Thyroid Hormone Negatively Regulates the Transcriptional Activity of the β-Amyloid Precursor Protein Gene*

The expression of the β-amyloid precursor protein (APP), which plays a key role in the development of Alzheimer’s disease, is regulated by a variety of cellular mediators in a cell-dependent manner. In the present study, we present evidence that thyroid hormones negatively regulate the expression of the APP gene in neuroblastoma cells. Transient transfection studies using plasmids that contain progressive deletions of the 5′ region of the gene demonstrate that triiodothyronine (T3), the more active form of the thyroid hormones, represses APP promoter activity by a mechanism that requires binding of the nuclear T3 receptor (TR) to a specific sequence located in the first exon. The unliganded receptor increases promoter activity, and T3 reverses that activity to basal levels. The repressive effect of T3 does not exhibit TR isoform specificity, and it is equally mediated by TRα and TRβ. Gel mobility shift assays using in vitro synthesized nuclear receptors and nuclear extracts led to the identification of a negative thyroid hormone response element, at nucleotide position +80/+96, that preferentially binds heterodimers of TR with the retinoid X receptor. Insertion of sequences containing this element confers negative regulation by T3 to a heterologous TK promoter, thus indicating the functionality of the element.

The ␤-amyloid protein, the major component of the Alzheimer-associated plaques, is derived from a set of alternatively spliced ␤-amyloid precursor proteins (APP), 1 which are encoded by a single gene located on human chromosome 21 (for a review see Ref. 1). Although at physiological levels APP appears to be involved in neurotrophic events (2), its overexpression might cause neuronal degeneration by a mechanism that probably involves an increased production of ␤-amyloid protein (3) and neurotoxicity (4). APP is ubiquitously expressed in mammalian tissues, and its expression can be regulated by a variety of stimuli, including nerve growth factor (5,6), phorbol esters (7,8), or retinoic acid (9), a ligand of the nuclear superfamily of steroid/thyroid hormone receptors.
An apparent relationship between thyroid status and Alzheimer's disease has been suggested. Thyroid hormones, in particular T3, are essential for normal brain maturation and function (10), and their deficiency causes neurologic symptoms that in a way resemble those observed in Alzheimer's patients. Moreover, although a strong link between thyroid hormones and Alzheimer has not been yet established, it has been suggested that a history of thyroid dysfunction may represent a risk factor for this pathology (11,12). In addition, data from our laboratory (13) indicate that T3 affects splicing and secretion of APP isoforms in neuroblastoma cells.
Most of the effects of the thyroid hormone are mediated by binding and activation of nuclear thyroid hormone receptors (TRs). TR functions as a ligand-inducible transcription factor to increase or decrease the transcription of target genes by binding to specific DNA sequences called thyroid hormone response elements (TREs), which consist of hexameric half-sites of the consensus sequence AGGTCA arranged as palindromes or direct repeats (for a review, see Ref. 14). Unliganded TRs mediate transcriptional repression of most positive TREs due to binding of nuclear corepressors (15). Ligand binding induces a conformational change that causes the release of corepressors and the recruitment of coactivators, which bind to the AF-2 C-terminal domain (16), and allows transactivation.
The mechanisms involved in T3-dependent transcriptional repression remain less well defined. Negative response elements (nTREs) are located close to, and often downstream from, the transcriptional start site (17)(18)(19)(20). The nTREs are frequently composed of more than one TR-binding site, each of them containing sequences that partially resemble the AG-GTCA sequence described for positive TREs (17,20). TRs can bind to these sequences as monomers, homodimers, or heterodimers with RXR (17,20,21). However, up to now, a consensus sequence for nTREs has not been yet established, and the precise role of RXR on the T3-induced repression remains unclear. In addition, TRs can also negatively affect the expression of certain genes, without requiring binding to DNA, by interfering with AP-1-induced transcriptional activation (22).
