Identification and Characterization of a Versatile Retinoid Response Element (Retinoic Acid Receptor Response Element-Retinoid X Receptor Response Element) in the Mouse Tissue Transglutaminase Gene Promoter*

Tissue transglutaminase (transglutaminase type II) is an intracellular protein cross-linking enzyme that accu-mulates in connective tissue and in cells undergoing apoptosis. Retinoids regulate the transcription of the mouse tissue transglutaminase gene via activation of regulatory elements contained within 4 kilobases of the 5 (cid:42) -end of the gene. Co-transfection studies with retinoid receptor expression vectors in CV-1 cells demonstrated that the mouse tissue transglutaminase promoter is activated by ligand activation of either retinoic acid re- ceptor-retinoid X receptor (RAR (cid:122) RXR) heterodimers or RXR homodimers. Optimal induction is achieved with retinoid receptor panagonists; partial activation can also be achieved with either RAR-specific or RXR-spe- cific retinoids. Retinoid-dependent activation of the tissue transglutaminase promoter depends on both a prox- imal regulatory region containing sequences highly conserved between the human and the mouse tissue transglutaminase promoters and a distal region that includes a 30-base pair retinoid response element (mTGRRE1).


Tissue transglutaminase (transglutaminase type II) is an intracellular protein cross-linking enzyme that accumulates in connective tissue and in cells undergoing apoptosis. Retinoids regulate the transcription of the mouse tissue transglutaminase gene via activation of regulatory elements contained within 4 kilobases of the 5-end of the gene. Co-transfection studies with retinoid receptor expression vectors in CV-1 cells demonstrated that the mouse tissue transglutaminase promoter is activated by ligand activation of either retinoic acid receptor-retinoid X receptor (RAR⅐RXR) heterodimers or RXR homodimers. Optimal induction is achieved with
retinoid receptor panagonists; partial activation can also be achieved with either RAR-specific or RXR-specific retinoids. Retinoid-dependent activation of the tissue transglutaminase promoter depends on both a proximal regulatory region containing sequences highly conserved between the human and the mouse tissue transglutaminase promoters and a distal region that includes a 30-base pair retinoid response element (mTGRRE1). mTGRRE1 contains three hexanucleotide half-sites (two canonical and one non-canonical) in a DR7/DR5 motif that bind both RAR⅐RXR heterodimers and RXR homodimers. These studies suggest that retinoid-dependent expression of the mouse tissue transglutaminase gene is mediated by a versatile tripartite retinoid response element located 1.7 kilobases upstream of the transcription start site.
Tissue transglutaminase is a multifunctional protein involved in a number of distinct biological processes (1,2). In addition to its well characterized activity as a catalyst of protein cross-linking reactions, recent studies have shown that tissue transglutaminase also can function as a GTP-dependent regulator of phospholipase C activity (3). A unique feature of tissue transglutaminase's cross-linking activity is that, unlike other transglutaminases, which are restricted in their activity, tissue transglutaminase cross-links proteins in both intracellular and extracellular compartments (2). Extracellular protein cross-linking activity occurrs in basement membranes (4), during chondrogenesis (5) and particularly during wound healing (6). Tissue transglutaminase-mediated cross-linking of intracellular proteins occurs in cells undergoing apoptotic or programmed cell death (7,8). It has been suggested that the induction and activation of tissue transglutaminase in apoptotic cells results in protein cross-linking that stabilizes the apoptotic bodies and prevents the leakage of cytosolic proteins into the extracellular space during cellular fragmentation (9).
The multiple activities of tissue transglutaminase suggest that its expression and activation must be tightly regulated. Tissue transglutaminase has a very selective pattern of tissue-specific expression, with the highest levels of the enzyme accumulating in endothelial cells, in erythrocytes, and in cells of the lens epithelium (10,11). Retinoids appear to be generalized regulators of tissue transglutaminase expression. Rats rendered vitamin A-deficient have a marked depression in the level of tissue transglutaminase activity in many tissues (12). This decrease is rapidly reversed by administration of all-trans retinoic acid (ATRA). 1 Administration of retinoids to rats with normal retinoid nutritional status causes increased transglutaminase activity in several tissues (12)(13)(14)(15). The induction of tissue transglutaminase in vivo can be reproduced in vitro. Exposure of macrophages to ATRA results in a dramatic increase in tissue transglutaminase activity (16), replicating the induction that occurs during macrophage activation (17). Retinoids also induce the enzyme in human promyelocytic leukemia (HL-60) cells (18,19), replicating the induction of the enzyme that occurs during myeloid cell apoptosis. TGF-␤ and retinoids induce transglutaminase activity in tracheal epithelial cells (20,21), and IL-6 induces the enzyme in hepatocytes (22).
The transcriptional effects of retinoids on gene expression are mediated by at least two families of retinoid receptors, retinoic acid receptors (RARs) and retinoid X receptors (RXRs), both of which belong the superfamily of nuclear receptors that mediate the transcriptional effects of the steroid hormones, vitamin D, and thyroid hormone (23,24). The complexity of retinoid signaling is further increased by the fact that there are three distinct genes encoding RARs (RAR␣, -␤, and -␥) and RXRs (RXR␣, -␤, and -␥) that give rise to homologous but distinct receptor proteins (25)(26)(27)(28)(29)(30)(31). Both the RAR and RXR receptor subtypes are expressed in specific temporal and spatial patterns during embryogenesis and in tissue-specific patterns in adult animals, suggesting that they each have distinctive biological activities (32). Vitamin A (all-trans retinol) does not bind to retinoid receptors, but its metabolites, such as ATRA and 9-cis-RA, do. ATRA only binds to RARs, whereas 9-cis-RA binds to and activates both RARs and RXRs (33,34). Retinoid receptors do not bind to DNA as monomers; they require dimerization for activity. RARs preferentially dimerize with RXRs to form RAR⅐RXR heterodimers that are thought to be obligatory intermediates in the effects of RAR ligands on gene expression (35,36). RXRs can associate with several other receptors including thyroid hormone receptors (TR), the vitamin D receptor (VDR), and the receptor for drugs that induce peroxisomal proliferation (PPARs) (23). RXRs also can homodimerize to give transcriptionally active complexes (37,38).
