent increase in the activity. These results indicate the existence of two independent RAresponsive regions (one localized at Ϫ1562 to Ϫ1492 bp and the other localized at Ϫ213 to Ϫ120 bp of the transcription site) in

FIG. 1. t-RA increases antigen and mRNA levels of TM in BxPC-3 cells, a human pancreas carcinoma cell line.
When BxPC-3 cells cultured with RPMI 1640 medium containing 10% FCS became confluent, t-RA was added to the culture medium at various concentrations for 24 h and TM antigen (A) and TM mRNA (B) levels of the cells were measured as described under "Experimental Procedures." C, nuclear run-on assays was performed on nuclei obtained from HUVE and BxPC-3 cells treated with 10 M t-RA or 0.1% Me 2 SO for 6 h as described under "Experimental Procedures." the TM promoter sequence.
To identify the RA-responsive sequences, mutations were introduced at DR4 of RARE in the pTM1562-CAT plasmid and at two consensus Sp1 sites in the pTM346-CAT plasmid (Fig.  4). As shown in Fig. 4A, mutation at DR4 reduced the t-RA-dependent increase in CAT activity of pTM1562-CAT without extensive reduction of the normal CAT activity, and the increase ratio became 2.4-fold, which is comparable to that in the case of pTM1491-CAT (Fig. 3). Each deletion at the third or second Sp1 motif in the TM promoter sequence of pTM346-CAT produced different responses (Fig. 4B). pTM346-CAT d(Ϫ208/ Ϫ201) showed a slight decrease in the CAT activity compared with that of the wild-type plasmid and exhibited a mild response to t-RA treatment, whereas pTM346-CAT d(Ϫ142/ Ϫ135) decreased the CAT activity, irrespective of t-RA treatment. The reduction with the plasmid deleted at the second Sp1 motif of the TM gene sequence reveals the significance of this region in the primary and t-RA-dependent enhancement of the transcriptional activity. The third Sp1 motif might contribute to cooperative interactions of other transcription factors with Sp1 protein. On the other hand, a plasmid deleted from Ϫ221 to Ϫ214 (pTM346-CAT d(Ϫ221/Ϫ214)) enhanced the CAT activity of the transfectant to 165% of that of pTM346-CAT, and further enhancement of the activity was observed after treatment with t-RA. As expected from the result in Fig. 2, this deleted region might act negatively, in concert with an unknown factor, on the TM promoter activity.
Expressions of RA Receptors Modulate Promoter Activity of TM-To examine the effect of changes in concentration of RA receptors on the promoter activity of TM in BxPC-3 cells, expression plasmids of RA receptors were cotransfected with pTM1562-CAT plasmid. Cotransfection of the pTM-CAT plasmid and each of pCMX-hRAR␣, pCMX-hRAR␤, and pCMX-hRAR␥ plasmid into BxPC-3 cells resulted in a significant increase in CAT expression in the absence of retinoid, especially when more than 0.1 g of pCMX-hRARs was cotransfected (Fig. 5). However, enhancement of the promoter activity of the cells by t-RA or 9C-RA treatment did not increase further after the cotransfection of each RAR expression plasmid. As shown in Fig. 5D, cotransfection of pCMX-hRXR␣ and pTM1562-CAT plasmids into the cells resulted in little increase in the promoter activity in the absence of retinoid treatment, and a remarkable decrease in the activity was seen in the cells transfected with more than 1 g of pCMX-RXR␣, despite treatment of the cells with t-RA or 9C-RA. The results suggest that the retinoid-dependent increase in TM promoter activity involves at least two different regulations, that is retinoid receptors-dependent and -independent pathways. RAR-RXR heterodimers activate transcription in response to t-RA or 9C-RA by binding to target gene response elements consisting of DR5 or DR2 (26,39,40), whereas RXR-RXR homodimers activate transcription in response to 9C-RA by binding to DR1 (28, 40 -42). Since the levels of activation of TM promoter activity were the same in the cells treated with t-RA and 9C-RA (Fig. 5), it appears that the homodimer of RXR␣ (RXR␣-RXR␣) does not function as an activator of the promoter activity. Furthermore, suppression of the promoter activity by RXR␣-RXR␣ homodimer may also be excluded, because RXR␣ did not bind to the DR4 element of the TM gene (see below).
