INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

The nuclear receptor superfamily makes up a large group of transcription factors that play diverse roles in cell growth, differentiation, development, and homeostasis (35, 37, 58). The superfamily is composed of type I receptors that mediate the actions of steroid hormones and type II receptors that mediate the actions of a variety of structurally diverse ligands such as thyroid hormone, retinoids, and 1, 25-(OH)₂ vitamin D₃, as well as a number of orphan receptors whose ligands (if any) remain unknown (34, 35). Members of the superfamily share similar domain structures. Generally, a receptor molecule consists of a highly variable N-terminal domain (region A/B), a highly conserved central DNA-binding domain (region C), and a C-terminal ligand-binding domain (LBD; region DEF) that is diverse in sequence but exhibits some conservation in overall structure (5, 18, 58, 61). Typically, a receptor harbors two activation functions: an AF1 within the A/B region and a ligand-dependent AF2 in the LBD (3, 13, 39, 43, 56).

Ligand binding is a critical event in receptor biology, since it induces a conformational change in the receptor that bears broad functional consequences. For example, in the absence of ligand, type I receptors are associated with heat shock protein chaperones and do not bind DNA (5, 44, 45, 46). Ligand binding dissociates these receptors from the chaperones, which enables both their binding to target DNA sequences and the recruitment of coactivators that leads to transactivation (37). In contrast, type II receptors are not associated with heat shock proteins and appear to bind DNA in the absence of their ligands (11, 17, 51). As a consequence, for some of the type II receptors (such as TR and RAR), their unliganded forms can function to repress transcription through the recruitment of corepressor complexes (6, 10, 25, 43). Thus, in such cases, ligand binding promotes both the dissociation of corepressors and recruitment of coactivators, resulting in receptor-mediated transactivation (20).

Efforts to gain deeper insights into the molecular mechanisms of receptor-mediated transactivation have led to the identification of a number of putative coactivators (20, 37), such as the p160 family (1, 9, 24, 28, 32, 41, 55, 57, 59), CBP/p300 (7, 22, 28), the TRAP/DRIP complexes (16, 27, 47, 48), PGC-1 (42), p/CAF (4), NRIF3 (31), and hNRC (33). For many of these coactivators, their interaction with liganded nuclear receptors appears to involve one or more LXXLL motifs contained within their receptor interacting domains (RIDs) (20, 23, 37). The combination of molecular biological and structural studies has indicated that ligand binding induces a conformational change in the receptor LBD that repositions helix 12 which, together with helices 3, 4, and 5, forms a hydrophobic cleft that serves as the docking site for the LXXLL motif contained within the RIDs of coactivators (12, 14, 40).

Members of the p160 family are among the best-characterized examples for coactivator-receptor interactions (20, 37). Their receptor interaction domains contain three LXXLL boxes (often referred as NR boxes), which are in turn differentially utilized by different nuclear receptors (12, 36). For example, coactivation of ER requires the second LXXLL box of SRC-1/NCoA-1, while the function of TR or RAR requires both the second and the third LXXLL boxes (36). Evidence from several studies has suggested that sequences surrounding the LXXLL core are important in determining the specific utilization of different LXXLL boxes by different nuclear receptors (12, 36). This notion is reinforced by a recent study with combinatorial libraries to isolate different LXXLL-containing peptides that exhibit diverse receptor interaction patterns (8). Interestingly, although the various LXXLL boxes of a p160 family member such as SRC-1 show differential receptor preferences, the entire coactivator molecule does not appear to exhibit receptor specificity, since it generally interacts with many nuclear receptors (37).

We recently reported the cloning of a novel coactivator (NRIF3) from a yeast two-hybrid screen (31). One of the unique properties of NRIF3 is its receptor specificity. Most known coactivators identified thus far (such as the p160 family) appear to interact with many nuclear receptors (37). NRIF3, in contrast, only interacts with liganded TR and RXR but not any other examined receptors (31), raising an interesting question about the mechanism of its receptor specificity. Our previous study identified an essential RID (referred to here as RID1) at the C terminus of NRIF3 which, interestingly, contains an LXXIL module, a variant of the canonical LXXLL motif (31). Computer modeling and mutagenesis analyses have indicated that RID1 docks to the hydrophobic groove of a liganded LBD through its LXXIL motif in a way similar to that of the canonical LXXLL module utilized by other coactivators (31). However, RID1 alone does not appear to harbor the same specificity as NRIF3. For example, while NRIF3 interacts with only TR and RXR but not RAR, RID1 was found to interact with all three receptors (31). Thus, the molecular mechanism underlying the receptor specificity of NRIF3 remained unclear.

