β-Trace Gene Expression Is Regulated by a Core Promoter and a Distal Thyroid Hormone Response Element*

We isolated and characterized the human β-Trace protein (βTP) gene promoter. βTP, also known as prostaglandin D2 synthase, is a lipocalin secreted from the choroid plexus and meninges into cerebrospinal fluid. Basal transcription of the βTP gene is directed from a core promoter found within the first 325 bases of the 5′-flanking sequence. The βTP gene promoter is responsive to thyroid hormone (3,3′,5-triiodothyronine, T3) and efficiently repressed by unliganded human thyroid hormone receptor β (TRβ). Functional analysis of the βTP promoter in TE671 cells revealed that responsiveness to T3 occurs in sequences 2.5 kilobase pairs 5′ of the start site. Within the hormone-responsive region we identified a thyroid hormone response element (TRE) located from −2576 to −2562 base pairs relative to the transcription start site. The βTP TRE is composed of two directly repeated consensus half-sites separated by a 3-base pair space (DR3). The βTP TRE forms specific complexes with TRβ. We have shown that a gene active in the choroid plexus and meninges is responsive to T3. T3 may play a role in the regulated transport of substances into the cerebrospinal fluid and ultimately the brain.

We isolated and characterized the human ␤-Trace protein (␤TP) gene promoter. ␤TP, also known as prostaglandin D 2 synthase, is a lipocalin secreted from the choroid plexus and meninges into cerebrospinal fluid. Basal transcription of the ␤TP gene is directed from a core promoter found within the first 325 bases of the 5-flanking sequence. The ␤TP gene promoter is responsive to thyroid hormone (3,3,5-triiodothyronine, T 3 ) and efficiently repressed by unliganded human thyroid hormone receptor ␤ (TR␤). Functional analysis of the ␤TP promoter in TE671 cells revealed that responsiveness to T 3 occurs in sequences 2.5 kilobase pairs 5 of the start site. Within the hormone-responsive region we identified a thyroid hormone response element (TRE) located from ؊2576 to ؊2562 base pairs relative to the transcription start site. The ␤TP TRE is composed of two directly repeated consensus half-sites separated by a 3-base pair space (DR3). The ␤TP TRE forms specific complexes with TR␤. We have shown that a gene active in the choroid plexus and meninges is responsive to T 3 . T 3 may play a role in the regulated transport of substances into the cerebrospinal fluid and ultimately the brain.
␤-Trace protein (␤TP) 1 is a component of human cerebrospinal fluid (CSF) and one of very few proteins found in CSF not also present in serum. In human CSF, ␤TP is present at 2.6 mg/dl, ranking it among the major CSF proteins (1). ␤TP, identified by Clausen in 1961 (2), is primarily expressed in the choroid plexus (CP). ␤TP is also expressed to a lesser extent in meninges and oligodendrocytes (3,4). Other than the CNS, the major site of ␤TP expression is the epididymis (4,5).
A protein with similar distribution to ␤TP has been identified as prostaglandin D 2 synthase (PDS) in rats (6,7). PDS catalyzes the conversion of prostaglandin H 2 to prostaglandin D 2 (PGD 2 ). A role for PGD 2 in regulation of sleep induction has been proposed (8,9). Recently, ␤TP and PDS were shown to be the same protein (10,11). In prior studies we have referred to ␤TP/PDS as PDS but, in deference to precedence, we now refer to it as ␤TP (12).
The human ␤TP message encodes a 180-residue polypeptide that is a member of the lipocalin superfamily. Lipocalins are secretory proteins that transport hydrophobic ligands (13,14). Lipocalin genes appear to have arisen by gene duplication, with most of them clustered in the q34 region of chromosome 9 in man and in the syntenic b-c region of chromosome 4 in the mouse (15). In previous work we localized the human ␤TP gene to the lipocalin gene cluster on 9q34. The ␤TP gene bears a striking resemblance to other lipocalin genes, suggesting a role for ␤TP in transport (12). CSF, primarily produced by the CP, can be viewed as an ultra filtrate of serum with protein levels approximately 0.5% those in serum. Exchange of proteins and other substances between CSF and the extracellular fluid of the brain is free (16). The CP secretes highly specialized transporters that carry essential substances into the CSF and then to the brain. The primary function of the meninges is the maintenance of the blood-CSF barrier, but it also contributes to CSF and many substances enter into CSF equally well from either the meninges or CP. Cultured meningeal cells secrete many of the same transport proteins as the CP (17). Several CSF transporters have been characterized including transthyretin, transferrin, and ceruloplasmin; they carry thyroxine, iron, and copper, respectively (18,19).
