TRAM-1, A Novel 160-kDa Thyroid Hormone Receptor Activator Molecule, Exhibits Distinct Properties from Steroid Receptor Coactivator-1*

Nuclear hormone receptors (NRs) are ligand-dependent transcription factors that regulate target gene transcription. We report the molecular cloning and characterization of a novel human cDNA encoding TRAM-1, athyroid hormone receptor activatormolecule, a ∼160-kDa protein homologous with SRC-1/TIF2, by far-Western-based expression screening. TRAM-1 binds to thyroid hormone receptor (TR) and other NRs in a ligand-dependent manner and enhances ligand-induced transcriptional activity of TR. The AF-2 region in NRs has been thought to play a critical role in mediating ligand-dependent transactivation by the interaction with coactivators. Surprisingly, TRAM-1 retains strong ligand-dependent interaction with an AF-2 mutant of TR (E457A), while SRC-1 fails to interact with this mutant. Furthermore, we identified a critical TRAM-1 binding site in rat TRβ1 outside of AF-2, as TRAM-1 shows weak ligand-dependent interaction with a helix 3 ligand binding domain TR mutant (K288A), compared with SRC-1. These results suggest that TRAM-1 is a coactivator that may exhibit its activity by interacting with subdomains of NRs other than the AF-2 region, in contrast to SRC-1/TIF2.

The N-terminal domain (AF-1) contains a ligand-independent activation function, whereas the extreme C-terminal region of the ligand binding domain (AF-2) exhibits ligand-dependent transactivation (1). The AF-2 region is conserved among NRs, and deletion or point mutations in this region impair transcriptional activation without changing ligand and DNA binding affinities (2)(3)(4). Recent x-ray crystallographic studies of the ligand binding domain of NRs revealed that the ligand induces a major conformational change in the AF-2 region (5-7), suggesting that this region may play a critical role in mediating transactivation by a ligand-dependent interaction with coactivators. Several putative coactivators of NRs have been identified, using either far-Western or yeast two-hybrid techniques, including SRC-1, RIP140, TIF1, TIF2, and CREB binding protein (CBP)/p300 (8 -14). As expected, these proteins fail to interact with AF-2 mutants of NRs (10 -14).
Using bacterially expressed thyroid hormone receptor (TR) as a probe, a cDNA expression library was screened, and a novel cDNA encoding TRAM-1, a putative thyroid hormone receptor activator molecule, was isolated. TRAM-1 belongs to a nuclear receptor coactivator (NCoA) gene family that includes SRC-1 and TIF2. TRAM-1 exhibits ligand-dependent, and unexpectedly, AF-2-independent interaction with TR, a biochemical feature distinct from SRC-1/TIF2. Our findings suggest that TRAM-1 may serve a novel pathway for transcriptional activation of NRs by interacting with subdomains other than the AF-2 region.
Isolation of TRAM-1-The expression cDNA library screen by far-Western blotting was performed as described previously (10). Human pituitary (Stratagene) and 293 kidney embryonic cell libraries (Dr. J. A. DeCaprio, Dana-Farber Cancer Institute, Boston, MA) in ZAPII were screened using 10 5 cpm/ml of 32 P-labeled ligand binding domain (LBD) of rat TR␤1 in the presence of 1 M T 3 . The remaining 5Ј-end sequence was cloned by PCR amplification from the human pituitary library with the pBS-specific sense primer, 5Ј-GGAAACAGCTATGACCATGAT-TACG-3Ј, and the cDNA-specific antisense primer of the original clone obtained by far-Western screening, 5Ј-AAACACTTGTGTTAACCAG-GTCCTCTTGCT-3Ј. A full-length cDNA (TRAM-1) was reconstructed in pBK-CMV from overlapping TRAM-1 clones. Nucleotide sequences of positive clones were determined using an Applied Biosystems 377 DNA sequencer.
Northern Blot Analysis-Northern blot analysis was performed using human multiple tissue poly(A) ϩ RNAs (2 g per lane; CLONTECH), according to the protocol of the manufacturer. The random primed 32 P-labeled DNA probes used were nucleotides 1153-2213 of TRAM-1 and nucleotides 1274 -1896 of F-SRC-1.