In this report, we extended our analysis of NRIF3 in a number of areas. First, we further analyzed the NRIF3-receptor interaction in order to gain additional insight into the receptor specificity of NRIF3. These results, together with our previous study, suggest a bivalent interaction model, in which a single NRIF3 molecule utilizes both the C-terminal LXXIL (RID1) and the N-terminal LXXLL (RID2) modules to cooperatively interact with the receptors (presumably a receptor dimer). The spacing between RID1 and RID2 appears to play a critical role in influencing the affinity of the interactions and thus is likely a determinant in the receptor specificity of NRIF3. Second, during the course of such analyses, we also uncovered and mapped a dimerization domain in the middle of the NRIF3 molecule, which may have functional implications in NRIF3 action(s). Third, by using Gal4 fusion constructs, we found that NRIF3 harbors both transactivation and transrepression functions. A transactivation domain (AD1) and a novel repression domain (RepD1) were mapped to residues 162 to 177 and residues 20 to 50, respectively. Fourth, we also analyzed the two alternatively spliced isoforms of NRIF3. Fluorescence microscopy showed that, like NRIF3, these two isoforms are also nucleus localized. Consistent with the fact that both isoforms contain the same RepD1 as NRIF3, Gal4 fusions of these two isoforms repress the transcription of a Gal4 reporter. Since these two isoforms do not interact with nuclear receptors because they both lack the essential RID1, our study raises the possibility that they may function as coregulators for other transcription factors. Taken together, these studies suggest that the NRIF3 gene encodes an interesting family of coregulators that, collectively, may play dual roles in mediating both positive and negative regulation of gene expression.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Plasmids for the yeast two-hybrid assay. All yeast plasmids expressing various LexA fusion proteins were constructed from a derivative of pEG202 (21) that contains a modified polylinker. Such plasmids include LexA-NRIF3, LexA-cTR LBD, LexA-hRXR LBD, and LexA-RID1. All yeast plasmids expressing various B42 fusions were derived from pJG4-5 (21). These plasmid include the following: B42-cTR LBD(120-408), B42-cTR LBD(120-398), B42-hRXR LBD, B42-hRAR LBD, B42-RID1, B42-NRIF3, B42-EnS, B42-EnL, B42-NRIF3(112-177), B42-NRIF3(112-161), B42-NRIF3 L9A, B42-NRIF3(87-111), B42-NRIF3 L89R, B42-NRIF3 L96R, and B42-NRIF3 DM.

Domain and mutagenesis analyses. The plasmid expressing LexA-RID1 has been described previously (as LexA-NCD) (31). LexA-NRIF3 was constructed by ligating the full-length NRIF3 fragment derived from pEx-NRIF3 (31) by NcoI and XhoI digestions into a pEG202 vector digested with the same pair of enzymes. To construct B42-RID1, synthetic oligonucleotides that encode the RID1 region of NRIF3 (residues 162 to 177) were annealed and ligated to a pJG4-5-derived vector digested with NcoI and XhoI. B42-NRIF3, B42-EnS, B42-EnL, and B42-NRIF3 L9A have been described previously (31). To construct B42-NRIF3(112-177), primers PNC1 (5'-CGC GAC GTG CCA TGG CTT TGG AGG GCA GTA GAG AGC-3') and NFCP2 (5'-CGC GAC GTG AGA TCT CGA GCT GGT ATT TAC TGG GCA G-3') were used to PCR amplify the cognate region of NRIF3. The PCR product was then digested with NcoI and BglII and cloned into a pJG4-5-derived vector digested with the same pair of enzymes. The resulting construct was confirmed by sequencing analysis. B42-NRIF3(112-161) was constructed in two steps. First, pEx-NRIF3 was digested with BstZ17I and BglII and ligated to annealed oligonucleotides encoding residues 162 to 177 of NRIF3 to generate pEx-NRIF3(112-161). pEx-NRIF3(112-161) was then digested with NcoI and XhoI, and the resulting NRIF3(112-161) fragment was ligated into a pJG4-5-derived vector digested with the same pair of enzymes. B42-NRIF3(87-111), B42-NRIF3 L89R, B42-NRIF3 L96R, and B42-NRIF3 DM were all constructed by PCR-based methods. The primers used for these PCRs were PNC3 (5'-CGC GAC GTG GAA TTC GCT TTG GAG GGC AGT AGA GAG C-3'), PNC4 (5'-GAA GTT GGT GCT CAT GGT GAG TGC-3'), PNC5 (5'-GAC AAT GAT GAA TTC ATG ATG AGA CTA TCA AAA GTT GAG AAA TTG TCA GAA G-3'), PNC6 (5'-ATG ATG AAT TCA TGA TGT TGC TAT CAA AAG TTG AGA AAA GAT CAG AAG AAA TCA TGG AG-3'), and PNC7 (5'-ATG ATG AAT TCA TGA TGA GAC TAT CAA AAG TTG AGA AAA GAT CAG AAG AAA TCA TGG AG-3'). PNC4 was the common downstream primer used for all PCRs. The upstream primers were as follows: PNC3 for B42-NRIF3(87-111), PNC5 for B42-NRIF3 L89R, PNC6 for B42-NRIF3 L96R, and PNC7 for B42-NRIF3 DM. For each of these PCRs, the resulting product was digested with EcoRI and ligated to a B42-NRIF3 vector that was digested with the same enzyme. All resulting constructs were confirmed by sequencing analysis. The suitable combinations of plasmids expressing various LexA or B42 fusion proteins were then used in a yeast two-hybrid assay to examine protein-protein interactions (31).