Effect of Concentration of Retinoid Receptor on Promoter Activity of TM-In the present study, RXR␣ was used as the representative of RXRs, because both HUVE and BxPC-3 cells express its mRNA and protein (see the data in Figs. 10 and 11), and because it has been reported that RXR isoforms such as RXR␣, RXR␤, and RXR␥ behave similarly in heterodimeric complex formation with RARs or TR (23,(43)(44)(45). It is also reported that 9C-RA has the highest affinity for RXR␣ among RXR isoforms (44). The promoter activity of pTM1562-CAT  plasmid in the case of simultaneous expressions of RAR␣ or RAR␤ and RXR␣ in the cells was examined (Fig. 6). The RAR␣and RAR␤-dependent increases in promoter activities were further enhanced by cotransfection of pCMX-RXR␣ (0.1 g) (Fig. 6, C and D). However, cotransfection with pCMX-RAR␣ (or pCMX-RAR␤) and a higher dose of pCMX-RXR␣ (1 g) drastically reduced the promoter activity, even when 0.1 g of pCMX-RARs was used (Fig. 6, E and F). These results demonstrate that TM promoter activity is controlled by the expressions of RARs and RXR␣ and that the extent of the expression of the latter is critical for gene expression of TM.
Sequence-specific Acceleration of TM Promoter Activity by RAR-RXR Expressions-To confirm that RA receptors modulate TM gene expression through the RARE element, we determined the promoter activity of BxPC-3 cells transfected with DR4-deleted pTM1562-CAT plasmid (pTM1562-CAT d(Ϫ1524/ Ϫ1521)) or pTM346-CAT plasmid instead of pTM1562-CAT plasmid (Fig. 7). The following observations were made. First, the promoter activity of the cells transfected with pTM1562-CAT plasmid was increased by the coexpression of RARs and RXR␣ in the absence of retinoid treatment, as already shown in Figs. 5 and 6, whereas the promoter activity of those transfected with pTM1562-CAT d(Ϫ1524/Ϫ1521) or pTM346-CAT was not increased under the same conditions. Second, the promoter activities of cells transfected with these pTM-CAT plasmids were increased by treatment with t-RA or 9C-RA under conditions where retinoid receptors were not expressed, but the increase ratio after retinoid treatment was the highest when pTM1562-CAT plasmid was used. Third, the 9C-RA-dependent increase in the promoter activity of pTM1562-CAT in the cells, which expressed RARs and a small amount of RXR␣, was not greater than the t-RA-dependent increase. These results suggest that 1) the interaction between heterodimer of RAR-RXR␣ and the DR4 element in TM gene participates in part in the retinoid-dependent increase in the promoter activity of TM; 2) ligand-independent dimerization of RARs-RXR␣ occurs, although we cannot exclude the possibility that a small but sufficient amount of t-RA exists in the cell in the absence of retinoid treatment; and 3) formation of RAR-RXR rather than RXR-RXR is significant when RXR␣ is not highly expressed. Interaction of RAR-RXR Heterodimer with DR4 Sequence of TM Gene-To test the idea, that RAR-RXR␣ heterodimer binds to the DR4 site of the TM gene, interaction between RARs, RXR␣, and a DNA probe containing the DR4 sequence of the TM gene was examined by EMSA (Fig. 8). When a nuclear fraction of BxPC-3 cells transfected with each pCMX-hRAR plasmid was incubated with the probe, a retarded band was observed, regardless of the isoform, although that of cells transfected with RXR␣ expression plasmid showed only a weak band (Fig. 8A). By using a nuclear fraction prepared from cells cotransfected with one RAR isoform and RXR␣ expression plas-mids, a more distinct band with the same mobility was observed. The same band on the EMSA was also obtained when the cDNA probe was incubated with a sample, which was prepared by mixing the nuclear fraction expressing RAR␣ (or RAR␤) with that expressing RXR␣ (Fig. 8B, lane 3). The binding specificity of these receptors to the radiolabeled probe corresponding to the DR4 sequence of TM gene was confirmed by competition with an excess amount of the unlabeled cDNA probe. Moreover, these retarded bands were further shifted in the presence of RAR␤ and RXR␣ antibodies (lanes 7 and 9), and therefore it was confirmed that RAR␤-RXR␣ heterodimer bound to the consensus sequence of the DR4 element in the TM gene. To exclude contamination with endogenous receptors, RA receptors synthesized with a cell-free reticulocyte lysate system were also subjected to EMSA using the same DNA probe (Fig. 8C). Since the RAR␣-or RAR␤-dependent retarded band was observed in the absence of endogenous RXR␣ in the cellfree system, it is possible that the homodimer of each RAR could bind to DR4 of the TM sequence. Furthermore, it became apparent that homodimer (or monomer) of RXR␣ did not bind to the DR4 site (lane 3). Specific binding of the heterodimer binding to this DR4 sequence was further confirmed using a scramble probe of the DR4 sequence as a competitor (lane 8).
