Direct Functional Interactions between Insulin-like Growth Factor-binding Protein-3 and Retinoid X Receptor-α Regulate Transcriptional Signaling and Apoptosis*

Insulin-like growth factor-binding protein (IGFBP)-3 regulates apoptosis in an IGF-independent fashion and has been shown to localize to nuclei. We cloned the nuclear receptor retinoid X receptor-α(RXR-α) as an IGFBP-3 protein partner in a yeast two-hybrid screen. Multiple methodologies showed that IGFBP-3 and RXR-α bind each other within the nucleus. IGFBP-3-induced apoptosis was abolished in RXR-α-knockout cells. IGFBP-3 and RXR ligands were additive in inducing apoptosis in prostate cancer cells. IGFBP-3 enhanced RXR response element and inhibited RARE signaling. Thus, RXR-α-IGFBP-3 interaction leads to modulation of the transcriptional activity of RXR-α and is essential for mediating the effects of IGFBP-3 on apoptosis.

The insulin-like growth factor (IGF) 1 -binding proteins (IGFBPs) are a family of proteins that bind IGFs with high affinity and specificity. They modulate IGF action by inhibiting or potentiating IGF binding to the IGF receptor. There are six high-affinity IGFBPs, of which IGFBP-3 is the most abundant in serum (1). Circulating IGFBP-3 is derived mainly from hepatic Kupffer cells, primarily under regulation by growth hormone, but IGFBP-3 is also produced locally in many tissues, where it serves important paracrine and autocrine roles in modulating cellular growth (2). In cells, IGFBP-3 is potently regulated by a number of factors including p53 (3), transforming growth factor-␤ (4), and hypoxia-induced factor-1 (5). IG-FBP-3 has been shown to directly induce apoptosis in prostate cancer cells (6), breast cancer cells (7), and other cell types. In these instances, IGFBP-3 acts directly, independently of the IGF-IGF receptor system, by binding to its own receptor(s), the nature of which is currently being unraveled.
The ability of IGFBP-3 to bind other molecules in addition to the IGFs has been previously demonstrated. IGFBP-3 binds to ALS, which together with IGF forms a stable ternary complex in serum (8). IGFBP-3 can undergo post-translational modification by proteases and can form an intermediary complex with plasmin and related enzymes (9). IGFBP-3 is noted to have a heparin-binding domain in its mid-region and is known to interact with heparin-containing molecules in the extracellular matrix (10) and with fibrin (11). Several groups, including our own, have demonstrated specific binding of IGFBP-3 to other uncharacterized proteins in serum (12), cell lysates (13), and cellular membranes (6). It has also been proposed that IG-FBP-3 may share a common receptor with transforming growth factor-␤ (14). It is of yet unclear what role these interactions play in mediating IGFBP-3 direct actions on cells.
IGFBP-3 has been observed in the nucleus of certain cells and contains a nuclear localization sequence that may facilitate shuttling into nuclei (15)(16)(17). The role of nuclear IGFBP-3 is currently unknown, but the cellular effects of IGFBP-3 appear to involve modulation of gene transcription, and thus, interactions of IGFBP-3 with transcription factors have been hypothesized.
The retinoid X receptor-␣ (RXR-␣) serves a key role in the regulation of gene transcription mediated by a variety of factors (18). RXR-␣ is an obligatory co-factor for the retinoic acid receptors, RARs (19), the peroxisome proliferator activating receptors (20), the thyroid receptors (21), and the vitamin D receptors (22). RXR-␣ can form heterodimers with these nuclear transcription factors and signal through specific DNA response elements such as the RARE, peroxisome proliferator receptor element, thyroid receptor element, and vitamin D receptor elements (23), or form homodimers and signal through RXRE (24). The activity of these transcriptional dimers is further regulated by a variety of transcriptional co-activators and co-inhibitors that modulate gene transcription in various states (25).
We report here the identification of a novel partner for IG-FBP-3 in the form of the nuclear receptor RXR-␣, confirm the validity of RXR-␣/IGFBP-3 binding through multiple independent in vitro methods, and demonstrate physiologically significant consequences of RXR-␣/IGFBP-3 binding on transcrip-tional signaling and cellular apoptosis. This unexpected interaction between the IGF/BP growth factor cascade and nuclear receptors represents a novel paradigm shift in our understanding of these systems.