Preparation of In Vitro Translated Proteins and GST Pull-down Assays-Full-length and fragments of TRAM-1 in pBK-CMV, and TR and RXR in pcDNA/AMP, were transcribed and translated in rabbit reticulocyte lysate (Promega) with 35 S-methionine according to the instructions of the manufacturer. The GST-fusion protein pull-down assay was performed as described previously (10).
Electrophoretic Mobility Shift Assay (EMSA)-Deoxyribonucleotides The poly Q region of TRAM-1 is located at the C terminus, compared with that of p/CIP. The C-terminal tail sequence of TRAM-1 bears no homology with that of p/CIP. D, nucleotide and deduced amino acid comparison between the C-terminal tails of TRAM-1 and p/CIP. TRAM-1 is highly conserved with p/CIP until amino acid residue 1321. One nucleotide deletion of p/CIP at codons 1296 -1297 (TAG) results in a protein sequence again homologous with TRAM-1, shown by shaded box.
containing F2 or DR4 TRE (16) were end-labeled with [␥-32 P]ATP by T4 polynucleotide kinase. Unlabeled in vitro translated receptor proteins, 20 ng of GST-fusion proteins, and a 50,000-cpm oligonucleotide probe were mixed and incubated together before being subjected to electrophoresis and autoradiography.
Transient Transfection Experiments-CV-1 cells were transiently transfected using the calcium phosphate coprecipitation method in 6-well plates with 1.7 g of reporter plasmid containing the F2 or DR4 TRE, fused to TK-LUC cDNA; 20 ng of Rous sarcoma virus (RSV)-␤galactosidase plasmid as a internal control; 100 ng of TR␤1; and various amounts of TRAM-1 or F-SRC-1 expression plasmids as detailed in the legend of Fig. 2. Empty expression vectors were added to equalize total transfected plasmid DNA concentrations. Cells were grown for 24 h in the absence or presence of 10 Ϫ7 M T 3 in serum-free media and harvested. Cell extracts were then analyzed for both luciferase and ␤galactosidase activities to correct for transfection efficiency. The corrected luciferase activities of untreated samples were normalized to the luciferase activities of samples containing the vector alone in the absence of the ligand (1-fold basal). The results shown are the mean Ϯ S.D. (n ϭ 3).

RESULTS AND DISCUSSION
Isolation, Expression and Function of TRAM-1-To identify TRAMs, we employed a far-Western approach previously used to clone a full-length version of human SRC-1 (F-SRC-1) cDNA (10). We screened cDNA expression libraries with a 32 P-labeled probe containing the LBD of rat TR␤1 in the presence of T 3 . Two positive clones, termed TRAM-1(293) and TRAM-1(PIT), from a 293 cell line and a human pituitary library, respectively, were identified. The nucleotide sequences of the coding regions and 3Ј-ends of these two clones were determined and found to be overlapping. TRAM-1(293) cDNA is ϳ6 kb in length, contains a poly(A) signal, and encodes an open reading frame (ORF) of 913 amino acids (aa), whereas TRAM-1(PIT) cDNA contains an ORF of 848 aa. The remaining 5Ј-sequence of TRAM-1 mRNA was cloned using 5Ј-rapid amplification of cDNA ends. A full-length version of the cDNA, designated as TRAM-1 cDNA, consists of a protein coding region of 4272 base pairs flanked by 191 base pairs of 5Ј-untranslated and ϳ3.5 kb of 3Ј-untranslated regions with a poly(A) tail. The ORF of TRAM-1 cDNA encodes 1424 aa with a calculated molecular mass size of TRAM-1 of 155.3 kDa. In TRAM-1(293) cDNA, the region corresponding to aa 904 -918 was missing, probably due to alternative splicing. Comparison of these sequences with those in the GenBank showed that TRAM-1 shares homology with SRC-1 (Fig. 1A) and TIF2 sequences (33.2 and 42.2% identity, respectively). Thus, TRAM-1 belongs to the 160-kDa protein subset of the NCoA gene family, which has a basic helix-loop-helix (bHLH)/PAS domain in the N-terminal, a nuclear receptor binding domain in the central, and a glutamine (Q) rich sequence in the C-terminal regions (Fig. 1C).
