Regulation of Major Histocompatibility Complex Class I Gene Expression in Thyroid Cells

The major histocompatibility complex (MHC) class I gene cAMP response element (CRE)-like site, −107 to −100 base pairs, is a critical component of a previously unrecognized silencer, −127 to −90 bp, important for thyrotropin (TSH)/cAMP-mediated repression in thyrocytes. TSH/cAMP induced-silencer activity is associated with the formation of novel complexes with the 38-base pair silencer, whose appearance requires the CRE and involves ubiquitous and thyroid-specific proteins as follows: the CRE-binding protein, a Y-box protein termed thyrotropin receptor (TSHR) suppressor element protein-1 (TSEP-1); thyroid transcription factor-1 (TTF-1); and Pax-8. TTF-1 is an enhancer of class I promoter activity; Pax-8 and TSEP-1 are suppressors. TSH/cAMP decreases TTF-1 complex formation with the silencer, thereby decreasing maximal class I expression; TSH/cAMP enhance TSEP-1 and Pax-8 complex formation in association with their repressive actions. Oligonucleotides that bind TSEP-1, not Pax-8, prevent formation of the TSH/cAMP-induced complexes associated with TSH-induced class I suppression, i.e. TSEP-1 appears to be the dominant repressor factor associated with TSH/cAMP-decreased class I activity and formation of the novel complexes. TSEP-1, TTF-1, and/or Pax-8 are involved in TSH/cAMP-induced negative regulation of the TSH receptor gene in thyrocytes, suppression of MHC class II, and up-regulation of thyroglobulin. TSH/cAMP coordinate regulation of common transcription factors may, therefore, be the basis for self-tolerance and the absence of autoimmunity in the face of TSHR-mediated increases in gene products that are important for thyroid growth and function but are able to act as autoantigens.

Thyrotropin (TSH) 1 suppresses major histocompatibility (MHC) class I gene expression in association with its action to increase the growth and function of rat FRTL-5 thyroid cells in continuous culture (1,2). Since enhanced class I expression has been demonstrated in thyrocytes from patients with autoimmune thyroid disease (ATD) (3), we proposed (1, 2, 4 -6) that TSH suppression of class I levels might be a normal mechanism to preserve self-tolerance in the face of increases in gene products associated with growth and function and that its loss or attenuation might cause ATD. The importance of suppressing class I to preserve self-tolerance and prevent autoimmunity is becoming clear in multiple disease states. For example, methimazole and iodide, agents used to treat patients with Graves' disease, one form of ATD, act in part by suppressing class I levels in thyrocytes (2,7), also methimazole prevents the development of a systemic lupus erythematosus syndrome or autoimmune blepharitis in experimental models in mice (8,9). Class I-deficient mice are resistant to developing these experimental diseases (9,10), and the action of methimazole mimics the class I-deficient state in these experimental diseases (8 -10).
TSH/cAMP coordinately decrease expression of the TSH receptor (TSHR) and class I genes (5,6), while increasing thyroglobulin (TG) and thyroid peroxidase (TPO) gene expression. We suggested that TSH-decreased MHC class I and TSHR gene expression might involve common transcription factors and that this allowed the cross-talk necessary for preserving selftolerance to gene products increased during TSHR-directed function and growth. Similarly, we considered the possibility that transcription factors involved in TSH/cAMP-increased TG and TPO gene expression might suppress class I, since TG and TPO are major thyroid autoantigens in ATD.
In this report, we show that the CRE-like site (TGACGCGA) at Ϫ107 to Ϫ100 bp in the class I promoter, which is homologous to a consensus CRE (TGACGTCA) (23,24), is critical for the activity of a hitherto unrecognized 38-bp (Ϫ127 to Ϫ90 bp) constitutive silencer of the class I promoter. TSH/cAMP induce the formation of specific and novel protein-DNA complexes with the silencer; the induced complexes reflect the ability of TSH/cAMP to regulate the interaction of multiple transcription factors with the silencer, the net result of which is suppression of class I gene expression. We show that TSH/cAMP-induced suppression of the class I, TG, and TSHR genes involves common transcription factors, as hypothesized (1, 2, 4 -6).

EXPERIMENTAL PROCEDURES
Materials-Highly purified bovine TSH was obtained from the hormone distribution program, NIDDKD, National Institutes of Health (NIDDK-bTSH; 30 units/mg), or was a previously described preparation, 26 Ϯ 3 units/mg, homogeneous by ultracentrifugation, about 27,500 in molecular weight, with the amino acid and carbohydrate composition of TSH (25). [␣-32 P]Deoxy-CTP (3000 Ci/mmol) and [ 14 C]chloramphenicol (50 mCi/mmol) were from NEN Life Science Products and [␥-32 P]ATP from Amersham Corp. Anti-CREB-327 or -activating transcription factor-2 (ATF-2), and their preimmune counterparts, was the gift of Dr. James P. Hoeffler, Invitrogen, San Diego, CA. Anti-CREB-2 was from Dr. J. M. Leiden (University of Michigan Medical Center, Ann Arbor, MI) and anti-mXBP from Dr. L. H. Glimcher (Harvard School of Public Health and Department of Medicine, Harvard University Medical School, Boston). The TTF-1 expression vector, pRcCMV-TTF-1, was that used previously (15,16,26) and was the kind gift of Dr. Roberto Di Lauro (Stazione Zoologica A. Dohrn, Villa Comunale, Naples, Italy); the pRcCMV plasmid used in its construction was from Invitrogen. The Y-box and Pax-8 expression vectors, pRcCMV-TSEP-1 and pRcCMV-Pax-8, were constructed by ligating their fulllength coding sequences with the pRc/CMV vector (19,21). All other materials were from the Sigma unless otherwise noted.
Construction of MHC Class I Promoter-CAT Chimeric Plasmids-The swine PD15Ј-flanking sequence-CAT chimeras, p(Ϫ1100)CAT, p(Ϫ203)CAT, and p(Ϫ127)CAT have been described (28 -32). Other PD1-CAT-chimeras were created by polymerase chain reaction (PCR; Ref. 33) using 25 ng of p(Ϫ203) or p(Ϫ127) CAT as template and 100 pmol each of a forward primer with a BamHI site on the 5Ј-end and a reverse primer having the PD1 sequence from Ϫ13 to ϩ1 bp of the transcription start site plus a HindIII site on the 3Ј-end. The reaction was performed at 94°C (1 min), 55°C (2 min), and 72°C (3 min) for 30 cycles; final extension was for 7 min at 72°C; and the amplified fragment was purified using 1.5% agarose gel electrophoresis. Mutants of p(Ϫ127)CAT were created by two-step, recombinant PCR (33,34). In the first step, two PCR products that overlap the sequence were created, both of which contain the same mutation introduced as part of the PCR primers. The second step PCR was performed using these overlapped PCR products as template and DNA sequence of the 5Ј-or 3Ј-end of the final products as primer. The PCR products were inserted into the multicloning site of pSV3CAT (28) 2 or pCAT-promoter or pCAT-Basic vectors purchased from Promega (Madison, WI). In the case of the pCAT vector, the CRE-like sequence and its mutants were created with a BamHI site on both ends of the primers.
The sequences of all constructs were confirmed (35), and DNA was prepared by CsCl gradient centrifugation (36). pSVGH, used to evaluate transfection efficiency, was a BamHI-EcoRI fragment encoding the human growth hormone (hGH) gene inserted into the BamHI-XbaI site of the pSG5 expression vector (Stratagene, La Jolla, CA) (11).
