A Nuclear Factor, ASC-2, as a Cancer-amplified Transcriptional Coactivator Essential for Ligand-dependent Transactivation by Nuclear Receptors in Vivo *

Many transcription coactivators interact with nuclear receptors in a ligand- and C-terminal transactivation function (AF2)-dependent manner. We isolated a nuclear factor (designated ASC-2) with such properties by using the ligand-binding domain of retinoid X receptor as a bait in a yeast two-hybrid screening. ASC-2 also interacted with other nuclear receptors, including retinoic acid receptor, thyroid hormone receptor, estrogen receptor α, and glucocorticoid receptor, basal factors TFIIA and TBP, and transcription integrators CBP/p300 and SRC-1. In transient cotransfections, ASC-2, either alone or in conjunction with CBP/p300 and SRC-1, stimulated ligand-dependent transactivation by wild type nuclear receptors but not mutant receptors lacking the AF2 domain. Consistent with an idea that ASC-2 is essential for the nuclear receptor function in vivo, microinjection of anti-ASC-2 antibody abrogated the ligand-dependent transactivation of retinoic acid receptor, and this repression was fully relieved by coinjection of ASC-2-expression vector. Surprisingly, ASC-2 was identical to a gene previously identified during a search for genes amplified and overexpressed in breast and other human cancers. From these results, we concluded that ASC-2 is a bona fidetranscription coactivator molecule of nuclear receptors, and its altered expression may contribute to the development of cancers.

The nuclear receptor superfamily is a group of ligand-dependent transcriptional regulatory proteins that function by binding to specific DNA sequences named hormone response elements in the promoters of target genes (for a review, see Ref. 1). The superfamily includes receptors for a variety of small hydrophobic ligands such as steroids, T3, 1 and retinoids as well as a large number of related proteins that do not have known ligands, referred to as orphan nuclear receptors (reviewed in Ref. 2). Functional analysis of nuclear receptors has shown that there are two major activation domains. The activation function-2 (AF-2) at the extreme C-terminal region of the ligandbinding domain (LBD) exhibits ligand-dependent transactivation, whereas the N-terminal activation function-1 contains a ligand-independent transactivation domain. The AF-2 region is conserved among nuclear receptors, and deletion or point mutations in this region impair transcriptional activation without changing ligand and DNA binding affinities. X-ray crystallographic studies of the LBD of nuclear receptors revealed that the ligand binding induces a major conformational change in the AF-2 region (3)(4)(5)(6)(7), suggesting that this region may play a critical role in mediating transactivation by a ligand-dependent interaction with coactivators. As expected, many coactivators fail to interact with AF-2 mutants of nuclear receptors (8 -10).
The regulation of gene expression by nuclear receptors has been postulated to involve targeted changes in chromatin structure (33)(34)(35). In particular, recent biochemical and genetic studies support the notion that hyperacetylation of core histones is characteristic to gene activation, whereas histone deacetylation is involved with transcriptional repression (36,37). SRC-1 (38) and its homologue ACTR (39), along with CBP/ p300 (40,41), were recently shown to contain potent histone acetyltransferase activities themselves and associate with the histone acetyltransferase protein P/CAF (42). CBP/p300 also forms a complex with SRC-1 (15)(16)(17)27). These results suggest that nuclear receptors target at least three different histone acetyltransferase activities (SRC-1 or related proteins CBP/ p300 and P/CAF) to promoters (39). In contrast, it was shown that SMRT and N-CoR, nuclear receptor corepressors, form complexes with Sin3 and histone deacetylase proteins, suggesting that chromatin remodeling by cofactors contributes to receptor-mediated transcriptional regulation (43)(44)(45)(46). Genomic instability in human cancers commonly results in gene amplification and the consequent overexpression of specific genes. Breast cancers frequently exhibit increased copy number of chromosomal segments that encode genes related to tumor growth such as ERBB-2, MYC, and cyclin D1 (47). The application of molecular cytogenetic technology to breast cancers has identified several regions of increased DNA copy number whose target genes remain unknown (48,49). In order to isolate candidate genes from one of these regions (20q), chromosome microdissection and hybrid selections were used to clone amplified cDNAs. In particular, three novel amplified genes of potential biological relevance to cancer progression (termed AIB1, AIB3, and AIB4) have been isolated from a cDNA library constructed from the 20q amplified breast cancer cell line BT-474 (50). These genes were mapped to 20q11 (AIB3 and AIB4) and 20q12 (AIB1) by fluorescence in situ hybridization. AIB1 was subsequently shown to be a member of the SRC-1 family and a coactivator of ER␣ (51).