The yeast two-hybrid and in vitro binding assays. The bait and prey plasmids used for the yeast two-hybrid assay in this study have been described above. Generally, the yeast strain EGY48 harboring the LacZ reporter plasmid (pSH18-34) (21) was transformed with appropriate bait and prey plasmids. For each transformation, 8 to 10 transformants were randomly selected and analyzed on appropriate X-Gal (5-bromo-4-chloro-3-indolyl--D-galactopyranoside) plates for preliminary evaluations. Typical colonies were then selected for quantitative -galactosidase assays as previously described (31). For in vitro binding assays, ³⁵S-labeled wild-type cTR and the mutant cTR L398R were generated by in vitro transcription and translation with a reticulocyte lysate system (Promega). The glutathione S-transferase (GST) control and GST-NRIF3 were expressed in Escherichia coli and affinity purified with glutathione-agarose beads (31). The in vitro binding assay was then carried out as previously described (31), with a slightly modified binding buffer (20 mM HEPES [pH 7.8], 1 mM MgCl₂, 1 mM dithiothreitol, 10% glycerol, 0.05% Triton X-100, 1 µM ZnCl₂, 150 mM KCl, 0.15 mg of bovine serum albumin/ml).

Transfection studies. The G5-tk-chloramphenicol acetyltransferase (CAT) and G5-SVB-CAT reporters used in this study have been described previously (43). Various Gal4 fusion constructs have been described above. Appropriate plasmids were transfected into HeLa cells by calcium phosphate coprecipitation, with typically 2 µg of G5-tk-CAT or 500 ng of G5-SVB-CAT. When appropriate, the Gal4 control or various Gal4 fusion vectors (400 ng to 1.2 µg) were cotransfected. After transfection, cells were incubated at 37°C for 42 h before being harvested. CAT assays were then carried out, and CAT activities were calculated as previously described (31). The experiments were repeated two to four times, with similar results. GH4C1 cells were transfected by using the Geneporter 2 reagent (Gene Therapy Systems). At 24 h before transfection, cells were set at 1.5 million per well in a six-well plate. Transfections were carried out according to the manufacturer's protocol. Typically, 400 to 750 ng of G5-tk-CAT and 250 to 500 ng of various Gal4 fusion constructs were used for each transfection. When appropriate, 2 µg of empty control vector or 2.5 µg of expression vectors for NRIF3, EnL, or EnS were cotransfected. About 48 h after transfection, cells were harvested for the determination of CAT activities. When applicable, the fold repression was calculated by comparing the CAT activity from the reporter transfected alone with that from the reporter cotransfected with the examined Gal4 fusion.

Fluorescence microscopy. The green fluorescent protein (GFP) fusion technique was used to study the subcellular location of examined proteins. Vectors expressing GFP-EnS (29) and GFP-EnL were provided by Sanford Shattil. The GFP control and GFP-NRIF3 vectors have been described previously (31). Each vector was transfected into HeLa cells by calcium phosphate coprecipitation. After transfection, cells were incubated at 37°C for 24 h before the examination with a fluorescence microscope to determine the subcellular location of the examined protein (31).

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

NRIF3 and its isoforms. We initially cloned NRIF3 in a yeast two-hybrid screen as a factor interacting with TR in a ligand-dependent manner (31). A search of the GenBank identified two highly related proteins, which are alternatively spliced products of the same gene (31). These two proteins were previously designated 3-endonexin short (referred to here as EnS) and long (referred to here as EnL) forms (54). EnS was cloned from a yeast two-hybrid screening with the cytoplasmic tail of 3-integrin as bait (54). EnL was then cloned as an alternatively spliced product of the same gene. However, EnL does not bind to integrin 3 (54). Interestingly, despite their extensive identity with NRIF3 (Fig. 1), our previous study indicated that EnS and EnL do not interact with nuclear receptors (31).

View larger version (12K): [in this window] [in a new window]   FIG. 1.   Domain organization of NRIF3 and its isoforms. NRIF3 consists of 177 amino acids. EnS consists of 111 amino acids and is 100% identical to the corresponding region of NRIF3. The first 161 amino acids of EnL are identical to those of NRIF3. EnL and NRIF3 differ in their unique C termini (9 residues in EnL, filled box; 16 residues in NRIF3, hatched box). The unique C terminus of NRIF3 (hatched box) harbors RID1, which contains an LXXIL motif. Another RID, RID2, is located at the N terminus of NRIF3 and contains the canonical LXXLL motif. A coiled-coil (dimerization) domain is mapped to the center of the NRIF3 molecule (residues 84 to 112, waved box) and is also found in EnS and EnL. This region also contains a putative leucine zipper-like motif. A transactivation domain (AD1) is mapped to the unique C terminus of NRIF3 (hatched box), a region that also harbors RID1. A transrepression domain (RepD1) is mapped in the N-terminal portion of NRIF3 (residues 20 to 50, dotted box), a region also common to EnS and EnL.

View larger version (92K): [in this window] [in a new window]   FIG. 2.   EnS and EnL are nucleus localized. HeLa cells were transfected with an expression vector for GFP-EnS or GFP-EnL. Cells were then incubated at 37°C for 24 h before being examined under a fluorescence microscope to determine the subcellular location of the fusion protein. GFP fusions of both EnS and EnL are localized in the nucleus. The control GFP alone was found to be distributed throughout the cell (data not shown). GFP-NRIF3 was used as another control, and its pattern was found to be similar to that of GFP-EnS or GFP-EnL (data not shown).