Homo-and heterodimeric retinoid receptor complexes bind to distinct retinoid response elements (RREs) embedded in the regulatory regions of retinoid-responsive genes (23,39). Although there is considerable variability in the sequence and structure of the RREs in retinoid-regulated genes, they conform to a general canonical sequence in which two directly repeated receptor-binding hexanucleotide motifs (consensus (A/G)G(G/T)TCA) are separated by a variable number of intervening nucleotides (39). RXR homodimers bind preferentially to hexanucleotide motifs separated by one nucleotide (DR1), whereas RXR⅐RAR heterodimers bind to motifs separated by one, two, or five nucleotides (DR1, DR2, or DR5). Other heterodimeric complexes (RXR⅐TR, RXR⅐VDR, and RXR⅐PPAR) bind to motifs with characteristic preferred spacings (39).
We have been interested in the molecular mechanisms that contribute to retinoid-regulated expression of the tissue transglutaminase gene. Studies in both normal and leukemic myeloid cells have shown that ATRA acts as an acute and direct activator of tissue transglutaminase gene expression (40). Studies with synthetic retinoids that activate only RARs or RXRs have suggested that ligand activation of both receptors can induce tissue transglutaminase expression, suggesting that both RAR and RXR-mediated signaling pathways may regulate expression of this enzyme (21,41). 2 Recently we cloned the promoter of the mouse tissue transglutaminase gene and have shown that, like the endogenous gene, this promoter is regulated by ligand activation of both RARs and RXRs. 2 In the studies reported here, we have isolated and characterized a retinoid response element embedded in this promoter fragment. We have found that this retinoid response element is versatile, capable of binding to and being activated by both RAR⅐RXR heterodimer and RXR homodimer receptor complexes. These observations suggest that induction of tissue transglutaminase expression may be a response to multiple pathways of retinoid signaling.

Materials
CV-1 cells were purchased from the American Type Culture Collection (ATCC, Rockville, MD). pBluescript SK II and pBC SK II were from Stratagene (La Jolla, CA). pCAT-Basic, pSV 2 -CAT, pSV2-␤-gal, pGL2-Basic, and pGEM-4Z were from Promega Biotech (Madison, WI). PLacF (43) was a gift from Dr. Richard Behringer, M. D. Anderson Cancer Center (Houston, TX). ATRA, 9-cis-RA, TTNPB, and LG100069 were obtained from the Allergan-Ligand Joint Venture in Retinoid Research (San Diego and Irvine, CA). Stock solutions of retinoids were prepared in ethanol and stored in the dark at Ϫ20°C. Oligonucleotides were purchased from Genosys (Houston, TX).

Enzyme Assays
Chloramphenicol Acetyltransferase-Chloramphenicol acetyltransferase (CAT) activity was assayed by the method of Gorman et al. (44). Cell extracts containing equal amounts of protein in 90 l were incubated with 0.25 Ci of [ 14 C]chloramphenicol, 35 l of Tris-HCl (pH 7.5), and 20 l of 4 mM acetyl-CoA for 2 h at 37°C. The reaction was terminated by extracting chloramphenicol with 1 ml of ethyl acetate. The acetylated products were removed, dried under vacuum, separated by thin layer chromatography (Whatman silica gel) in chloroform:methanol (95:5), and detected by autoradiography. Regions of the chromatogram containing acetylated chloramphenicol were cut out, and radioactivity was determined by liquid scintillation counting. Assays for testing the deletion constructs derived from pmTG1.8-CAT were performed by using a CAT-enzyme-linked immunosorbent assay kit (Boehringer Mannheim), following the manufacturer's suggested protocol.
␤-Galactosidase-␤-galactosidase activity in transfected cells was assayed using the Promega ␤-Galactosidase Enzyme Assay System. Cell extracts (100 l) were incubated with 120 mM Na 2 HPO 4 , 80 mM NaH 2 PO 4 , 2 mM MgCl 2 , 100 mM ␤-mercaptoethanol, 1.33 mg/ml O-nitrophenyl-␤-D-galactopyranoside for 2 h. The reaction was terminated by the addition of 300 l of 1 M sodium carbonate, and the absorbance of the reaction was measured at 420 nm in a spectrophotometer.

Transient Transfections
CV-1 Cells-Co-transfection assays were performed in CV-1 cells using a 12-well dish format (45). CV-1 cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, L-glutamine, and gentamycin sulfate. Cells were allowed to attach to the plate for at least 4 h before transfection. Constitutive RSV promoter-based expression vectors for retinoid receptors (hRAR␣, hRAR␤, hRAR␥, hRXR␣, mRXR␤, and mRXR␥) were introduced into CV-1 cells with luciferase-based reporter constructs as indicated in the figure legends. The quantity of plasmids per well was as follows: 0.1 g of receptor expression vector, 0.5 g of reporter construct, 0.5 g of RSV-␤-gal, and 0.8 g of carrier DNA (pGEM-4Z). Following overnight transfection with calcium phosphate, retinoids were added to the cells for 20 h, and then cells were harvested and lysates were prepared (46,47). The lysates were assayed for luciferase and ␤-galactosidase activity, and luciferase activity was normalized for transfection efficiency (␤-galactosidase activity). Values calculated are the mean of determinations from triplicate wells.