In this study, we report the isolation of a novel nuclear protein-activating signal cointegrator 2 (ASC-2) based on its interaction with retinoid X receptor (RXR) which on sequence analysis is identical to AIB3. Consistent with specific interactions with basal factors TFIIA and TBP, ASC-2 exhibits an autonomous transactivation function in yeast. ASC-2 also interacts with transcription integrators SRC-1 and CBP and stimulates transactivation by nuclear receptors in conjunction with these proteins. Remarkably, microinjection of anti-ASC-2 antibody blocks the ligand-dependent transactivation of nuclear receptors, and this repression is fully relieved by coinjection of ASC-2-expression vector. Thus, ASC-2 is a novel transcriptional coactivator molecule required for the liganddependent transactivation of nuclear receptors in vivo, and its altered expression may contribute to the development of cancers.
Yeast Two-hybrid Screening and Yeast ␤-Galactosidase Assay-Gal4-DBD/RXR-LBD was used as bait to screen a Xenopus oocyte cDNA library in pGAD10 vector (CLONTECH) for RXR-interacting proteins, according to the manufacturer's protocols. However, the screening was executed either in the presence or absence of 1 M 9-cis-RA as described (52). The library plasmids from positive clones that expressed both HIS3 and lacZ reporters were rescued and retransformed into yeast cells, together with the original or other bait constructs, to test specificity of the interactions. The cotransformation and ␤-galactosidase assays in yeast were performed as described (52). For each experiment, at least three independently derived colonies expressing chimeric receptors were tested.
GST Pull Down Assays-The GST fusions or GST alone was expressed in Escherichia coli, bound to glutathione-Sepharose-4B beads (Amersham Pharmacia Biotech), and incubated with labeled proteins expressed by in vitro translation by using the TNT-coupled transcription-translation system, with conditions as described by the manufacturer (Promega, Madison, WI). Specifically bound proteins were eluted from beads with 40 mM reduced glutathione in 50 mM Tris (pH 8.0) and analyzed by SDS-PAGE and autoradiography as described (56).
Cell Culture and Transfection-CV-1 or HeLa cells were grown in 24-well plates with medium supplemented with 10% charcoal-stripped serum. After 24 h incubation, cells were transfected with 100 ng of ␤-galactosidase expression vector pRSV-␤-gal and 100 ng of an indicated reporter gene, along with hRXR␣-, ER␣, ER␣⌬AF2, RXR⌬AF-, ASC-2-, SRC-1-, or p300-expression vectors. Total amounts of expression vectors were kept constant by adding decreasing amounts of the CDM8 expression vector (i.e. pcDNA3) to transfections. Twelve hours later, cells were washed and refed with Dulbecco's modified Eagle's medium containing 10% charcoal-stripped fetal bovine serum. After 12 h, cells were left unstimulated or stimulated with 0.1 M ligand. Cells were harvested 24 h later, and luciferase activity was assayed as described (57), and the results were normalized to the ␤-galactosidase expression. Consistent results were obtained in more than two similar experiments.