In this report, we present evidence that T3 reduces APP transcripts in N2a-␤ neuroblastoma cells. Transient transfection studies with different fragments of the 5Ј region of the APP gene, together with gel mobility shift assays, show that the negative effect of T3 requires binding of TR to sequences located within the first exon of the APP gene, between positions ϩ80 and ϩ96. These sequences preferentially bind TR/RXR heterodimers, whereas monomers or homodimers are either not bound or bound with a very low affinity. Insertion of the ϩ55/ϩ102 region of the APP gene, which contains the ϩ80/ϩ96 element, confers negative regulation by T3 to a heterologous promoter. Taken together, our results reveal the existence of a * This work was supported in part by Dirección General de Investigación Científica y Técnica Grant PB93-0135, Comisión Interministerial de Ciencia y Tecnología Grant SAF 97-0183, and Fondo de Investigación Sanitaria Grant 94/0272. 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. ‡ These two authors contributed equally to this work. § Recipient of a fellowship from the Departamento de Educación y Cultura del Gobierno de Navarra.
¶ Supported by a fellowship from SmithKline-Beecham and Consejo Superior de Investigaciones Científicas.

EXPERIMENTAL PROCEDURES
Cell Culture-Murine N2a neuroblastoma cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum as described previously by Ortiz-Caro et al. (23). Previous to the experiments, the culture medium was replaced with a similar medium containing serum depleted of thyroid hormone by treatment with resin AG1X8 as described by Samuels et al. (24), and the cells were then incubated in this medium for an additional 24-h period before the beginning of the experiments. N2a-␤ cells, a subclone that constitutively expresses the ␤-isoform of TR (TR␤), were grown as described previously by Lebel et al. (25), and the experiments were carried out in the same medium containing 0.5% thyroid hormone-depleted fetal calf serum.
RNA Extraction and Hybridization-Total RNA was extracted from the cell cultures by the guanidine thiocyanate method (26). The RNA (30 g) was run in 1% formaldehyde-agarose gels and transferred to nylon-nitrocellulose membranes (Nytran) for Northern blot analysis. The RNA was stained with 0.02% methylene blue. The blots were hybridized, as described by Church and Gilbert (27), with a plasmid containing a human APP cDNA labeled by random oligonucleotide priming. Hybridizations were at 65°C in PSE buffer (0.3 M sodium phosphate, pH 7.2, 7% SDS, 1 mM EDTA). Quantification of mRNA levels was carried out by densitometric scan of the autoradiograms. The values obtained were corrected by the amount of RNA applied in each lane, which was determined by densitometry of the stained membranes.
Reporter Plasmids and Expression Vectors-The chloramphenicol acetyl transferase (CAT) reporter plasmid containing the Ϫ1099/ϩ105 fragment of the human APP gene has been previously described (28). Progressive 5Ј deletions to Ϫ487, Ϫ307, Ϫ15, ϩ55, and ϩ75 bp were prepared by polymerase chain reaction and subcloned into the BamHI site of pBL-CAT8, a plasmid that lacks the AP-1 binding site present in the pUC backbone. The cDNAs encoding TR␣, RXR␣, and v-ErbA as well as the cDNAs encoding TR mutants E401Q, E401K, and C-1 are inserted into the EcoI site of the expression vector pSG5, which contains the SV40 early promoter (29). The reporter construct TRE APP /TK-CAT consists of a single copy of the ϩ55/ϩ102 fragment of APP inserted in front of a TK promoter driving the expression of the CAT gene.
DNA Transfection-N2a cells were transfected by the calcium phosphate coprecipitation method with 1 g of reporter plasmids and carrier DNA. One hundred nanograms of a luciferase reference vector was simultaneously used as an internal control of the transfection efficiency. In cotransfection experiments, 1 g of reporter plasmid and 1 g of the corresponding receptor expression vector were used. After 16 h of incubation in the presence of calcium phosphate, a new medium containing 0.5% thyroid hormone-depleted serum was added, and the cells were then incubated for an additional period of 48 h in the presence or absence of 5 nM T3. Each treatment was performed in duplicate cultures that normally showed less than 5-15% variation in CAT activity, which was determined by incubation of [ 14 C]chloramphenicol with cell lysate protein. Each experiment was repeated at least two or three times with similar relative differences in regulated expression.