EXPERIMENTAL PROCEDURES
Materials-Protigen (Mountain View, CA) provided recombinant human IGFBP-3 and the NLS mutant IGFBP-3. Amersham Pharmacia Biotech (Sweden) provided recombinant human IGF-I. Ligand (San Diego, CA) and SRI International (Menlo Park, CA) provided the RXRspecific ligands LG1069 and SR11235, respectively. These ligands bind the retinoid receptor X exclusively with maximal binding achieved at 1 M (26). 125 I-IGFBP-3 and anti-human IGFBP-3 antibodies, which were affinity purified on an IGFBP-3 column, were purchased from DSL (Webster, TX). Anti-human RXR-␣ antibodies, RAR antibodies, and HeLa nuclear extracts were purchased from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA). IGFBP-3 blocking peptides were purchased from Genemed Synthesis (South San Francisco, CA). Retinoic acid, dimethyl sulfoxide, and Igepal CA-630 were purchased from Sigma. Tris (crystallized free base) was purchased from Fisher (Fair Lawn, NJ). SDSpolyacrylamide gel electrophoresis (PAGE) reagents, Tween, and fatfree milk were purchased from Bio-Rad. Yeast two-hybrid screening kits were purchased from CLONTECH (Palo Alto, CA). The HeLa cDNA library was purchased from Stratagene (La Jolla, CA).
Yeast Two-hybrid Screening-The yeast strain Saccharomyces cerevisiae HF7c (MATa, ura3-52, his3-200, lys2-801, ade2-101, trp1-901, leu2-3, 112, gal4-542, gal80-538, LYS2::GAL1-HIS3, URA3::(GAL4 17mers) 3 -CYC1-lacZ)) was purchased from CLONTECH. The fusion gene IGFBP-3/BD was constructed by splicing cDNA encoding the human IGFBP-3 gene into the plasmid pGBT9 directly 5Ј to and in-phase with the gene encoding the binding domain (BD) of the yeast transcriptional activator GAL4 (CLONTECH). A HeLa cDNA library with the activation domain (AD) of the GAL4 gene was purchased from Stratagene and screened by co-transforming yeast with both plasmids. Yeast colonies were made competent for transformations according to the manufacturer's instructions (CLONTECH Matchmaker protocol handbook). Positive co-transformants were selected by growth on histidine-deficient agar media and assayed for ␤-galactosidase activity according to the manufacturer's instructions (CLONTECH Matchmaker protocol handbook). All results were reproducible in at least two independent assays. Genes encoding IGFBP-3-binding proteins identified through this method were isolated by plasmid recovery, amplified using polymerase chain reaction, sequenced using the GAL4 activation domain sequencing primer, and compared with known sequences in GeneBank using the MacVector TM software program (Oxford Molecular Ltd., Williamstown, MA).
Yeast Mating Assays-IGFBP-3/BD was co-transformed with the candidate gene/AD construct into yeast, using the protocol for the two-hybrid system. The transformants were streaked on the SD agar plates without Leu and Trp or His, and incubated for 3 days at 30°C. Expression of the LacZ gene by the co-transformants was determined by assaying for ␤-galactosidase activity, CLONTECH.
Ligand Blots-Reverse Western ligand blots were used to assess the binding of IGFBP-3 to RXR-␣. GST, GST-RXR-␣, or IGF-I were carefully dot-blotted directly onto nitrocellulose (2 l at a time) and allowed to dry completely. The nitrocellulose was buffered in Tris-buffered saline (TBS), 3% Igepal CA-630 for 30 min. The membranes were blocked for 3 h with TBS, 1% bovine serum albumin, and then incubated overnight with 125 I-IGFBP-3 (10 6 cpm) in TBS, 0.1% Tween, 1% bovine serum albumin. In some experiments, the incubation was performed in the presence of peptides analogous to domains of the IGFBP-3 molecule. The peptides used included: 1) an N-terminal-domain 20-mer peptide, corresponding to amino acids 33-52 in the IGFBP-3 protein: TELVREPGCGCCLTCALREG. 2) A peptide corresponding the nuclear localization sequence (NLS) within the heparin-binding domain (HBD). This 18-mer peptide corresponded to amino acids 215-232 in the IGFBP-3 protein: KKGFYKKKQCRPSKGRKR. The nitrocellulose was washed four times with TBS, 0.1% Tween and TBS. Autoradiography and phosphorimaging visualized the resulting bands. Experiments were repeated three times. Values of densitometrically analyzed values are expressed as mean Ϯ S.D.