RAR␥-3T3 Cells-Balb/C-3T3 cells stably transfected with a human RAR␥ expression vector 3 (RAR␥-3T3) were plated at a density of 5 ϫ 10 5 cells/60-mm plate 24 h prior to transfection. At the time of transfection cells were washed twice with Dulbecco's modified Eagle's medium, and 30 g of Lipofectin reagent (Life Technologies, Inc.) was used to transfect 2.5 g of test plasmid along with 1.5 g of pSV2-␤-gal (Promega). Cells were transfected for 24 h in Opti-MEM (Life Technologies, Inc.). The medium was then replaced with Dulbecco's modified Eagle's medium and 10% fetal bovine serum with or without 1.0 M ATRA and cultured for another 24 h. Cells were then washed with phosphatebuffered normal saline, scraped into 250 l of 0.25 M Tris-HCl (pH 7.5), and lysed by three rounds of freeze-thawing. The extracts were centrifuged, and the supernatants were assayed for CAT and ␤-galactosidase activity. Protein concentration was determined by the Bio-Rad protein assay (Bio-Rad Chemical Division, Richmond, CA). and annealed to form double-stranded DNA. Double-stranded oligonucleotides (1 g) were radiolabeled in a mixture containing 5 l of 10X-T4 kinase buffer, 5 l of [␥-32 P]ATP (10 Ci/l), and 2.5 l of T 4 polynucleotide kinase (10 units/l). Specific activity was 1.5-3 ϫ 10 7 cpm/g DNA. Electrophoretic mobility shift assay was carried out by incubating 1 g of cellular extract with 30,000 cpm of end-labeled double-stranded oligonucleotide in the presence or absence of unlabeled competitor in the indicated molar excess, 1 l of poly(dI-dC) (5 g/l stock), and 10 l of 100 mM Tris, pH 7.9, 25 mM MgCl 2 , 40% glycerol, 2 mM dithiothreitol, 50 mM KCl. This mixture was incubated for 10 min at room temperature and electrophoresed on a 5% polyacrylamide gel. The gel was dried and exposed to film.

Plasmid Constructs
pmTG0.2-CAT-A 5.8-kb SacI fragment (pmTG5.8) of a mouse tissue transglutaminase genomic DNA -phage clone (mTG-GC3) was cloned into pBluescript SK II. 2 pmTG0.2-CAT was synthesized by polymerase chain reaction amplifying a 289-bp fragment (Ϫ256 to ϩ33) of pmTG5.8, using as primer pairs oligonucleotides mTG-267s (and mTGϩ16a). 2 The amplified polymerase chain reaction fragment was cloned into the XbaI and the HindIII sites of pCAT-Basic. The resulting chimeric gene contains the proximal 0.25 kb of the transglutaminase promoter fused to the CAT structural gene.
pmTG1.8-CAT-pBCmTG2.7 was constructed by cloning a 2.7-kb SacI fragment of pmTG5.8 (from the SacI site in the cloning vector to ϩ93 of the transglutaminase gene) into pBC SK II. A 616-bp fragment of the 5Ј end of the insert in pBCmTG2.7 was deleted by PstI digestion and religation to generate pBCmTG2.1. pmTG1.8-CAT was constructed by cloning an EagI-HindIII fragment of pBCmTG2.1 (from the HindIII site in the cloning cassette to Ϫ42 of the transglutaminase insert) into the EagI-HindIII sites of pmTG0.2-CAT.
pmTGRRE1-TK-Luc and pmTGRRE1-MUTA1-TK-Luc-Sense and antisense strand oligonucleotides of mTGRRE1 and mTGRRE1-MUTA1 (see Fig. 7A) were synthesized, annealed, and cloned into the BamHI site of the "enhancer trap" vector pTK-Luc. Insert orientation and multiplicity were confirmed by sequence analysis.

Isolation of the Mouse Tissue Transglutaminase Promoter
Previous studies from our laboratory 2 have resulted in the isolation of the 5Ј end of the mouse tissue transglutaminase gene and its promoter (Fig. 1A). A 3.8-kb fragment of the promoter that included 66 nucleotides of 5Ј-untranslated cDNA and approximately 3.8 kb of 5Ј-flanking DNA was cloned upstream of a ␤-galactosidase (LacF) reporter gene in pUC18 (Fig.  1A). This transglutaminase promoter/reporter gene construct (pmTG3.8-LacF) was transiently transfected into HeLa cells, and the retinoid responsiveness of the reporter gene was assayed by measuring the induction of ␤-galactosidase activity in the presence of 1 M ATRA (Fig. 1B). ATRA induced a 10-fold increase in ␤-galactosidase activity in cells transfected with pmTG3.8-LacF, whereas it had no effect on the activity of cells transfected with the control vector, pBasic-LacF.

