INTRODUCTION

Depolarization of skeletal muscle triggers the release of Ca²⁺ from the sarcoplasmic reticulum (SR),¹ activating the contractile machinery. Relaxation requires the re-uptake of Ca²⁺, and this is mediated by the SR Ca²⁺-ATPase (EC 3.1.6.38). Three different genes (SERCA1, -2, and -3) encode multiple isoforms of this Ca²⁺ pump (1, 2, 3, 4) of which SERCA1 and SERCA2a are the predominant isoforms found in skeletal muscle (5, 6, 7, 8, 9, 10). SERCA1 is expressed at a level up to 8-fold higher than SERCA2a (8, 11, 12, 13, 14) and is consequently part of the fast phenotype of skeletal muscle fibers, characterized by high rates of relaxation and contraction. In contrast, SERCA2a is the predominant isoform found in slow skeletal muscle fibers (7, 9, 10).

It is now well established that thyroid hormone (T₃) plays an important role in the development and maintenance of the fast phenotype. During neonatal development of fast muscle, T₃ stimulates a massive increase in SERCA1 expression (15), as well as the expression of fast isoforms of a number of contractile proteins (16, 17, 18, 19). Furthermore, in adult slow skeletal muscle sufficient levels of T₃ can induce a shift from a slow to a fast phenotype (7, 20, 21, 22, 23). For example in rat soleus muscle, T₃ stimulation (hypo- to hyperthyroidism (13, 14, 24, 25, 26, 27)) produces a 7-fold increase in total SR Ca²⁺-ATPase, almost all due to induction or stimulation of expression of SERCA1 in all fibers of this muscle (5, 7). Therefore, the SERCA1 isoform is the primary target for T₃ in the development of fast relaxation properties in skeletal muscle. Subsequent work indicated that T₃ acts at a pretranslational level, both in neonatal development and in adult muscle (5, 28), and a direct action of T₃ was indicated by studies using the skeletal muscle cell line L6 (29, 30). Finally, run-on analysis established that T₃ enhances the rate of transcription of the SERCA1 gene (31).

Stimulation of transcription by T₃ is mediated by nuclear T₃ receptors (TR) (32, 33, 34, 35) that recognize and bind specific DNA sequences called thyroid hormone response elements (TRE) (36, 37, 38). TREs act as enhancers and will confer T₃ responsiveness when inserted in a heterologous promoter. Unlike the target sequences of other trans-acting factors, naturally occurring TREs are remarkably variable in structure. Analysis of the TREs in several promoters has yielded the consensus hexamer element (A/G)GGT(C/G)A as the optimal binding site for TR, although specific flanking nucleotides further increase affinity (38, 39, 40, 41, 42, 43). Naturally occurring TREs typically consist of multiple elements (half-sites) arranged as either a direct repeat spaced by 4 base pairs (DR + 4), as an inverted repeat spaced by 0 (palindrome), 1, or 2 base pairs (IR + 1/+ 2), as an inverted palindrome with a 6- or 7-base pair gap (IP + 6/+ 7), or as a combination of these in TREs composed of three or four elements (38, 44, 45, 46, 47, 48, 49). Furthermore, a promoter may contain several TREs contributing to the overall T₃ responsiveness (50, 51, 52). The multiple half-site arrangement within a TRE allows for cooperative receptor binding and dimerization or multimerization, which positively correlates with transactivation (44, 53, 54). Deviations from the optimal hexamer sequence and, to a lesser extent, from the optimal spacing between them are common in naturally occurring TREs and are responsible for large differences in TR binding affinity and transactivation potency (44). Depending on the structure of a TRE, TR binding and transactivation can be increased by heterodimerization with a class of apparently ubiquitous nuclear proteins of which the retinoid X receptor (RXR) is a major representative. RXR augments TR binding to most TREs and potentiates T₃ induction with no requirement for its natural ligand (9-cis-retinoic acid) (55, 56, 57, 58, 59, 60). Finally, interactions of other transcription factors with single TRs bound to single half-sites have been described, further increasing the potential of T₃ to modulate gene transcription (55, 61).

