Activation Function-1 Domain of Androgen Receptor Contributes to the Interaction between Subnuclear Splicing Factor Compartment and Nuclear Receptor Compartment

In the androgen receptor (AR), most of its transactivation activity is mediated via the activation function-1 (AF-1). By employing yeast two-hybrid assay, we isolated a cDNA sequence encoding a protein binding to AR-AF-1. This protein, named ANT-1 (AR N-terminal domain transactivating protein-1), enhanced the ligand-independent autonomous AF-1 transactivation function of AR or glucocorticoid receptor but did not enhance that of estrogen receptor α. In contrast, the ANT-1 did not enhance any ligand-dependent AF-2 activities. Furthermore, the ligand-independent interaction between AR-AF-1 and ANT-1 was confirmedin vivo and in vitro. The ANT-1 sequence was identical to that of a protein that binds to U5 small nuclear ribonucleoprotein particle, a human homologue of yeast splicing factor Prp6p, involved in spliceosome. ANT-1 was compartmentalized into 20–40 coarse splicing factor compartment speckles against the background of the diffuse reticular distribution. AR colocalized with ANT-1 only in the diffusely distributed area, whereas the ANT-1 speckles were spatially distinct from but surrounded by the AR compartments. The active gene transcription has been shown to couple simultaneously with pre-mRNA processing at the periphery of the splicing factor compartment. The molecular interaction between two spatially distinct subnuclear compartments mediated by ANT-1 may therefore recruit AR into the transcription-splicing-coupling machinery.

Steroid hormone receptors contain the following three functional domains: a variable N-terminal transactivating domain, a highly conserved DNA binding domain, and a moderately well conserved C-terminal ligand binding domain (1,2). AR, 1 like other members of the superfamily, harbors the following two transcription activation functions (AF): constitutively active AF-1 located in the N-terminal transactivating domain of the receptor, and a ligand-dependent AF-2 within the ligand binding domain (3). A number of transcriptional cofactors (coactivator or corepressor) have been identified that associate, in a ligand-dependent fashion, with the AF-2 regions of the steroid hormone receptors. They include the p160 family (4 -7), CBP/p300 (8), PCAF/GCN5 (9), and TRAPs/DRIPs (10). These transcriptional cofactors are organized in multiprotein complexes and facilitate the access of nuclear receptors and the RNA polymerase II core machinery to their target DNA sequences by chromatin remodeling and histone modification. In addition to these transcription cofactors interacting with many nuclear receptors, HBO1 (11), ARA54, ARA55, ARA70, (12)(13)(14)(15)(16), p21-activated kinase PAK6 (17), Caveolin 1 (18), Tip60 (19), and FHL2 (20) have been identified as ligand-dependent ARassociated cofactors.
AR has been thought to be quite unique among the nuclear receptor superfamily members, because most, if not all, of its activities are mediated via the ligand-independent constitutive activity of AF-1 function (21,22). This is in strong contrast to estrogen receptor ␣ (ER␣), in which the overall transactivation capacities are primarily dependent on AF-2 (23). In addition, the interaction of the N-and C-terminal domain is important for exerting the full AR transactivation capacity (21,24,25). The AR shares the hormone-response element sequences on the DNA with the receptors for glucocorticoid (GR), mineralocorticoid, and progesterone (26). In this regard, the N-terminal region, which varies among these receptors, is also thought to be responsible for the cell-and ligand-specific regulation of their target genes (27). In addition to p300/CBP (8) and SRC-1 (21), which interact with both AF-1 and AF-2, ARA24 (12), ARA160 (16), cdk-activating kinase (28), , breast cancer susceptibility gene 1 (BRCA1) (30), SRA (31), and cyclin E (32) have also been proposed to bind to AR N-terminal transactivating domain. Among these cofactors, cyclin E is known to interact with U2 Small nuclear ribonucleoprotein particle (snRNP) (33), thus suggesting that AR-AF-1 might be involved in the pre-mRNA splicing mechanisms. The fundamental role of AR-AF-1 was also further supported by our recent clinical finding (34) showing that the absence of an AR-AF-1-specific transcription coactivator resulted in androgen insensitivity syndrome.