Co-immunoprecipitation and Western Immunoblots-HeLa nuclear extracts were immunoprecipitated with anti-RXR-␣ or anti-IGFBP-3 antibodies. Briefly, 250 l of protein A-agarose were incubated overnight at 4°C with 5 l of anti-human IGFBP-3 antibodies or 5 l of anti-RXR-␣ antibodies. 125 l of each antibody-treated protein A-agarose were added to 10 g of HeLa nuclear extract and incubated for 3 h at 4°C with shaking. Immunoprecipitated proteins were pelleted by centrifugation and washed 3 times with 100 l of SACI buffer. 200-l sample buffer (ϫ1) were added to each sample and vortexed vigorously. Samples were boiled and vortexed again to release protein-antibody complexes from the protein A-agarose. The protein A-agarose was then separated from the immunoprecipitated complexes by centrifugation. The supernatants were saved, and the immunoprecipated proteins were separated by nonreducing SDS-PAGE (8%) at constant voltage overnight, then transferred to nitrocellulose for 4 h at 170 mA. The nitrocellulose was immersed in blocking solution (5% non-fat milk/TBS) for 45 min, washed with TBS, 0.1% Tween, and incubated with primary anti-human RXR-␣ or anti-human IGFBP-3 antibody (1:4,000) for 2 h. After washing off any unbound antibodies, the nitrocellulose was incubated with a secondary antibody (1:10,000) for 1 h. The membrane was washed 4 times with TBS, 0.1% Tween and TBS. Bands were visualized using the peroxidase-linked enhanced chemiluminescence detection system (ECL, Amersham Pharmacia Biotech). Experiments were repeated three times.
Immunoprecipitation and Autoradiography-125 I-IGFBP-3 was incubated with 10 g of HeLa nuclear extracts for 3 h at room temperature, and then with 5 l of IGFBP-3, RXR-␣, or RAR antibodies overnight. Complexes were then immunoprecipitated with protein A-agarose as above. Samples were subjected to separation on 12.5% nonreducing SDS-PAGE gel and autoradiography. Experiments were repeated three times.
GST Pull-down Assays-The GST-RXR fusion vector encoded the full-length RXR molecule and was the generous gift of Dr. D. J. Mangelsdorf and has been previously described (27). GST-RXR-␣ fusion protein was produced in Escherichia coli DH5␣, transformed with a GST-RXR-␣ construct, which were lysed and loaded on glutathione -Sepharose 4B beads from Sigma. 10 g of purified GST-RXR-␣ bound to beads were incubated with 5 g of recombinant IGFBP-3 protein or IGFBP-3 mutants and then separated by centrifugation. The bound proteins were analyzed by nonreducing SDS-PAGE followed by Western blotting using anti-IGFBP-3 antibody. Experiments were repeated three times.
Gel Mobility Shift Assays-[␥-32 P]ATP-labeled RXRE or RARE were incubated with HeLa nuclear extracts in 25 l of binding buffer containing 10 mM Tris, pH 7.5, 50 mM KCl, 1 mM EDTA, 0.5 mM dithiothreitol, 0.2% Nonidet P-40, 20 g of bovine serum albumin, 36 g of salmon sperm DNA, and 10% glycerol at 25°C for 20 min. Mixtures were incubated with or without IGFBP-3 or RXR-␣ antibodies for an additional 30 min. Incubations were carried out with or without unlabeled competitors as indicated. Protein-DNA complexes were separated from free probe on a 4.5% polyacrylamide gel in 1 ϫ TGE at 12 V/cm for 3 h, and visualized by autoradiography. Experiments were repeated three times.
Transient Transfection Analysis-Cells grown in 24-well culture plates were transfected with the appropriate combination of expression plasmids using LipofectAMINE (Life Technologies, Inc.). The total amount of plasmid DNA was adjusted to 500 ng/plate. The transfection solution was removed after 6 h of transfection, and the cells were cultured for 24 h. Cells were then incubated with or without 1 M retinoic acid or RXR ligand overnight. Cells were subjected to luciferase assays as described previously (30). Luciferase activities were expressed as the means and standard deviations of five identical wells. Luciferase reporter plasmid constructs contained either the DR1 RXRE (AGGTCA), or the DR5 RARE (AGTTCA) in a direct repeat separated by 5 nucleotides in the context of a thymidine kinase heterologous promoter, or a control plasmid containing the thymidine kinase promoter but no response element (TK-LUC) (31). The full-length IGFBP3 cDNA was constructed in the expression vector (pKG3226) and cotransfected when indicated. Experiments were repeated three times. Values are expressed as mean Ϯ S.D.
Cell Viability Assays-Cells were plated at a starting density of 250 cells/well in 96-well plates and treated with varying doses of IGFBP-3 for 3 days. Cell number was determined as follows: the fluorescent dye calcein AM (1 mM) was added to the medium for 30 min prior to termination of the assay. The plate was then read on a Biolumen 960 fluorescence plate reader (Molecular Dynamics, Inc.). The excitation and emission wavelengths are 485 and 530 nm, respectively, for calcein AM. The intensity of fluorescence of enzymatically cleaved calcein AM is a positive measure of cell number. Values are expressed as mean Ϯ S.D.