We present an analysis of the first 962 base pairs of the 5-upstream region of the rat SERCA1 gene, which constitutes a functional, T₃-inducible promoter. The T₃ responsiveness appears to be mediated by two weak and one strong TRE. A detailed analysis of the strong TRE indicates an unusual array of half-sites, four of which are required for optimal transactivation. This complex arrangement has important implications for T₃ gene regulation.

MATERIALS AND METHODS

Isolation and Characterization of the 5-Flanking Region of the SERCA1 Gene

Unless otherwise indicated all molecular and biological assays were carried out using standard methods (62). A rat (Sprague-Dawley, adult) genomic library in EMBL-3 (Clontech), consisting of 2.3 × 10⁶ independent clones, was screened with the ApaI/KpnI fragment from rabbit SERCA1 cDNA (nt 21 to +635) (gift from Dr. Jonathan Lytton, Brigham and Women's Hospital, Boston, MA). The fragment was labeled by random priming in the presence of [-³²P]dATP, and a screening of approximately 1 × 10⁶ clones yielded two independent clones with inserts of 14 and 15 kb. Secondary screening of the clones using -³²P-end-labeled synthetic oligonucleotides corresponding to the 5-nontranslated sequences of the cDNAs of rat SERCA2a (5-GAGTGCAGGCGGAGGCGAGGAGGC-3) (63) and the non-muscle Ca²⁺ pump isoform SERCA3 (5-TCAAGTTCGCAGCATTCTGCACAGT-3) (1) proved negative. The clones were characterized by Southern analysis of restriction fragments using various rat SERCA1 cDNA fragments as probes and an oligonucleotide complementary to the 5-nontranslated sequence of the SERCA1 cDNA (5-GGGGGTGATGTATTCCTTCTTA-3). A 3.2-kb SacI-SacI fragment that hybridized to both a 5-cDNA fragment (65 to +118 (SacI), +1 is the translation initiation site) and the oligonucleotide was cloned into pTZ19U (pTZ19U/Sac3.2) and analyzed by restriction enzyme digestions. A 3 SphI-SacI fragment of 1.5 kb was subcloned and sequenced on both strands by the dideoxy chain termination method using synthetic oligonucleotides as primers.

A synthetic oligonucleotide (5GGGGGTGATGTATTCCTTCTTA-3) derived from the 5-noncoding sequence of the rat SERCA1 cDNA, 44 nt upstream of the translation initiation site, was used in the primer extension assay. The primer was end-labeled with [-³²P]dATP and T₄ kinase (Pharmacia Biotech Inc.) and hybridized at 55 °C overnight to approximately 15 µg of total RNA, isolated from gastrocnemius muscle of 6-week-old rats by the guanidinium thiocyanate method (64). In this muscle, SERCA1 is expressed at a high level and its message comprises over 95% of total SR Ca²⁺-ATPase mRNA (5). Primer extension using reversed transcriptase (Life Technologies, Inc.) was performed for 1 h at 37 °C in the presence of actinomycin D and analyzed on a 5% denaturing polyacrylamide sequencing gel.

The probe used in the mung bean nuclease method was synthesized from a single-stranded M13 DNA template containing an approximately 900-nt RNA-like PstI (335 nt from the ATG start site) -SacI fragment subcloned from the SERCA1 genomic clone pTZ19U/Sac3.2 (see above). The above-described synthetic oligonucleotide was used as a primer in the synthesis of the complementary DNA strand in the presence of [-³²P]dATP. After digestion with PstI, the single-stranded primer PstI fragment was separated on a 5% denaturing polyacrylamide gel. The position of the fragment was determined by autoradiography, and the fragment was cut from the gel and electro-eluted. Of this fragment, 2-4 × 10⁵ cpm was used as a probe and hybridized overnight with approximately 15 µg of total RNA at 40 °C in 70% formamide. The nuclease reaction was performed using 30, 100, and 300 units of mung bean nuclease (Promega) for 2 h at 37 °C, and fragments were analyzed on a 5% denaturing polyacrylamide sequencing gel.