Integrity of the receptor AF2 helix is required for the NRIF3-TR interaction. Our previous study indicated that a unique C-terminal domain in NRIF3 (referred to here as RID1; see Fig. 1) is essential for its interaction with liganded TR and RXR (31). The importance of RID1 is highlighted by the fact that EnS and EnL, the two alternatively spliced isoforms of NRIF3 that lack the RID1 (Fig. 1), do not interact with TR or RXR (31). RID1 contains an LXXIL motif (Fig. 1), a variant of the canonical LXXLL motif. On the basis of a combination of computer modeling and subsequent experimental analysis, we proposed a model where RID1 docks to the hydrophobic groove of the liganded TR or RXR LBD via its LXXIL motif in a fashion similar to that of the canonical LXXLL motif (31). In support of this model, we found that LexA-RID1 can directly interact with the liganded LBDs, and such interaction is completely abolished when the conserved hydrophobic residues of the LXXIL core are mutated (31).

View larger version (23K): [in this window] [in a new window]   FIG. 3.   Integrity of helix 12 (the AF2 helix) is essential for the NRIF3-TR interaction. (A) 35S-labeled wild-type TR (WT) or the mutant TR (L398R) was generated by in vitro translation. The labeled receptors were then examined for binding to purified GST control or GST-NRIF3 in the presence (+) or absence (-) of T3 as described in Materials and Methods. (B) Yeast two-hybrid assay. LexA-NRIF3 was examined for interaction with the control B42 alone, B42-TR LBD(120-408), or B42-TR LBD(120-398) that deletes 10 amino acid residues from helix 12. -Galactosidase activities were determined in the presence (shaded columns) or absence (open columns) of T3.

Interactions of the NRIF3 RID1 with receptor LBDs in yeast are fusion partner dependent. Yeast two-hybrid assays were used to further characterize NRIF3-receptor interactions and to map an essential receptor interacting surface of NRIF3 (RID1) to its C terminus (31) (Fig. 1). As shown in Fig. 4, LexA-RID1 interacts with the LBDs of TR and RXR (fused to B42) in a ligand-dependent manner, while LexA alone exhibits no such interactions (data not shown). However, in a reciprocal experiment with LexA-TR (or RXR) LBD as the bait and B42-RID1 as the prey, we did not observe any interactions with or without ligand (Fig. 4). As a positive control, LexA-TR LBD or LexA-RXR LBD was found to interact with full-length B42-NRIF3 in a ligand-dependent manner (Fig. 4). Therefore, the interactions of RID1 with receptor LBDs appear to depend on the identity of its fusion partner, with the LexA-RID1 proficient and B42-RID1 deficient in the interaction.

View larger version (44K): [in this window] [in a new window]   FIG. 4.   Interactions of RID1 with liganded LBDs in yeast are fusion partner dependent. The indicated baits LexA-RID1, LexA-TR LBD, and LexA-RXR LBD were examined for interactions with the indicated preys (as B42 fusions) in a yeast two-hybrid assay as described in Materials and Methods. -Galactosidase activities were determined in the presence (shaded columns) or absence (open columns) of cognate ligands: T3 for TR, 9-cis RA for RXR.

NRIF3-NRIF3 interaction. A plausible explanation for the difference in the abilities of LexA-RID1 and B42-RID1 to interact with the receptor LBDs is that two copies of RID1 are required for efficient association with the liganded LBDs. However, we cannot exclude the possibility that the B42-RID1 fusion protein might be unstable or fold incorrectly. LexA-RID1 can potentially provide two copies of RID1 for interaction since the LexA DNA-binding domain (DBD) binds to its cognate operator sequences as a dimer (30, 38). In contrast, the B42 activation domain is not known to dimerize and therefore the B42-RID1 monomer would be expected to be monovalent for RID1. Thus, the functional differences found between LexA-RID1 and B42-RID1 raise the possibility that bivalency might be required for a productive interaction between the RID(s) of NRIF3 and the liganded receptor LBDs.

View larger version (22K): [in this window] [in a new window]   FIG. 5.   Characterization of the NRIF3-NRIF3 interaction. LexA-NRIF3 was studied in a yeast two-hybrid assay for interactions with various B42 fusions as depicted in the figure. The region of the coiled-coil domain is indicated as a waved box.

A central coiled-coil domain is essential for the NRIF3-NRIF3 interaction. We further defined the NRIF3 dimerization surface in order to generate an appropriate mutant(s) to examine whether dimerization of NRIF3 plays a role in its interaction with TR or RXR. Structural analysis of NRIF3 by computer modeling (52) revealed the presence of a putative coiled-coil domain in the middle of the NRIF3 molecule that spans amino acid residues 84 to 112 (Fig. 1 and 6A). Close inspection of this putative coiled-coil domain identified a leucine zipper-like motif (Fig. 6A). Interestingly, virtually the entire portion of this putative coiled-coil domain is contained within the region of NRIF3 found to be involved in its dimerization (Fig. 5).