In the second procedure, FRTL-5 cells were grown to 80% confluency and then maintained 6 days in 5H medium plus 5% calf serum. Cells were returned to 6H medium for 12 h and transfected with the CAT chimeras as described above. Twelve h later, fresh 5H medium with 5% calf serum was added, supplemented or not with 10 Ϫ10 M TSH or 10 M forskolin. CAT activity was assayed 36 h later and conversion rates were normalized to hGH levels and protein.
CAT activity was measured as described (1, 11-13, 15-20, 32, 37), using 10 -30 g of cell lysate and a 130-l assay volume. Incubation was at 37°C for 4 h with acetyl-CoA supplementation (20 l of a 3.5 mg/ml solution) after 2 h. Acetylated chloramphenicol was separated by thin layer chromatography and autoradiographed; spots were quantitated in a scintillation spectrometer. Protein concentration was determined by Bradford's method (Bio-Rad); recrystallized bovine serum albumin was the standard.
Cellular Extracts-Cell extracts were made by a modification of the method of Dignam et al. (38). Briefly, FRTL-5 cells were harvested by scraping, after being washed twice in ice-cold phosphate-buffered saline and pelleted. The pellet was resuspended in 2 volumes of Dignam buffer C (20 mM HEPES at pH 7.9, 1.5 mM MgCl 2 , 0.42 M NaCl, 25% glycerol, 0.5 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride, 1 g/ml leupeptin, 1 g/ml pepstatin). The final NaCl concentration was adjusted on the basis of cell pellet volume to 0.42 M. Cells were lysed by repeated cycles of freezing and thawing. Extracts were centrifuged at 100,000 ϫ g and at 4°C for 20 min. The supernatant was recovered, aliquoted, and stored at Ϫ70°C.
Electrophoretic mobility shift assays were performed as described previously (12, 13, 14 -20, 32, 40). Binding reactions, in a volume of 20 l, were for 20 min at room temperature. Reaction mixtures contained 1.5 fmol of [ 32 P]DNA, 3 g of cell extract, and 0.5 or 3 g of poly(dI-dC) in 10 mM Tris-Cl at pH 7.9, 1 mM MgCl 2 , 1 mM dithiothreitol, 1 mM EDTA, and 5% glycerol. Where indicated, unlabeled double-or singlestranded oligonucleotides were added to the binding reaction as competitors and incubated with the extract for 20 min prior the addition of labeled DNA. In experiments using antisera, extracts were incubated with the serum in the same buffer for 1 h at 20°C before being processed as above. Reaction mixes were electrophoresed on 4 or 5% native polyacrylamide gels at 160 V in 1 ϫ TBE at 4°C. Gels were dried and autoradiographed.
The 1,10-Phenanthroline-Copper Ion Footprinting Procedure-Footprinting, using 1,10-phenanthroline-copper ion, was carried out essentially as described (41). After a scaled-up EMSA using an end-labeled fragment, Fr168, comprising Ϫ168 through Ϫ1 bp of the PD1 promoter, the gel was immersed in 200 ml of 50 mM Tris-HCl at pH 8.0 and 20 ml of the following solutions were added: 2 mM 1,10-ortho-phenanthroline, 0.45 mM CuSO 4 , and 58 mM 3-mercaptopropionic acid. After 15 min at room temperature, 20 ml of 28 mM 2,9-dimethyl ortho-phenanthroline was used to quench the reaction; 2 min later, the gel was rinsed in distilled H 2 O and autoradiographed for 40 min at 4°C, until the retarded bands were visible. Bands were excised and eluted overnight at 37°C in 0.5 M ammonium acetate containing 0.1% sodium dodecyl sulfate and 10 mM magnesium acetate. The eluted DNA was ethanolprecipitated and resuspended in distilled H 2 O. Equal numbers of counts from each sample were dried, resuspended in 98% formamide containing 10 mM EDTA, 0.025% bromphenol blue, and 0.025% xylene cyanol, and then separated on an 8% sequencing gel along with G ϩ A and C ϩ T Maxam-Gilbert sequence reactions (42) performed using the same probe. Autoradiography was at Ϫ80°C overnight.
Statistical Significance-All experiment were repeated at least three times with different batches of cells. Values are the mean Ϯ S.E. of these experiments where noted. Significance between experimental values was determined by two-way analysis of variance and are significant if p values were Ͻ0.05 when data from all experiments were considered.

The CRE-like Sequence between Ϫ107 and Ϫ100 bp Functions as a Constitutive Silencer and Is a Target for TSH/cAMP-mediated Repression of the Class I Promoter
The ability of TSH/cAMP to repress MHC class I transcription has been mapped to within 127 bp of initiation of transcription (1). Examination of this 128-bp DNA segment revealed the presence of an 8-bp sequence, Ϫ107 to Ϫ100 bp (Fig.  1A), with homology to characterized CREs (23,24). To determine whether this element functioned to regulate class I promoter activity, a set of derivative constructs was generated from a parental construct containing 127 bp of 5Ј-flanking sequence p(Ϫ127CAT). In one derivative, the 8-bp CRE-like sequence was deleted; in the other, we substituted a nonpalindromic mutation of the CRE-like octamer (Fig. 1B). Both constructs displayed increased promoter activity, relative to the parental construct, when transfected into FRTL-5 cells maintained in the absence of TSH (Fig. 1B).
The ability of the CRE-like element to silence a heterologous promoter was also assessed by introducing a 38-bp DNA segment, spanning Ϫ127 to Ϫ90 bp, downstream of an SV40 minimal promoter (Fig. 1C). The choice of a 38-bp segment was derived from the experiments described below. When placed in a 5Ј to 3Ј orientation, a single copy of this DNA segment was able to significantly reduce SV40 promoter activity, and the magnitude of the effect increased with the number of copies of the 38-bp segment inserted (Fig. 1C). When placed in a 3Ј to 5Ј orientation, two copies of this DNA segment were also able to significantly reduce SV40 promoter activity (Fig. 1C). Derivatives of the 38-bp segment, containing either a deletion (⌬CRE 1 (ϩ)) or nonpalindromic (NP CRE 1 (ϩ)) mutation of the CRE-like element, did not similarly decrease SV40 promoter activity (Fig. 1C).
TSH or forskolin significantly decreased the activity of the p(Ϫ127)CAT construct in FRTL-5 cells relative to untreated controls (Table I, A, columns 3-5), i.e. TSH enhanced the silencer activity. Deletion or mutation of the CRE, which increases the constitutive level of promoter activity in the absence of TSH, diminished the repressive response to TSH (Table I, A, columns [3][4][5]. The role of the CRE-like site in conferring TSH/cAMP responsiveness was confirmed in studies using the 38-bp silencer linked to the heterologous promoter (Table I, B). Although the SV40 promoter alone did not respond to TSH or forskolin, the promoter activities of a construct containing a single copy of the CRE 1 in a sense orientation (CRE 1 (ϩ)) or of constructs containing one or two copies of CRE 1 in a 3Ј to 5Ј orientation (CRE 1 (Ϫ), CRE 2 (Ϫ)), but not their nonpalindromic mutations, were significantly reduced by TSH or forskolin (Table I, B, columns [3][4][5]. From these data, it is concluded that the 8-bp CRE-like site is important for the function of a constitutive silencer located in a 38-bp fragment of the class I 5Ј-flanking region, Ϫ127 to Ϫ90 bp from the start of transcription. The 38-bp silencer is responsive to TSH or its cAMP signal; the CRE-like element within it is necessary for this functional response. The residual suppressive effect of TSH in p(Ϫ127)CAT chimeras containing a CRE deletion or mutation (Table I, A) suggests, nevertheless, that this may not be the sole site of TSH/cAMP action and that additional sites downstream of Ϫ90 bp might be TSH/cAMP-responsive.