We isolated a cDNA sequence that encoded a protein named ANT-1 (androgen receptor N-terminal domain transactivating protein-1) as the AR-AF-1-binding protein. The ANT-1 was identical to a protein that binds to a human splicing factor U5 snRNP, a human homologue of yeast splicing factor prp6p (35,36) compartmentalized with the yeast spliceosomal penta-snRNP (37). By using a three-dimensional reconstruction of the confocal microscopic images, the spatial interrelationship between two distinct subnuclear compartments, one formed by steroid hormone receptors and another by splicing factors, was elucidated.
Isolation of ANT-1 by Swapped Yeast Two-hybrid Screening and Northern Blot-MatchMaker Plus (CLONTECH) was used for the yeast two-hybrid screening. Poly(A) RNA was isolated from primary cultured human skin fibroblasts. cDNA library was constructed using Time-Saver cDNA Synthesis Kit (Amersham Biosciences) with random primers (Amersham Biosciences) and was inserted into pLexA-BD included in the MatchMaker Plus kit. A cDNA fragment, used for bait, encoding N-terminal transactivating domain (1-532-aa residue) of human AR was ligated in-frame into pB42-AD, thus creating pB42-AD-AF1-expressing AR-AF-1 fused to the GAL1 activation domain. The yeast EGY48 strain was transformed with the pLexA-BD carrying the human fibroblast cDNA libraries and with the pB42-AD-AF1 according to the manufacturer's protocol, and thereafter the transformants were selected on an appropriate nutrition medium. Positive candidate plasmids for AR-AF-1-binding proteins were recovered from the yeast, and the nucleotide sequences were determined using the ABI PRISM 377 DNA Sequencer (PerkinElmer Life Sciences). The specificity of interaction was further confirmed by a liquid galactosidase assay. To obtain fulllength ANT-1 cDNA, the partial cDNA fragment encoding 78 -495-aa residues of ANT-1, obtained by the two-hybrid screening, was 32 Plabeled as a probe for the screening of human prostate cDNA library carried by a gt10 phage vector (CLONTECH). The full-length ANT-1 cDNA was ligated into pcDNA3 to create pcDNA3-ANT-1. For the Northern blot analysis, MTN blots were purchased from CLONTECH. A DNA fragment covering 463-703 amino acids of ANT-1 was labeled as a probe with [␣-32 P]dCTP by the random priming method. Prehybridization and hybridization were performed according to the manufacturer's protocol.
Cell Culture, Transient Transfection Assay-COS-7 and ALVA-41 prostatic cancer cells (5 ϫ 10 5 cells per well in 6-well plate), cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum, were transiently transfected using a Superfect Transfection Kit (Invitrogen). Generally, 2.5 g of plasmid DNA (0.2-1.0 g of pcDNA3-ANT-1, 0.2 g of pCMVhAR, 1.0 g of pMMTV-luc) per well were used for the transfection, whereas the total amounts of the transfected DNA were kept constant by adding pcDNA3 plasmid. At 16 h post-transfection, the cells were rinsed and re-fed with medium containing 10% charcoal-stripped fetal calf serum with or without steroid hormones (10 Ϫ8 M dihydrotestosterone (DHT) or 17␤-estradiol, and 10 Ϫ7 M dexamethasone, respectively). After an additional 18 h, the cells were harvested and assayed for luciferase activities using the Dual-Luciferase Reporter Assay System (Promega).
Immunoprecipitations-For the immunocoprecipitation analysis, COS-7 cells were transfected with plasmids expressing Myc-tagged ANT-1 and full-length or truncated AR and were maintained with or without 10 Ϫ8 M DHT. Whole cell lysates were prepared by lysing cells in a buffer (1.0% Nonidet P-40, 50 mM Tris-Cl, pH 7.8, 150 mM NaCl, 1 mM DTT, 1 tablet of protease inhibitor mixture). In one experiment, nuclear lysates were prepared as described previously (39). The lysates were incubated at 4°C for 1 h with the antibody against c-myc (Santa Cruz Biotechnology) in immunoprecipitation (IP) buffer (0.5% Nonidet P-40, 1 mM EDTA, 50 mM Tris-Cl, pH 7.8, 200 mM NaCl, 1 mM DTT, 1 tablet of protease inhibitor mixture) and then were further incubated with protein-A-Sepharose beads (Amersham Biosciences) at 4°C for 2 h. A Western blot analysis was performed using either antibody N-20 (Santa Cruz Biotechnology) to detect the full-length or N-terminal transactivating domain fragment of AR or using antibody C-19 (Santa Cruz Biotechnology) to detect the C-terminal fragment.