Northern blot analysis of poly(A) ϩ RNAs from human tissues (Fig. 1B) indicated that the major TRAM-1 mRNA is ϳ9 kb; a minor species (ϳ5.5 kb) is also seen in several tissues. Although TRAM-1 is ubiquitously expressed, the expression pattern of TRAM-1 (top panel) is different from that of SRC-1 (bottom panel). TRAM-1 is highly expressed in placenta, whereas SRC-1 is most abundant in brain. This result suggests that the expression of TRAM-1 and SRC-1 may be differentially regulated and thus may play specific roles in NR-mediated gene expression in different target tissues.
To investigate the coactivator activity of TRAM-1, cotrans- fection studies in CV-1 cells were performed with TR and TRAM-1 as well as SRC-1 expression plasmids and reporter plasmids containing F2 or DR4 TRE. As shown in Fig. 2, cotransfection of TRAM-1 enhances T 3 -mediated transactivation on both F2 and DR4 TREs, similar to SRC-1, although TRAM-1 seems to increase basal activity as well. Thus, TRAM-1 has coactivator activity for TR function.
While our work was in progress, Torchia et al. (17) reported a mouse cDNA clone, p/CIP, encoding a 152-kDa coactivator protein. The amino acid sequence of TRAM-1 has 74.2% identity with that of p/CIP. Therefore, TRAM-1 may be a human homolog of p/CIP. Of note, there are two major differences as illustrated in Fig. 1C. One is the different position of the poly Q region. The poly Q stretch of TRAM-1 is located ϳ300 aa downstream of a comparable region in p/CIP. The other difference exists at the C terminus. There is no significant homology between TRAM-1 and p/CIP in the ϳ100 aa of C-terminal tail, despite shared homology of TRAM-1 and SRC-1/TIF2 in this region. Comparison of TRAM-1 and p/CIP DNA sequences revealed that a single nucleotide insertion present in p/CIP cDNA at codons 1296 -1297 causes a putative frameshift, as nucleotide deletion of this codon in p/CIP results in an alternate coding sequence that shows 80.4% identity with that of TRAM-1 (Fig. 1D). In cotransfection experiments, p/CIP has not shown any significant coactivator function (17), whereas TRAM-1 has such activity. Thus, the apparent functional differences between TRAM-1 and p/CIP may be due to different C-terminal tails, although a sequencing error of p/CIP cannot be ruled out.
In Vitro Binding Assay-Binding of TRAM-1 to NRs was tested using GST-fusion proteins. As shown in Fig. 3A, in vitro translation of TRAM-1 mRNA generated a major full-length 35 S-labeled product (ϳ160 kDa) (lane 1). In the presence of cognate ligands, 35 S-labeled TRAM-1 protein showed increased binding with GST-TR (lanes 2 and 3), GST-ER (lanes 4 and 5), GST-RAR (lanes 6 and 7), and GST-RXR (lanes 8 and 9) fusion proteins. Thus, TRAM-1 can interact with several members of NRs in a ligand-dependent manner.
Next, we generated GST-TRAM-1A (aa 577-800) and GST-TRAM-1B (aa 800 -1215) fusion proteins, which contain regions A and B, respectively, and tested their interactions with in vitro translated 35 S-labeled TR and RXR. As shown in Fig.  3C, TR showed ligand-dependent interaction with TRAM-1A (lanes 2 and 3) and little or no ligand-dependent interaction with TRAM-1B (lanes 4 and 5). RXR showed ligand-dependent interaction with TRAM-1A (lanes 7 and 8) but not TRAM-1B (lanes 9 and 10). This result confirms the recent reports that the function of an LXXLL motif also depends on other sequences (17,18).
Region A, which contains three LXXLL motifs, is highly conserved among members of the NCoA (SRC-1, TIF2, p/CIP, and TRAM-1) gene family. On the other hand, the LXXLL motif in region B is not present in SRC-1, TIF2, and p/CIP. A transfection study using an SRC-1 mutated in the LXXLL motif in region A has shown that the mutant fails to enhance ligandinduced RAR-or ER-mediated transactivation (17,18). Therefore, region A of TRAM-1 may play a more important role for ligand-dependent interaction with NRs than region B.