Fluorescence in Situ Hybridization (FISH)-AIB3-specific oligonucleotide primers, G3F1 (TTTATTCACTGGTCCATTTCTACA) and G3R2 (AGTTTTCACATTTCCTAAGC), were used to screen a human genomic BAC library (Research Genetics). A positive BAC clone was labeled with Spectrum Orange-dUTP (Vysis) using the BioPrime DNA labeling system (Life Technologies, Inc.), purified through a Bio-Spin 6 chromatography column (Bio-Rad), and ethanol-precipitated. Hybridizations of breast cancer cell lines were performed as described (58). The tumor microarray technique has been previously described (59). For FISH analysis of tumors, array sections were deparaffinized, air-dried, and dehydrated in an ethanol series followed by denaturation for 4 min at 72°C in 70% formamide, 2ϫ SSC. The hybridization mixture contained 200 ng of probe and 15 g of COT1 DNA and was incubated overnight. Slides were washed and counterstained with 4,6-diamidino-2-phenylindole in antifade. Images were captured with a non-laser confocal fluorescence optical module (CARV optical module, Atto Instruments, Rockville, MD) connected to an Axioplan 2 microscope (Zeiss, Jena, Germany). For cell lines, the modal ASC-2 copy number was determined in counts of 50 nuclei. Tumors were scored in three groups according to the number of ASC-2 signals observed: low (Յ4), moderate (4 -6), and high (Ͼ6). Only high copy number increases are considered evidence of gene amplification. Cancers and numbers of specimens included in the multicancer microarray were colon (32), stomach (12), renal cell (24), testis (19), non-small cell lung (78), prostate (19), bladder (34), endometrial (8), ovarian (22), head/neck (15) Antibodies, Western Analyses, Immunofluorescence, and Single Cell Microinjection Assay-The nuclear receptor-binding domain of ASC-2 (i.e. the ASC-2 residues 719 -860) was expressed and used as an antigen to raise a polyclonal rabbit antibody, which was affinity purified using standard procedures (60). Similarly, a monoclonal antibody (A3C1) was raised against the C-terminal peptide epitope of ASC-2 and used to probe the ASC-2 protein expressions in various cancer cell lines, and a rabbit anti-tubulin antibody (ICN, Costa Mesa, CA) was used as a loading control. Western analyses were essentially done as described previously (57). Rat-1 fibroblast cells were immunostained with anti-ASC-2 antibody and rhodamine-conjugated antibodies. Insulin-responsive Rat-1 fibroblasts, made quiescent by incubating in serum-free medium for 24 h, were microinjected with either preimmune IgG or the affinity purified anti-ASC-2 IgG along with ␤RARE-SV40-lacZ reporter constructs (25 g/ml). About 1 h after injection, cells were stimulated, where indicated, with 0.1 M 9-cis-RA. After overnight incubation, cells were fixed and stained to detect injected IgG by using FITC-conjugated antibodies and examined for ␤-galactosidase expression as described previously (61). The image was photographed with Zeiss AxioplanII microscope equipped with a PIXERA camera.

Molecular Cloning and Expression of ASC-2-
To identify potential transcriptional coactivators of nuclear receptors, we screened a Xenopus oocyte cDNA library using the LBD of RXR as a bait in the yeast two-hybrid system (62). Among 11 isolates that showed 9-cis-RA-and AF-2-dependent interactions, two were found to encode xSRC-3, a new member of the SRC-1 family (54). Four independent isolates encoded various lengths of an identical protein that showed ability to interact functionally with nuclear receptors (this paper) and a series of distinct transcription factors. 2 Accordingly, we named this protein ASC-2 (for activating signal cointegrator). ASC-2 is encoded by an approximately 8,500 nucleotide messenger RNA and highly expressed in only early stages of Xenopus oocyte development (Fig. 1). However, ASC-2 is relatively widely expressed in various human adult tissues examined (data not shown). Data base search revealed the existence of a highly homologous human cDNA of approximately 6,500 nucleotides (approximately 50% identity and 65% similarity in amino acid level) (GenBank TM accession number D80003). The 5Ј 325-base pair fragment of this human clone was used as a probe to screen a few different human cDNA libraries. A number of full-length cDNA clones encoding 2,063 amino acids were obtained in which the initiator methionine is preceded by a few in-frame stop codons. The amino acid sequence (Fig. 2) shows that ASC-2 contains an acidic domain at the N terminus, two glutamine-rich domains, and a serine/threonine-rich domain at the C terminus. It also contains two separate basic domains, potential nuclear localization signals, and two copies of a sequence motif LXXLL (where L and X denote leucine and any amino acid, respectively), recently shown to function in liganddependent interaction with the AF-2 domain of nuclear receptors (28,63).