FIG. 1. Effect of modifications of the CRE-like element on the activity of the p(؊127)CAT promoter (B) or on the CRE-like element linked to a heterologous promoter (C).
A, the sequence of the MHC class I gene between Ϫ127 and Ϫ90 bp is presented to show the position of the CRE-like site. The numbering system used is that in our previous report (32). B, the 8-bp CRE-like element in p(Ϫ127)CAT was either deleted (⌬CRE) or mutated to a nonpalindromic sequence (NP CRE) as noted. The CAT activities of these derivative constructs were compared with that of the parental p(Ϫ127)CAT following transfection into FRTL-5 cells and their incubation in 5H medium plus 5% calf serum for 2 days. Conversion rates were normalized to hGH levels and protein; then CAT activities were expressed relative to the parental p(Ϫ127)CAT, which averaged 2-fold higher than the pSV0-CAT control. Statistically significant differences (p Ͻ 0.05) from p(Ϫ127)CAT are noted by the asterisk. C, the sequence between Ϫ127 and Ϫ90 bp (designated CRE) was introduced at the 3Ј-end of constructs containing the SV40 promoter ligated to the CAT gene and then transfected into FRTL-5 cells ("Experimental Procedures"). CAT activities were measured after maintaining the cells for 2 days in 5H medium plus 5% calf serum. Constructs are diagrammatically represented on the left of the figure; arrows depict the orientation and the number of CRE copies. CAT activities are presented relative to the parental promoter construct, pCAT Promoter, which contains a minimal SV40 promoter; statistically significant decreases are indicated as p Ͻ 0.05 (*) or p Ͻ 0.01 (**). Data in B and C are the mean Ϯ S.E. for three separate experiments; results in both panels were duplicated with cells maintained in 3H medium plus 5% calf serum. For example, in B, when cells were incubated in 3H medium plus 5% calf serum for 2 days after transfection, CAT activities were 100%, 196 Ϯ 9%, and 208 Ϯ 8% for p(Ϫ127)CAT, p(Ϫ127)⌬CRE CAT, and p(Ϫ127) NP CRE CAT, respectively, in three separate experiments (mean Ϯ S.E.).

TSH/cAMP Induces Novel Complexes Whose Formation Depends on the CRE-like Element
When a DNA fragment encompassing the silencer, Ϫ168 to Ϫ1 bp (termed Fr168 ( Fig. 2A, bottom), was used in gel mobility shift assays with extracts derived from FRTL-5 cells cultured with or without TSH, a multiplicity of protein-DNA complexes was formed with either extract ( Fig. 2A). Whereas protein-DNA complexes A to E were common to both extracts, TSH treatment of the FRTL-5 cells induced the appearance of two novel complexes, F and G ( Fig. 2A, lane 2 versus 1). Formation of the TSH-induced complexes was specific, since their appearance could be prevented by unlabeled Fr168 (Fig. 2A, lane 3). More importantly, formation of only the F and G complexes could be prevented by the 38-bp silencer fragment, Ϫ127 to Ϫ90 bp, containing the CRE-like site and termed CRE-1 ( Fig. 2A, Forskolin (10 M) could substitute for TSH to induce the formation of the F and G complexes with the Fr168 probe ( Fig.  2B). Moreover, with either forskolin- (Fig. 2B) or TSH-treated cells (data not shown), formation of the F and G complexes could be prevented by a derivative 38-bp fragment, Ϫ127 to Ϫ90 bp, in which a consensus CRE sequence (CON CRE) was substituted for the native CRE-like sequence (Fig. 2B, lanes 3 and 4 versus 2) but not by derivative oligonucleotides from which the CRE-like element had been deleted (⌬CRE) or mutated to a nonpalindromic (NP CRE) substitution ( Phenanthroline-copper ion footprint analysis of the TSHinduced F or G complex identified a protected region, Ϫ131 to Ϫ95 bp, bounded by two strong hypersensitive sites (Fig. 3) that encompasses the CRE-like site, Ϫ107 to Ϫ100 bp. A less prominent hypersensitive band at Ϫ110 bp suggests this region may bind more than one factor as will be shown below and in a separate report. 3 Although not the only protected region in the footprint, these data, together with the data in Fig. 2, established that coincident with TSH/forskolin-induced suppression of class I RNA levels (1, 2) and TSH/forskolin-activated silencer activity ( Fig. 1; Table I), TSH/forskolin-induced the appearance of novel complexes with the class I promoter, whose formation required the CRE-like sequence, as did silencer activity. The data ( Fig. 2) additionally suggested that sequences flanking the CRE-like site are involved in complex formation, consistent with the extended footprint (Fig. 3).

The 38-bp DNA Fragment with CRE-dependent Silencer Activity Forms CRE-dependent Complexes with Multiple Proteins: the Effect of TSH on These Complexes and the Functional Role of the Proteins Involved
Identification of a Multiplicity of Transcription Factors That Interact with the 38-bp DNA Fragment Exhibiting CRE-dependent Silencer Activity-To characterize proteins capable of interacting with the 38-bp silencer element, we radiolabeled a double-stranded oligonucleotide spanning the segment Ϫ127 to Ϫ90 bp (CRE-1) and used it in gel shift assays with extracts from FRTL-5 cells maintained in the absence of TSH (Fig. 4); subsequently (see below), we evaluated the effect of TSH on the characterized complexes.
Four sets of complexes were observed (Fig. 4, A-D) with extracts from cells without TSH, all of which appeared to be specific, since their formation was prevented by competition with unlabeled CRE-1 (Fig. 4A, lanes 3-6 versus 2), albeit with different affinities. More importantly, formation of all the complexes was dependent on the CRE-like element. Thus, whereas their formation was inhibited by the native 38-bp CRE-1 silencer, no comparable competition was evident using the silencer fragment in which the CRE was deleted (Fig. 4A, ⌬CRE-1, lanes 7-9, versus 2). An oligonucleotide derived from the somatostatin receptor (Promega CRE), which contains a TABLE I Effect of TSH or forskolin (Fsk) on p(Ϫ127)CAT or pCAT promoter constructs with an intact, deleted, or mutated (nonpalindromic) CRE A, the 8-bp CRE-like element in the p(Ϫ127)CAT promoter construct was either deleted (⌬CRE) or mutated to a nonpalindromic sequence (NP CRE) as noted (Fig. 1B). The CAT activities of these derivative constructs were compared with that of the parental p(Ϫ127)CAT following transfection into FRTL-5 cells and their incubation in 5H medium plus 5% calf serum with or without 1 ϫ 10 Ϫ10 M TSH or 10 M forskolin for 36 h. B, class I DNA sequence between Ϫ127 and Ϫ90 bp (designated CRE) were introduced at the 3Ј end of constructs containing the SV40 promoter ligated to the CAT gene (Fig. 1C) and transfected into FRTL-5 cells as in Fig. 1. CAT activities were measured after maintaining the cells for 36 h in 5H medium plus 5% calf serum with or without TSH or forskolin. A and B, conversion rates were normalized to hGH levels and protein ("Experimental Procedures"); CAT activities were then expressed relative to the parental p(Ϫ127)CAT or pCAT promoter control. Data are the mean Ϯ S.E. for three separate experiments.