Seven fragments with consecutive 5-ends at 962, 584, 490, 327, 269, 220, and 141 and a common 3-end at +91 (+1 is the transcription initiation site) were obtained from pTZ19U/Sac3.2 using available restriction sites in combination with polymerase chain reaction using synthetic oligonucleotide primers with attached restriction sites. These were cloned upstream of the chloramphenicol acetyltransferase (CAT) gene in the polylinker site of the promoterless expression vector pOCAT2 (65). The constructs were sequenced to check the fidelity of the polymerase chain reactions. Synthetic double-stranded oligonucleotides with either BamHI or XbaI ends and comprising wild type or mutated TRE sequences were inserted in the appropriate restriction site upstream of the thymidine kinase promoter in the CAT expression vector pUTKAT3 (65). Transfections of COS7 cells using the calcium-phosphate precipitation method were carried out as described previously (66). Transfections were done in pairs, and a mix for two plates (6 cm diameter) included 20 µg of CAT expression plasmid and 6 µg of pTKGH or pCMV-Gal. The latter two plasmids constitutively express human growth hormone and -galactosidase, respectively, and were used to correct for differences in transfection efficiency between plates. Where appropriate, expression plasmids (CDM8) for mouse TR1 and human RXR (gift of Drs. D. Mangelsdorf and R. Evans, Salk Institute, La Jolla, CA) or mouse RXR (gift from Dr. Keiko Ozato, NIH, Bethesda, MD) were present at 1 µg. Cell culture medium with or without 50 nM T₃ was made with charcoal-stripped serum. CAT activities (overnight incubations) were determined with the phase extraction method (67) and expressed relative to the level of human growth hormone or -galactosidase activity. The presented T₃ induction ratio's are the mean CAT/human growth hormone (or Gal) levels of T₃-treated to untreated plates, performed in duplicate, of at least four independent transfections. Statistical significance of differences between groups was calculated using Student's two-sided t test.

Gel shift experiments were carried out with purified chicken TR (cTR). The isolation of receptor following overexpression in Escherichia coli BL21 DE3 pLYSs and conditions for gel shift experiments have been described before (53). The double-stranded oligonucleotides comprising TREs, or fragments thereof, used in the construction of pUTKAT3 plasmids were labeled by Klenow fill-in reaction using [³²P]dTTP and gel-purified. Labeled probe (15-25,000 cpm, 4.5-7.5 fmol) was incubated with cTR (15-200 fmol) in 30 µl containing 100 ng of poly(dI-dC), 88 mM KCl, 10% glycerol, 25 mM Tris-HCl, 500 µM EDTA, 0.05% Triton X-100, 10 mM -mercaptoethanol, 20 nM T₃, and 5 µg of bovine serum albumin. The binding reactions were incubated for 30 min at room temperature and analyzed at 4 °C on a 5% nondenaturing polyacrylamide gel in low ionic strength buffer (10 mM Tris-HCl, 7.5 mM glacial acetic acid, and 40 mM EDTA, pH 7.8) and electrophoresed at 500 V in the same buffer with constant circulation. Gels were dried under vacuum and autoradiographed for 5-15 h. Receptor binding was quantified by laser densitometry (Pharmacia) of autoradiographs exposed for various periods. The assignment of monomer and dimer bands for a given TRE (fragment) was done by comparison with the characterized patterns of shifts of rGH or lysozyme silencer TREs run on the same gel. The fraction of TRE bound by TR monomer, dimer, or oligomer was calculated by dividing the density of a given band by that of the sum of all retarded bands and free probe. The data shown are representative of at least two independent experiments.

RESULTS

Isolation and Characterization of the 5-Flanking Region of the SERCA1 Gene

A rat genomic DNA library in EMBL3 was screened with a rabbit SERCA1 cDNA fragment, as described under "Materials and Methods," and two independent clones (8 and 17) with 14- and 15-kb SERCA1 inserts were isolated. These inserts overlapped significantly, with the most 5-situated insert (8) containing greater than 6 kb of sequence upstream of the transcription initiation site (Fig. 1A). A 1.5-kb SphI-SacI fragment was subcloned and sequenced. Comparison with rat cDNA and rabbit SERCA1 genomic sequences (68) revealed that this fragment contained part of the first intron, the first exon, and 962 base pairs of sequence upstream of the transcription initiation site.

Characterization of the 5-flanking sequence of the rat SERCA1 gene.