Effect of Ratio of RAR/RXR on Binding of RAR-RXR to DR4 Sequence-Since an RXR␣ concentration-dependent decrease in the promoter activity was observed in Fig. 6, the effect of the ratio of RAR␤ to RXR␣ on interaction of the heterodimer with the DR4 sequence was investigated (Fig. 9). Nuclear samples of BxPC cells cotransfected with RAR␤ and RXR␣ at various ratios were prepared, and the binding ability was examined  using DR4 probe. The intensity of the band corresponding to the retardate complex with RAR␤-RXR␣ heterodimer initially increased with the increase in the expression of RXR␣, but a further increase in the RXR␣ expression apparently decreased the intensity and no retardate band was observed at 5-fold excess of RXR␣ over RAR␤ (Fig. 9B). When RAR␤ and RXR␣, prepared in a reticulocyte lysate system instead of by cotransfection of cells with both plasmids, were mixed in the reaction mixture at various ratios, virtually the same results were obtained (Fig. 9B), indicating rapid formation of RAR␤-RXR␣ dimer. These results suggest that the ratio of RARs to RXR␣ is critical for not only formation of RAR-RXR heterodimer but also binding of the heterodimer to the DR4 sequence. Since the maximum binding was observed when a higher ratio of RXR␣ compared with RAR␤ was applied (Fig. 8B), the formation of RXR␣-RXR␣ heterodimer and/or the presence of RXR␣ monomer may be possible under the cell-free condition.
Effect of t-RA or 9C-RA on mRNA and Protein Levels of Retinoid Receptors-Changes in mRNA levels of retinoid receptors in BxPC-3 cells after treatment with t-RA or 9C-RA were measured (Fig. 10). The mRNA level of TM increased in a time-dependent manner, which was compatible with the results in Fig. 1. BxPC-3 cells contained mRNA of all four RA receptors, and RAR␤ mRNA was markedly augmented after the treatment with t-RA. The level of RAR␣ mRNA also increased within 6 h after t-RA treatment, whereas RAR␥ mRNA increased only slightly up to 24 h after the treatment. On the other hand, the RXR␣ mRNA level decreased with increase in the time after t-RA treatment. Changes in the mRNA levels of these four RA receptors in HUVE cells after treatment with t-RA were also examined, and quite similar results to those in BxPC-3 cells were obtained (data not shown). The presence of RAR␣ and RAR␤ mRNAs in normal HUVE cells has already been reported by Fesus et al. (46), and Kooistra et al. (47) observed the expression of transcripts of three RAR and two RXR subtypes in normal HUVE cells. They also found induction of RAR␤ and RXR␣ mRNA after exposure of the cells to t-RA. Although there was definitive difference in changes in RXR␣ level after t-RA treatment between their results and ours, similar patterns of increase in RAR␤ mRNA level and decrease in RXR␣ mRNA level in BxPC-3 cells were observed when 9C-RA instead of t-RA was used (Fig. 10, C and D). Western blot analysis showed that these changes in mRNA levels of RAR␤ and RXR␣ reflected those of the protein levels in the nuclear fraction of the cells after treatment with t-RA (Fig.  11). Thus, treatment with t-RA gradually increased and decreased the amounts of RAR␤ and RXR␣ in the nuclear fraction of BxPC-3 cells, respectively.