CAT activity relative to p(Ϫ127)CAT or pCAT promoter control (%) consensus CRE but otherwise unrelated flanking nucleotides, was unable to duplicate the action of CRE-1 as a competitor (Fig. 4A, lanes 10 -12 versus 2-6). These data are consistent with interpretation that multiple proteins interact with the 38-bp silencer, that the CRE is a critical element required for their binding, but that sequences flanking the CRE are involved in the binding of most.
As is the case for the TSHR CRE (11)(12)(13), one protein interacting with the CRE-like site in the 38-bp silencer is CREB-327 or an immunologically related CRE binding protein (23,24,(43)(44)(45). Thus, the CRE-containing oligonucleotide derived from the somatostatin receptor, which contains a consensus CRE but otherwise unrelated flanking nucleotides, appeared to inhibit the formation of one component within the A complex (Fig. 4A,  lanes 10 -12 versus 2). Moreover, one of components in the A complex could be super-shifted (Fig. 4B, lane 4) by an antibody to CREB-327 (43) but not by anti-CREB2, anti-mXBP, or antiactivating transcription factor-2 (ATF2-BR).
Three points should be noted with respect to these data. First, in the A complex region, proteins other than CREB are likely to interact with the CRE-like sequences, given the limited effect of anti-CREB antibody in supershifting the complexes. One will be identified as Pax-8 below; however, other CRE binding protein analogs of CREB (23, 24) may be compo-nents of the A complex and remain to be identified. Second, the CRE-like element is a functional CRE, since it interacts with CREB and is TSH/forskolin-responsive (11-13, 23, 24, 43-45). Third, the consensus CRE from Promega, while decreasing the formation of a component in the A complex region, appears to enhance the formation of one or more components in the C region (Fig. 4, lane 12). This observation may be relevant to data that reveal a similar action of anti-CREB (see below).
We pursued the identity of other factors forming complexes with the 38-bp silencer. As noted in the introduction, we had hypothesized that transcription factors that negatively regulate class I promoter activity might be the same as some involved in TSH/cAMP-induced negative regulation of the TSHR and/or positive regulation of the TG or TPO promoters (4 -6). Two obvious candidates are the tissue-specific factors, TTF-1 and Pax-8. Maximal TSHR gene expression in the absence of TSH is associated with the binding of TTF-1; additionally, TTF-1 acts synergistically with proteins such as CREB, which bind the CRE-like sequence of the TSHR (15). In the TG and TPO promoters, some TTF-1 sites also interact with Pax-8, i.e. the oligo(C) site in the TG promoter (21,22,26). TSH/cAMP decreases TTF-1 complex formation with the TSHR and TG promoters but increases Pax-8 complex formation with the TG promoter, coincident with increased TG gene expression (15,  4). Protein-DNA complexes are denoted by the letters A to G; F and G represent complexes present in the TSH-treated cell extracts only. B, EMSA were performed using the radiolabeled Fr168 probe incubated with extracts from FRTL-5 cells maintained in the presence or absence of 10 M forskolin for 48 h, after 6 days in medium with no TSH plus 5% serum. The probe was incubated with extract alone (lanes 1 and 2) or with extract plus a 100-fold excess of the cold competitors whose sequences are diagrammatically presented below the EMSA and which represent derivatives of the 38-bp silencer region termed CRE-1. As indicated, the F and G protein-DNA complexes induced by TSH in A are also induced by forskolin. 16). We evaluated the possibility that TTF-1 and/or Pax-8 might interact with the class I 38-bp silencer.
DNA binding sites for both TTF-1 and Pax-8 are contained within oligonucleotide C derived from the TG promoter (Fig.  5C). Oligonucleotide C is able to completely inhibit formation of complex B and reduce the amount of complex A, whereas a mutant of oligonucleotide C, which loses the ability to bind TTF-1 or Pax-8 (21,22,26), also loses its inhibitory effect (Fig.  5A, lanes 4 and 5 versus 2; Fig. 5B, lanes 3 and 4 versus 2). This suggests that either TTF-1 or Pax-8 (or related factors) contribute to these complexes.
To distinguish between TTF-1 and Pax-8 binding, an oligonucleotide that binds only TTF-1, oligonucleotide TTF-1 from the TSHR promoter, or its mutated counterpart (Fig. 5C) which loses TTF-1 binding and activity (15,16), was substituted in the competition assays (Fig. 5B). Whereas oligonucleotide C affected both complexes A and B, oligonucleotide TTF-1 completely inhibited complex B but did not affect complex A (Fig.  5B, lane 5 versus 2-4). In contrast, the mutant derivative of oligonucleotide TTF-1 does not compete for B complex formation (Fig. 5B, lane 6 versus 2-4). Taken together, these data suggest that complex B contains TTF-1 and that Pax 8 is a component of complex A, in addition to CREB.
Two additional points should be noted in these experiments. First, both the Pax-8 and TTF-1 complexes require CRE-dependent interactions with the 38-bp silencer. Thus, in the same experiment (Fig. 5A) the 38-bp CRE-1 oligonucleotide, but not its ⌬CRE-1 derivative, prevented the formation of both A and B complexes (Fig. 5A, lanes 1 and 3, respectively, versus 2).
Second, the concentration of poly(dI-dC) in the assay could change the appearance of the complexes (Fig. 5, A versus B). Higher poly(dI-dC) concentrations (3 g/assay) were observed to significantly enhance formation of the A and B complexes (Fig. 5, A versus B) but attenuate formation of the C complexes (Fig. 5, B versus Fig. 4). Since TTF-1, CREB, and Pax-8 are double strand-specific in their DNA interactions (12-16, 21, 22, 26), one possibility to explain this phenomenon was that higher concentrations of poly(dI-dC) were suppressing the binding of proteins with lesser affinity or specificity, such as single strand DNA-binding proteins, and that these comprised the protein/ DNA adducts in complex C.
To assess this possibility, each of the component single strands of the 38-bp silencer was tested for its ability to inhibit complex formation in low poly(dI-dC) conditions (Fig. 6A). Neither strand affected the formation of complexes A or B; however, both single strand oligonucleotides reduced the formation of complex C, the non-coding strand sequence more efficiently than the coding strand (Fig. 6A). Thus, whereas the A and B complexes involved double strand DNA interactions, formation of the components of complex C appeared to result from the interaction of factors that can bind to individual strands of the 38-bp CRE-1 silencer.
Two proteins involved in the regulation of the TSHR gene are known to be single-strand DNA binding proteins. One is a single strand binding protein (SSBP-1) which was identified by its ability to bind to the noncoding strand of the TSHR promoter, immediately 5Ј to and contiguous with the TTF-1 binding site (17,20). The second is a Y-box protein that we cloned and termed TSEP-1 (TSHR suppressor element protein-1), because of its ability to suppress the constitutive enhancer activity of the TSHR CRE (19). TSEP-1, like other Y-box family proteins, binds to both single or double strand DNA (13, 19, 46 -48). The ability of the cognate DNA binding sites of these two factors on the TSHR (Fig. 6C) to inhibit formation of complex C bands was assessed (Fig. 6B).