View larger version (32K): [in this window] [in a new window]   FIG. 6.   The central coiled-coil domain is essential for the NRIF3-NRIF3 interaction. (A) Amino acid sequence of the coiled-coil domain of NRIF3 (residues 84 to 112). The putative leucine zipper-like structure is shown in boldface and underlined, with the occurrence of Leu, Leu, Ile, and Ile at every seventh position between residues 89 and 110. Leu89 and Leu96 are marked with asterisks. (B) LexA-NRIF3 (WT) was examined in a yeast two-hybrid assay for interactions with the following preys (as B42 fusions), respectively: wild-type NRIF3 (WT), mutant NRIF3 with an internal deletion of the coiled-coil domain (1), and mutant NRIF3s L89R, L96R, and DM.

Dimerization of NRIF3 is not required for its interaction with TR or RXR. If the NRIF3-receptor association involves a bivalent interaction, one possible model is that dimerization of NRIF3 facilitates the interaction of two copies of RID1 with the receptor LBDs. To test this possibility, we examined the three NRIF3 mutantsL89R, L96R, and DMfor their interactions with TR and RXR. As shown in Fig. 7, L89R and L96R, which are still capable of dimerization, remain proficient in interacting with TR and RXR LBDs in a ligand-dependent manner. In addition, the DM mutant that is deficient in NRIF3-NRIF3 dimerization exhibited interactions with liganded LBDs of TR and RXR which were equal to or greater than those of wild-type NRIF3 (Fig. 7), suggesting that NRIF3 dimerization is not required for NRIF3-receptor interactions.

View larger version (46K): [in this window] [in a new window]   FIG. 7.   Dimerization of NRIF3 is not required for receptor interactions. The baits LexA-TR LBD and LexA-RXR LBD were examined in a yeast two-hybrid assay for interactions with the following preys (as B42 fusions), respectively: wild-type NRIF3 (WT), NRIF3 L89R, NRIF3 L96R, and the double mutant (DM). -Galactosidase activities were determined in the presence (shaded columns) or absence (open columns) of cognate ligands: T3 for TR, 9-cis RA for RXR.

Spacing between RID1 and RID2 influences NRIF3-receptor interactions. Since dimerization of NRIF3 appears to be dispensable for TR or RXR interactions, it seems unlikely that such interactions would involve two copies of RID1 contributed by a NRIF3 dimer. We previously reported that the N-terminal LXXLL motif is required for optimum NRIF3-receptor interactions, since a mutation in one of the core leucine residues reduces NRIF3 interactions with TR or RXR (31). Thus, an alternative possibility is that NRIF3-receptor interactions involve and require both the C-terminal RID1 and the N-terminal LXXLL motif (referred to here as RID2). This model is consistent with our findings that (i) RID1 is essential and RID2 is important for optimum NRIF3-receptor interactions (31); (ii) a monovalent version of RID1(B42-RID1; see Fig. 4), or EnS and EnL molecules that contain only RID2 (31), fail to interact with receptors; and (iii) dimerization of NRIF3 is dispensable for receptor interactions.

View larger version (20K): [in this window] [in a new window]   FIG. 8.   NRIF3-receptor interaction requires regions of both RID1 and RID2 and is influenced by the spacing between RID1 and RID2. The baits LexA-TR LBD and LexA-RXR LBD were examined in a yeast two-hybrid assay for interactions with the following preys (as B42 fusions), respectively: full-length NRIF3 (WT), the N-terminal portion of NRIF3(1-111) that contains RID2, the C-terminal portion of NRIF3(112-177) that contains RID1, and mutant NRIF3s that delete residues 87 to 111 (1), or residues 112 to 161 (2). -Galactosidase activities were determined in the presence (shaded columns) or absence (open columns) of cognate ligands: T3 for TR, 9-cis RA for RXR.

The C terminus of NRIF3 contains an autonomous transactivation domain. To gain further insights into NRIF3-mediated coactivation, cDNAs encoding full-length NRIF3, its derived fragments, or its related isoforms were fused in frame with the Gal4 DBD. The resulting Gal4 fusion constructs were transfected into HeLa cells with G5-tk-CAT, a reporter under the control of the basal tk promoter linked to five copies of a Gal4 response sequence. As shown in Fig. 9, G5-tk-CAT exhibited a relatively low basal activity, and cotransfection of the Gal4 DBD resulted in a modest activation (~2-fold) that is commonly observed with this reporter. In contrast, cotransfection of Gal4-NRIF3(162-177) (also referred to as Gal4-AD1) resulted in a 10-fold activation (Fig. 9), suggesting that this region (the very C terminus) of NRIF3 harbors an autonomous transactivation domain (AD1). However, all other Gal4 fusions, including that of full-length NRIF3(1-177), NRIF3(1-111)/EnS, and the alternatively spliced isoform EnL, failed to activate G5-tk-CAT.

View larger version (29K): [in this window] [in a new window]   FIG. 9.   The C-terminal region of NRIF3 contains an autonomous transactivation domain (AD1). HeLa cells were transfected with the G5-tk-CAT reporter either alone or together with one of the following constructs: Gal4, Gal4-NRIF3 (full-length, 1 to 177), Gal4-EnS (equivalent to residues 1 to 111 of NRIF3, see Fig. 1), Gal4-EnL (full length; Fig. 1), and Gal4-AD1 (residues 162 to 177 of NRIF3, see Fig. 1). Cells were harvested 42 h after transfection, and CAT activities were then determined as described in Materials and Methods.