Amplification and Overexpression of ASC-2 in Human Cancers-Remarkably, ASC-2 is identical to AIB3, a gene previously identified during a search for amplified and overexpressed genes mapping to chromosome 20q in breast cancers (50). ASC-2 maps to 20q11 and defines a distinct region of amplification (64). To determine the frequency of ASC-2 amplification in breast cancer, we examined 335 specimens by FISH using the recently developed tissue microarray technology (59) (Fig. 3A). In this series, ASC-2 copy number was increased to a moderate level (4 -6 copies) in 14/335 (4.2%) cases and to a high level (Ͼ6 copies) in 15/335 (4.5%) cases (Table I). We further extended our search for ASC-2 amplification to 284 specimens of various cancer types using tissue microarrays. A pattern of recurrent ASC-2 amplification was also observed in colon and non-small cell lung cancer (Table I) with the frequency of increased copy number in colon cancer being the highest observed in any cancer studied. The presence of amplification in colon, breast, and lung cancers suggests that increased quantities of ASC-2 may contribute to tumor growth in these diseases. In addition, ASC-2 mRNA expression was detected in each of 11 breast cancer cell lines with the highest level of expression in BT-474 (Fig. 3B). Western analyses also showed that ASC-2 protein was differentially expressed in various breast cancer cell lines (Fig. 3C). A relatively high expression of ASC-2 in MDA-MB-231 cells and a low level expression in HT29 or HeLa cells were also observed (data not shown). Interestingly, two protein bands appear to exist, which could have resulted from post-translational modification such as phosphorylation or represent distinct isoforms. It is also notable that high levels of expression are not strictly correlated with ER␣ positivity. Although ASC-2 is expressed in three of four ER␣-positive cell lines, significant expression is apparent in two ER␣-negative cell lines BT20 and MDA-MB-453 and with very low expression in ER␣-positive MCF7 cell lines.
Interactions of ASC-2 with Nuclear Receptors-The original Xenopus ASC-2 isolates P1, P4, and P5 were isolated from the yeast two-hybrid screening in the presence of the RXR ligand 9-cis-RA, whereas M8, identical to P1, was independently isolated in the absence of 9-cis-RA ( Fig. 2A). All of these proteins showed a strong ligand dependence in interactions, i.e. their interactions with RXR were significantly stimulated in the presence of 9-cis-RA (data not shown). Consistent with these results, activation domain fusions to the ASC-2 amino acids 438 -1041 (ASC-2c) showed strong ligand-dependent interactions with LexA-RXR␣/LBD in yeast (Fig. 4A) 586 -860 (ASC-2b) readily interacted with TR␣ and TR␤/LBD but not with RXR␣/LBD and RAR␣. These yeast results were further confirmed by direct in vitro binding (56). ASC-2 and ASC-2⌬C that includes only ASC-2 amino acids 1-928 interacted relatively weakly with a full-length RXR␣-glutathione S-transferase fusion protein (GST-RXR␣) (Fig. 4B) but did not bind luciferase and other control proteins (data not shown). These weak interactions were significantly stimulated by 9-cis-RA. A similar ligand-dependent interaction was also observed with a GST fusion to the LBD of RXR (GST-RXR␣/LBD), which was completely abolished by a mutation in the AF-2 domain (55). In addition, ASC-2 and ASC-2⌬C independently interacted with the N-terminal ABC domains of RXR␣. Similar in vitro binding was also observed with a series of other receptors including TR, ER␣, and RAR (Fig. 4B). Overall, these results, along with the yeast results, define the ASC-2 amino acids 586 -860 as a minimum receptor interaction domain and suggest that the receptor-ASC-2 interactions involve the AF-2 domain of receptors.
ASC-2 Exhibits Autonomous Transactivation Function-One of the characteristic properties of transcription coactivators is an autonomous transactivation function. To probe for such activities associated with ASC-2, we expressed a series of ASC-2 fragments fused to a heterologous DNA-binding protein LexA in yeast, along with a lacZ reporter construct controlled by upstream LexA-binding sites (57). As shown in Fig. 5A, ASC2-1 consisting of the ASC-2 residues 1-557 (as depicted in Fig. 2A) exhibited a very strong transactivation function. In addition, ASC2-2 (the ASC-2 residues 391-1057) showed a relatively strong transactivation activity, whereas more C-terminal fragments of ASC-2 were transcriptionally inert. These results suggest that ASC-2 may directly associate with RNA polymerase II pre-initiation complex by recruiting basal transcription factors such as TBP and TFIIA. Consistent with this idea, both ASC-2 and ASC-2⌬C readily interacted with TBP and TFIIA in vitro, whereas SRC-1 interacted only with TBP, and IB␤ did not interact with either protein (Fig. 5B). Thus, we concluded that the N-terminal subregions of ASC-2 have autonomous transactivation function, at least in yeast, probably through recruitment of basal transcription factors TBP and TFIIA.