Co-transfection Assays
Receptor Specificity-The preceding studies demonstrated that the endogenous retinoid receptors in HeLa cells were capable of mediating ligand-dependent activation of the mouse tissue transglutaminase promoter. A limitation of these experiments was that HeLa cells contain multiple retinoid receptors (RAR␣, RAR␥, RXR␤), 2 and the ligand used in these studies (ATRA) can activate both classes of receptors (it binds to RARs directly and can be isomerized in vivo to 9-cis-RA, which binds to RXRs). Thus it is difficult using ATRA to identify the retinoid receptors directly involved in the activation of the transglutaminase promoter. To address this issue we co-transfected tissue transglutaminase promoter/reporter constructs and retinoid receptor expression vectors into CV-1 cells. CV-1 cells were selected because the level of endogenous retinoid receptors is too low to support retinoid-dependent transcription of transglutaminase promoter/reporter constructs (data not shown). In initial experiments, cells were co-transfected with pmTG3.8-Luc (containing 3.8 kb of the mouse transglutaminase promoter linked to a luciferase reporter gene) and expression vectors for RAR␥, RXR␣, or both. The increase of luciferase activity in response to 9-cis-RA, an activator of both RARs and RXRs, was then used to measure the retinoid-dependent induction of transglutaminase promoter activity (Fig. 2). Co-transfection of the reporter construct and RAR␥ and RXR␣ expression vectors resulted in a 5-fold increase in 9-cis-RA-dependent luciferase activity. Co-transfection of the reporter construct and an RXR␣ expression vector alone resulted in 9-cis-RA-dependent transcriptional activity that was significantly greater than control (2.9-fold, n ϭ 5, p Ͻ 0.001) but was less than the induction in cells transfected with the combination of RAR and RXR receptors. Co-transfection of the cells with the reporter construct and an RAR␥ expression vector alone resulted in no significant retinoid-dependent promoter activity.
Both RAR and RXR receptors are encoded by multigene families (23). To assess the role of the individual RAR and RXR subtypes in retinoid-dependent transcription of the tissue transglutaminase gene, CV-1 cells were co-transfected with the transglutaminase promoter/reporter construct and combinations of receptor expression vectors ( Table I). None of the RAR receptors alone induced significant retinoid-dependent promoter activation (1.6, 1.1, and 1.8-fold activations for RAR␣, -␤, and -␥, respectively). RXR␣ and RXR␥ increased promoter activity 2.9-and 2.8-fold, respectively, whereas RXR␤ was inactive (1.2-fold). Cells transfected with RXRs plus RARs had higher levels of promoter activation than cells transfected with RXRs alone.The lack of retinoid-dependent activity of the transglutaminase promoter in RXR␤-transfected cells was promoter-specific since co-transfection of the same cells with the RXR␤ expression vector and another RXR-responsive reporter construct, RXRE-TK-luciferase, showed 9-cis-RA-dependent transcriptional activation (data not shown).
Ligand Specificity-9-cis-RA is a bifunctional retinoid that activates both RARs and RXRs (33,34,47). Thus, results obtained in the preceding studies did not distinguish whether the activation of the transglutaminase promoter was due to ligand activation of the RXR or the RAR component of receptor heterodimers. To address this issue we compared the luciferaseinducing activity of 9-cis-RA with that of two receptor-selective retinoids, TTNPB, an RAR-selective agonist (Table II and Refs.  49 and 50) and LGD1069, an RXR-selective compound (Table II and Refs. 51 and 52), in cells co-transfected with the pmTG3.8-Luc and RAR␥ and RXR␣ expression vectors (Fig. 3). TTNPB caused a dose-dependent and saturable increase in promoter activity. The EC 50 for this effect, 1 nM, is lower than the IC 50 for the binding of TTNPB to RAR␥, 26 nM (Table II). LGD1069 also induced a dose-dependent increase in promoter activity; its EC 50 , 20 nM, was similar to its IC 50 for binding to RXR␣, 32 nM (Table II). Although the LGD1069 dose-response curve did not saturate within the experimentally tolerable range, it is evident from the data in Fig. 3 that its maximal efficacy in activating the transglutaminase promoter is comparable with or greater than that of TTNPB. 9-cis-RA is more active than either TTNPB or LGD1069 in activating the tissue transglutaminase promoter. The EC 50 for this effect, 20 nM, is close to its IC 50 for binding to both RAR␥ and RXR ␣, 17 and 32 nM, respectively (Table II). These results suggest that ligand activation of either the RAR or the RXR components of RAR⅐RXR heterodimers can increase the transcriptional activity of the tissue transglutaminase promoter.   Fig. 1. Transfected cells were treated with 9-cis RA or control (Ͻ0.1% ethanol) media for 20 h, and luciferase activity was assayed. The 9-cis RA-dependent increase in promoter activity was calculated as the ratio of the normalized luciferase activity in cells treated with retinoid to that in cells treated with control media. Values shown are the mean Ϯ S.D. of triplicate determinations. b ND, not determined.

Deletion Mapping of the Promoter
In our initial characterization of the mouse tissue transglutaminase promoter we identified a core promoter region (Ϫ256 to Ϫ1) with high basal transcriptional activity 3 and a 3.8-kb fragment that was retinoid-inducible (Fig. 1B). To identify regions of this promoter responsible for its regulation by retinoids, we generated restriction fragments of the 3.8-kb promoter construct, linked them to a CAT reporter gene and evaluated their retinoid-dependent activity in an RAR␥-transfected 3T3 cell line (RAR␥-3T3). We first deleted the most distal 2 kb of the 3.8-kb promoter (from approximately Ϫ3800 to Ϫ1831) and tested the ability of the residual 1.8-kb fragment (pmTG1.8-CAT) to respond to ATRA (Fig. 4). The basal activity of pmTG1.8-CAT was similar to the promoterless control vector, pCAT-Basic. ATRA increased the activity of pmTG1.8-CAT 7-fold, a level of induction comparable with the induction of RARE-TK-CAT in these cells (4.5-fold).
To locate the sites of retinoid regulation within pmTG1.8-CAT we made several deletions in the reporter construct (Fig. 5A). Removal of a 1-kb segment of the promoter immediately upstream of the core promoter region (pmTG1.8-⌬SmaI-CAT) did not alter its retinoid-dependent activity. The removal of an additional 439 bp (pmTG1.8-⌬SmaI⌬SphI-CAT) also had no effect on the retinoid inducibility of the construct. This residual retinoid-responsive reporter construct contained only three small fragments of the 3.8-kb tissue transglutaminase promoter, the 60-bp core promoter, a 113-bp SphI-SmaI fragment, and a 125-bp PstI-SphI fragment. Removal of either of the two fragments upstream of the core promoter resulted in a complete loss of retinoid-inducible activity. These observations suggested that retinoid-dependent activation of the core tissue transglutaminase promoter depended on the cooperation of two small upstream regulatory regions.