García-Ferná ndez et al. (20) found that levels of ␤TP mRNA in the CNS of adult rats decrease following chemically induced hypothyroidism. The mechanism by which thyroid hormone (T 3 ) influences ␤TP gene expression is unknown. T 3 exerts its effects through binding to thyroid hormone receptors (TR), which are widely distributed in the CNS (21). In the CP, T 3 augments transport function; hypothyroid rats have reduced Na ϩ -K ϩ -ATPase activity, a marker for transport processes (22).
To better understand mechanisms of ␤TP gene regulation, we subcloned the human ␤TP gene promoter and analyzed its expression in the human rhabdomyosarcoma cell line TE671. We identify a small core promoter that directs basal gene transcription at high levels and a distal element that determines T 3 responsiveness.

MATERIALS AND METHODS
Isolation and Sequencing of the ␤TP Promoter-The 3.8-kb XhoI-XbaI fragment from the ␤TP genomic clone pG4CS86 (12) was inserted * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
into the SmaI site of pBSKSϩ, and approximately 1 kb of 3Ј sequence was excised using the Exo III/mung bean nuclease system (Stratagene). The resulting fragment, spanning from Ϫ2759 to ϩ65 bp, was subcloned into the CAT vector pJFCAT1 (23) to generate clone pCAT2759 (Fig. 1). Exo-and endonuclease deletions of pCAT2759 produced clones with successive 5Ј deletions. Sequence was analyzed as described previously (12).
␤TP-Thymidine Kinase (TK) Promoter Fusions-Clone pCAT235 was produced by subcloning the region between Ϫ2759 and Ϫ2080 bp of the ␤TP promoter into pBSKSϩ. Small internal deletions were introduced into pCAT235 by digestion with StyI and EcoNI followed by Klenow fill-in or mung bean nuclease digestion to remove one or both of the ␤TP TRE half-sites, respectively. To liberate the inserts from the pBSKSϩ vector, the constructs were opened with BamHI, made blunt-ended with Klenow, and subsequently digested with SalI. Gel-purified fragments were subcloned into the pBLCAT2 vector that had been opened first at the HindIII site and made blunt-ended with Klenow enzyme followed by digestion with SalI. The three clones thus produced were as follows: 1) pTK680F, with the 680-bp fragment of pCAT235 in the forward orientation, 2) pTK⌬3Ј with a 104-base internal deletion which removes the 3Ј half-site of the TRE, and 3) pTK⌬5Ј ϩ 3Ј, with a 108-base internal deletion which removes both half-sites of the TRE.
Clone pTK300, spanning bases basesϪ2759 to Ϫ2464 of the ␤TP promoter, was produced by deleting 385 bp of 3Ј sequence from clone pTK680F by double digestion with EcoNI and SalI. Clone pTK680R was produced by subcloning the 680-bp BamHI-SalI fragment of pCAT235 into the corresponding sites of pBLCAT2. Clone pTK⌬DR3 was produced from overlapping oligonucleotides cloned into the HindIII-XhoI sites of pBLCAT2. Clone pTK100 was produced by PCR amplification of bases Ϫ2620 to Ϫ2518 bp of the ␤TP promoter using pCAT2759 as template and oligonucleotides that introduced a 5Ј HindIII site and a 3Ј XhoI site. PCR product was digested with HindIII and XhoI and subcloned into the corresponding sites in pBLCAT2.
Northern Blot Analysis-Total RNA was isolated from adult rat brain or TE671 cells using the method of Chomczynski et al. (24). Total RNA was electrophoresed through 1% agarose gels containing 3% formaldehyde, capillary blotted onto a GeneScreen nylon membrane (Du-Pont NEN), and probed as described previously (12).