The rat SERCA1 5-genomic sequence from 962 to +106 (sequenced on both strands) is shown (Fig. 1B). The transcription initiation site, or cap site of the mRNA, was mapped to a thymidine at 184, rather than to the guanidine two positions downstream as reported for the rabbit (68). Mapping was performed by primer extension analysis using a primer derived from the 5-noncoding sequence, as well as by mung bean nuclease mapping using the same primer (see "Materials and Methods"). The product of a sequence analysis of the genomic fragment, using the 5-noncoding primer, was run alongside the product of both the primer extension and the nuclease mapping in an electrophoretic gel. This allowed unequivocal identification of the T-residue at 184 as the 5-end (Fig. 2), and adds 18 bases to the previously published 5-untranslated region of the rat SERCA1 message (6). The first 200 base pairs of the rat SERCA1 5-flanking region were 90% identical to the same region in the rabbit gene (68). A TATA-like sequence (CATAA) was found at position 28 and a CCAAT box at position 78. Further sequence inspection revealed a possible NF-1 binding site (TGG(N)₇CCA) at position 192, several E-boxes (CANNTG), which bind myogenic factors like MyoD and myogenin (reviewed in Ref. 69), and five TR-binding hexamers corresponding to the consensus sequence (A/G)GGT(C/G)A (Fig. 1B).

Mapping of the 5-end of rat SERCA1 mRNA.

A series of fragments with progressive 5-deletions and a common 3-end (at position +91), comprising between 962 and 141 base pairs of 5-sequence (see Fig. 1B), were cloned upstream of the chloramphenicol acetyltransferase (CAT) gene in the promoterless expression vector pOCAT2, placing the CAT gene under control of the SERCA1 promoter fragment. COS cells were used in the subsequent transient transfection assays, since these cells constitute a well characterized and sensitive system for the analysis of T₃ responsiveness. Transfection assays indicated that the minimal construct (141), containing the CCAAT and TATA box, constitutes a functional promoter as it directed 4-fold higher CAT expression as compared with the pOCAT2 control plasmid. Constructs including additional 5-sequence displayed similar or somewhat lower activities. None of the constructs showed T₃ responsiveness in the functionally TR-deficient COS cells. Co-transfection of a mouse TR expression plasmid, however, induced T₃ responsiveness while reducing basal CAT expression. We observed an average 65% reduction of basal CAT activity in all constructs with co-transfection of TR, except for the shortest (141), where there was 37% inhibition (pOCAT2 basal activity was not affected by TR co-transfection). Basal activity of T₃ responsive promoters has previously been shown to be repressed by binding of unliganded TR (70, 71). CAT expression driven by the full-length promoter construct (962) was stimulated almost 4-fold by 50 nM T₃ and deletion to 584 did not significantly reduce T₃ induction. However, T₃ responsiveness was reduced in three discrete steps upon deletion of sequences between 584 and 490, between 269 and 220, and between 220 and 141 (Fig. 3). The low residual T₃ induction of the 141 construct of 1.27 was not significantly different from the pOCAT control.

We inspected the sequences in the fragments of the 5-flanking region that conferred T₃ induction (Fig. 3). All three fragments contained regions that closely matched known TR-binding sequences with spacing configurations found in functional TREs. These three regions (R1, R2, and R3) are depicted in Fig. 4, and some putative half-site hexamers forming DR, IR, or IP structures are indicated by arrows. We tested TR binding to the three regions in gel-shift assays. The specificity of binding was confirmed by competition for bands with excess unlabeled oligo but not with nonspecific DNA (-actin). Fig. 5A shows specific receptor binding for each region, with prominent TR homodimer bands on the R1 and R2 elements. R2 also shows multiple higher order oligomers, and these structures are the predominant species in R3. For both R2 and R3, double-stranded probe is shifted at this receptor concentration, yet a significant amount of R1 probe is not, suggesting a lower receptor affinity for this sequence.

Binding of TR to putative TRE's R1, R2, and R3.

To obtain a better estimate of the relative affinity for TR of the three regions, labeled R3 probe was incubated with TR in the presence of increasing amounts of unlabeled competitor probe, either R3, R2, or R1. The results depicted in Fig. 5B suggest that R2 is a stronger competitor than R3, whereas R1 is less potent (note the disappearance of the dimer band). The order of receptor affinity of the three regions is therefore R2 > R3 > R1.