Sequence analysis of the two fragments of the tissue transglutaminase promoter necessary for retinoid-dependent activation (Fig. 5B) suggests that they both may have important regulatory functions. The proximal regulatory sequence (SphI-SmaI fragment, Ϫ1207 to Ϫ1094) contains an unusual set of DNA motifs that is highly conserved between the human and mouse tissue transglutaminase promoters (53). 2 This 67-bp segment of DNA (homology region I), which is located far upstream in both the mouse (Ϫ1087 to Ϫ1152) and the human a RARs and RXRs from nuclear extracts of baculovirus-infected Sf21 cells were incubated with [ 3 H]9-cis-RA and varying concentrations of retinoids as described previously (46). The IC 50  (Ϫ1597 to Ϫ1682) promoters, includes 45/67 nucleotides that are identical in the two promoters (Fig. 5B). The function of this segment is unknown, but its conservation between species and its role in the retinoid-dependent activation of the transglutaminase promoter suggests that it may have an important regulatory role.
The distal regulatory element (PstI-SphI) includes three directly repeated hexanucleotide motifs (highlighted in Fig. 5B), separated by 7 and 5 nucleotides (DR7/DR5). Two of these motifs (the A and C motifs in Fig. 5B) conform to the canonical retinoid receptor binding motif, (A/G)G(G/T)TCA), and one (the B motif) differs in the substitution of a C residue for an A in position Ϫ1734. This distal regulatory fragment of the mouse transglutaminase promoter is a potential tissue transglutaminase retinoid response element (mTGRRE1).
To assess the contribution of mTGRRE1 to retinoid regulation of the tissue transglutaminase promoter we deleted it from the 3.8-kb promoter fragment and compared the retinoid-dependent activation of the modified promoter (pmTG3.8-⌬mT-GRRE1-Luc) with that of the intact promoter (pmTG3.8-Luc) (Fig. 6). The intact and modified reporter constructs were co-transfected with RAR␥ and RXR␣ expression vectors into CV-1 cells. In cells treated with either 9-cis-RA or equimolar concentrations of TTNPB and LGD1069, deletion of the DR7/DR5 response element caused a much lower retinoid-dependent induction of luciferase activity than in cells transfected with the intact promoter (Fig. 6A). To determine if the loss in response to 9-cis-RA reflected a selective loss in ligand activation of either RAR or RXR receptors, we also measured the induction of luciferase activity in cells treated with either TTNPB or LGD1069 alone (Fig. 6B). TTNPB alone increased luciferase activity to the same degree regardless of whether the promoter did or did not include the DR7/DR5 element. The same was true for LGD1069. Although the LGD1069 dose-response curve was shifted to the right of the TTNPB curve, it induced a similar modest increase in promoter activity in both the intact and modified reporter constructs. The induction by either TT-NPB alone or LGD1069 alone was similar to the induction by either 9-cis-RA or TTNPB plus LGD1069 in cells transfected with pmTG3.8-⌬mTGRRE1-Luc. This induction was much lower than the maximal induction of pmTG3.8-Luc by 9-cis-RA or TTNPB plus LGD1069 (compare Fig. 6, A and B). These A, retinoid responsiveness of deletion constructs. Fragments of the mouse tissue transglutaminase promoter linked to CAT were generated as described under "Experimental Procedures." The deletion constructs were transiently transfected into RAR␥-3T3 cells, and the ATRA-inducible CAT activity was measured as described in the legend to Fig. 3. Constructs giving a reproducible Ͼ2-fold induction of CAT activity were scored as responsive (ϩ), and those with consistently lower levels of activation were scored as unresponsive (Ϫ). B, nucleotide sequence of segments of the mouse tissue transglutaminase promoter necessary for retinoid-inducible activation. HR-1, homology region I. Shown is a comparison of the nucleotides of mouse (upper sequence, Ϫ1088 to Ϫ1152) and human (lower sequence, Ϫ1597 to Ϫ1662). Identical nucleotides are boxed. mTGRRE1, mouse tissue transglutaminase retinoid response element 1. Shown are both strands of mT-GRRE1 (Ϫ1703 to Ϫ1751); candidate hexanucleotide half-sites are highlighted (identified as A, B, and C); and the orientation and spacing of the half-sites are identified by the arrows and spacers. DR7 is direct repeat with a 7-nucleotide spacer, and DR5 is a direct repeat with a 5-nucleotide spacer. studies suggest that the small activation of the mouse transglutaminase promoter generated by either TTNPB or LGD1069 alone does not depend on the presence of the DR7/DR5 response element, whereas the significantly larger stimulation of activity generated by either 9-cis-RA or TTNPB plus LGD1069 does.