On the day preceding transfection, 4 ϫ 10 5 cells were seeded into 60-mm culture dishes. Plasmid DNA was transfected into cells using the calcium phosphate co-precipitation method (27). For each dish, 1 pmol of CAT construct was co-transfected with 2 g of ␤-galactosidase (␤GAL) expression vector pRSV-␤GAL (28) as an internal control. For TR cotransfections 2 g of hTR␤1 expression vector was used (29). pBSKSϩ was added to bring the total DNA in each dish to 12 g. Medium was replaced 18 h after transfection. Where necessary, T 3 or hormone vehicle were introduced into the fresh medium at a final concentration of 100 nM. After 48 h the cells were harvested. pSV 2 CAT (30) was used throughout as a positive control vector for CAT expression, and pTK83 was used for T 3 /TR responses (31). The negative control for CAT expression was the promoterless CAT vector pJFCAT1 and for T 3 responses, pBLCAT2, which contains the TK promoter (32).
Reporter Gene Assays-To correct for variations in transfection efficiency, cell extracts were assayed for ␤GAL activity (33). After adjusting for ␤GAL levels, CAT activity was determined using a variation of the diffusion assay (34). All transfections were repeated at least four times. Data, reported as mean Ϯ S.E., except where noted, are from three separate transfections.

RESULTS
Isolation and Identification of the ␤TP Core Promoter-The ␤TP promoter was isolated from the genomic clone pG4CS86 FIG. 1. Deletion analysis of the ␤TP promoter. Schematic on the left represents the promoter deletion constructs fused to the CAT gene. Stippling denotes the first 65 bases of the ␤TP 5Ј-untranslated region fused to the CAT gene. The start site of each construct relative to the start site of transcription is shown to the left. pCAT495R has the first 500 bp of the ␤TP promoter cloned in the reverse orientation. Restriction sites used to generate clones are marked with down arrows. Bar graphs on the right represent the CAT activity in TE671 cells of the clones shown schematically on the left. Data are expressed as cpm/min/unit ␤-GAL activity and are reported as the mean Ϯ S.E. from three separate transfections. The CAT activity of all clones was significantly different from the fulllength clone, pCAT2759 (p Ͻ 0.05). (12). To localize regions of the promoter important to ␤TP transcription, a set of 10 promoter-CAT gene fusion constructs with increasing 5Ј deletions was produced, the 5Ј termini ranging from Ϫ2759 to ϩ 16 bp in the untranslated region (Fig. 1). The human rhabdomyosarcoma cell line TE671 (25,26) expresses ␤TP mRNA at high levels and transfects efficiently (Fig. 2). Parallel transfections of the 10 deletion constructs into TE671 cells revealed the ␤TP gene promoter to be highly active, generating CAT activity at a level comparable to the positive control vector pSV 2 CAT (30). The ␤TP core promoter region is small, deletions from Ϫ2759 bp to Ϫ595 bp had minimal effects on ␤TP promoter activity. Deleting the bases between Ϫ595 and Ϫ325 actually increased ␤TP promoter activity 1.8-fold. Deletions within the 325-bp core promoter region results in major loss of activity (Fig. 1). The Ϫ80-bp clone is inactive, which localizes the sequences necessary for maximal basal activation of the ␤TP gene between Ϫ325 and Ϫ80 bp of the promoter.
Nucleotide Sequence of the ␤TP Gene Promoter-The sequence of the core promoter is presented in Fig. 3. The region from Ϫ227 to Ϫ180 bp of the ␤TP promoter has high sequence identity with regions of the human luteinizing hormone subunit ␤ promoter (LH-␤) (68% identity, bases Ϫ239 to Ϫ192) (38) and the human insulin-like growth factor II P4 promoter (80% identity, bases Ϫ318 to Ϫ287) (IGF-II) (39). The IGF-II P4 promoter is active in the CP (40). A 20-bp near-perfect palindrome (PAL I, bases Ϫ176 to Ϫ157) bears extended homology to the AP4 site originally identified in the SV40 enhancer (41) and to the cAMP response element, ENKCRE-2, found within the proenkephalin gene promoter (42). However, the ␤TP promoter is only mildly responsive to forskolin (data not shown).