We then tested whether the R1, R2, and R3 elements conferred T₃ responsiveness to a heterologous promoter. Each region was cloned upstream of the viral thymidine kinase promoter driving the CAT gene in pUTKAT3. Fig. 4 summarizes the T₃ induction ratios of these constructs. The pUTKAT3 control displayed a 2-fold background induction in agreement with previous observations (44, 66). Insertion of the R1 or the R2 elements resulted in a significant, yet small increase in T₃ responsiveness to 3-fold. In contrast, R3 conferred an almost 6-fold T₃ induction ratio. The order of potency of transactivation of the three regions is therefore R3  R2 = R1.

The most potent of the three TREs, R3, was analyzed in more detail. To identify the half-sites that are essential for maximal activity of R3, we first tested the receptor-binding characteristics and transactivating properties of R3 and six partially overlapping fragments of this element. Fig. 6 lists R3 and its fragments which were named according to their half-site composition. It should be noted that the six half-sites A-F were chosen as likely receptor binding sites but that additional hexamer sequences are possible. Analysis of the affinity of TR for R3 showed that at the lowest receptor concentration (15 fmol) approximately 50% of the R3 probe (ABCDEF) was bound and migrated mostly as a dimer complex (Fig. 7, panel A). Some higher order complexes were already visible, and these increased in intensity at higher concentrations of TR where three oligomeric species could be discerned. The relative contribution of the various complexes to the total amount of probe bound is shown in Fig. 7, panel B. Similar analysis of the 5-half of R3, comprising the putative half-sites ABCD, indicated high affinity dimer binding but no oligomeric species, even at the highest concentration of TR (Fig. 8. panel A). Further truncation of this sequence, leaving BCD, greatly reduced receptor affinity, but dimers were still formed. Analysis of dimer binding to the ABCD and BCD fragments is depicted in panel B of Fig. 8 showing the cooperative nature of binding (sigmoidal curves) and the higher affinity with inclusion of half-site A. These results suggest that the IR + 2 formed by B and D allows for low affinity dimer binding. High affinity receptor binding is supported by the 5-end of R3, possibly by dimer formation on the IP + 6 formed by A and C. However, alternative dimer complexes, possibly involving the ultimate 5-hexamer sequence of R3, could not be ruled out. The T₃ induction ratio conferred by fragments ABCD and BCD was the same and significantly lower than R3 (Fig. 6). Both truncated R3 sequences can therefore function as a TRE, but the 3-half of R3 is clearly required for maximal activity.

Binding of TR to Region 3. 

Binding of TR to fragments of Region 3.

Analysis of receptor binding to the downstream half of R3 (fragment EF) showed low affinity but cooperative dimer formation (Fig. 8, panels A and B). Half-sites E and F are most likely responsible for the observed binding of receptors, and the DR + 2 formed by E and F constitutes a functional TRE conferring a 3.5-fold T₃ induction ratio (Fig. 6). R3 therefore consists of an upstream and a downstream TRE contained in fragments ABCD and EF, respectively. Analysis of the transactivating activity of fragment BCDEF (Fig. 6) showed that the sequence 5 of half-site B is essential for maximal transactivating activity. Fragments that partially overlap both domains did not confer T₃ responsiveness above background (fragments DE and CDE, Fig. 6).

A mutational analysis was then done to test which of the putative half-sites are required for transactivating activity of the complete R3 element. The core G residues in TR-binding half-sites have been shown to be essential receptor contact points, and mutation of these abolishes binding and functional activity (39, 40, 42, 43, 45, 52). The data in Fig. 9 show that such mutations in either hexamer A (mut-A) or C (mut-C) abolished the function of the upstream domain, i.e. the remaining T₃ induction ratio was not different from that conferred by fragment EF (Fig. 6). Construct mut-BD showed that half-sites B and D are not required for activity (Fig. 9). Finally, constructs mut-E and mut-F showed that hexamers E and F are equally essential for activity of the downstream domain, and consequently for maximal activity of R3.