Microscopy and Imaging Analysis-The COS-7 cells were divided into 35-mm glass-bottom dishes (MatTek Corp.) and then were transfected with 0.5 g of pAR-CFP and pANT-1-YFP using 2.5 l/dish of Superfect reagents (Qiagen). Six to 18 h post-transfection, the culture medium was replaced with a fresh Dulbecco's modified Eagle's medium containing 10 Ϫ8 M DHT. One hour after adding DHT, the cells were scanned using a Leica TCS-SP system (Leica Microsystems, Heidelberg, Germany) as described previously (40). The cells were imaged for yellow or cyan fluorescence by excitation with the 514 and 450 nm line, respectively, from an argon laser. The emissions were viewed through either a 530 -590-nm bandpass filter for yellow fluorescence protein (YFP) or a 470 -500-nm bandpass filter for cyan fluorescence protein (CFP). The nuclei were stained with Hoechst 33342 (2 g/ml) and were imaged by excitation with the 350-nm line from an ultraviolet laser, and the emission was viewed through a 400 -450-nm bandpass filter. A series of 30 -50 images were collected for each single nucleus. In each plane, the cyan, yellow, and ultraviolet fluorescence were consecutively collected using serial scanning methods equipped in Leica TCS-SP system. Three-dimensional image reconstruction was performed using either the three-dimensional analysis TRI Graphics Program software package (Ratoc System Engineering, Tokyo) (40) or the deconvolution method (nearest neighbors).

Isolation of ANT-1 with Swapped Yeast Two-hybrid Screen-
ing-To identify the cDNA encoding proteins binding to AR-AF-1 sequence, we performed a yeast two-hybrid screening. Because AR-AF-1 possesses a strong autonomous transactivation capacity, we could not use pLexA-BD, which is regularly used as the bait protein plasmid vector. Instead, we inserted AR-AF-1 cDNA into pB42-AD, and cDNA libraries representing human skin fibroblast mRNAs were carried by the pLexA-BD. By employing this "swapped" modification, the background false-positive blue colonies representing the transactivation capacities of the bait protein in yeast strain EGY48 became very weak, thus allowing us to perform the two-hybrid screening. Although during the second and the third screens retesting the specific ability of the candidate clones was critical, about 2 ϫ 10 5 independent clones were subjected to the screening. We identified one positive clone harboring ϳ1.3 kb of the open reading frame, and then this fragment was used to obtain the full-length cDNA. The translation of the coding sequence revealed that the putative protein consisted of 941-aa residues with a predicted molecular mass of 102 kDa, which was named ANT-1. The homology search of the known amino acid sequences revealed that ANT-1 was identical to a nucleoprotein that binds to a human splicing factor U5 snRNP (GenBank TM accession number AF221842) (35,36). ANT-1 contains 19 tetratricopeptide repeats (TPR) elements, two LXXLL motifs (41), and one leucine zipper motif (Fig. 1a) (42). Typically, the TPR motif appears in a tandem array and thus provides the scaffolds to mediate protein-protein interactions (43). A Northern blot analysis revealed the ANT-1 sequence to be ubiquitously expressed among the tissues examined (Fig. 1b).
Protein-Protein Interactions between ANT-1 and AR, GR, and ER-To confirm the binding capacities of ANT-1 with AR-AF-1, a GST pull-down analysis was performed using a series of GST-fused fragments of AR (Fig. 2a), GR, or ER␣ (Fig.  2b). Ligand-independent binding was observed, not only for full-length AR-(1-919) but also for AR-(1-532) covering AF-1, AR-(622-919) covering the ligand binding domain, or AR-(564 -919) covering DNA binding domain and ligand binding domain. Furthermore, ANT-1 interacted with GR. The truncated GR fragments, such as GR-(14 -438) covering AF-1, GR-(486 -777) covering ligand binding domain, and GR-(421-777) covering ligand binding domain and DNA binding domain, were bound to ANT-1 without ligand. No significant difference was observed between the incubations with or without specific ligands. In contrast, two fragments covering ER-(29 -180) for the AF-1 of ER␣ or ER-(282-595) for the AF-2 region failed to bind to ANT-1. The addition of 17␤-estradiol did not promote such binding (data not shown).