Mobility Shift Assays-Synthetic oligonucleotides containing the TRbinding sequences of the human APP promoter, were end-labeled with [ 32 P]ATP using T4-polynucleotide kinase and then incubated with in vitro translated receptors or with nuclear extracts obtained from N2a-␤ cells. cDNAs for TR and RXR in pSG5 were transcribed and translated in vitro with the TNT kit (Promega) following the manufacturer's recommendations. The nuclear extracts were obtained by the method of Andrews and Faller (30). For gel retardation assays, translated receptors (2 l) or nuclear extracts (5 g) were incubated on ice for 15 min in a buffer (20 mM Tris HCl, pH 7.5, 75 mM KCl, 1 mM dithiothreitol, 5 g/ml bovine serum albumin, 13% glycerol) containing 3 g of poly(dI-dC) and then were incubated for 15-20 min at room temperature with approximately 70,000 cpm of the double-stranded labeled oligonucleotide. Unprogrammed reticulocyte lysate was used as a control for nonspecific binding. For competition experiments, increasing concentrations of unlabeled double-stranded oligonucleotide or an oligonucleotide containing the consensus sequence TRE PAL (5Ј-AGGTCATGACCT-3Ј), were added to the binding reaction mixture. For gel supershift, the reaction mixtures were incubated with 1 l of specific anti-TR or anti-RXR antibodies for 30 min at 4°C. DNA-protein complexes were resolved on 7% nondenaturing polyacrylamide gels containing 0.5% TBE buffer. The gels were dried and autoradiographed at Ϫ70°C.

Negative Regulation of the APP mRNA Levels by Thyroid
Hormone in N2a-␤ Cells-APP mRNA levels were determined in N2a-␤ cells, which express high levels of TR␤ (25), after treatment with 5 nM T3. Fig. 1 illustrates the results obtained in a representative Northern blot carried out with 30 g of total RNA. A single band of 3.4 kilobases, which corresponds to APP mRNA was detected. Densitometric scanning of the bands showed that T3 decreased APP mRNA levels in a time-dependent manner. The negative effect of T3 was already detected at 24 h and became more evident after 48 h of treatment. No effects of T3 on APP mRNA levels were observed in parental N2a cells, which express very low levels of TR (23) (data not shown).
TR-mediated Repression of APP Promoter Activity-Transient transfection assays were carried out to determine whether or not T3 affects the transcriptional activity of the APP gene in neuroblastoma cells. N2a-␤ cells were transiently transfected with a chimeric plasmid containing the Ϫ1099 to ϩ105 bp fragment of the human APP gene linked to the CAT reporter gene and then incubated for 48 h in the presence or absence of 5 nM T3. As shown in Fig. 2 (left panel), CAT activity was significantly inhibited in N2a-␤ cells incubated in the presence of T3, whereas this hormone did not affect the activity of the APP gene promoter in the parental N2a cell line.
To confirm the role of TR in this negative response, the unresponsive parental N2a cells were co-transfected with the APP promoter-CAT construct and a vector expressing the ␣-isoform of the thyroid hormone receptor (TR␣). As shown in Fig. 2 (right panel), the unliganded TR increased basal promoter activity, and T3 effectively reversed this effect. The negative effect of T3 in N2a cells transiently transfected with TR␣ was very similar to that observed in N2a-␤ cells, thus showing that the inhibitory effect of T3 does not exhibit TR isoform specificity.
Identification of DNA Regions Mediating the Negative Regulation of APP Transcriptional Activity-To map the DNA sequences of the 5Ј-flanking region of the human APP gene involved in the T3-induced response, progressively deleted fragments (Ϫ1099, Ϫ487, Ϫ307, and Ϫ15) of the promoter were linked to the upstream region of the reporter CAT gene and transfected into N2a-␤ cells. As shown in Fig. 3, the negative effect of T3 was maintained even in cells transfected with the shortest (Ϫ15/ϩ102 bp) fragment, which lacks the AP-1 sites. Similar results were obtained in N2a cells transfected with TR␣ (data not shown). These results strongly suggest the existence of a TR-binding site in the Ϫ15 to ϩ102 bp region of the gene.