Mobility Shift Assays: Receptor and Response Element Specificity
The preceding studies suggested that the 30-bp segment of the transglutaminase promoter containing the three hexanucleotide motifs (DR7/DR5 element) was a possible site through which retinoid receptors might exert transcriptional control. To determine if retinoid receptors could be bound by this region, we prepared a radiolabeled synthetic doublestranded oligonucleotide (mTGRRE1, Fig. 8A) and measured the effect of equivalent amounts of RAR␥ and RXR␣ to alter its electrophoretic mobility (electrophoretic mobility shift assay, Fig. 7A). Lane 1 contains radiolabeled probe alone (open arrow). Addition of RAR␥and RXR␣-containing extracts (lane 2) resulted in the formation of a shifted band, indicating that the receptors were bound by the radiolabeled nucleotide. This binding was saturable, since a 100-fold molar excess of unlabeled mTGRRE1 blocked its formation (lane 4) whereas an equivalent excess of a nonspecific oligonucleotide did not (lane 3). To determine the contribution of the individual receptors to the band-shifted complex, we compared the ability of RAR␣ alone and varying ratios of RAR␣ and RXR␣ to produce the same band shift (Fig. 7B). RAR␣ alone produced no evidence of a band shift (lane 2); however, the combination of RXR␣ and RAR␣ resulted in a concentration-dependent band shift (lanes [3][4][5]. The addition of RXR␣ alone (Fig. 7C) resulted in the formation of a complex (lane 2) that was competed for by an excess of mTGRRE1 (lane 4) and not by a nonspecific oligonucleotide (lane 3). These studies demonstrate that both RAR⅐RXR heterodimeric complexes and RXR⅐RXR homodimers can bind to mTGRRE1.
Heterodimers of RAR and RXR bind with high affinity to directly repeated consensus hexanucleotide motifs separated by one, two, or five nucleotides (39). To compare the affinity of RAR⅐RXR heterodimers for mTGRRE1 with their affinity for either DR1 or DR5 consensus RREs (Fig. 8A), we measured the competitive inhibition of radiolabeled mTGRRE1 binding to RAR␥⅐RXR␣ by unlabeled oligonucleotides, mTGRRE1, DR1, DR4, and DR5 (Fig. 8B). Both DR1 and DR5 competed equivalently with mTGRRE1 for binding to RAR␥⅐RXR␣, whereas DR4 and the nonspecific oligonucleotide did not compete.
The oligonucleotide mTGRRE1 is comprised of three hexanucleotide half-sites (A, B, and C) separated by seven and five nucleotides (Fig. 8A). The consensus retinoid receptor binding motifs GGGTCA (sites A and C) and the variant motif AG-GTCC (site B) are located on the "antisense" DNA strand, such that A (GGGTCA) is 7 bp upstream of B (AGGTCC), which is in turn 5 bp upstream of C (GGGTCA). To assess the contribution of each half-site to receptor binding we synthesized oligonucleotides in which nucleotides within the A, B, or C half-sites were rearranged (Fig. 8A). We then compared the ability of these mutant oligonucleotides to compete with mTGRRE1 for binding to RAR␥⅐RXR␣ (Fig. 8C). Rearrangement of nucleotides in the A half-site (RRE-MUT A1) gave an oligonucleotide that competed very well for RAR⅐RXR binding to mTGRRE1. The affinity of receptors for this oligonucleotide appears to be greater than their affinity for the mTGRRE1, since the competition curve of the RRE-MUT A1 oligonucleotide is shifted to the left of the competition curve of unlabeled mTGRRE1 (Fig.  8C). Rearrangement of the nucleotides in the C half-site (RRE-MUT C1) gave an oligonucleotide that competes weakly with mTGRRE1 for receptor binding. A point mutation in the B half-site was made by replacing the C residue in the position corresponding to nucleotide Ϫ1734 with an A. This mutation converts the non-canonical hexanucleotide motif AGGTCC to a canonical retinoid receptor binding motif AGGTCA (Fig. 8A). This mutant oligonucleotide (RRE-MUT B1) was a very poor competitor of receptor binding to mTGRRE1. These studies suggested that perturbations in the B and C hexanucleotide motifs disrupted RAR␥⅐RXR␣ binding to the mTGRRE1, whereas perturbations in the A motif had no deleterious effect.
To confirm the conclusions reached from the competitive binding assays we also measured the ability of the RAR␥⅐RXR␣ receptor combination to bind directly to the oligonucleotides containing modifications in the A, B, or C half-sites (Fig. 9). The receptors bound very well to the RRE-MUT A1 oligonucleotide, producing a shifted band with the same electrophoretic mobility as the complex formed on the mTGRRE1 (Fig.  9, lane 5 versus lane 2). The increased intensity of the shifted band in the sample containing the RRE-MUT A1 oligonucleotide supports the conclusion from the competitive binding studies that the RAR␥⅐RXR␣ receptor complexes have a higher affinity for this oligonucleotide than they do for mTGRRE1. The same band-shifted complex is not seen when receptors are combined with either the RRE-MUT C1 or RRE-MUT B1 oligonucleotides (lanes 8 and 11). Some low affinity complexes with anomalous mobility are formed on these oligonucleotides, but the single RAR⅐RXR-dependent complex formed on the mTGRRE1 and RRE-MUT A1 oligonucleotides is not detected. These studies support the conclusion of the competitive binding studies that RAR⅐RXR complexes preferentially bind the oligonucleotides in which the B and C sites are intact. It appears that replacement of a single nucleotide in the B hexanucleotide half-site is sufficient to disrupt receptor binding to mTGRRE1.
Transactivation Assays: Response Element Specificity-To test whether mTGRRE1 had the properties of a ligand-dependent enhancer element, we inserted two copies of the element upstream of a heterologous promoter, the thymidine kinase (TK) promoter, linked to a luciferase reporter gene (pmTG-RRE1 2 TK-Luc). Co-transfection of pmTG-RRE1 2 TK-Luc into CV-1 cells with expression vectors for RAR␥ and RXR␣ resulted in a 4-fold 9-cis-RA-dependent stimulation of transcriptional activity. (Table III). This construct was also retinoid-inducible in cells co-transfected with the reporter and RXR␣ alone (4.3fold induction). Insertion of a single mTGRRE1 element upstream of the TK-Luc reporter in either the wild type (pmT-GRRE1-TK-LUC) or inverted orientation (pmTGRRE R -TK-Luc) gave a modest degree (1.5-2.5-fold) of stimulation in cells transfected with either the RAR␥⅐RXR␣ receptor combination or RXR␣ alone (Table III).