The ␤TP Gene Promoter Is Responsive to Thyroid Hormone (T 3

) and Human Thyroid Hormone Receptor ␤ (TR␤)-In vivo
analysis has shown that ␤TP mRNA expression is regulated by T 3 . To determine if the human promoter mounted transcriptional responses to T 3 and TR␤, we studied T 3 effects on the ␤TP promoter in TE671 cells. Results from reporter constructs (Fig. 4, top) and Western blot analysis (data not shown) indicate that TE671 cells do not express thyroid hormone receptors. Thus to determine if the human ␤TP gene responds to T 3 and TR␤, the full-length clone, pCAT2759, was cotransfected with a TR␤ expression vector (29) into TE671 cells and cultured in the presence or absence of 100 nM T 3 . The results reveal that the ␤TP promoter is strongly regulated by TR␤ in a T 3 -dependent manner (Fig. 4). ␤TP transcription is elevated 4-fold over basal levels in the presence of T 3 and TR␤. Unliganded TR␤ (no T 3 ) represses the activity of the ␤TP promoter 12-fold compared with basal levels. When both effects are considered, ␤TP promoter activity is stimulated 45-fold by T 3 over the level observed with unliganded TR␤ alone. Fig. 4 also shows that the response of pCAT2759 to T 3 and TR␤ is in the range observed with the strong TRE of the rME-positive control, pTK83 (31). Varying the amount of TR␤ expression vector cotransfected with CAT constructs did not alter the result (data not shown). As shown in Fig. 4, two heterologous promoter-CAT constructs, pSV 2 CAT and pBLCAT2, have responses to TR␤ and T 3 different from those observed for the ␤TP promoter, indicating that ␤TP promoter responses do not result from effects on cell viability or transcriptional competence.
T 3 -responsive Region of the ␤TP Gene Promoter-To identify the T 3 -responsive region of the ␤TP promoter, the deletion constructs ( Fig. 1) were cotransfected with a TR␤ expression vector or pBSKS sham control and cultured with 100 nM T 3 . Only the full-length clone, pCAT2759, shows activation by T 3 and TR␤, localizing the T 3 responsive region to the sequence between Ϫ2759 and Ϫ2018 bp (Fig. 5A). There was no significant activation by T 3 alone (pBSKS sham) over basal levels for any of the deletion constructs examined (data not shown). To account for the repressive effects from unliganded TR␤, the activity of the deletion constructs in the presence of T 3 and TR␤ was compared with that from TR␤ alone. The results, shown in Fig. 5B, again demonstrate that only the Ϫ2759-bp clone possesses major responses to T 3 .
Similar experiments were performed to identify the region responsible for TR␤-mediated repression. Deletion of the region responsible for T 3 activation (Ϫ2759 to Ϫ2018 bp) barely altered repression by unliganded TR␤ (Fig. 5C). Repression by unliganded TR␤ was alleviated by deletions inward from Ϫ1423 bp. Thus, repression occurs at alternative or additional sites to those responsible for T 3 -mediated activation.
Identification of a Thyroid Hormone-responsive Element (TRE) in the ␤TP Gene Promoter-To characterize further the T 3 -responsive region of the ␤TP promoter, 680 bp of upstream sequence (Ϫ2759 bp to Ϫ2080 bp) was cloned upstream of the minimal thymidine kinase (TK) promoter fused to the CAT gene as contained in the pBLCAT2 vector. T 3 stimulates an 8-fold increase of CAT activity when the 680-bp fragment was cloned in the forward direction and 9.5-fold when cloned in the reverse orientation (Fig. 6). Thus the ␤TP promoter T 3 -responsive region also confers strong T 3 induction on the heterologous TK promoter in an orientation independent manner. To delineate the T 3 -responsive region of the upstream fragment, two constructs were produced that successively removed 5Ј and 3Ј sequence. In the first construct a deletion was introduced on the 3Ј end of the 680-bp fragment, leaving the 295 bp of 5Ј sequence (pTK300 in Fig. 6). The construct, pTK300, is nearly as active as the original 680-bp fragment (7.6-versus 8-fold activation), indicating that the 3Ј sequence makes a negligible contribution to the T 3 response. The second construct further narrowed the sequence on both 5Ј and 3Ј ends of the 300-bp construct, encompassing 102 bp from Ϫ2620 to Ϫ2518 bp of the ␤TP promoter (pTK100 in Fig. 6). The 102-bp construct is as effective as the 295-bp construct (7.5-versus 7.6-fold) and nearly as effective as the 680-bp construct, indicating that the T 3 -responsive region is located within sequence spanning Ϫ2620 to Ϫ2518 bp.