The Repressive Effect of T3 Requires the AF-2 Domain of the Thyroid Hormone Receptor-Ligand-dependent transactivation function of the TR is associated with an autonomous and highly conserved C-terminal region of the receptor referred to as AF-2 (29). To determine whether this transcriptional domain could also play a role in the ligand-dependent repression of the APP promoter, we examined the response to T3 in N2a cells transfected with v-ErbA, a natural AF-2 mutant of TR␣ that fails to bind ligand, as well as with several TR␣ mutants affecting the AF-2 domain. The C-1 mutant carries a 9-amino acid C-terminal deletion like that found in v-ErbA, and the E401Q and E401K mutants contain a point mutation of the Glu (E) residue at position 401. As illustrated in Fig. 4, the wildtype receptor and the AF-2 mutants increased with a similar potency the activity of the Ϫ1099/ϩ105 APP promoter construct in the absence of ligand. T3 effectively reversed the promoter activation induced by the wild-type TR, but it was unable to reduce significantly the constitutive induction caused by the AF-2 mutant receptors. The C-1 and E401K mutants show a strongly reduced ligand binding affinity (29), which, as in the case of v-ErbA, could explain the lack of response to T3. However, the effect of T3 was also abolished in the E401Q mutant, in which ligand binding affinity remains unaltered (29).
Identification of TR-binding Elements in the Ϫ15 to ϩ102 region of the APP Gene-For a more detailed study of the DNA sequences involved in this T3-induced response, we further analyzed the Ϫ15/ϩ102 bp region of the gene. As depicted in Fig. 5 (upper panel), a computer-assisted study of the nucleotide sequence of this region revealed the existence of three potential thyroid hormone response elements located at the nucleotide positions Ϫ4 to ϩ9 (E1), ϩ20 to ϩ30 (E2), and ϩ80 to ϩ96 (E3). The first motif overlaps the major transcriptional start point, and the other two, E2 and E3, are located within the first exon of the gene.
To determine whether the potential response elements of the APP gene are able to bind TR, we conducted gel mobility shift assays with in vitro translated receptors and oligonucleotides containing the E1, E2, and E3 sequences. As illustrated in Fig.  5, no specific retarded bands were detected when TR or RXR were used separately, and only the oligonucleotide (ϩ75/ϩ101) containing the E3 motif was able to specifically bind TR/RXR heterodimers. No detectable complexes were established between the heterodimer and the probes containing the E1 (Ϫ10/ ϩ15) or E2 (ϩ13/ϩ37) sequences.
To further analyze the E3 domain, the only motif that effectively bound TR/RXR, we performed new gel mobility shift assays, using both in vitro translated receptors and nuclear extracts obtained from control and T3-treated N2a-␤ cells. 6A shows binding of the in vitro translated protein preparations. A specific band running in a position that is compatible with a mobility complex containing the heterodimer TR/RXR was detected. In agreement with the results observed in Fig. 5, bands corresponding to monomers or homodimers of TR or RXR were not detected. TR/RXR binding to the oligonucleotide containing the E3 motif appeared not to be significantly affected by T3, which only induced a slight increase in the mobility of the retarded band. The retarded band was competed by the unlabeled E3-containing oligonucleotide as well as by the thyroid hormone consensus response element TREpal. No competition was observed when an unrelated oligonucleotide was used. Fig.  6B shows the results of a representative gel mobility shift assay carried out with nuclear extracts of N2a-␤ cells. Several retarded bands were detected. The most prominent band presented a mobility identical to that obtained with the in vitro translated TR/RXR heterodimer (lane 3). The mobility of this band was again slightly increased in the presence of T3. This band was specifically competed by an excess of both unlabeled E3 oligonucleotide or TREpal but not by an unrelated oligonucleotide. The presence of TR and RXR in the complex was tested with specific antibodies against both receptor proteins. The retarded band was inhibited by the TR antibody (Fig. 6, ␣TR) and supershifted by the RXR antibody (␣RXR). These data show that the E3 element binds the endogenous TR/RXR heterodimers present in N2a-␤ cell nuclei. In addition to the major TR/RXR-containing complex, nuclear extracts from N2a-␤ cells caused the formation of a minor retarded band of slower mobility. This unidentified band contains neither TR nor heterodimers of RXR with other nuclear receptors, since it was unaffected by the antibodies.