Studies of receptor binding to the native and mutant mTGRRE1s demonstrated that disruption of the A hexanucleotide half-site increased receptor binding. To determine if this increased affinity for receptors improved the transcriptional enhancer properties of the oligonucleotide, we measured its transcriptional activity in the CV-1 cell co-transfection assay (Table III). The mutant enhancer element (mTGRRE1-MUTA1-TK-Luc) showed no retinoid-dependent transcriptional activation in cells transfected with either RAR␥⅐RXR␣ or RXR␣ alone. Thus, although this mutated oligonucleotide binds receptors better than mTGRRE1, it is much less efficient in its ligand-dependent transcriptional enhancer activity.

Regulatory
Regions of the Promoter-The goal of these studies has been to extend our investigations of the molecular mechanisms involved in the induction of the enzyme tissue transglutaminase. Previous studies from our laboratory have identified retinoic acid as an important regulator of tissue transglutaminase gene expression (16, 54 -56). 2 This regulation is at the transcriptional level and was attributable to sequences embedded within the DNA flanking the 5Ј-end of the tissue transglutaminase gene (57). 2 We have shown that in HeLa cells the expression of a reporter gene linked to a large (3.8-kb) fragment of the tissue transglutaminase promoter was optimally induced by retinoid receptor panagonists. Furthermore both RAR-specific and RXR-specific retinoids stimulated the promoter's transcriptional activity. This pattern of dual regulation replicated the regulation of the endogenous tissue transglutaminase gene. The current studies were undertaken to refine our understanding of this induction by identifying specific cis-regulatory sequences within the mouse tissue transglutaminase promoter that contributed to retinoid-dependent transcriptional activation.
Our approach has been to reduce the size of the transglutaminase promoter and then to identify residual segments essential for retinoid-dependent activation. Using this approach we have identified two regions of the promoter critical for this activation. One of these, the most upstream, contains a triplicate retinoid receptor binding motif that functions as a liganddependent enhancer element. Although this retinoid-response element (mTGRRE1) can activate a heterologous promoter such as the thymidine kinase promoter, it cannot by itself confer ligand-dependent transcriptional activity on the mouse tissue transglutaminase promoter. For this activity it must be coupled with a second short DNA segment (HR-1) that is located 1 kb upstream from the transcription start site. This segment of DNA is unusual because it is one of the very few segments of DNA that is highly conserved between the human and the mouse tissue transglutaminase promoters (53). 2 Within the 67-bp span of HR-1 the sequence homology is striking; not only are there no gaps in the optimal nucleotide alignment, but homologous sequences are grouped into colinear segments of 5-15 nucleotides. This conservation of sequence appears to reflect important functional properties of the promoter, since combination of this element with the upstream retinoid-responsive motif is permissive for retinoid-dependent transcriptional activation.
The observation that retinoid-dependent promoter activity requires the cooperation of two distinct regulatory elements resembles the situation in the mouse CRABP II promoter (58). In both promoters two distant and distinct cis-regulatory elements are required for efficient ligand dependent transactivation. In the case of the mouse CRABP II promoter, both elements bind retinoid receptors. In the mouse transglutaminase  Fig. 6, A and B.   FIG. 9. Binding of RAR␥⅐RXR␣ to mTGRRE1 and mutated oligonucleotides. Radiolabeled mTGRRE1 (lanes 1-3), MUT A1 (lanes 4 -6), MUT B1 (lanes 7-9) and MUT C1 (lanes 10 -12) were incubated with RAR␥⅐RXR␣ nuclear extracts (1:1 ratio) alone (lanes 2, 5,8,11) or in the presence of a 100-fold molar excess of unlabeled oligonucleotide (Sp; lanes 3, 6, 9, and 12) under conditions described under "Experimental Procedures" and in the legend to Fig. 6. The radiolabeled oligonucleotides alone were run in lanes 1, 4, 7, 10. a Reporter constructs are as follows: mTGRRE1, a single copy of the DR7/DR5 element in its normal orientation; mTGRRE1 2 , two copies of the DR7/DR5 element in their normal orientation; mTGRRE1 R , a single copy of the DR7/DR5 element in the reverse orientation; mTGRRE1-MUT A1, a single copy of the DR7/DR5 element containing a mutation in half-site A as shown in Fig. 7A.

Retinoid-dependent activation of mTGRRE 1 constructs in a TK-Luc enhancer trap vector: effects of configuration and multimerization
b Reporter constructs were transiently cotransfected into CV-1 cells with expression vectors for either RAR␥ and RXR␣ or RXR␣ alone under conditions described under "Experimental Procedures" and in the legend to Fig. 1. Transfected cells were treated with 9-cis-RA or control (Ͻ0.1% ethanol) media for 20 h, and luciferase activity was assayed. The 9-cis-RA-dependent increase in promoter activity was calculated as the ratio of the normalized luciferase activity in cells treated with retinoid to that in cells treated with control media. Values shown are the mean Ϯ S.D. of triplicate determinations. promoter the upstream element binds retinoid receptors; the downstream element does not contain sequences that match a canonical retinoid receptor binding motif, although recent studies suggest that it too may have retinoid receptor binding activity. 4 Detailed characterization of the proteins associated with both the upstream and downstream regulatory elements will be required to explain the cooperation between these two regulatory regions.