ASC-2 Associates with CBP/p300 and SRC-1-ASC-2 shows properties that are consistent with its putative role as a transcriptional coactivator of nuclear receptors, i.e. ligand-and AF-2-dependent interactions with nuclear receptors as well as an autonomous transactivation function. Recently, SRC-1 has been shown to functionally associate with CBP/p300 to coactivate the nuclear receptor transactivation (15)(16)(17)27). These results led us to explore the possibility of direct interactions of ASC-2 with these transcription integrator molecules. Indeed, a LexA fusion to ASC-2 amino acids 1-557 (LexA-ASC2-1) interacted with activation domain fusions to SRC-D and SRC-E in yeast (Fig. 6A). A LexA fusion to the ASC-2 amino acids 391-1057 (i.e. LexA-ASC2-2) also interacted with SRC-E, whereas  (c and d). Biopsy specimens were analyzed by tumor microarray FISH (59). A portion of a hematoxylin and eosin-stained tumor microarray section (b) and individual tumors with high (c) or low (d) ASC-2 copy numbers are illustrated. FISH was performed using a Spectrum Orange-labeled ASC-2 BAC. B, expression levels relative to MCF-10 cells and modal ASC-2 gene copy numbers are indicated. Total RNA (15 g) from established breast cancer cell lines was size-fractionated by gel electrophoresis, transferred to a nylon membrane, and hybridized with a 4.2-kilobase pair ASC-2 cDNA. A ␤-actin probe was used to normalize loading. C, ASC-2 protein expression levels were examined by Western analysis using a monoclonal antibody raised against the C-terminal peptide epitope of ASC-2. Nuclear extracts from different breast cancer cell lines have been examined as indicated. Tubulin expression was examined as a loading control. a LexA fusion to the ASC-2 amino acids 586 -1310 (i.e. LexA-ASC2-3) interacted only with SRC-D. In contrast, more Cterminal ASC-2 fragments (i.e. ASC2-4 and ASC2-5) didn't show interactions with none of the SRC-1 fragments tested. Consistent with these results, ASC-2 and ASC-2⌬C specifically bound to GST fusions to SRC-M (i.e. the SRC-1 residues 782-1139) and SRC-C (i.e. the SRC-1 residues 1107-1441) but not to SRC-N (i.e. the SRC-1 residues 1-415), as shown by the in vitro GST-pull down assays (Fig. 6B). In contrast, c-Jun interacted only with GST-SRC-C, as we recently reported (30). These results indicate that multiple interactions occur between the ASC-2 amino acids 1-1310 and various SRC-1 subregions. Similarly, LexA-ASC2-1 specifically interacted with CBP-A in yeast, whereas LexA-ASC2-2 interacted with CBP-A and CBP-E. In contrast, LexA fusions to ASC2-3, ASC2-4, and ASC2-5 didn't interact with any of the CBP fragments examined (Fig. 6C). ASC-2 also interacted with GST fusions to specific subregions of CBP, including CBP1, CBP3, and CBP4 that consist of the CBP amino acids 1-450, 1069 -1459, and 1459 -1891, respectively (Fig. 6D). In contrast, ASC-2⌬C interacted only with GST-CBP1 and GST-CBP3. As a control, SRC-M was shown to interact only with CBP5, as previously reported (15)(16)(17)27). It is notable that there are some discrepancies between the yeast and the GST-pull down results. The reasons for these differences are not currently understood.
Overall, these results suggest that multiple interactions occur between the ASC-2 amino acids 1-1057 and various CBP subregions.