Next, immunocoprecipitation experiments using either whole cell extracts (Fig. 3, a, c, and d) or nuclear extracts (Fig.  3b) were performed to test whether or not ANT-1 binds to AR in living cells. A plasmid expressing Myc-tagged ANT-1 was transfected into COS-7 cells together with expression plasmid for the full-length or truncated mutants of AR, and then the cells were maintained with or without 10 Ϫ8 M DHT. The fulllength AR was specifically precipitated with Myc-tagged ANT-1 in the presence of DHT (Fig. 3, a and b). In addition, AR-AF-1, which exists in the nucleus without DHT (44), was precipitated with Myc-tagged ANT-1 in a ligand-independent fashion (Fig. 3c). In contrast to the in vitro binding in the GST The immunoprecipitation was performed using an antibody against Myc, and the precipitate was subjected to a Western blot analysis using the antibody N-20 for the full-length or AR-AF-1 and the antibody C-19 for the AR-AF-2. In each panel, the middle and the bottom blot represent the Western blot findings of the lysates used for the immunocoprecipitation, as controls. a, immunocoprecipitation of the full-length AR. For the immunocoprecipitation using non-immune normal rabbit serum for lane 6, the lysates for the precipitation was the same as those used for lane 5. NIS, non-immune normal rabbit serum. b, immunocoprecipitation using nuclear lysates. The cells were treated with 10 Ϫ8 M DHT, and the nuclear lysates were prepared previously as described (39). Lanes 1-4 are nuclear (NL) or whole cell (WCL) lysates before the immunocoprecipitation. For the immunocoprecipitation using nuclear lysates (IP/NL), the lysates were the same as those used for lane 2. c, immunocoprecipitation of AR-AF-1 with the Myctagged ANT-1. For lanes 6 and 7, the same whole cell lysates were used. d, immunocoprecipitation of AR-AF-2 with the Myc-tagged ANT-1. For lanes 6 and 7, the same whole cell lysates were used. pull-down experiments, ANT-1 did not bind to AR-AF-2 in living cells (Fig. 3d).
ANT-1 Enhances AR-and GR-mediated but Not ER-mediated Transactivation Activities-To examine the effect of ANT-1 on the transactivation function of AR, GR, or ER␣, we cotransfected an expression plasmid for each receptor together with the plasmids expressing ANT-1 and an appropriate reporter plasmid (pMMTV-luc for AR or GR, pERE2-tk109-luc for ER␣, respectively) into ALVA-41 human prostatic cancer cells or COS-7 cells. The reporter gene lucϩ (CLONTECH) harbored in pMMTV-luc does not contain any intronic sequences. As shown in Fig. 4a, ANT-1 further enhanced the ligand-induced transactivation function of full-length AR and GR (10 Ϫ8 M DHT and 10 Ϫ7 M dexamethasone, respectively). When the plasmid expressing the truncated AR-AF-1 (aa residues 1-622) or AR-AF-2 (aa residues 563-919) was transfected, ANT-1 enhanced the constitutive transactivation function mediated by AR-AF-1 by from 2-to 3-fold. However, no enhancement of the liganddependent transactivation mediated by AR-AF-2 was observed. A similar profile of the domain-specific transactivation was also observed for GR. In contrast, ANT-1 did not enhance the ER␣-dependent transactivation (Fig. 4b). Whereas the ANT-1 is identical to the human homologue (the p102 U5 snRNP-binding protein) of the yeast splicing factor prp6p, involved in the spliceosome (37,45), the transfection of ANT-1 did not enhance the basal promoter activities of reporter plasmids measured by luciferase activity (data not shown). In addition, the reporter gene lucϩ (pGL3-luc from CLONTECH) harbored in pMMTV-luc does not contain any intronic sequences. Therefore, it was suggested that the observed enhancements of the reporter gene activities were not mediated by the nonspecific enhancement of the splicing activities but were mediated through AR-or GR-dependent transcription. Together with the findings in the immunocoprecipitation experiments, we concluded that ANT-1 is primarily an AF-1 interacting transcriptional coactivator for AR or GR but not for ER␣.