Interaction of Nuclear Sp1 Protein with Sp1-dependent Sequence in TM Gene-Next we focused on Sp1 as a candidate factor to interact with the GC box of the TM gene, and EMSA was performed after incubation of 32 P-labeled TM cDNA probes containing these Sp1 sites with Sp1 or AP2 (the latter is also a transcription factor that binds to GC-rich sequences (50) and has been shown to play a role in some retinoid-affected morphogenetic processes (51)). The promoter activity of pTM346-CAT was not influenced by coexpression of RARs and/or RXR␣ (Fig. 7), but recently Suzuki et al. (52) showed that RAR-RXR interacts physically with Sp1 and that the complex of RAR-RXR-Sp1 binds to the GC box of the urokinase promoter sequence. Therefore, the probe was incubated with recombinant Sp1 in the presence or absence of both RAR and RXR prepared in a reticulocyte lysate system, followed by electrophoresis. Fig.  12A shows that Sp1 could form a retarded complex with labeled probes containing an Sp1-binding site in the presence of RAR␤ and RXR␣. The intensity of the retarded bands of cDNA probes containing the three Sp1-binding sites were in the following order: second Ͼ third Ն first Sp1 sites (lanes 2, 5, and 8, respectively). However, these Sp1 sites were not target sequences for AP2 (data not shown). As shown in Fig. 4, the significance of the second Sp1-binding site among the these sites in the TM promoter region was evaluated by measuring the promoter activity of a TM-CAT plasmid in which this second Sp1 site was deleted. The shifted bands with slower and faster migrations were Sp1-dependent and -independent complexes, respectively, since the lower band with fast migration appeared in the absence of Sp1 in the reaction mixture (lanes 1,  4, and 7). The most intense shifted band was seen with a cDNA probe containing tandem Sp1 sites with 6 spaces corresponding to the second and first Sp1-binding sites of the TM gene (lane 11).
The effect of RA receptors on the Sp1 binding to this tandem Sp1 site was further investigated by EMSA (Fig. 12B), because a curious result was obtained as described below. The Sp1-dependent retarded band with slow migration was detected on EMSA when this probe was incubated with Sp1 and one retinoid receptor, such as RAR␣, RAR␤, or RXR␣ (lanes 1, 3 and 5), and all of the corresponding bands disappeared upon addition of excess competitor (lanes 2, 4, and 6). However, the coexistence of RARs and RXR␣ did not increase the intensity of the band in the presence of Sp1 (lanes 7 and 9), and the bands were not supershifted by the addition of the respective retinoid receptor antibody (lanes 12-14). The supershifted band was restricted to the case of Sp1 antibody (lanes 11, 15, and 16). Surprisingly, as shown in lane 21, the Sp1-dependent retarded band was detected in the absence of RARs and RXR␣ but in the presence of reticulocyte lysate solution as the control for all the RA receptors. These results suggest that interaction between the Sp1 and Sp1 sites of the TM promoter sequence is dependent on the presence of unknown factor(s) in the reticulocyte lysate solution and that the interaction is independent of formation of the multicomplex of Sp1-RAR-RXR␣. Therefore, we concluded that homodimer and heterodimer of retinoid receptors do not participate directly or indirectly in the interaction between Sp1 and the tandem Sp1 sites of TM gene. This is consistent with the finding in Fig. 7 that expressions of RARs and/or RXR␣ did not induce transactivation activity of pTM346-CAT, which contains all four Sp1 sites of the TM gene. Nevertheless, it is true that Sp1 binds specifically to these Sp1  11. Changes in antigen levels of RAR␤ and RXR␣ in BxPC-3 cells treated with t-RA. BxPC-3 cells cultured in 100-mm dishes were treated with 10 M t-RA for 0, 6, and 24 h, and nuclear extracts were prepared as described under "Experimental Procedures." After SDS-PAGE, protein was transferred to a nitrocellulose membrane and incubated with polyclonal antibodies to RAR␤ and RXR␣. A, four isoforms of RAR␤ corresponding 56 to 47 kDa (48), as shown by an arrow, were recognized, and a low molecular band (arrowhead) seems to be a degradation product of RAR␤. B, RXR␣ corresponding 53 kDa (49) was recognized. Detection was carried out using antibodies to RAR␤ and RXR␣ followed by biotinylated anti-rabbit IgG and horseradish peroxidase-conjugated streptavidin, and the light emission developed was captured on an autoradiography film. Open arrowheads in A and B indicated nonspecific bands.