An oligonucleotide corresponding to the SSBP binding site, Ϫ194 to Ϫ169 bp on the non-coding strand of the TSHR minimal promoter, competed for complex C entirely, without affecting either complexes A or B (Fig. 6B, lanes 3 and 4 versus 2). A second oligonucleotide containing the TSEP-1 binding site de- FIG. 3. 1,10-Phenanthroline-copper ion footprinting of the TSH-induced G complex. Radiolabeled Fr168 (see Fig. 2A) was incubated with extract from TSH-treated FRTL-5 cells and EMSA performed as in Fig. 2. Gels were treated with 1,10-ortho-phenanthroline containing CuSO 4 as described under "Experimental Procedures." Protein-DNA complex G (Fig. 2) was excised from the gel along with a region containing unbound probe. After both sections of gel were eluted overnight, the eluted DNA was precipitated with ethanol, resuspended in H 2 O, and equal amounts of radioactivity separated on an 8% sequencing gel along with G ϩ A and C ϩ T Maxam-Gilbert sequence reactions performed using the same probe. Autoradiography was at Ϫ80°C overnight. Hypersensitive sites delineating or within the protected region, including the CRE-like sequence, are noted at the top. The sequence of the protected area between Ϫ131 and Ϫ95 bp is noted on the bottom; circled residues are the hypersensitive sites.
FIG. 4. In the absence of TSH, the 38-bp silencer region containing the CRE-like sequence forms multiple protein-DNA complexes with extracts from FRTL-5 cells, one of which appears to be CREB; their formation depends on the CRE-like sequence, ؊107 to ؊100 bp, and on sequences flanking the CRE. The radiolabeled double-stranded 38-bp DNA fragment, Ϫ127 to Ϫ90 bp, termed CRE-1, was incubated with extracts from FRTL-5 cells maintained in 3H medium plus 5% calf serum for 6 days, and complexes were analyzed by EMSA, as described under "Experimental Procedures." A, complex formation was evaluated in the presence or absence of the noted unlabeled double-stranded oligonucleotides: CRE-1, ⌬CRE-1 with the CRE-like sequence deleted, and an oligonucleotide from Promega (Promega CRE) which contains the consensus CRE and flanking residues from the somatostatin promoter. The amount of each competitor in fold excess over probe is noted at the top of each set of gels, along with a diagrammatic representation of the structure of the competitor. B, incubations were performed in the presence of 2 l of rabbit antiserum against CREB2, mXBP, ATF2, or CREB-327, as indicated (lanes 1-4, respectively); the dashed line notes the supershifted complex resultant from incubation with the CREB-327 antibody. Letters represent groups of complexes formed by the extract. The improved separation of the A complex region in this experiment was achieved using a lower gel concentration and longer period during the separation.
rived from the coding strand of the TSHR promoter also competed for complex C but not A or B (Fig. 6B, lanes 6 and 7 versus  5). In contrast to oligonucleotide SSBP, however, oligonucleotide TSEP-1 inhibited formation of only the major center band within complex C, sparing adjacent weaker bands. Association of TSEP-1 was, therefore, clearly specific, since only one of the complexes was competed; it was unclear why oligonucleotide SSBP competed for all of the bands within complex C, although we speculated that its binding to the noncoding strand might exclude the binding of TSEP-1 to the coding strand. To further assess the specificity of these complexes and characterize their single strand binding ability, we performed the following experiments using radiolabeled single strand oligonucleotides with the sequence of the 38-bp silencer termed CRE-1.
Using a radiolabeled oligonucleotide with the sequence of the CRE-1 coding strand on the class I promoter, we showed that formation of the complex with the slowest mobility in Fig. 6A was inhibited by the presence of excess unlabeled wild type oligonucleotide TSEP-1 (Fig. 6C). TSEP-1 binds to a CCTC motif on the terminus of the 5Ј-tandem repeat (Fig. 6C, TR); mutation of the CCTC motif, but not a mutation not involving the CCTC motif, resulted in a loss of this competition. These data were consistent with the interpretation that TSEP-1 binding to the 38-bp silencer is one component of the C complex.
SSBP sites on the TSHR are on the noncoding strand of the TSHR, 5Ј and contiguous with an upstream and downstream TTF-1 double strand binding site; each has a GXXXXG binding motif (17,20), the G nucleotides of which are underlined in Fig.  6C. Mutation of the 5Ј and 3Ј terminal G nucleotides in the SSBP sites results in a loss of SSBP binding and function (17,20). Using the CRE-1 noncoding strand as radiolabeled probe, we showed that formation of one of the protein complexes with the noncoding strand of class I 38-bp silencer (Fig. 6A) was reduced by including an excess of wild type, single strand oligonucleotide able to bind SSBP from the TSHR (Fig. 6C) but much less so by the SSBP oligonucleotide with its GXXXXG site mutated. These data were consistent with the interpretation that SSBP-1 or a related protein binds to the 38-bp silencer and is another component of the C complex.
We believe it is reasonable to conclude from these data that complex C includes two major components formed by the interaction of single strand binding proteins, SSBP-1 and TSEP-1, which interact with the TSHR and that the binding of either requires the CRE-like site. We conclude from this series of experiments (Figs. 4 -6) that the CRE is part of a constitutive silencer that associates with a multiplicity of transcription factors, of which CREB appears to be a minor component. The formation of the complexes depends on both the CRE-like element, as well as additional flanking sequences. Among the factors that interact with this silencer are factors that interact with the TSHR (TTF-1, SSBP-1, TSEP-1, and CREB) as well as TG or TPO promoters (TTF-1 and Pax-8). They include, therefore, factors that exhibit tissue specificity (TTF-1 and Pax-8), as well as factors that are ubiquitous (CREB, TSEP-1, and SSBP-1).

Effect of TSH on the Multiplicity of Protein-DNA Complexes Formed with the 38-bp DNA Fragment with CRE-dependent
Silencer Activity-TSH treatment of FRTL-5 thyroid cells decreases class I promoter activity, in part by acting through a 38-bp silencer whose activity is dependent on a CRE-like element, as evidenced above ( Fig. 1 and Table I). This is consistent with the observation that TSH treatment for as little as 12-18 h causes a significant decrease in class I RNA levels (1,2). We now show that extracts from cells treated with TSH for this period alter the amount and composition of the protein-DNA complexes formed with the 38-bp silencer region whose activity and binding depends on the CRE (Fig. 7).
Thus, TSH treatment of the cells results in markedly diminished formation of the A and B complexes, which contain CREB, Pax-8, and TTF-1 (Fig. 7, lane 4 versus lane 2). The decreased CREB interaction is evidenced by a diminished ability of anti-CREB-327 to supershift the A complex (Fig. 7, lane  5 versus 3, dashed arrow). The residual component of the A complex involves Pax-8, as evidenced by inhibition of its for- FIG. 5. The 38-bp silencer region containing the CRE-like sequence forms complexes with both TTF-1 and Pax-8 in addition to CREB. The double-stranded, radiolabeled 38-bp DNA fragment, Ϫ127 to Ϫ90 bp (CRE-1), was incubated with extracts from FRTL-5 cells maintained in 5H medium plus 5% calf serum for 6 days; complexes were analyzed by EMSA in 3.0 (B), as well as 0.5 (A) g/ml poly(dI-dC), as described under "Experimental Procedures." A, complex formation was evaluated in the presence or absence of the noted unlabeled, doublestranded oligonucleotides: CRE-1, ⌬CRE-1 with the CRE-like sequence deleted, an oligonucleotide (oligonucleotide C) containing the TTF-1/Pax-8 element in the TG promoter, and a mutant thereof which loses TTF-1/Pax-8 binding and activity in the FRTL-5 cell. The sequence of oligonucleotide C and its mutant, with the TTF-1/Pax-8 site underlined, is presented in C; their properties have been characterized previously (15,16,21,22,26). Mutated nucleotides are starred and included within the boxed area. B, complex formation was again evaluated in the presence or absence of CRE-1, an oligonucleotide containing the TTF-1 element in the TSHR (15)(16)(17)(18)20), and a mutant thereof which loses TTF-1 binding and activity in the FRTL-5 cell (15)(16)(17)(18)20), in addition to oligonucleotide C and its mutant. The TSHR TTF-1 site does not bind Pax-8 (14 -16). The sequences of the oligonucleotides containing the TSHR TTF-1 element and its mutant are presented in C; their properties have been characterized previously (15)(16)(17)(18)20). Mutated nucleotides are again starred and included within the boxed area. The amount of each competitor was 100-fold in excess of probe. Letters represent groups of complexes formed by the extract; the TTF-1 and Pax-8 containing complexes with the 38-bp silencer, which are noted, are based on the inhibition data as discussed in the text. mation by oligonucleotide C of the TG promoter (data not shown). The simultaneous decrease of the TTF-1 and CREB complexes may be related, since CREB and homeodomaincontaining binding proteins such as TTF-1 are known to act synergistically on the minimal promoter of the TSHR (15) and somatostatin receptor (49 -51).