The ANT-1-mediated transactivation was further confirmed in a reporter assay using natural authentic promoter (P450 aromatase exon Ib upstream promoter (38)) transactivated by the endogenous GR in primary-cultured human skin fibroblast. In this experiment, the expression of ANT-1 further enhanced the GR-mediated aromatase gene promoter activities (Fig. 5a). Furthermore, ANT-1 exhibited an independent additive effect on the transactivation function of cyclin E (32) or SRC-1␣ (21), which is reported to interact with AR-AF-1 (Fig. 5b). When either ANT-1-truncated mutant covering the N-terminal half (78 -495 aa), ANT (N), or ANT (C) covering the C-terminal half (495-941 aa) including one leucine zipper motif together with two LXXLL (Fig. 1a) motifs was expressed, ANT (N) alone enhanced the AR-mediated transactivation almost fully (90%) (Fig. 5c). This suggested that the transactivation domain(s) of ANT-1 exist within 78 -495-aa residues.
Subcellular Localization of ANT-1-By using a high resolution three-dimensional reconstruction of the confocal microscopic images, we previously showed that transcriptionally active AR produces 250 -400 small subnuclear speckles (Fig. 6a, applying a surface method (40)), which are shared with GR or ER. The liganded AR recruited the transcriptional cofactors such as SRC-1␣, TIF-II, and p300/CBP into these subnuclear speckles (44). We were interested in the subnuclear spatial interrelation between nuclear receptor compartment, colocalizing with p160 members and p300/CBP, and splicing factor compartment. Therefore, the spatial interrelation of AR-CFP with ANT-1-YFP was explored in detail using a three-dimensional image analysis. A volume method in three-dimensional reconstruction showed the nuclear receptor speckles (fluorescence foci) and revealed many small spatial "pockets" where no cyan fluorescence was observed as in the nucleolus (Fig. 6b, cyan fluorescence was digitally converted into red as pseudocolor). The subnuclear localization of ANT-1 was clearly distinct from that of AR (Fig. 6c). In a good agreement with the subnuclear distribution of prp6p (yeast homologue of ANT-1) (45), the ANT-1 distribution was identical to the known distribution pattern of splicing factors (46 -48). The transfected ANT-1-YFP distributed in the nucleus in two distinct patterns as follows: a diffuse fine reticular distribution throughout nucleus, devoid of a nucleolus, and a coarsely clustered distribution (speckles) known as the splicing factor compartment, both of which were exclusively in the euchromatin region where the Hoechst 33342 staining was less dense (Fig. 6d). When images of AR-CFP (Fig. 6b) and ANT-1-YFP (Fig. 6c) were spatially merged, the CFP and YFP fluorescence was colocalized only in the diffuse fine distribution (Fig. 6e, merged area is represented in orange). To focus on the spatial interrelation between ANT-1 speckles (splicing factor compartment) and AR-CFP (nuclear receptor speckles), the diffuse fine reticular distribution of ANT-1-YFP was cut off and expressed as a blank image (Fig. 6f). As a result, it became clear that the ANT-1 speckles fall into the small spatial pockets of the cyan fluorescence volume (Fig. 6, b, g, and h). When the image was digitally magnified (Fig. 6, i-l), the ANT-1 volume was shown to possess a rough surface, which was surrounded by a spatial mass representing transcriptionally active ARs without merging with each other (Fig. 6, i and k, surface views; Fig. 6, j and l, tomographic views).

DISCUSSION
By employing a swapped yeast two-hybrid system using AR-AF-1 as the bait, we identified a novel coactivator of AR-AF-1, named ANT-1. Functional reporter assays and the immunocoprecipitation experiments demonstrated that the ANT-1 is essentially the AF-1-interacting transcriptional coactivator. The sequence of ANT-1 contains 19 TPR elements. The TPR motif consists of 34-amino acid residues containing the 8 loosely conserved sequence: WLGYAFAP (43). So far, more than 25 proteins possessing the TPR motif have been identified, and we have been shown to play important roles in the diverse biological functions, including transcriptional repression (49), a phosphorylation of nucleoproteins (50), and pre-mRNA splicing (51). The results shown in Fig. 5c suggest that the first 6 TPR motifs alone and/or upstream N-terminal sequence play a fundamental role in the ANT-1 function.