probes, because all the retarded bands were eliminated by an excess of each unlabeled probe. Further increase in the intensity of the retarded band on EMSA was directly proportional to the concentration of Sp1 in the presence of a small amount of reticulocyte lysate solution (Fig. 12C, lanes 3-8 and 9 -14). In addition, the retarded band with slower migration was super-shifted in the presence of the specific antibody of Sp1 (lane 16), and it was also confirmed that AP2 was unable to form a retarded complex with the same cDNA probe (lanes 18). Binding of Sp1 to Sp1 sites of the TM sequence was also independent of the presence of t-RA in EMSA (data not shown).
RA-dependent Changes in mRNA and Protein Levels of The closed and open arrowheads indicate the Sp1-dependent and -independent DNA-protein complexes, respectively. A, lanes 1-3, probe (Ϫ217 to Ϫ193) containing the third Sp1 site; lanes 4 -6, probe (Ϫ154 to Ϫ130) containing the second Sp1 site; lanes 7-9, probe (Ϫ135 to Ϫ111) containing the first Sp1 site; lanes 10 -12, probe (Ϫ145 to Ϫ121) containing the tandem Sp1 site corresponding to the second and the first Sp1 sites. Probes were incubated with recombinant Sp1 (lanes 2, 3, 5, 6, 8, 9, 11, and 12) in the presence of in vitro translation products from RAR␤ and RXR␣ mRNAs by a reticulocyte lysate system. As a competitor, 100-fold unlabeled probe was added to the incubation mixture (lanes 3, 6, 9, and 12). B, probe containing a tandem Sp1 site incorporating the second and the first Sp1 sites of the TM promoter sequence was incubated with recombinant Sp1 (lanes 1-16, 18, 21, and 22) in the presence of in vitro translation products of RAR␣ (lanes 1, 2, 7, and 8), RAR␤ (lanes 3 and 4, and 9 -18), and RXR␣ (lanes [5][6][7][8][9][10][11][12][13][14][15][16][17][18] or the reticulocyte lysate solution (RL) as a reference for these retinoid receptors (lanes 19 -22). Anti-Sp1 IgG (lanes 11, 15, and  16), anti-RAR␤ IgG (lanes 12, 14 and 15) and anti-RXR␣ IgG (lanes 13, 14 and 16) were added at 15 min before electrophoresis. As a competitor, 100-fold unlabeled probe was added to the incubation mixture (lanes 2, 4, 6, 8, 10, 20, and 22). C, probe containing the tandem Sp1 site was incubated with various concentrations of recombinant Sp1 (lanes 4 -8 and 10 -17) or AP2 (lanes 18 and 19) in the presence of the reticulocyte lysate solution (RL) (lanes 2-8 (1 l) and lanes 9 -19 (0.05 l)). As a competitor, 100-fold unlabeled probe was added to the incubation mixture (lanes 2, 15, 17, and 19). Anti-Sp1 IgG was added in lane 16. Sp1-The time course of changes in mRNA level of Sp1 after treatment of BxPC-3 and HUVE cells with t-RA was examined (Fig. 13). The Sp1 mRNA was expressed not only in BxPC-3 cells but also in HUVE cells. Sp1 mRNA increased rapidly after treatment of BxPC-3 cells with t-RA, and the maximum increase was ϳ2.5-fold from the control level. The increase in the ratio and the time course of change of each mRNA level in HUVE cells after treatment with t-RA were almost the same as those in BxPC-3 cells. The time course of changes in Sp1 content after treatment of BxPC-3 cells with t-RA was examined (Fig. 13C). The appearance of a 105-kDa band, a phosphorylated form of Sp1 (53), was observed, in addition to a rapid increase in the 95-kDa Sp1, the dephosphorylated form (53). Upon alkaline phosphatase treatment of a sample, disappearance of the 105-kDa band followed by increase in the 95-kDa band was observed, indicating the existence of the two forms in the cells. These results suggest that treatment of human cells with