In contrast to the A or B complex, the intensity of the C complex region is enhanced by TSH treatment, and the complexes become evident even in the presence of high poly(dI-dC) concentrations (Fig. 7, lane 4 versus 2). Since the C complex is now seen, despite the high levels of poly(dI-dC), this suggests that the specificity and/or affinity of binding by proteins within the C complex are enhanced, also that one or both single strand binding proteins, SSBP-1 or TSEP-1, is now better able to bind double strand DNA.
Interestingly, the addition of anti-CREB-327 in vitro mimics the in vivo effect of TSH treatment of FRTL-5 cells in its ability to increase C complex formation (Fig. 7, lane 3 versus 4 by  comparison to lane 2). This action is specific since other antisera, i.e. anti-CREB2, anti-mXBP, or ATF2-BR, do not similarly increase the amount of complex C (data not shown). Furthermore, anti-CREB-327 does not change B complex formation (Fig. 7, lane 3 versus 2 or 4 versus 3). The TSHinduced increase in C complex may, therefore, reflect a TSHinduced decrease in CREB binding mimicked by anti-CREB-327 in vitro. Nevertheless, it seems clear that TSH increases the binding of TSEP-1 and/or SSBP-1 (complex C) while decreasing CREB and TTF-1, but not Pax-8 binding, to the 38-bp silencer (complexes A and B).
These data raised the possibility that the novel complexes, which are evident when TSH/forskolin-treated extracts are incubated with Fr168 of the 5Ј-flanking region of the class I promoter, reflect the TSH-induced change in complexes with the 38-bp silencer, particularly the binding of TSEP-1 and/or Pax-8, since both are suppressors of class I gene expression as will be shown below. To evaluate this possibility, we tested the ability of oligonucleotides containing these binding sites on the TSHR or TG promoters to compete for the TSH-induced complexes formed with radiolabeled, double strand Fr168 or Fr127 (Fig. 8).
Oligonucleotide TIF from the TSHR insulin response element (18), like oligonucleotide TSEP-1, contains a CCTC motif and TSEP-1 binding site (Fig. 8C). In the presence of excess, unlabeled single strand oligonucleotide TIF(Ϫ), formation of the TSH-induced complexes with either double strand Fr168 or Fr127 are inhibited (Fig. 8A, lane 1 versus 2 and Fig. 8B, lane  3 versus 2, respectively), as is the case for the positive control, double strand oligonucleotide CRE-1, the 38-bp silencer itself (Fig. 8B, lane 7 versus 2). In contrast, a single strand form of oligonucleotide TIF(Ϫ), with a mutation in the CCTC motif that is important for TSEP-1 binding (mutant 2), did not compete (Fig. 8A, lane 3 versus 2; Fig. 8B, lane 5 versus 2), whereas the oligonucleotide TIF(Ϫ)-derivative with a mutation that does not involve the CCTC motif, mutant 1, remained a competitor (Fig. 8A, lane 4 versus 2; Fig. 8B, lane 6 versus 2). These results were duplicated by wild type and mutant single strand oligonucleotide TSEP-1(ϩ) and oligonucleotide S-box(ϩ) from the TSHR (Fig. 8C), which also have TSEP-1 sites (data not FIG. 6. The C complexes formed with the 38-bp silencer appear to involve proteins able to bind either its coding or noncoding strands (A); these appear to involve two single strand binding proteins important in TSH/cAMP suppression of TSHR gene expression in FRTL-5 thyroid cells: TSEP-1 and SSBP-1 (17,19,20). The double-stranded radiolabeled 38-bp DNA fragment, Ϫ127 to Ϫ90 bp, termed CRE-1, was incubated with extracts from FRTL-5 cells maintained in 5H medium plus 5% calf serum for 6 days; complexes were analyzed by EMSA in 0.5 g/ml poly(dI-dC) as described under "Experimental Procedures." A, complex formation was evaluated in the presence or absence of a 100-fold excess over probe of the unlabeled single strand oligonucleotides comprising the coding and noncoding stand of CRE-1. B, complex formation was evaluated in the presence or absence of a single strand oligonucleotide from the noncoding strand of the TSHR, which binds a single strand binding protein termed SSBP-1 (17,20), and a single strand oligonucleotide from the coding strand of the TSHR which binds a Y-box protein termed thyrotropin receptor suppressor element protein-1 (TSEP-1) (19). These are termed oligonucleotide SSBP and oligonucleotide TSEP-1, respectively; their sequences are presented in C. The amount of each unlabeled oligonucleotide, in fold-excess over probe, is noted.

FIG. 7. TSH treatment of FRTL-5 cells decreases CREB and TTF-1, but not Pax-8, binding within the A and B complexes formed by the class I 38-bp silencer containing the CRE-like sequence; it also causes an increase in C complex formation, which includes protein-DNA complexes with TSEP-1 and the SSBP-1.
The radiolabeled double-stranded 38-bp DNA fragment, Ϫ127 to Ϫ90 bp, termed CRE-1, was incubated with extracts from FRTL-5 cells maintained in 5H medium plus 5% calf serum for 6 days and then treated for 16 additional h with the same medium or the same medium plus 1 ϫ 10 Ϫ10 M TSH. Incubations were performed in the presence of 3 g/ml poly(dI-dC) and in the presence or absence of 2 l of rabbit antiserum against CREB-327. Complexes were analyzed by EMSA, as described under "Experimental Procedures." The A, B, and C complex areas are noted (see Figs. 4 -6); the A region contains complexes with CREB and Pax-8, the B with TTF-1, and the C with TSEP-1 and SSBP-1 (Figs. 4 -6). shown).
Oligonucleotide C, an oligonucleotide that binds the doublestranded binding proteins TTF-1 and Pax-8 (Fig. 5), did not affect the TSH-induced complex (Fig. 8B, lane 4 versus 2). The oligonucleotide with the sequence of the SSBP site also did not inhibit formation of the TSH-induced complex (data not shown).
We conclude from these data (Fig. 8), that TSEP-1, a Y-box protein involved in TSH/cAMP suppression of TSHR gene expression in FRTL-5 cells, is a dominant component in the formation of the TSH/cAMP-induced novel complex with the CRE-like silencer of the MHC class I gene. Formation of this novel complex is associated, therefore, not only with TSH/ cAMP-induced suppression of class I gene expression ( Fig. 1; Table I), which requires an intact CRE-like site, Ϫ107 to Ϫ100 bp, but also with the ability of TSH to increase the binding of the Y-box protein to the 38-bp silencer in a CRE-dependent manner (Fig. 7).