Concerning the interaction between the functional domains in R3, the data suggest that the downstream domain (EF), because of its considerably lower affinity (Fig. 8), will not contribute to functional activity at limiting receptor concentration. However, we noticed that at the lowest receptor concentration tested, where the upstream domain ABCD should be responsible for all binding, the complete element R3 consistently bound twice as much dimer complex (compare binding data in Fig. 7B and Fig. 8B). This indicates some cooperative interaction between the two domains. We therefore tested receptor binding to fragments DE and CDE, which span the central region and partially overlap each domain. The fragment comprising half-sites DE bound less than 14% of the probe at the highest receptor concentration; however, this was all bound as dimers, suggesting a high degree of cooperativity at a low overall affinity (Fig. 10). Surprisingly, the 5-extension of this fragment with 3 base pairs, reconstituting half-site C, resulted in a similar cooperative formation of dimers, but with much higher apparent affinity (Fig. 10). The binding data are also depicted in Fig. 8, panel B, showing cooperativity of receptor complex formation, especially in case of fragment CDE.

Binding of TR to fragments of Region 3.

Finally, we tested whether the T₃ responsiveness of the SERCA1 promoter could be potentiated by RXR. Co-transfection of RXR or - did not affect basal promoter activity, either in the presence or absence of TR or T₃, and both isoforms augmented T₃ induction to the same extent (data not shown). As shown in Fig. 11, co-transfection of an RXR expression plasmid led to an almost doubling of the T₃ induction ratio for the two longest constructs, i.e. 575 and 962. Deletion of R3 in construct 490 reduced the RXR potentiation to +50%. Surprisingly, this potentiation of the T₃ induction mediated by R1 and R2 was lost upon deletion of sequences immediately upstream of element R2, i.e. between 327 and 269.

DISCUSSION

Cloning and sequencing of the upstream region of the rat SERCA1 gene indicated that the 5-nontranslated and initial 200 base pairs of the putative promoter are virtually identical to the rabbit SERCA1 sequence, except that the transcription initiation site is located 2 base pairs upstream to that reported for the rabbit (68). Comparison with the published 559 base pairs of promoter sequence of the SERCA2 gene of the rat (72) reveals no substantial sequence similarity, although three TREs are also thought to be responsible for T₃ responsiveness in this gene (51). The marked difference in promoter sequence between SERCA1 and SERCA2 is in line with the different modes of transcriptional regulation of these phenotype-specific isoforms, both in adult and in neonatal skeletal muscle. T₃ initially stimulates expression of SERCA1 as well as SERCA2 but then represses SERCA2 in certain muscle fibers (7, 28). Clearly, the availability of two differentially regulated, but otherwise similar, genes allows the skeletal muscle cell to use a broader spectrum of regulatory pathways in governing the important Ca²⁺ uptake function of the SR.

The SERCA1 promoter contains multiple E-boxes, capable of binding myogenic transcription factors of the MyoD family (69), but no other known cis-elements that confer or enhance muscle-specific expression, such as m-CAT, CArG, GArG, MHox, and MEF-2. Myogenic factors are not essential for function of the SERCA1 promoter used in this study, given its activity in COS cells. However, initial experiments with L6 myoblasts showed a severalfold higher normalized basal expression of the full-length promoter construct as compared with COS cells, suggesting a potential effect of muscle-specific transcription factors. Nevertheless, the observed T₃ induction ratio in these experiments was similar to that found in COS cells.² We chose COS cells to analyze the T₃ responsiveness of the promoter in detail, because the low transfection efficiencies and low absolute levels of CAT expression obtained with L6 muscle cells precluded an accurate analysis in these cells.