In pull-down experiments, region A also displayed ligand-de- pendent interaction with the AF-2 mutant, E457A. To compare the interaction properties between TRAM-1 and SRC-1 with TR and its mutants, we generated a GST-SRC-1A fusion protein (aa 595-780) containing a homologous region of GST-TRAM-1A, and we used GST-TRAM-1A and GST-SRC-1A for EMSA. As shown in Fig. 4A, wild-type TR forms homodimers (lane 1) and heterodimers with RXR (lane 7) on an inverted palindrome TRE, F2. Addition of T 3 decreased homodimer formation (lane 2) as noted previously (19). Incubation of TR and TR/RXR with GST-SRC-1A produced a ligand-dependent TR/ SRC-1A (lane 4) and TR/RXR/SRC-1A (lane 10) complex, respectively. Similarly, GST-TRAM-1A showed ligand-dependent interaction with TR (lane 6) and TR/RXR (lane 12). The mobilities of TR/RXR/SRC-1A and TR/RXR/TRAM-1A were slightly less than those of TR/SRC-1A and TR/TRAM-1A, respectively. In addition, a RXR antibody could supershift TR/RXR/SRC-1A and TR/RXR/TRAM-1A complexes (data not shown), indicating that both SRC-1 and TRAM-1 can form a ligand-induced complex with TR/RXR heterodimer on F2 TRE. Interestingly, GST-TRAM-1B did not show any interaction with TR or TR/RXR on EMSA (data not shown), consistent with region A being more important for ligand-dependent interaction with NRs than region B.
Next, we compared the interaction of an AF-2 mutant of TR, E457A, with TRAM-1 and SRC-1. As shown in Fig. 4B, E457A showed similar properties of homo-and heterodimers formation as wild-type TR. While SRC-1A fails to show ligand-dependent interaction with E457A (lane 4) and E457A/RXR (lane 10), TRAM-1A still exhibits strong ligand-dependent interaction with E457A (lane 6) and E457A/RXR (lane 12). We speculated that TRAM-1 may interact with subdomains of NRs other than AF-2, in contrast to SRC-1. Previously, O'Donnell and Koenig (20) reported that TR␤1 point mutants of aa 288 -300 (helix 3 region in the LBD of TR), which are conserved among NRs, have impaired transcriptional activation without altered ligand and DNA binding, similar to AF-2 mutants. X-ray crys-tallographic studies have shown that the AF-2 region comes into close contact with the helix 3 region after ligand binding (5)(6)(7). In EMSA (Fig. 4C), K288A also showed similar properties of homo-and heterodimers formation as wild-type TR.
Interestingly, TRAM-1A shows negligible ligand-dependent interaction with K288A and K288A/RXR, compared with SRC-1. Similar results were obtained when we used the direct repeat TRE DR4 (data not shown). These EMSA studies indicate that although both TRAM-1 and SRC-1 interact with liganded TR, SRC-1 may preferentially bind to the AF-2 region; on the other hand, TRAM-1 may preferentially bind to helix 3 in the LBD of TR.
The C-terminal region of CBP has been reported to interact with SRC-1 (9) and P/CAF, which is a p300/CBP-interacting protein with histone acetylase activity (21). In addition, the progesterone receptor has been shown to interact directly with P/CAF (22). Thus, we tested whether CBP or P/CAF interacts with TRAM-1. As shown in Fig. 5, both CBP (aa 1626 -2260) and P/CAF were significantly retained on GST-TRAM-1B compared with GST-TRAM-1A. This observation suggests that TR, TRAM-1, CBP/p300, and P/CAF can potentially form a liganddependent complex to mediate ligand-induced transactivation.
In summary, we have isolated cDNA, encoding a putative thyroid hormone receptor coactivator TRAM-1. TRAM-1 belongs to a 160-kDa NCoA family that includes SRC-1 and TIF2. However, TRAM-1 exhibits expression and interaction properties distinct from SRC-1. As such, TRAM-1 may be another component of a TR coactivator complex critical for TR action.