Differential Effects of Retinoids on Promoter Activation-Retinoids regulate the expression of many genes (59). In most experimental studies, retinoid-dependent regulation of gene expression has been established using ATRA as a ligand for retinoid receptors. ATRA binds to and can activate RARs; however, under physiological conditions it can isomerize to 9-cis-RA, a ligand for both RARs and RXRs (33,34,47,60). Thus, under physiological conditions it has been difficult to know whether the biological effects of ATRA are due to ligand activation of RAR-dependent processes, RXR-dependent processes, or processes that depend upon activation of both receptors. The development of synthetic retinoids with restricted receptor specificity has facilitated analysis of this issue (41,(61)(62)(63)(64)(65). In many retinoid-dependent biological responses such as the differentiation of F-9 cells (50), the inhibition of squamous differentiation of T epithelial cells (21) the activity of all-trans RA can be replicated by retinoids, such as TTNPB, that bind only to RARs. This specificity is replicated in the regulation of the transcriptional activity of the promoters of many retinoid-responsive genes. Activation of the RAR␤ promoter is optimal with ligands that bind to and activate RARs (58,66). RXR ligands are inactive, and the combination of RAR and RXR ligands is no more active than RAR ligands alone (58,63). RXR-specific retinoids do not even bind to RAR⅐RXR heterodimeric complexes bound to the RAR␤ RARE (52). While some promoters appear to be completely activated by RAR ligands alone, other promoters are synergistically activated by ligands that activate both RARs and RXRs. The mouse CRABP II gene can be induced by RAR-selective ligands, but it is much better induced by ligands that bind to both RAR and RXR receptors (58). The tissue transglutaminase promoter, like the CRABP II promoter, also appears to be versatile in its response to retinoids. Not only can it be activated by RAR-specific or RXR-specific retinoids, but its activity is optimally induced by retinoid receptor panagonists that bind to both receptors.
Activation of the tissue transglutaminase promoter by retinoid receptor panagonists appears to be linked to the activity of a specific retinoid response element (mTGRRE1) that is located 1.7 kb upstream of the transcription start site. In promoter constructs from which this element was excised, RAR-or RXRagonists could induce a slight activation of the promoter (suggesting residual or cryptic retinoid response elements within the promoter) but there was no greater activation of the promoter if the RAR-and RXR-specific retinoids were combined or if the promoter was stimulated with a panagonist such as 9-cis-RA. In promoter constructs in which the mTGRRE1 was intact, RAR-and RXR-specific retinoids induced some activation but it was much less than that achieved either by panagonists or by combinations of receptor-selective retinoids. It ap-pears that mTGRRE1 has the pharmacologic properties of a retinoid receptor panagonist response element.
Configuration of mTGRRE1-mTGRRE1, like other hormone response elements, is a ligand-dependent enhancer element. It functions as an activator of transcription in the context of heterologous promoters such as TK (Table III) and the SV-40 promoter. 5 It is equivalently active in both the sense and antisense orientation, and multimerization increases its transcriptional activity in the context of both the TK (Table III) and SV-40 promoters. 5 The configuration of mTGRRE1 is complex and includes three directly repeated receptor-binding motifs (half-sites). This is a pattern recognizable in a number of response elements from hormone-regulated genes (67)(68)(69)(70). The configuration of the MTGRRE1 is strikingly similar to the thyroid hormone response element (TRE) of the Moloney leukemia virus long terminal repeat promoter (Fig. 10, MoMLV-LTR) (42,72). Both response elements include an upstream canonical halfsite located either 7 or 11 nucleotides upstream from a core receptor binding sequence. The MoMLV-LTR TRE element includes a GGTCA motif separated by four nucleotides from an AGGTCC motif. In mTGRRE1 the same half-sites are present, but their order is reversed. The AGGTCC motif is located five nucleotides upstream of the canonical GGTCA halfsite. The MoMLV-LTR TRE binds thyroid hormone receptors with high affinity and does not bind retinoid receptors (72). The mTGRRE1 core motif binds retinoid receptors (both RAR⅐RXR heterodimers and RXR homodimers), but it does not bind thyroid hormone receptors. 5 In both cases the sequence of the non-canonical half-site plays a critical role in receptor binding. In the context of the MoMLV-LTR-TRE, conversion of the sixth position C into an A (AGGTCC to AGGTCA) increases retinoid receptor binding (42), whereas in the context of mTGRRE1 the same conversion of the same nucleotide abolishes retinoid receptor binding. It appears that both the spacing and nucleotide sequence of the receptor binding half-sites dictate receptor specificity in these naturally occurring hormone response elements.
The multiplicity of half-sites in mTGRRE1 is a common feature of naturally occurring hormone response elements (23). Furthermore, there is evidence that cooperation between these half-sites contributes to the transactivation process. The TRE of the rat growth hormone promoter contains three half-sites. Mutation of any of these half-sites abolishes the transcriptional activity of the entire element (70). Similarly the half-sites of the rat CRBPI RXRE cooperate (69). mTGRRE1 appears also to conform to this type of regulation. Even though the distal two half-sites (B and C) are sufficient for retinoid receptor binding, the intact tripartite element is required for efficient transcriptional activity. We do not know the molecular basis for the cooperation between the half-sites in mTGRRE1. The upstream A motif could contribute an additional site for retinoid receptor binding in vivo or it could be a site of binding for one or more additional transcription activating factors.
Conclusions-Analysis of the mouse tissue transglutaminase promoter suggests that its activation by retinoids depends upon ligand activation of RAR⅐RXR heterodimeric complexes associated with a complex retinoid response element located 1.7 kb upstream of the transcription start site (Fig. 11). It appears that ligand activation of this complex is not sufficient to fully activate the promoter; an additional cooperating segment of DNA (HR-1) is required for full activation of the homologous promoter. Ligand-dependent stimulation of this complex transcriptional assembly is required for initiating the induction of tissue transglutaminase expression in cells. This model raises interesting possibilities for explaining the differential regulation of tissue transglutaminase expression in connective tissue cells and in cells undergoing programmed cell death.