Whereas most of the transcriptional cofactors so far reported are known to interact with AF-2, the number of transcriptional cofactors specifically interacting with the AR-AF-1 sequence is limited. Interestingly, molecules that have been shown to bind to AR-AF-1 are apparently unique. They include SRA, ARA24 (12), ARA160 (16), BRCA1 (30), cdk-activating kinase (28), 5. a, the enhancement of the human aromatase exon Ib promoter activity. The reporter plasmid pAROM-luc driven by the human aromatase exon Ib promoter (Ϫ1101 to ϩ60) was cotransfected into the primary-cultured human forearm skin fibroblasts with expression plasmids for ANT-1. 10 Ϫ7 M dexamethasone was added to activate the gene by the endogenous GR binding to the glucocorticoid-response element located from Ϫ333 to Ϫ319 of the promoter Ib sequence. b, additive effects of ANT-1 with cyclin E or SRC-1a on the AR-dependent transactivation. The cells were transfected with the AR expression plasmid (0.2 g) and also underwent the cotransfection of plasmids expressing 1.0 g of ANT-1, cyclin E, SRC-1a, or the combination of the plasmids as shown. The total amount of the transfected DNA was kept constant (3.2 g) by adding pcDNA3 empty vector plasmid. The cells were cultured with or without 10 Ϫ8 M DHT (dashed bar, and open bar, respectively). c, the N-terminal 78 -495-aa residues of ANT-1 is essential for the transactivating function of ANT-1. The two truncated mutants shown in Fig. 1a were used for the transfection experiments.
ART-27 (29), and cyclin E (32). For example, SRA (31) is an RNA molecule that has been shown to enhance the transactivation functions of the progesterone receptor, ER, GR, and AR by binding to AF-1. BRCAI, the mutation of which may play a role in the carcinogenesis of breast or ovary, also is thought to have implications regarding the proliferation of normal and malignant androgen-regulated tissues. In addition to these molecules, we presented evidence that ANT-1, p102 U5 snRNPbinding protein, binds to AR-AF-1. Together with the finding that cyclin E also interacts with U2 snRNP, AR-AF-1 may be involved in the spliceosomal machinery.
Transcripts synthesized by RNA polymerase II should undergo specific and extensive processing, including capping at the 5Ј end, addition of the poly(A) tail, and the removal of the intervening sequences called splicing, before being transported into the cytoplasm. Especially in view of the splicing step, snRNPs play critical roles in composing the multiprotein-RNA complex called spliceosome. The splicing snRNPs (U1, U2, U4/ U6, and U5) associate with pre-mRNA and with each other in an ordered sequence to form spliceosome, during which U5 snRNA base-pairs with exon sequences flanking the split sites. Furthermore, a component of U5 snRNP, called U5-200-kDa protein, which shares homologies with the DEAD box families of RNA helicases, was shown to unwind U4/U6 RNA duplexes (52). Interestingly, p72/68, a subfamily of the RNA helicase possessing DEAD box, functions as a novel class of ER␣ coac-tivator by enhancing the AF-1 transactivation capacity (39). It has been hypothesized that the active gene transcription simultaneously coupling with pre-mRNA processing may occur at the periphery of splicing factor compartment, and this process is called cotranscriptional splicing or "transcription-splicing coupling" (53)(54)(55). We presented further data supporting that transcription/splicing coupling may be evoked in a nuclear receptor-dependent fashion. This was first described in a peroxisome proliferator-activated receptor-␥-coactivator-1 (PGC-1), which possesses a pre-mRNA splicing activity itself (56). In AR, in addition to cyclin E-U2 snRNP interaction (33), FHL2, a tissue-specific AR cofactor (20), has been shown to interact with the polypyrimidine tract binding protein-associated splicing factor (57). However, FHL2 binds neither AR-AF-1 fragment nor AF-2 fragment alone but instead the full-length of the complete structure of AR is required for such binding. When the splicing activity of ANT-1 per se was assessed using an artificial luciferase minigene (kindly provided by Dr. S. Kato, University of Tokyo), we observed an increase in the minigene transcripts without enhancement of the splicing efficiency (data not shown). One possible explanation is that ANT-1 per se does not possess any splicing activity, but instead it functions as a transcriptional coactivator recruiting the active AR or GR into the active site of gene transcription coupled with mRNA splicing (54,58). In this regard, it is shown that specific sets of proteins recruited as transcription coactivators function as FIG. 6. A three-dimensional analysis of the subnuclear compartmentalization of AR and ANT-1. COS-7 cells were transfected with plasmids expressing AR-CFP fusion, ANT-1-YFP fusion, or both, treated