The N-terminal portion of NRIF3 harbors a transrepression function. Although NRIF3 contains AD1, we found that full-length NRIF3 fused to Gal4 did not activate G5-tk-CAT (Fig. 9). In fact, the reporter activity appears to be lowered in the presence of Gal4-NRIF3, suggesting possible repression by this fusion protein (Fig. 9, columns 1 to 3). Since the basal activity of G5-tk-CAT is quite low in our transfected HeLa cells (Fig. 9), we used the G5-SVB-CAT reporter, which exhibits higher basal activity (Fig. 10), to further evaluate the potential repression function of Gal4-NRIF3. G5-SVB-CAT was transfected into HeLa cells alone, with the Gal4 DBD control, or with one of the Gal4 DBD fusions depicted in Fig. 10. As shown in Fig. 10, while Gal4 DBD alone had little effect on the reporter activity, Gal4-NRIF3 (full length, 1 to 177) significantly repressed the reporter activity. A Gal4 fusion of the N-terminal portion of NRIF3 (residues 1 to 111, which is also equivalent to EnS) exhibited a similar extent of repression (Fig. 10). In contrast, a Gal4 fusion with the C-terminal portion of NRIF3 (residues 112 to 177) failed to repress the reporter activity. These results suggest that the N-terminal portion of NRIF3 (residues 1 to 111) harbors a transrepression function. Thus, NRIF3 appears to harbor both transactivation and transrepression functions. In the context of a Gal4 DBD fusion with full-length NRIF3 [Gal4-NRIF3(1-177)], the transrepression function appears to be dominant. We also examined the Gal4 DBD fusion of EnL, an alternatively spliced isoform of NRIF3, and found it also mediated repression (Fig. 10).

View larger version (51K): [in this window] [in a new window]   FIG. 10.   The N-terminal portion of NRIF3 harbors a transrepression function. HeLa cells were transfected with the G5-SVB-CAT reporter either alone or together with one of the following constructs: Gal4, Gal4-NRIF3 (full length, 1 to 177), Gal4-EnS (equivalent to residues 1 to 111 of NRIF3; Fig. 1), Gal4-EnL (full length; Fig. 1), and Gal4-(112-177) (residues 112 to 177 of NRIF3). Cells were harvested 42 h after transfection, and CAT activities were then determined as described in Materials and Methods.

The repression function of NRIF3 does not require its coiled-coil domain. The repression function of NRIF3 is localized to its N-terminal region which also contains the coiled-coil domain essential for mediating dimerization (Fig. 1). Since the coiled-coil domain and its leucine zipper-like motif are also candidate structures for mediating protein-protein interactions, we tested whether the coiled-coil domain may function as a surface for a docking factor(s) responsible for repression. To this end, we constructed Gal4 fusions of two mutant forms of NRIF3: Gal4-NRIF3(87-111), which deletes essentially the entire coiled-coil domain, and Gal4-NRIF3 DM, which contains mutations in the first and second leucine residues of the putative leucine zipper-like motif. These two fusions were then evaluated for the ability to repress G5-SVB-CAT in transfected HeLa cells. As shown in Fig. 11, both mutants exhibited levels of repression similar to that of wild-type Gal4-NRIF3, indicating that the coiled-coil domain is not required for the repression function of NRIF3 and that the repression function of NRIF3 appears to reside in amino acid residues 1 to 86, which are also common to EnS and EnL.

View larger version (33K): [in this window] [in a new window]   FIG. 11.   The coiled-coil domain is not required for the transrepression function of NRIF3. HeLa cells were transfected with the G5-SVB-CAT reporter, together with one of the following constructs: Gal4, Gal4-NRIF3 (wild type), Gal4-(87-111) (a mutant NRIF3 that deletes the coiled-coil domain), and Gal4-DM (a mutant NRIF3 containing double mutations L89R and L96R). Cells were harvested 42 h after transfection, and CAT activities were then determined as described in Materials and Methods.

Gal4 fusions of NRIF3 and its isoforms function as potent repressors in GH4C1 cells. Our study of Gal4 fusions of NRIF3 and its isoforms was first carried out with transfected HeLa cells. While these experiments reproducibly showed that Gal4-NRIF3, Gal4-EnS, and Gal4-EnL function as repressors (Fig. 10), the extent of repression is nevertheless moderate (about two- to threefold). To test whether different cell types might affect the transcriptional activity of Gal4-NRIF3 (or its isoforms), we performed similar studies with another cell line, GH4C1, which has been used extensively to study nuclear receptor actions (50). Since a simian virus 40 promoter-controlled reporter exhibits minimal activity in GH4C1 cells, we used the G5-tk-CAT reporter, which showed high basal activity under our transfection conditions (Fig. 12). As shown in Fig. 12, Gal4-NRIF3, Gal4-EnS, and Gal4-EnL all function as potent repressors in GH4C1 cells. The extent of repression is between 10- and 20-fold, compared to the 2- to 3-fold repression observed in HeLa cells. We also examined several NRIF3 derivatives. Gal4-(112-177) did not repress in HeLa cells (Fig. 10, column 6) and also showed no repression in GH4C1 cells (Fig. 12, column 2). The coiled-coil domain deletion mutant Gal4-(87-111) and the leucine zipper mutant Gal4-DM exhibited a repression similar to that of Gal4-NRIF3 in HeLa cells (Fig. 11), and both were found to function as potent repressors in GH4C1 cells as well. Thus, the overall repression pattern of Gal4 fusions of NRIF3 or its isoforms or its derivatives is conserved between HeLa and GH4C1 cells, despite the fact that the extent of repression is greater in GH4C1 cells.