Functional Role of Proteins That Bind to the 38-bp DNA Fragment with CRE-dependent Silencer Activity-We evaluated the functional effect of TSEP-1, TTF-1, and Pax-8 on the activity of the construct containing 127 bp of 5Ј-flanking sequence, p(Ϫ127)CAT, and on a derivative, in which the 8-bp CRE-like sequence was deleted p(Ϫ127 ⌬-CRE)CAT. Each was cotransfected into FRTL-5 cells, maintained in the absence of TSH, along with plasmids containing the TSEP-1, TTF-1, and/or Pax-8 cDNAs: pRcCMV-TSEP-1, pRcCMV-TTF-1, pRc-CMV-Pax-8, respectively (Fig. 9). Transfections with pRcCMV, the vector's only construct, served as the control in each case.
Cotransfection of pRcCMV-TSEP-1 decreased the promoter activity of p(Ϫ127)CAT but not the activity p(Ϫ127 ⌬-CRE) CAT (Fig. 9A); the same result was obtained comparing the effect of pRcCMV-TSEP-1 on p(Ϫ127)CAT versus p(Ϫ127 NP CRE)CAT, which has a nonpalindromic mutation of the CRE as described in Fig. 1B. In contrast to pRcCMV-TSEP-1, cotransfection of pRcCMV-TTF-1 increased the promoter activity of p(Ϫ127)CAT but not the activity p(Ϫ127 ⌬-CRE)CAT (Fig. 9B). pRcCMV-Pax-8 decreased the promoter activity of p(Ϫ127)CAT but not the activity p(Ϫ127 ⌬-CRE)CAT (Fig. 9C), and its presence prevented the enhancer activity of pRcCMV-TTF-1 (Fig. 9D). The effect of Pax-8 to reduce p(Ϫ127)CAT activity in Fig. 9C is ascribed to its action on endogenous TTF-1, whose levels are maximally expressed in FRTL-5 cells maintained in 5H medium with no TSH (15). In each case, the activity was not duplicated by the vector control, pRcCMV.
These data indicated that TSH, which increased the binding  1; B, lane 3), a TSEP-1 binding site on the TSHR, or mutants thereof. Oligonucleotide TIF is one of three TSEP-1 binding sites of the TSHR, termed oligonucleotide TSEP-1(ϩ), oligonucleotide S-box(ϩ), and oligonucleotide TIF(Ϫ) (C); the sequences of the competitor oligonucleotides and their location in the TSHR 5Ј-flanking region are noted. The dark bars represent the CCTC motif which is important for TSEP-1 binding in the TSHR (19). In each case the mutant 1 (Mut. 1) oligonucleotide binds TSEP-1, whereas the mutant 2 (Mut. 2) form loses binding activity (19). Oligonucleotide TIF Mut-2, which loses TSEP-1 binding activity because of a mutation in the CCTC binding motif, does not act as a competitor (A, lane 3;, B, lane 5), whereas oligonucleotide TIF Mut-1, which does not affect TSEP-1 binding activity, retains the competitive effect of the wild type oligonucleotide (A, lane 4; B, lane 6). B, we additionally show that double strand oligonucleotide C from the TG promoter, which can bind TTF-1 or Pax-8, does not prevent formation of the TSH-induced complex, whereas double strand CRE-1 does inhibit formation of the TSH-induced complexes.
of TSEP-1 but decreased the binding of TTF-1 to the 38-bp silencer, was increasing the binding of a suppressor of class I activity (TSEP-1) while decreasing the binding of an enhancer (TTF-1). Moreover, Pax-8, whose binding to the 38-bp silencer is not decreased by TSH, also suppresses the enhancing activity of TTF-1. Since TSH-induced decreases in TTF-1 RNA levels and complex formation in the TSHR are only partial (15,16), the Pax-8 action may explain the apparent complete loss of TTF-1 complex formation associated with TSH effects on the silencer complex of class I (Fig. 8). DISCUSSION MHC class I molecules are expressed on the cell surface and are vehicles for the presentation of antigenic peptides to immune cells; quantitative and qualitative variations in class I expression play important roles in determining the T-cell response (52,53). MHC class I can be positively or negatively regulated in response to virus infections or lymphokines (54); recently we showed that class I levels could also be hormonally regulated and suggested this hormonal regulation might be important to suppress autoimmunity during hormonally induced changes in growth and function which resulted in altered levels of gene products known to be autoantigens (1, 2, 4 -6).
The present experiments were aimed at characterizing the molecular mechanisms governing class I transcription in thyrocytes and those mediating the TSH/cAMP-induced decrease in transcription. With respect to the former, we show, for the first time, that a CRE-like sequence, Ϫ107 to Ϫ100 bp from the start of transcription, functions as a constitutive silencer of the class I gene. Thus, not only does its deletion or mutation in the homologous class I promoter result in increased activity, its insertion within a 38-bp surrounding region into a heterologous promoter decreases promoter activity as a function of copy number, so long as the CRE is intact. With respect to the latter, we also show that TSH activates the silencer, that the TSH action requires an intact CRE-like sequence, that this is one mechanism by which the hormone decreases class I expression, and that the action of TSH is cAMP-mediated, since it is duplicated by forskolin.
The CRE-like sequence appears to be a functional CRE and can bind CREB, although the binding of other CRE binding proteins as homo-or heterodimers is not excluded and is even likely. The ability of TSH to decrease CREB interactions with the silencer, in association with decreased class I transcription, suggests that CREB binding to the silencer normally functions as an enhancer of class I gene activity, in accord with its known mode of action in other genes (23,24). One possibility for the action of TSH to decrease CREB binding is to alter its homo-or heterodimer structure. This is evidenced in our separately submitted report. 3 The ability of anti-CREB to alter the CREB complex, and thereby duplicate the in vivo effect of TSH/cAMPtreatment of cells to increase the CRE-dependent formation of complex C and its component TSEP-1 suppressor, supports the important role of CREB binding. The decrease in CREB binding appears to be an important component of the TSH-increased binding of TSEP-1 and the TSH/cAMP-induced class I repressive effect. Since CREB and TSEP-1 are ubiquitous proteins, this result may be relevant to other tissues where hormones regulate the growth or function of a cell via the cAMP signal transduction system. TSH/forskolin treatment of FRTL-5 cells induce the formation of a novel complex with the class I promoter, whose existence is dependent on the 38-bp silencer and its CRE. Thus, its formation with class I promoter fragments containing 168 or 127 bp of 5Ј-flanking region is prevented by an unlabeled oligonucleotide with the sequence of the 38-bp silencer containing the CRE but not by the same unlabeled oligonucleotide wherein the CRE is mutated or deleted. We show that in the absence of TSH, the 38-bp silencer binds a multiplicity of proteins in a CRE-dependent fashion in addition to CREB: TTF-1; a Y-box protein termed TSEP-1 which is homologous to human YB-1; a single strand binding protein, SSBP-1 (or a related protein), which binds to the TSHR; and Pax-8 which binds to the TG and TPO promoters. In addition to decreasing the interaction of CREB with the silencer, TSH treatment of the cells decreases the formation of a complex which we associate with the binding of TTF-1, but retains or enhances, at least relatively, the binding of complexes associated with TSEP-1 and Pax-8 binding. At this time, we have not determined its effect on complexes associated with SSBP-1 binding, although we anticipate these might decrease, consistent with TSH/cAMP-induced decreases in SSBP-1 RNA levels and SSBP-1 complex formation with the TSHR (20).