The TR-dependent, almost 4-fold (8-fold in the presence of RXR) T₃ induction ratio found for the full-length (962) SERCA1 promoter construct confirms that at least part of the transcriptional activation by T₃ is mediated by upstream sequences. The SERCA1 promoter contains several single optimal half-sites for TR binding in the first 327 base pairs. Such binding of receptor to the single perfect half-sites immediately downstream of the CCAAT box (60) and 22 base pairs downstream of the transcription initiation site (+25) might be expected to interfere with the transcriptional machinery and could be responsible for the TR-dependent, 37% reduction of basal activity of the minimal (141) promoter, which was not reversed by T₃. Although typical TREs consist of two or more half-sites, TR monomers are capable of transactivation (41, 42) also through interaction with other trans-acting factors, such as NF-1, Sp-1, COUP-TF, and CCAAT-binding protein (55, 61). In particular, T₃ induction conferred to the thymidine kinase promoter by a single TR-binding half-site was strongly potentiated by an NF-1 site inserted 45 base pairs downstream of this half-site (61). In the SERCA1 promoter, the NF-1 site at 192, which disrupts R1, is located 40 base pairs downstream of a perfect half-site, and another such half-site is located between the NF-1 site and the CCAAT box. Although we assume that regions R1 and R2 are primarily responsible for the T₃ response in constructs up to 490, it is possible that additional interactions between TR monomers and other trans-acting factors are involved in this response. Clearly, however, R3 is responsible for the additional T₃ responsiveness of the longer constructs.

The identification of regions R1, R2, and R3 as functional TREs was based on specific binding of TR dimers and T₃-dependent transactivation of a heterologous promoter. Comparison of the order of affinity with potency of transactivation shows that these parameters are not necessarily correlated, as has been previously noted for other TREs (44, 53, 73). The high affinity binding to R2 is most likely explained by the presence of a perfect hexamer GGGTCA, as well as an almost perfect copy of the extended half-site TGAGGTAACT (central T is a C in R2), shown to be a strong TR binder (39). However, the 9-base pair gap separating these high affinity sites makes a functional interaction less likely (41), although a DR + 12 TRE has recently been identified in the human type 1 deiodinase gene (52). Williams et al. (44) described the transactivating properties of a large number of natural response elements in the context of the TK promoter and comparison with our data would rank R1 and R2 with the bGH, ADH3, and estrogen response elements. All of these are weak T₃ responders. R3, on the other hand, ranks with relatively strong elements, predominantly regulated by T₃, such as that of the  myosin heavy chain gene and the rGH intronic TRE.

Whether all three TREs contribute to the total T₃ response is not known. The 4-fold T₃ induction of the promoter more than compensates for the TR-dependent repression of activity and is therefore not the result of hormone-induced dissociation of receptor dimers (74). Presumably, the mechanism of transactivation involves a direct interaction of the receptor complex with the transcription-initiation complex, possibly mediated by adaptor proteins, as has been shown for several members of the steroid hormone receptor superfamily, including TR (75, 76). That such a mechanism may put a limit on the number of these interactions is suggested by previous work in which two copies of the rGH TRE in tandem more than doubled the T₃ induction of the TK promoter, but a third copy did not have an additional effect (77). Multiple TREs have been identified in natural promoters (50, 51, 52), and in two cases deletion studies have shown that two TREs contribute to the total T₃ responsiveness (50, 52). In view of this we suggest that the T₃ response of the full-length SERCA1 promoter construct is the result of R3, possibly in combination with either R1 or R2, rather than of all three regions. This conclusion is supported by a comparison of the rat and rabbit promoter structures (68). The stimulation of SERCA1 expression by T₃ is similar in rat and rabbit, but the 5-sequences of the rabbit and rat SERCA1 gene strongly diverge upstream of position 200. Up to position 450 in the rabbit gene, the overall similarity is less than 60% and, notably, the sequences of R1 and R2 are not conserved. However, between positions 475 and 522, the rabbit sequence contains an almost perfect copy of R3, except for a 3-base pair insertion in the sequence separating the upstream and downstream domains. The essential half-sites A, C, E, and F (see "Discussion") are either fully conserved or mutated to a similar (half-site C) or a more optimal sequence (half-site F): The virtually complete conservation of R3, but not of R1 and R2, suggests a significant role for R3 as a regulatory element in the promoter of the SERCA1 gene.