Functional Analysis of the E3 Element-The nucleotide sequence of the E3 element contains three motifs ϩ80/ϩ85, ϩ85/ ϩ90, and ϩ91/ϩ96, each resembling the consensus core sequence AGGTCA established for the positive TREs. To confirm the functionality of this potential nTRE, we studied whether this element is able to confer T3 responsiveness to a heterologous nonresponsive promoter. For this purpose, N2a cells were transiently cotransfected with a TK-CAT construct or the TRE APP /TK-CAT (a chimeric plasmid containing the ϩ55 to ϩ102 bp fragment of the human APP gene linked to the TK- FIG. 5. Gel retardation analysis of TR␣, RXR, and TR␣/RXR binding to nucleotide sequences of the ؊15 to ؉102 region of the APP promoter. A, nucleotide sequence of the Ϫ15 to ϩ102 bp region. Potential TR-binding sites are indicated by boldface type. Nucleotide identity with the consensus sequence (AGGTCA) is indicated by small black squares above the letters. The arrows indicate the position and orientation of potential half-sites. The oligonucleotides used in the gel retardation assays (Ϫ10/ϩ15, ϩ13/ϩ37, and ϩ75/ϩ101) are also indicated. B, the results obtained in a representative assay carried out with in vitro translated TR and RXR. The first and second lanes of each group are controls obtained with the corresponding unlabeled probe or lysate alone, respectively.
FIG. 6. Characterization of the E3 TR-binding site of the APP gene. In the upper panels, gel mobility shift assays were carried out with a radiolabeled probe (ϩ75/ϩ101) containing the binding site and surrounding nucleotides and in vitro translated TR and RXR (2 l). When indicated, the proteins were incubated with 200 nM T3. A 5-and 50-fold excess of the unlabeled probe, the palindromic element TREpal, or an unrelated oligonucleotide was used for competition studies. In the lower panels, the assays were performed with the labeled oligonucleotide and 5 g of nuclear extracts obtained from control N2a-␤ cells and from cells treated with T3. The composition of the retarded bands was tested by incubation with 1 l of anti-TR (␣TR) or anti-RXR (␣RXR) antibodies. Competition experiments were carried out in the presence of a 50-fold excess of the same oligonucleotides used in the upper panels. The mobilities of the TR/RXR heterodimers as well as of a band not identified (n.i.) are indicated by arrows.
CAT reporter gene) and an expression vector for TR␣. As shown in Fig. 7, the activity of the TK promoter was not affected by TR either in the presence or in the absence of T3. In contrast, the insertion of the APP promoter fragment into the upstream region of the construct conferred T3 responsiveness to the TK promoter. The regulation was identical to that found with the APP promoter, where unliganded TR significantly increased CAT activity and T3 reverted the increased activity to the uninduced basal levels.

DISCUSSION
In this report, we present evidence that T3 negatively regulates APP gene expression in a rat neuroblastoma cell line. The repressive effect of T3 was observed in N2a-␤ cells, a subclone that constitutively expresses high levels of TR␤, but not in the parental cell line, which contains low levels of thyroid hormone receptors. This suggests that the negative effect of T3 on the APP gene will occur specifically in cells and brain regions expressing high TR levels.
Transient transfection studies demonstrated that T3 represses APP promoter activity in N2a-␤ cells. Several mechanisms, involving either binding to specific nTREs or interference with other positive transcription factors, in particular the AP-1 complex (22), have been described as mediating negative regulation of gene expression by T3. The APP promoter has the typical structure of a housekeeping gene, lacking the TATA and CAAT elements but containing two AP-1 sites, which are located upstream of the transcription start point, at nucleotide positions Ϫ345 and Ϫ38 (31)(32)(33)(34). Our results show that the progressive removal of the AP-1 sites did not significantly affect the repressive effect of T3 on the APP promoter, thus ruling out the participation of a mechanism involving transcriptional antagonism between TR and the AP-1 complex. These results, together with the requirement for TR, suggested a mechanism in which the negative effect of T3 could be directly mediated by receptor binding to specific DNA sequences.