with 10 Ϫ8 M DHT, and then were stained with Hoechst 33342 (2 g/ml) to visualize the chromatin structures (blue). A three-dimensional reconstruction was performed as described previously (40) for a, whereas deconvolution methods (nearest neighbors) were applied for other images. The AR-CFP is visualized in red as pseudocolor for all panels. For the chromatin images, less densely stained areas (namely euchromatin region) were shown as blank images, and densely stained areas (heterochromatin region) were shown as blue. a, the surface view of spatial merge of AR-CFP with chromatin structures. b, the surface view of AR-CFP. Note that in addition to the two nucleoli (nc), many small spatial pockets can be observed. c, the surface view of ANT-1-YFP compartments. d, the surface view of spatial merge of ANT-1-YFP with chromatin structures. e, the surface view of spatial merge of AR-CFP with ANT-1-YFP. f, the surface view of ANT-1-YFP compartments. To highlight the ANT-1 speckles, the diffusely distributed fine reticular network found in c was cut off and is shown as blank image. g, the surface view of the spatial merge of AR-CFP with ANT-1-YFP (f). The ANT-1 spatial mass falls into the several pockets found in b. h, the surface view of the spatial merge of AR-CFP, ANT-1-YFP, and chromatin images. i, the surface view of the spatial merge of AR-CFP, ANT-1-YFP, and chromatin images expressed in the volume method. The area indicated in the white rectangle in h is magnified in this view. j, the tomographic view of i. k, the magnified image highlighted with dashed rectangle in i. l, the tomographic view of k. The spatial mass representing the AR speckle does not merge with the YFP speckle, but instead it comes in contact with the YFP speckle at the periphery of the YFP speckle. bridge proteins coupling the transcription factor and mRNA processing (59). However, further studies using the actual ARtarget gene in place of the luciferase minigene still need to be performed.
In the nucleus, there exist different sets of functional compartments often called "foci" or "speckles," which include SFC that consists of nearly 20 -50 large speckles (55), and nuclear receptor speckles possibly associated with the nuclear matrix structures (40, 44, 60 -63). In contrast to the cytoplasmic compartments, the subnuclear compartments are not sequestered by the membrane structures, thus allowing the rapid movement of component proteins across the compartment. Splicing factor compartments consist of many protein complexes including snRNPs. We first visualized the spatial relationship between the steroid hormone receptor compartment and splicing factor compartment, thus revealing where the molecules locating in two distinct subnuclear compartments interact with each other. One is a nuclear receptor speckle, whereas another is the splicing factor compartment. We showed previously that the transcriptionally active ARs recruit chiefly AF-2-interacting transcriptional cofactors, such as TIF-II, with the functional interaction between AF-1 and AF-2, thus producing the fine subnuclear speckles in a p300/CBP-dependent fashion (44). In the three-dimensional image reconstruction, AR colocalized with ANT-1 only in diffusely distributed areas, whereas ANT-1 speckles representing splicing factor compartment were spatially distinct from but surrounded by the AR compartments. Recent studies (48,64) have shown that the splicing factor compartment may represent the site for the storage and/or assembly of the splicing factors and that splicing factors can be rapidly recruited from the splicing factor compartment into the active sites of transcription. Furthermore, this high mobility is not restricted to the splicing factors but is a general feature of nuclear proteins, namely many nuclear proteins roam the cell nucleus in an energy-independent fashion (65). In this regard, the merging of the diffuse ANT-1 distribution with AR speckles near the splicing factor compartment may represent where the ANT-1 or ANT-1-snRNP complex meets the active AR-cofactor complex. The binding of ANT-1 with AR or GR might thus play a key role in the interaction between these two distinct sets of transcription factors located at distinct subnuclear compartments, namely the recruitment of transcriptionally active AR from the nuclear receptor compartment colocalizing with p160 and p300/CBP coactivators into the transcriptionsplicing coupling machinery formed near the periphery of the splicing factor compartment. Furthermore, ANT-1 may selectively recruit AR or GR, whereas ER, in which AF-1 transactivation is much weaker than that of AR or GR (23), is not recruited. Different sets of nuclear receptors, such as ER or peroxisome proliferator-activated receptor-␥, may therefore possess different compartment to compartment interaction mechanisms.