In these experiments, we show that TTF-1 is an enhancer of CRE-dependent class I promoter activity, that TSEP-1 is a suppressor, and that Pax-8 acts as a suppressor of TTF-1-dependent enhancer activity. Additionally, we show that oligonucleotides that interact with TSEP-1, but not those interacting with Pax-8, inhibit the formation of the TSH/forskolin-induced novel complexes with the class I promoter. TSEP-1 is, therefore, the dominant component of the TSH-induced novel complex; since TSEP-1 is a suppressor, it is not surprising class I expression is decreased.
Together with previous observations, the above results can be incorporated into a reasonable model of TSH/cAMP control of class I gene expression via the novel silencer described herein. In previous experiments in FRTL-5 cells, we have shown (15,16) that TSH/cAMP decrease TTF-1 RNA levels and The CAT activities of the p(Ϫ127)CAT promoter construct were compared with that of its derivative with the CRE-deleted p(Ϫ127)⌬CRE following cotransfection into FRTL-5 cells with plasmids containing the TSEP-1, TTF-1, and/or Pax-8 cDNAs or their common vector only control, pRcCMV. In the case of the double transfections with pRcCMV-TTF-1 and pRcCMV-Pax-8 (D), the control (ϩpRcCMV) included twice the amount of pRcCMV to compensate for the pRcCMV-Pax-8. Cells were transfected as described in Fig. 1 and "Experimental Procedures" and then incubated in 5H medium plus 5% calf serum for 36 h. Conversion rates were normalized to protein and CAT activities expressed relative to the parental p(Ϫ127)CAT, which averaged 3-fold higher CAT activity than the pSV0 control chimera in these experiments. Data are the mean Ϯ S.E. for three separate experiments. Statistically significant decreases (p Ͻ 0.02) from the p(Ϫ127)CAT control are noted by one asterisk, statistically significant increases (p Ͻ 0.05) noted by two asterisks; a statistically significant decrease (p Ͻ 0.05) from the effect of pRcCMV-TTF-1 on p(Ϫ127)CAT is noted by three asterisks. that this decrease can account for a decrease in complex formation with TTF-1 binding sites. Similarly, we have previously shown (17,20) that TSH/cAMP decreases SSBP-1 RNA levels and that this decrease can account for a decrease in complex formation with SSBP binding sites. We have shown in FRTL-5 cells that TSH/cAMP, in contrast to decreasing TTF-1 complex formation with the TTF-1 sites, increases Pax-8 complex formation to sites that bind both TTF-1 and Pax-8 (15,16). Finally, we have shown in FRTL-5 cells that TSEP-1 binding can be increased by protein kinase A pretreatment of extracts from cells incubated without TSH, i.e. phosphorylation can increase TSEP-1 binding (19). This is consistent with studies of Y-box proteins in oocyte development (55).
A reasonable model to interpret the data in this and our previous reports is, therefore, as follows. The 38-bp silencer binds a multiplicity of proteins. In the absence of TSH, TTF-1 and CREB form prominent complexes and are enhancers that maximize class I expression in the cells. In the presence of TSH and its activated cAMP signal, CREB and TTF-1 binding are decreased, the former by an unknown mechanism possibly related to altered heterodimer formation, the latter reflecting a TSH-induced decrease in TTF-1 RNA and protein levels. This results in decreased maximal class I expression. We suggest that TSH/cAMP, simultaneously with decreased maximal class I expression, cause a protein kinase A-mediated phosphorylation of TSEP-1. TSEP-1 acts as a suppressor of class I activity. Pax-8 binding is preserved or relatively enhanced, thereby additionally minimizing TTF-1 enhancer binding and activity. The net result of the changes in the interaction of these proteins (CREB, TTF-1, Pax-8, and TSEP-1) with the 38-bp silencer is the formation of new complexes with the class I promoter and class I suppression. Formation of the complexes is dependent on the CRE-like element, Ϫ107 to Ϫ100 bp, and appears to reflect a dominant role of TSEP-1 and an ancillary one for Pax-8.
TSH up-regulates the function and growth of the thyroid cell; during the early stages of the cell cycle, it coordinately decreases TSHR and MHC class I gene expression (1,2,56). We have hypothesized that this is necessary to maintain self-tolerance in the face of the increase in proteins associated with TSH-induced growth and function (1, 2, 4 -6). The present data are consistent with this hypothesis. First, Pax-8 suppresses TTF-1-induced class I promoter activity, while simultaneously acting as an enhancer of TG or TPO gene expression (21,22,26). Pax-8 is, therefore, a transcription factor whose binding activity is increased by TSH/cAMP (15,16), whose activity is associated with increased thyroid function, TG synthesis, or TPO activity, and yet whose activity simultaneously helps suppress class I gene expression. Second, human Y-box homologs of TSEP-1 are able to bind to c-myc (46,47,57,58) and to other genes associated with activation of growth: the epidermal growth factor receptor and c-Ki-ras (59 -61). Sabath et al. (62) have suggested that TSEP-1 might stimulate the transcription of numerous growth-associated genes. TSEP-1, therefore, functions not only in the TSH/cAMP-induced suppression of TSHR and MHC class I, but also, simultaneously, is likely to be a positive regulator of the growth of FRTL-5 cells.
In sum, common transcription factors are involved in TSH/ cAMP-induced negative regulation of class I and the TSHR as hypothesized (1, 2, 4 -6) and in TSH/cAMP-induced positive regulation of TG and TPO regulation. The net result is crosstalk which suppresses class I and preserves self-tolerance, while allowing TSH-induced thyroid growth and function to proceed.
Y box proteins are known to suppress MHC class II gene expression and are important in cAMP-induced down-regula-tion of class II (63)(64)(65). Since aberrant class II, as well as increased class I expression, is associated with thyroid autoimmunity (3,6), coordinate suppression of class I and class II by a common transcription factor would additionally preserve selftolerance in the thyroid cell challenged with TSH (6).
There is a feature of the Y-box protein family members that may be relevant to understanding the break down of tolerance and the development of autoimmunity. Infections by hepatitis C, retro-, and foamy viruses have been associated with the appearance of thyroid autoantibodies and frank thyroid autoimmunity of various types (66 -73). All are single strand RNA viruses; foamy is a retrovirus family member with a similar long terminal repeat (74). We have identified a binding site for TSEP-1 on the hepatitis C virus upstream region, which is important in strand replication. 4 10 Studies by Kashanchi et al. (75) reveal the involvement of Y-box proteins in the transcription of human T-cell lymphotrophic virus type-1. Thus, the downstream regulatory element 1 in the long terminal repeat of human T-cell lymphotrophic virus type-1, as well as the site A region of human immunodeficiency virus, both contain Y-box elements. In addition, cotransfection of Jurkat T-cells with a YB-1 expression vector and wild type or mutant viral promoter-CAT constructs demonstrated that the Y-box sequence was essential for efficient transactivation. Kashanchi et al. (75) suggest that interleukin-2 induction of YB-1, as described earlier for T-cell proliferation, might lead to stimulation of viral gene expression and viral replication (75). We speculate that autoimmunity may be precipitated by the action of positive strand RNA viruses infecting cells, "capturing" TSEP-1, and using it, and/or other single strand binding proteins such as SSBP-1 which also regulates class I and the TSHR, to promote viral replication. This results in the loss of normal negative regulation of host genes important for self-tolerance.
Concern can be expressed that the present data and studies in thyrocytes in culture are not relevant to the whole animal. In this respect, previous work in transgenic animals has shown that downstream sequences including the silencer region can influence the pattern of tissue-specific class I expression and the ability of the gene to respond to ␥-interferon (76). The present report is consistent with the former point; our preliminary report, 5 which shows the CRE is a critical element in the ␥-interferon response, further supports this.