The analysis of R3 identified this sequence as an unusual, composite TRE, consisting of an upstream IP domain and a downstream DR domain, both of which can function as independent TREs. This combination resembles the structure of the laminin B1 (LamB1) TRE in which four half-sites are arranged as an upstream IP + 7, touching a DR + 3 (46), although R3 appears to be a stronger TRE, at least when analyzed in COS cells (78). The upstream half of R3, with multiple potential half-sites, can in principle accommodate three likely dimer configurations, i.e. an IR + 2 (B-D), as found in the MHC TRE (47), an IP + 6/7 (A-C), as also found in the lysozyme silencer (Lys-F2) (48) and myelin basic protein TREs (49), and another IP + 6 between B and the ultimate 5-hexamer of R3. The analysis of receptor binding to R3 fragments (Fig. 8) indicated that the apparent affinity of dimer binding to the B-D domain is approximately 5-fold lower than for the IP + 6/7 domain contained in fragment ABCD. This is in agreement with several studies showing optimal spacing for TR binding for the IP configurations of 6-9 but 0 for the IR (palindrome) (37, 43, 73, 79, 80, 81). The mutational analysis of putative receptor contact sites (Fig. 9) then showed that half-sites B and D are not required for functional activity of the upstream domain but that the core G residues in both A and C are essential. Half-site C, including the two preceding base pairs (TAAGGTTA), is an almost perfect copy of the octamer TAAGGTCA shown to be an optimal TR-binding site (40). It has been shown that the PydPuo nucleotides 5 of the hexamer half-site sequence contribute considerably to affinity and functional potency of a TRE (39, 40, 41, 42, 43, 52), and given the high affinity dimer binding to A and C we suggest that this element is an IP + 7 (43) with an octamer sequence of half-site A of TAGGGAGG.

Further analysis of receptor binding and transactivation by R3 mutants identified hexamers E and F as essential half-sites forming the downstream TRE (DR + 2) in R3. The similar effect of mutating either A, C, E, or F on the potency of R3 (Fig. 9) indicates that functional interaction between both domains depends on dimer formation on each. The affinity of dimer formation for the downstream DR + 2 TRE is much lower than for the A-C element, in agreement with previous studies (43, 82), and it is questionable whether this element by itself would be of relevance. The role of the downstream element appears to depend on its capacity to interact with the upstream element. Such potentiation of dimer binding by a third half-site is well documented for the rGH TRE (53, 54), and this cooperativity is also observed for the four half-sites in the LamB1 TRE, all of which are required for maximal activity (73). The data in Fig. 8 indicate that half-sites C and E may be responsible for the interaction of the upstream and downstream domains. Although the receptor dimer complex formed on E and C (IR + 10) has no transactivating properties (Fig. 6), the relatively high affinity, cooperative binding of receptors to these half-sites could contribute to the formation of functional complexes on both domains in the intact R3. Although dimer binding on an IR + 10 was reported before (83), cooperativity was not observed, and it was much weaker compared with the (optimal) palindrome structure. However, we noticed that the heptamer CATCTGC, separating C and E, is identical to the sequence separating both TR-binding domains in the rat thyrotropin- TRE (84). This sequence was shown not to bind TR itself but to be required for optimal TR binding and function of the element. We therefore suggest that the gap sequence between half-site C and E is responsible for the unusual high affinity of this IR + 10, thus affecting the overall TR-binding affinity of R3.

Heterodimerization of TR with RXR has been shown to increase overall binding affinity for a TRE, but potentiation of T₃ induction appears to be dependent on the structure of the TRE (78). Nevertheless, because the RXR's are ubiquitous, they are considered to be the principal partner of TR in vivo. Hsu et al. (78) analyzed several natural TREs in COS cells and showed that only the rat GH and laminin B1 TREs show augmentation of the T₃ induction ratio with RXR. They concluded that TREs with more than 2 half-sites support potentiation of transactivation by RXR. Our data, showing a doubling of the T₃ induction ratio of the SERCA1 promoter with RXR, are in line with this conclusion, given the complex half-site composition of R3.

We conclude that the T₃ responsiveness of the 962-base pair SERCA1 promoter of the rat is primarily mediated by R3. The presence of this TRE in the promoter of the rat SERCA1 gene provides a mechanism for the T₃-stimulated expression of this isoform of the SR Ca²⁺-ATPase in rat skeletal muscle. The complex arrangement of the half-sites in the R3 TRE may be required to mediate the developmental and tissue-specific actions of T₃ on SERCA1 gene expression.