t-RA modulates the TM gene promoter activity in part through increasing Sp1 content and phosphorylation in the nuclear fraction. DISCUSSION In this paper we have shown that t-RA-dependent enhancement of human TM gene expression is regulated by two sites in the 5Ј-flanking region. The t-RA-responsive distal region of the TM gene contains an RARE composed of TGGTCANNNNAG-GTCA. Our transient expression studies revealed that the TM promoter could be transactivated by the human RARs, especially by RAR␣ and RAR␤. On the basis of transient expression assay using F9 embryonal carcinoma cells, Weiler-Guettler et al. (54) suggested that the TM promoter could be activated by RAR␤. By the use of an RAR␣ antagonist, on the other hand, Shibakura et al. (55) obtained results indicating a major role of RAR␣ in TM up-regulation by retinoids in leukemia and HUVE cells. Treatment with t-RA (or 9C-RA) in the present study, however, caused a much greater increase in the RAR␤ level compared with that of RAR␣, and no difference was observed between RAR␣-and RAR␤-dependent transactivation of TM promoter activity. Furthermore, the binding ability of RAR␣ and RAR␤ with the DR4 sequence of the TM gene in the presence or absence of RXR␣ was the same when equivalent amounts of both RAR receptors were used. Therefore, the in-crease in the total amount of all the RARs might be a dominant factor for the up-regulation of transcriptional activity of TM, rather than the changes in the concentrations of individual RAR subtypes. Coexpression of RXR␣ with RAR␣ or RAR␤ showed a different mode of promoter activity compared with that in the case of expression of a single RAR receptor. When RXR␣ was highly expressed in BxPC-3 cells, the increase in the transactivation activity by RAR␣ or RAR␤ was eliminated, whereas it was augmented at a lower expression level of the RXR␣ receptor. Furthermore, the binding of RAR␤-RXR␣ to RARE of the TM gene was closely related to the proportion of RAR␤/RXR␣, and the levels of RARs increased in both BxPC-3 and HUVE cells after t-RA or 9C-RA treatment, whereas that of RXR␣ decreased under the same treatment. It seems very probable that t-RA and 9C-RA contributed to the acceleration of TM promoter activity through both an increase in RARs and a decrease in RXR␣, followed by appropriate formation of RARs-RXR␣ heterodimers that can bind to the DR4 site of the TM gene. In this connection, we have examined the changes in the levels of RARs and RXR␣ of HUVE cells after treatment with a synthetic RARs-selective retinoid, Ch55. The results supported the above finding that an increase in TM promoter activity is closely related to the increasing ratio of RARs/RXR␣.
To access the action of t-RA on the DR4 site-dependent enhancement of promoter activity of TM, a ligand-dependent change in the binding of RAR␤-RXR␣ to the DR4 site of the TM gene was investigated. In our EMSA experiment, essentially no increase in the intensity of the retarded band of the DR4 probe was observed in the presence of t-RA and/or 9C-RA in the reaction mixture, although a small increase in the intensity was observed at higher concentrations (10 Ϫ6 and 10 Ϫ5 M) (data not shown). Therefore, it is plausible that t-RA and 9C-RA are not stimulators of the interaction between the heterodimer and DR4 of the TM gene, in a similar manner to the Sp1 and Sp1 sites of the TM gene (see below), although we cannot exclude the possibility that both retinoids act as accelerators of the heterodimer formation of RAR-RXR, regardless of the presence or absence of the template DNA.