Negative TREs are normally located in the proximal promoter or even within the first exon, downstream of the transcriptional start site. The finding that T3 effectively repressed the activity of a minimal APP promoter fragment suggested the existence of a negative TRE in the proximity of the main transcriptional start site. Sequence analysis of the promoter region located between nucleotides Ϫ15 and ϩ102 revealed the presence of three potential TR binding sites. The promoter fragments Ϫ4/ϩ9 and ϩ20/ϩ30 contain sequences that strongly resemble the previously established AGGTCA consensus TRE half-site. However, analysis of the binding properties of these elements revealed that they are unable to bind TR or TR/RXR. Only the sequence located between nucleotides ϩ80 and ϩ96 binds the receptor significantly. Therefore, although the possibility that the more upstream sequences could contribute to modulate TR binding to the more downstream element cannot be dismissed, our results strongly suggest that the negative effect of T3 on the APP gene might be mediated by a unique nTRE present within the first exon at positions ϩ80/ϩ96.
The putative nTRE of the APP gene contains at least two hemisites, each resembling the consensus TR binding motifs. A central core of nucleotides might be interpreted as a third hemisite or alternatively as a spacer. It is well known that orientation and spacing as well as the primary nucleotide sequence of core DNA-binding motifs strongly contribute to dictate the selective positive or negative transcriptional actions induced by the ligand-dependent family of transcription factors (35). The APP element binds heterodimers of TR with RXR but does not bind TR monomers or homodimers. This result is different from that obtained with other negative responses to T3, where the monomeric TR forms appear to play an essential role (17,18,20). The preferential binding of TR/RXR heterodimers firmly supports a structure containing two hemisites spaced by five or, more likely, four base pairs. In both cases the downstream hemisite (ϩ91/ϩ96 bp) is identical to the sequence found by Desvergne et al. in the rat malic enzyme TRE (36). The first hemisite, in turn, can be defined either at nucleotides positions ϩ80/ϩ85 (5-bp spacer model) or ϩ81/ϩ86 (4-bp spacer), both resembling the consensus sequences described in the rat growth hormone TRE by Brent et al. Binding of unliganded receptors to nTREs can lead to a constitutive activation of gene transcription (39 -41). In N2a cells, unliganded TR␣ stimulates APP promoter activity, and as occurs with a number of negatively regulated genes (17,20,41), T3 reverses the activation induced by the transfected unliganded receptor. The molecular mechanisms responsible for constitutive activation are still unknown. However, it has been recently reported that unliganded TR might stimulate the basal activity of negatively regulated promoter in a manner that requires the association of corepressors (42). The addition of T3 might reverse basal activity, perhaps by dissociation of those corepressors. Experiments with AF-2-defective mutants show that this receptor region is required for the T3-dependent repression of the APP promoter in N2a cells. A point mutation, E401Q, which affects T3-dependent activation without altering T3 binding (29), severely impaired the repressive activity of T3 on the APP promoter. These data indicate that the AF-2 function is required not only for ligand-dependent stimulatory responses mediated by positive response elements but also for the repressive effect of T3 on the APP promoter. Since the AF-2 region appears to be mainly involved in the binding of coactivators (43), our results strongly suggest that these proteins might play a role in the regulation of the APP gene by T3. In contrast, TR mutants lacking the AF-2 function activated transcription in a ligand-independent manner as efficiently as wildtype TR, indicating that the C-terminal region of TR is dispensable for T3-independent transactivation of the APP promoter and that different domains are required for the ligand-dependent and ligand-independent actions of TR on this promoter. Finally, our data show that the TR/RXR-binding element of the APP promoter was able to confer negative regulation by T3 to the heterologous thymidine kinase promoter. Furthermore, this element also conferred stimulation by the unliganded receptor. These results confirm the functionality of the binding element and strongly suggest that this nTRE alone is sufficient to mediate both ligand-dependent and ligand-independent actions of the thyroid hormone receptor on the APP gene.
In summary, this report describes a negative regulation of APP gene expression by the thyroid hormone in neuroblastoma cells. This effect requires a nTRE, which binds TR/RXR heterodimers, located in the first exon downstream of the main transcriptional start site and is equally mediated by TR␣ and TR␤. In addition to the repressive effect of the thyroid hormone on APP mRNA levels and APP promoter activity, we have previously demonstrated that thyroid hormones specifically alter the pattern of intracellular and secreted APP isoforms (13). Our results together with the reduction of thyroid hormone receptor levels observed in Alzheimer hippocampal cells (44) and the effects exerted by estradiol or retinoic acid on APP expression and metabolism (9,45) strongly suggest that members of the nuclear superfamily of receptors and their ligands might play an important role in Alzheimer's disease.