ASC-2, as a Transcription Coactivator of Nuclear
Receptors-To assess the functional consequences of these interactions, ASC-2 was cotransfected into CV-1 cells along with a reporter construct ERE-TK-LUC. Increasing amounts of cotransfected ASC-2 enhanced the E2-dependent transactivation of this reporter in a dose-dependent manner, with cotransfection of 200 ng of ASC-2 increasing the fold activation approximately 3-fold (Fig. 7A). Interestingly, the ASC-2 enhancement of reporter gene expression was further stimulated with cotransfected SRC-1 or p300. Consistent with the protein interaction data, however, ASC-2 was not able to coactivate ER␣⌬AF-driven transactivations. Similar results were also obtained with a series of different reporter constructs responsive to GR, progesterone receptor, RXRs, TRs, or RARs in various FIG. 4. Interactions of ASC-2 with nuclear receptors. A, the indicated activation domain and LexA plasmids were transformed into a yeast strain containing an appropriate lacZ reporter gene. At least six separate transformants from each transformation were transferred to indicator plates containing X-gal, and reproducible results were obtained using colonies from a separate transformation. ϩϩϩ, strong blue colonies after 2 days of incubation; ϩϩ, light blue colonies after 2 days of incubation; ϩ, light blue colonies after more than 2 days of incubation; Ϫ, white colonies. B, bacterially produced GST fusion proteins were bound to a glutathione-agarose column and incubated with equivalent amounts of the indicated 35 S-labeled ASC-2, ASC-2⌬C, or luciferase produced by in vitro translation, as indicated. Specifically bound proteins were released by glutathione and resolved by SDS-PAGE. Approximately 15% of the labeled proteins used in the binding reactions were loaded as inputs.

FIG. 5. Autonomous transactivation function and interactions with basal transcription factors.
A, the indicated LexA plasmids were transformed into a yeast strain containing an appropriate lacZ reporter gene, as described (57). The results are expressed as induction fold (n-fold) over the value obtained with LexA/Ϫ, which was given an arbitrary value of 1. The data are representative of at least two similar experiments, and the standard deviations were less than 5%. B, bacterially produced GST-TBP, GST-TFIIA, or GST alone was bound to a glutathione-agarose column and incubated with equivalent amounts of the indicated 35 S-labeled ASC-2, ASC-2⌬C, IB␤, SRC-1, or luciferase produced by in vitro translation, as indicated. Specifically bound proteins were released by glutathione and resolved by SDS-PAGE. Approximately 15% of the labeled proteins used in the binding reactions were loaded as inputs.
cell types (Fig. 7, B-D, and data not shown). In contrast, ASC-2 had minimal effects on the basal expression of the reporter in the absence of ligand, expression of the control plasmids pRSV-␤-gal or TK-LUC, or GAL4-VP16-mediated transactivation of the GAL4-TK-LUC reporter construct ( Fig. 7 and  data not shown). These results strongly support the notion that ASC-2 is a bona fide transcription coactivator of nuclear receptors.

ASC-2 Is Required for the Nuclear Receptor Function in
Vivo-We raised and affinity purified a rabbit polyclonal antibody against the nuclear receptor binding domain of ASC-2, which specifically detected, in Western analyses, either bacterially expressed and purified ASC-2 or endogenous/cotransfected ASC-2 in mammalian cells (Fig. 8A and data not shown). In contrast, this ASC-2 antibody was not able to detect GST alone, GST/ASC-1, a novel transcription coactivator we have recently reported (65), or CBP2 (i.e. the CBP residues 451-1009) (Fig. 8A). Immunostaining of Rat-1 fibroblast cells with this antibody revealed that ASC-2 is a nuclear protein, as expected (Fig. 8B). Microinjection techniques were further utilized to investigate the function of ASC-2 in vivo (61). Reporter genes were placed under the control of an SV40 minimal promoter containing 9-cis-RA-responsive ␤RARE sites (16). Remarkably, microinjection of anti-ASC-2 IgG almost completely prevented 9-cis-RA from activating an RAR-dependent transcription unit ( Fig. 9) but had no effect on a promoter under the control of the cytomegalovirus promoter (data not shown). The percentage of cells that expressed the lacZ reporter (i.e. blue cells) was not observed among cells