View larger version (33K): [in this window] [in a new window]   FIG. 12.   Gal4 fusions of NRIF3, EnL, and EnS function as potent repressors in GH4C1 cells. Constructs expressing Gal4-NRIF3, Gal4-EnL, or Gal4-EnS were transfected into GH4C1 cells, together with the G5-tk-CAT reporter, to evaluate the potential repression function by the Gal4 fusion proteins. Gal4-(112-177) (which shows no repression and was included as a control), Gal4-(87-111 (a mutant NRIF3 that deletes the coiled-coil domain), and Gal4-DM (a mutant NRIF3 containing double mutations L89R and L96R) were used to examine the potential role of the coiled-coil domain.

A novel repression domain maps to residues 20 to 50 of NRIF3. Our transfection studies indicated that the major transrepression function of NRIF3 is contained in its N terminus (residues 1 to 111), since Gal4-EnS (which is equivalent to the region including residues 1 to 111 of NRIF3 [Fig. 1]) showed an extent of repression similar to that of full-length Gal4-NRIF3 in both HeLa and GH4C1 cells (Fig. 10 and 12). Moreover, deletion of the bulk of the coiled-coil domain (residues 87 to 111) appeared to have little effect on repression (Fig. 11 and 12). From these results, we inferred that the repression function of NRIF3 is probably contained within residues 1 to 86. To confirm this, we constructed Gal4-NRIF3(1-86) and found that it indeed mediates potent repression in GH4C1 cells (Fig. 13). To further define the domain(s) responsible for repression, a series of Gal4 fusions that contain various deletions of NRIF3 were constructed, including Gal4-(1-20), Gal4-(1-50), Gal4-(20-86), and Gal4-(47-86) (see Fig. 13 for schematic drawings of these constructs). These Gal4 fusions were designed based on a secondary structure analysis of NRIF3 so that the junction points of various deletions are in the predicted nonstructured regions of NRIF3 in order to minimize the potential conformational disruption to the resulting deletion molecules. As shown in Fig. 13, while Gal4-(1-50) and Gal4-(20-86) were found to be potent repressors in GH4C1 cells, repression is largely abolished for Gal4-(1-20) and Gal4-(47-86), suggesting that the region spanning residues 20 to 50 is the essential domain (termed RepD1) required for repression. RepD1 shares no homology with known domains in the database. Secondary structure analysis suggests that RepD1 is composed of two short -strands separated by an unstructured linker.

View larger version (27K): [in this window] [in a new window]   FIG. 13.   An essential repression domain (RepD1) is mapped to residues 20 to 50 of NRIF3. Constructs expressing Gal4 fusions of various regions of NRIF3 as illustrated were transfected into GH4C1 cells, together with the G5-tk-CAT reporter to evaluate repression. Gal4-(112-177), which shows no repression, was included as a control (construct 1 [not drawn]). The fold repression was calculated as described in Materials and Methods. The mapped RepD1 region spanning residues 20 to 50 is shown as a dotted box.

A single amino acid substitution (Ser28 to Ala) in RepD1 abolishes its transrepression function. The nested deletion experiment shown in Fig. 13 suggested that the transrepression domain of NRIF3 is located within residues 20 to 50. To confirm that this region can indeed mediate repression, we constructed Gal4-NRIF3(20-50) and examined its effect on the G5-tk-CAT reporter in transfected GH4C1 cells. As shown in Fig. 14, Gal4-NRIF3(20-50) was found to mediate potent repression in GH4C1 cells. Thus, the RepD1 region (residues 20 to 50) is both necessary and sufficient for mediating repression. As discussed earlier, RepD1 shares no homology with known domains in the database. Moreover, our cotransfection study argues against a corepressor recruitment model for NRIF3 family-mediated repression. To shed some more light on RepD1 function, we performed a motif search with an online bioinformatics tool (62) by using the full-length NRIF3 as a query. The search predicted a potential phosphorylation site (Ser28) in NRIF3 which, interestingly, is located within the RepD1 region. To test the potential role of Ser28 in RepD1 function, we constructed Gal4-NRIF3(20-50, S28A), which contains a single amino acid substitution (Ser28 to Ala) in RepD1. Remarkably, while wild-type RepD1 elicits potent repression (more than 35-fold in this experiment), the ability to repress is virtually abolished for the mutant RepD1 (about 1.3-fold [Fig. 14]). This result suggests that phosphorylation at Ser28 is essential for RepD1 function in vivo and raises the interesting possibility that cellular signaling may influence the regulatory property of the NRIF3 family through the regulation of Ser28 modification.