The sequence between Ϫ2620 and Ϫ2518 bp was searched for half-sites that conformed to the general consensus 5Ј-PuGG(A/T)CPu-3Ј (where Pu indicates a purine nucleoside) and that possessed the number and spacing of half-sites consistent with known TREs. Using this approach a TRE was identified between bases Ϫ2576 and Ϫ2562 bp (Fig. 7A), which is composed of two directly repeated half-sites separated by 3 bp (DR3). To test the role of the ␤TP TRE in directing the T 3 responses, deletion analysis was used to remove the 3Ј half-site and subsequently both half-sites of the TRE from the 680-bp fragment. Deletion of the 3Ј site and 3Ј-flanking sequence results in a drop in activation by T 3 and TR␤ from 8.0-to 2.9-fold (pTK⌬3Ј in Fig. 6). A similar deletion of both half-sites of the ␤TP TRE results in the loss of T 3 induction (pTK⌬5Јϩ3Ј in Fig.  6). Thus deletion of the ␤TP TRE results in the loss of the T 3 responses identified in the upstream fragment of the ␤TP promoter.
Gel shift assays were used to determine if the ␤TP TRE formed specific complexes with TR␤. The binding of TR␤ to the ␤TP TRE was compared with that of the DR4 type TRE from the rME promoter (43). An IR1 type element, between bases Ϫ2110 and Ϫ2088 bp, which was determined not to contribute to T 3 responses (data not shown), was used as a negative control (Fig. 7A). The ␤TP TRE binds and shifts with TR␤ homodimers and more intensely when the RXR␣ accessory protein is present (Fig. 7B). The shifted bands are at levels of intensity similar to those obtained when the rME element is used, indicating formation of high affinity complexes between the ␤TP TRE and TR␤/RXR␣. As expected, the IR1 element failed to bind TR␤ and shifted only faintly in the presence of TR␤/RXR␣ (Fig. 7C).
To further characterize the ␤TP TRE, cold competitions were performed using the ␤TP TRE element itself or the rME element. As expected, unlabeled ␤TP TRE competes with labeled ␤TP TRE (Fig. 8A). The rME element competes effectively with the ␤TP TRE element for binding to TR␤ indicating that ␤TP TRE forms complexes in the same fashion as rME (Fig. 8B).
Within TRE half-sites, loss of one or both of the two conserved G nucleotides substantially reduces TR binding and TRE function (44,45). To test the role of these nucleotides, a mutant ␤TP TRE was constructed (⌬DR3) in which the G residues at bases Ϫ2573-74 and Ϫ2564 -63 were changed to T or A, respectively (Fig. 7A). The reconfigured element, ⌬DR3, failed to bind TR␤ homodimers and bound TR␤/RXR␣ heterodimers only faintly (Fig. 8C) proving that residues known to be crucial for TR binding in other TREs are also necessary for TR binding to the ␤TP TRE. Confirming the results from gel shift, placement of the mutant ␤TP TRE element upstream of the TK promoter within the pBLCAT2 vector failed to confer T 3 responses to the TK-CAT construct in TE671 cells (construct pTK⌬DR3 in Fig. 6). Thus, the intact sequence of the ␤TP TRE is necessary to bind TR␤ and to activate transcription in response to T 3 and TR␤.