microinjected with control IgG in the absence of ligand but increased to approximately 70% in the presence of 9-cis-RA. However, only approximately 10% of cells turned blue even in the presence of 9-cis-RA, when microinjected with anti-ASC-2 IgG (Fig. 9). When coinjected with ASC-2-expression vector, however, anti-ASC-2 IgG was not able to prevent 9-cis-RA-dependent transactivation. In contrast, coinjection of pcDNA3 or SRC-1-expression vector was without any significant effects. However, coinjection of p300 expression vector resulted in approximately 40% of blue cells. These results suggest that p300 but not SRC-1 is functional with receptor-mediated ligand-dependent transactivation in the absence of ASC-2 and also implicate multiple or redundant activation pathways in transactivating nuclear receptors, in which distinct sets of coactivator molecules are employed. Similar results were also obtained with RXR response elements (data not shown). Taken together with the transient transfection data, we concluded that ASC-2 is a molecule essential for the function of nuclear receptors in vivo. DISCUSSION In this report, we have described the initial characterization of a novel coactivator protein ASC-2, which shows a strong ligand and AF-2 dependence in interactions with nuclear receptors (Fig. 4) and an autonomous transactivation function (Fig. 5A). Transcriptional activation of nuclear receptors involves at least two classes of cofactors, corepressors and coactivators (reviewed in Ref. 66). Corepressors that associate with unliganded nuclear receptors mediate repression, whereas coactivators are recruited upon ligand binding and concomitant dissociation of the corepressors. Currently, at least two mechanisms have been proposed to describe the function of these coactivators. First, they are postulated to function to transmit the signal of ligand-induced conformational change to the basal transcription machinery. Second, they have been associated with targeted alterations of chromatin structure (33)(34)(35).
Several groups of different macromolecular complexes containing transcription coactivators have been described. First, the TAF components of TFIID (reviewed in Ref. 67) and the SRB/MED components associated with polymerase II (reviewed in Ref. 68) comprise those that are ultimately associated with the general transcription machinery. The CBP-p300-SRC-1 coactivator complex defines the second coactivator complex that directly binds and coactivates a wide spectrum of different transcription factors. In particular, CBP/p300 was recently found to be complexed with a series of cellular proteins with relative molecular masses ranging from 44 to 270 kDa (69). Purification and analysis of various proteins in this group revealed that they are components of the human SWI-SNF complex and that p270 is an integral member of this complex.
Interestingly, heterogeneity appears to exist among the CBP-p300-containing complexes. Different classes of mammalian transcription factors (nuclear receptors, CREB, and STATs) were recently shown to require functionally distinct components of the CBP-p300 coactivator complex, based on their platform or assembly properties (70). RAR, CREB, and STATs were further demonstrated to require different histone acetyltransferase activities within the CBP/p300 complex to activate transcription. In addition, p300 and CBP, despite their similarities, have been recently shown to have distinct functions during retinoid-induced differentiation of embryonic carcinoma F9 cells (71). Finally, B cell-specific OCA-B (72), a group of distinct nuclear proteins termed thyroid hormone receptorassociated proteins (TRAPs) (73), and a transcriptionally active nuclear complex that interacts only with liganded vitamin D receptor (VDR) (DRIPs) (74) define a group of coactivator complexes with rather specific functions. TRAPs purified from HeLa cells grown in the presence of thyroid hormone (T3) were found to markedly activate transcription by liganded TR in vitro, whereas DRIPs consisting of a complex of at least 10 different proteins ranging from 65 to 250 kDa were found to coactivate the VDR-dependent transactivation. DRIPs and TRAPs are similar to each other, sharing common components, but distinct from the CBP-p300 complex, although like these coactivators, their interaction also required the AF-2 transactivation motif of VDR and TR (73,74).