On the other hand, the protein that interacts with the proximal sequence for transcriptional up-regulation appeared to be Sp1. Among the four Sp1 sites in the 5Ј region from the TATA box of the gene, the major role of the second Sp1 site, rather than the other sites, was found. Although it has been shown that one Sp1 site is sufficient for the Sp1-dependent augmentation of the promoter activities of cytochrome P450IA1 (56) and ␣2(I) collagen (57), there was a significant cooperative role of the second Sp1 site with the other adjacent Sp1 sites of the TM gene, especially the first Sp1 site. It is probable that multiple or adjacent Sp1 sites in the TM gene are essential for the maximal activation of the promoter, as reported for other genes (32, 58 -61). It is known that the binding and subsequent transactivation activities of Sp1 can be modulated by posttranslational modification, i.e. glycosylation (62,63) and phosphorylation (53, 64 -66). After t-RA treatment, the phosphorylated form of Sp1 (105 kDa) was increased (Fig. 13), and the binding activity of Sp1 to the consensus Sp1 site of the TM gene was lost when reticulocyte lysate in the reaction mixture was previously heat-inactivated (data not shown). These results suggest that the phosphorylation of Sp1 is important for the interaction between Sp1 and these Sp1 sites of the TM sequence. In our preliminary experiment, the binding of Sp1 to the Sp1 sites of the TM gene was increased or decreased after Sp1 was phosphorylated or dephosphorylated, respectively. Further studies on the functional regulation of TM promoter activity by such modifications of the Sp1 molecule in HUVE cells are in progress in our laboratory. In addition, the effects of A and B, BxPC-3 and HUVE cells cultured in 100-mm diameter dishes were treated with 10 M t-RA for 0, 3, 6, and 24 h. Total RNA was prepared, and Northern blot hybridization was performed using Sp1 and ␤-actin probes after 1% denatured agarose gel electrophoresis. Two transcripts for Sp1 (Ϸ8.0 and Ϸ5.5 kb) were detected. C, nuclear fraction of BxPC-3 cells treated with t-RA until 24 h was prepared, and SDS-PAGE was performed. Proteins transferred to a nitrocellulose membrane were incubated with polyclonal antibody of Sp1 as described under "Experimental Procedures." Two bands corresponding to 95 and 105 kDa are the dephosphorylated and phosphorylated forms, respectively, of Sp1 (53). Sp1-related factors, which recognize GC boxes, on the promoter activity should be determined, because some papers indicated that Sp1-mediated promoter activation was accelerated and attenuated by Sp1-related family members (58,67,69).
It could be assumed that interaction between Sp1 and RARs-RXR␣ does not serve as a cooperative process for Sp1 site-dependent promoter activity of TM. However, we cannot exclude the possible formation of a complex between Sp1 and RAR-RXR, which binds to sites other than Sp1 sequences of the TM gene. Synergistic action of Sp1 with glucocorticoid receptor (70) or estrogen receptor (71,72) has been reported. Krey et al. (61) demonstrated that Sp1 was not able to interact synergistically with peroxisome proliferator-activated receptor-RXR, although these three nuclear proteins could bind simultaneously to the acyl-CoA oxidase gene. However, in our experiment, an increase in the TM transcription activity by 9C-RA was not further elevated by the synchronous treatment of HUVE cells with 9C-RA and TR activator (T 3 ) or activators of PPAR subtypes such as Wy-14,643 and BRL49653 (data not shown). RXR can form heterodimers with many partners, including orphan receptors (for reviews, see Refs. [73][74][75], and thus it has been accepted that RXR is a key regulator of various cellular events in receptor-dependent signaling. Since the total amount of all the RXR subtypes was larger than that of all the RAR subtypes (68), the transcriptional activity of TM might be affected by changes in the amounts of various nuclear proteins that could form heterodimers with RXRs under various conditions.