To probe the composition of the DNA-protein complexes, anti-hTR␤ antibody was used to super-shift DNA-protein complexes that had formed with TR␤. ␤TP TRE-protein complexes, formed in the presence of TR␤ or TR␤/RXR␣, were shifted by antibody, demonstrating that the complexes with ␤TP TRE were formed by TR␤ binding (Fig. 8D). DISCUSSION Basal transcription of the ␤-Trace gene is directed from a small and highly active core promoter. The core promoter is found within the first 325 bp of upstream sequence and directs CAT gene expression in TE671 cells at a level similar to the pSV 2 CAT-positive control vector. Regions of the core promoter bear striking sequence identity to the P4 promoter of the IGF-II gene (39), which is active in the choroid plexus (40), and to the ␤-LH gene, which is active in the CNS (38).
The human ␤TP gene is regulated by TR␤ in a T 3 -dependent manner. T 3 and TR␤ substantially elevate ␤TP promoter activity, whereas unliganded TR␤ effectively represses the promoter. The level of T 3 -dependent activation observed is comparable to that observed using a classical TRE from the rME promoter (43), indicating that the overall response of the ␤TP promoter to T 3 is strong (Fig. 4).
Deletion analyses indicate that the ␤TP thyroid hormoneresponsive region lies between Ϫ2759 and Ϫ2018 bp. When placed upstream of the TK minimal promoter in either orientation, this region confers T 3 regulation on the heterologous TK promoter. Further deletion analysis of ␤TP-TK promoter fusions allowed the T 3 -responsive region to be localized to the sequence between Ϫ2620 and Ϫ2518 bp, a region in which we have identified a TRE composed of two consensus half-sites separated by a 3-bp spacer (DR3-type). The 3Ј half-site of the TRE exactly matches the general consensus half-site, and its deletion results in substantial although not complete loss of T 3 induction. Deletion of both half-sites completely abolishes T 3 induction. As with other TREs, T 3 induction from the ␤TP TRE is lost with mutation of the two conserved G nucleotides within each half-site.
Gel shift experiments demonstrate that the ␤TP TRE forms specific complexes with both TR␤ homodimers and TR␤⅐RXR␣ heterodimers. We used cold competitions with the rME TRE, mutagenesis of the ␤TP TRE, and super-shifts with anti-TR␤ specific antibodies to demonstrate that ␤TP TRE binds to and interacts with TR␤ in a manner consistent with other TRE sequences.
The ␤TP TRE is well upstream of the core promoter. This The TRE from the ␤TP promoter binds TR␤. A, the sequence of the ␤TP TRE is presented as is that of the IR1-negative control. Potential TRE half-sites are overlined by arrows. Shown below the ␤TP TRE sequence are the nucleotide changes introduced to produce the ⌬DR3 element. B and C, gel shift analysis. The ␤TP TRE and IR1 regions shown in A were labeled with 32 P. The rME promoter TRE probe was similarly labeled and served as the positive control for receptor binding. 10 fmol of labeled probe was incubated with purified recombinant TR␤, RXR␣, or both. Protein from a sham receptor preparation (Mock) was used to control for nonspecific protein binding. The rME TRE complexes were resolved on polyacrylamide gels alongside the ␤TP TRE and IR1 complexes. H, denotes TR␤1/RXR␣ heterodimers, D, denotes TR␤ homodimers, and F denotes unbound probe.
organization places the ␤TP promoter in a growing family of genes distinguished by TRE elements distal to the core promoter. These include the human insulin gene where the TRE is located at Ϫ1 kb (46), the rat S 14 gene which has multiple TREs located in a 200-bp region around Ϫ2.6 kb (47), and the rat ucp gene which has two TREs located in the region around Ϫ2.3 kb (48,49). Both the ␤TP TRE and one of the ucp TREs, the downstream TRE, are composed of two directly repeated halfsites separated by a 3-bp spacer. Directly repeated half-sites with three base pair separations are commonly associated with vitamin D receptors in accordance with the 3,4,5 rule (44,50). However, the rat ucp downstream TRE is unresponsive to induction by vitamin D receptors, indicating that DR3-type TREs are capable of T 3 -specific responses (48). Additionally, in vitro analyses have shown that DR3 elements can bind TR and direct T 3 responses at or near the level observed with the more common DR4 spacing (51). The sequences flanking the halfsites may prove to be more important in honing the T 3 response of the ␤TP TRE than the spacings between the half-sites. Koenig et al. (52) have identified an extended consensus halfsite sequence which functions as an equally strong TRE regardless of whether the spacing between the half-sites is 3, 4, or 5 bp. A similar dependence on half-site and flanking sequence, independent of half-site spacing, has been noted for the closely related retinoic acid receptor elements (53).