Based on its direct interaction with CBP and SRC-1 (Fig. 6), it is possible that ASC-2 could be included in the putative CBP-p300-SRC-1 complex. Consistent with this proposal, ASC-2 cooperated with SRC-1 and p300 to coactivate nuclear receptor transactivation (Fig. 7). Furthermore, ASC-2 also forms a complex with ASC-1 both in vitro and in vivo, 3 a novel transcription coactivator that in turn forms a complex with CBP and SRC-1 in vivo (65). The microinjection results in which anti-ASC-2 antibody completely abolished the 9-cis-RAdependent transactivation (Fig. 9) suggest that ASC-2 should be integral to sustain the function of this putative transcriptional coactivator complex. The associations with multi-functional CBP/p300/SRC-1 (Fig. 6) also suggest that ASC-2 may function with transcription factors other than nuclear receptors. Indeed, we recently found that ASC-2 is required for transactivation by a group of other transcription factors. 2 Alternatively, ASC-2 may form a distinct coactivator complex in vivo and functionally associates/communicates with other coactivator complexes such as CBP-p300, SRC-1, and ASC-1. Consistent with this notion, our recent preliminary data indicated that ASC-1, ASC-2, CBP/p300, or SRC-1 appears to elute as a distinct complex of proteins from Superose 6 gel filtration column. 4 It is also interesting to note that ASC-2, CBP/p300, and SRC-1 target identical receptor sites (i.e. the AF-2 domains), posing an interesting problem whether all of these factors co-occupy the same sites or assemble in an orderly fashion. The former possibility seems unlikely considering the relatively confined structure of the AF-2 core domain (3-7). Thus, it will be interesting to unravel their putative assembly order or the spatial relations between all of these distinct coactivator molecules in mediating transactivation of nuclear receptors.
Overexpression of amplified genes may provide a selective advantage for tumor growth. Recently, AIB1, an SRC-1 family member, was identified as a gene amplified and overexpressed in breast cancer (51), along with genes designated AIB3 and AIB4 (50). In this study, we found that ASC-2 is identical to AIB3, demonstrating that two distinct coactivator molecules of nuclear receptors can be co-amplified in cancer cells. ASC-2 maps to 20q11, substantially centromeric to AIB1 (which maps to 20q12) (51). Therefore, it is important to note that ASC-2 defines an independent region of amplification in tumors that have been surveyed for their pattern of amplification with probes mapping along 20q (64). Nonetheless, co-amplification of ASC-2 with AIB1 as observed in BT-474 is remarkable considering the related function of these proteins. This is consistent with a model in which the intrachromosomal amplicons (contained within homogeneously staining regions characteristically found in breast cancers) evolve from large chromosomal regions by a process favoring retention of target genes and deletion of biologically irrelevant intervening segments. Such a co-selection process may favor genes, which impinge on the same cellular processes. An analogous situation has been observed in amplicons from 12q which contain two genes, CDK4 and MDM2, which affect the G 1 to S phase transition of the cell cycle without amplification of the region between these genes (75). Co-amplification of AIB1 and ASC-2 may result in significant effects on transcriptional regulation within tumor cells. Overexpression of these multifunctional coactivators could potentially perturb signal integration by these proteins and affect multiple signal transduction pathways. It will be of considerable interest to identify the putative transcription factors, 3  which may be specifically targeted by increased levels of these coactivator proteins in vivo to sustain tumor growth.
Finally, it is notable that the LXXLL motifs (28,63) were not included in the minimum receptor-interaction domain mapped (Fig. 4). This was surprising since ASC-2 appeared to bind specifically the AF-2 domains of receptors in a ligand-dependent manner, also substantiated from the results in which these interactions were entirely abolished with mutations in the AF-2 domain (Fig. 4). In particular, the RXR mutant deleted for the AF-2 domain (i.e. RXR⌬AF) was previously shown to bind 9-cis-RA with wild type affinity (55), excluding the possibility of other nonspecific effects such as inhibiting ligand binding. However, ASC-2 may contain other subregions that could independently associate with receptors. In particular, RXR appeared to weakly bind ASC2-4 (ASC-2 residues 1172-1729), which contains a singly copy of the LXXLL motif, in a 9-cis-RA-dependent manner, as demonstrated in the GST-pull down assays. 5 Thus, this motif could still be involved with the receptor interactions under certain conditions. More detailed analyses to map these interactions are currently under way.
In conclusion, we identified a novel coactivator ASC-2, which is essential for the receptor function in vivo. ASC-2 is amplified and overexpressed in various human cancers and displays various properties of transcriptional coactivator, including an autonomous transactivation potential, the capacity for ligand-dependent interactions with the receptors, as well as interactions with the basal transcription factors and transcription integrators SRC-1 and CBP/p300. Further characterization of ASC-2 should provide important insights into the molecular mechanisms by which nuclear receptors modulate transcriptions as well as the tumorigenesis processes.