View larger version (20K): [in this window] [in a new window]   FIG. 14.   A single amino acid substitution (Ser28 to Ala) abolishes RepD1-mediated repression. Gal4 fusions of the wild-type RepD1 [Gal4-(20-50) WT] or the mutant RepD1 [Gal4-(20-50) S28A] were examined for the ability to repress the G5-tk-CAT reporter in transfected GH4C1 cells. The fold repression was calculated as described in Materials and Methods.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

Domain structure of the NRIF3 gene family. A summary of the domain organization of NRIF3 and its isoforms, EnS and EnL, is shown in Fig. 1. NRIF3 harbors two RIDs (RID1 and RID2; Fig. 1). The C terminus of NRIF3 harbors an autonomous transactivation domain (AD1), which is moderately active when fused to the heterologous Gal4 DBD (Fig. 9). A novel repression domain (RepD1) is mapped to the N-terminal part of NRIF3 (residues 20 to 50) (Fig. 1, 13, and 14). In addition, the central region of NRIF3 contains a coiled-coil (dimerization) domain and a putative leucine zipper-like structure (Fig. 1 and 6). The gene that encodes NRIF3 also expresses two other alternatively spliced isoforms, EnS and EnL (Fig. 1). EnS is 100% identical to the corresponding portions (residues 1 to 111) of NRIF3 and EnL (Fig. 1). EnL and NRIF3 share 100% identity in their first 161 amino acid residues but differ in their small C-terminal portions (Fig. 1). EnS and EnL do not interact with nuclear receptors, while the unique C-terminal RID1 (which contains an LXXIL motif) of NRIF3 allows for a high-affinity ligand-dependent interaction with the TRs and RXRs (31). Thus, EnS and EnL lack the RID1 and AD1 region but contain the RepD1 and the coiled-coil (dimerization) domain (Fig. 1). Fluorescence microscopy of GFP fusions indicates that, like NRIF3 (31), EnS and EnL localize to the cell nucleus (Fig. 2). Our finding that EnS localizes only to the cell nucleus is somewhat different from a report by Kashiwagi et al. (29), which showed that the GFP fusion of EnS was found in both the cytoplasm and the nucleus. This discrepancy may arise from the differences in the cell lines used for transfection and/or the expression levels of the fusion proteins. Nevertheless, we have consistently observed the nuclear localization of EnS and EnL under the described experimental conditions.

Receptor specificity of NRIF3. One of the unique features of NRIF3 is its receptor specificity. While most known coactivators exhibit interactions with a broad array of receptors, NRIF3 interacts only with liganded TRs and RXRs and not with other examined nuclear receptors (31). Our previous study mapped an essential RID (i.e., RID1) to the unique C terminus of NRIF3 (31; also Fig. 1). Interestingly, RID1 contains an LXXIL motif, a variant of the canonical LXXLL motif. Computer simulation suggested that RID1 docks to the hydrophobic cleft of liganded receptor LBDs via its LXXIL motif in a similar fashion as to the canonical LXXLL motif utilized by the p160 family members (31). This conclusion is supported by several lines of experimental evidence. First, RID1 is essential for receptor interactions (31). Second, LexA-RID1 interacts with the TR and RXR LBDs in a ligand-dependent fashion (31; also Fig. 4), and such interactions are abolished by mutations in the core leucine residues of the LXXIL motif (31). Third, the integrity of helix 12 (the AF2 helix), a critical part in the formation of the hydrophobic cleft in the liganded LBDs, is essential for the NRIF3-receptor interaction (Fig. 3).

Repression mediated by the NRIF3 family is cell type dependent. Although we initially observed the transrepression function of Gal4 fusions of the NRIF3 family in transfected HeLa cells, the relative magnitude of repression is only moderate (Fig. 10). In contrast, these Gal4 fusion proteins were found to be potent repressors in GH4C1 cells (Fig. 12). The precise mechanism underlying such an apparent cell type specificity in repression by the NRIF3 family is unclear. A previous study with the orphan nuclear receptor Rev-erb showed that it can elicit different levels of repression in different cell types depending on the presence or absence of the corepressor N-CoR (63). Thus, one potential explanation for the cell type specificity in repression by the NRIF3 family would be that they mediate repression through the recruitment of an additional corepressor(s), whose abundance or availability or activity may vary in HeLa and GH4C1 cells. Although we cannot exclude this possibility, our finding that transrepression by Gal4 fusions of the NRIF3 family is not relieved by cotransfection of the corresponding wild-type proteins argues against this model.

Potential dual regulatory roles of NRIF3 and its isoforms. NRIF3 represents an unusual example of a coregulator that harbors both a transactivation and a transrepression domain. Although AD1 coresides in the same region with RID1 (Fig. 1), transactivation by AD1 appears to be independent of receptor binding. We previously showed that, in transfected HeLa cells, NRIF3 enhances wild-type TR- or RXR-mediated transactivation of reporters controlled by their cognate hormone response elements. Thus, it is intriguing that NRIF3 also contains a transrepression domain (RepD1) (Fig. 1, 13, and 14). In the context of Gal4-NRIF3, RepD1 appears to be dominant over AD1, since Gal4-NRIF3 was found to repress transcription in both HeLa and GH4C1 cells.