The ␤TP promoter shows considerable T 3 -independent repression by TR␤ (54). Repression appears to be specific for the ␤TP promoter as two heterologous promoters used in this study (SV 2 and TK) are only mildly affected by unliganded TR␤. Deletion of the sequences responsible for T 3 activation does not alleviate repression. Repression may result from TR binding to TRE half-sites located elsewhere in the promoter. TRE halfsites can comprise a functional element in TRE-mediated repression (55,56). Additionally, unliganded TR might disrupt protein-protein interactions necessary for transcription as has been observed in the human glycoprotein hormone ␣ gene (57). TE671 cells express ␤TP but not TR␤. These properties permitted the examination of basal ␤TP promoter activity in the absence of the potentially confounding effects of TR␤. Subse-quent expression of TR␤ in TE671 cells allowed dissection of the repressive effects of TR␤ on ␤TP basal transcription from the more familiar role of TR␤ in activation. The basal activity of the ␤TP promoter observed in TE671 cells in the absence of TR␤ may be physiologically relevant. Although TR expression is widespread it is not universal (37,58).
The increase in the activity of the ␤TP promoter by T 3 and TR␤ and its repression by TR␤ alone, as shown here, agrees with and provides a dual mechanism for the reduced ␤TP mRNA levels observed in thyroidectomized rats (20). The thyroid hormone responsiveness of the ␤TP promoter may also imply T 3 control of PGD 2 synthesis. However, the high levels of ␤TP and ␤TP promoter activity observed should be contrasted to the low levels of PGD 2 observed in human CSF (59). Perhaps ␤TP functions as a ligand transporter within the CNS, as is the case for other proteins secreted by the CP and meninges. The structural similarity between ␤TP and other lipocalin transporters provides indirect support for a role in lipid transport. Further support for the role of ␤TP in transport processes has recently been provided by the work of Hoffmann et al. (60) who, in a careful in situ analysis of ␤TP expression in the developing mouse, have observed ␤TP expression at or near a number of blood-tissue barriers, hinting at a role for ␤TP in transport across these barriers.
The present study indicates that T 3 exerts a measure of control over ␤TP gene expression. If ␤TP functions to transport a specific ligand into CSF then T 3 potentially exerts a general level of control over the availability of the ligand in the CNS. Important questions regarding the regulation of ␤TP transcription remain to be addressed. Foremost among them is whether the expression of ␤TP is regulated in a tissue-specific manner, implying the use of different enhancer elements within the ␤TP promoter or different combinations of tissue-specific transcription factors. Additionally, the role of T 3 on ␤TP expression must be examined in the context of the other tissues in which it is expressed.
Acknowledgment-We thank Dr. Mark A. Jensen for critically reading the manuscript. Lanes marked C identify labeled rMEpositive controls, and lanes marked M identify proteins isolated from a sham receptor preparation. A gives the result of competing cold ␤TP TRE against labeled ␤TP TRE; B, the result of competing cold rME TRE against labeled ␤TP TRE. C, binding of TR␤ and RXR␣ to the ␤TP TRE probe is compared with a mutated ␤TP TRE probe (⌬DR3) in which two G nucleotides within each of the TRE half-sites have been changed to T or A. The location of the mutated bases are as shown below the sequence of the ␤TP TRE element in Fig. 7A. D, Super-shifts. Homo-or heterodimeric complexes of TR␤ and RXR␣ were formed with the ␤TP TRE probe and super-shifted with antibody specific to TR␤. Antibody shifted complexes are marked with SS. In all panels, B marks probe-protein complexes, and F marks unbound probe. Gel shifts were performed as described in Fig. 7 using 10 fmol of 32 P-labeled ␤TP TRE probe.