Apoptosis Enzyme-linked Immunosorbent Assay Assays-Photometric cell death detection enzyme-linked immunosorbent assay (Roche Molecular Biochemicals, Indianapolis, IN) was performed to quantitate the apoptotic index by detecting the histone-associated DNA fragments (mono-and oligonucleosomes) generated by the apoptotic cells. The assay is based on the quantitative sandwich-enzyme immunoassay principle using mouse monoclonal antibodies directed against DNA and histones, respectively, for the specific determination of these nucleosomes in the cytoplasmic fraction of cell lysates. In brief, an equal number of LAPC-4 cells were plated in 24-well culture plates (1 ϫ 10 4 /cm 2 ) in serum-supplemented medium, and grown to confluency for 72 h. At that time the confluent cells were washed with PBS and treated with IGFBP-3, an RXR ligand, and a combination thereof. The cells were dissociated gently (PBS with 0.1 M EDTA) and pelleted along with the floating cells (mostly apoptotic cells) collected from the conditioned media. The cell pellets were used to prepare the cytosol fractions, which contained the smaller fragments of DNA. Equal volumes of these cytosolic fractions were incubated in anti-histone antibody-coated wells (96-well plates) and the histones of the DNA fragments were allowed to bind to the anti-histone antibodies. The peroxidase-labeled mouse monoclonal DNA antibodies were used to localize and detect the bound fragmented DNA using photometric detection with 2,2Ј-azino-di-(3-ethylbenzathiazoline sulfonate) as the substrate. Calcium ionophoretreated conditions were used as positive controls. SFM-treated conditions were used as negative controls. Each experimental condition was performed with four samples and was repeated three times. The reaction products in each 96-well plate were read using a Bio-Rad microplate reader (model 3550-UV). Averages of the values Ϯ S.D. from double absorbance measurements of the samples were plotted.
Immunofluorescence Confocal Microscopy-1 ϫ 10 4 LAPC-4 cells were plated on coverglass in serum containing media for 2 days. The cells were then incubated in serum-free media with or without an RXR ligand (10 Ϫ6 M) for 24 h before staining for immunofluorescence. After three washes in PBS, fixation and permeabilization of the cells were performed with 1% paraformaldehyde in PBS for 15 min at room temperature and 0.2% Triton X-100 in PBS for 15 min on ice, and cells were washed twice with PBS. IGFBP-3 protein localization was detected using the DSL hIGFBP-3 goat polyclonal antibody (which was previously purified an IGFBP-3 column), diluted 1:200, followed by fluorescein anti-goat antibody from Vector (Burlingame, CA). RXR-␣ protein was detected using an RXR-␣-specific rabbit polyclonal antibody, diluted 1:150, followed by Texas Red anti-rabbit IgG from Sigma. Specimens were incubated with primary antibodies in PBS for 1 h at room temperature, with secondary antibodies in PBS for 40 min at room temperature, and then incubated with Hoechst from Electron Microscopy Sciences (Ft. Washington, PA) for 2 min. Samples were analyzed using the Inverted Confocal Microscope (Leica, Inc., Germany), equipped by digital camera Himamatsu (Japan), and operated by QEDimage software.

RESULTS
Cloning of RXR-␣ as an IGFBP-3 Partner-A yeast twohybrid system was used to identify novel IGFBP-3 partners. We utilized the yeast strain Hf7c (which is unable to grow in histidine, leucine, or tryptophan-deficient media) with a histidine-selection marker and ␤-galactosidase marker under control by the GAL1 promoter. We performed co-transformations with: 1) pGBT9 plasmid containing the fusion gene IGFBP-3/ GAL4 BD and a histidine-selection marker plus; 2) pGAD424 plasmid containing a HeLa cell cDNA library/GAL4 AD fusion gene and a tryptophan-selection marker. Positive co-transformants were isolated by growth on tryptophan-, leucine-, and histidine-deficient media, and colonies with ␤-galactosidase activity were harvested to recover library plasmids. Library fragments were amplified using PCR and sequenced. We isolated a 1200-base pair cDNA fragment encoding the C-terminal portion of the human RXR-␣ gene followed by the 3Ј-untranslated region of the human RXR-␣ cDNA. Yeast mating experiments confirmed the interaction between IGFBP-3 and RXR-␣ and co-transfected colonies disclosed potent ␤-galactosidase activity, which was completely absent in single transformants or in co-transformations of either hybrid with an empty vector. The RXR clone was one of 5 positive clones that reproducibly bound IGFBP-3 in the yeast two-hybrid screen. It was cloned out in four separate experiments.
Verification and Characterization of IGFBP-3-RXR-␣ Binding-In order to investigate the specificity of RXR-␣ binding to IGFBP-3, we performed GST pull-down experiments using RXR-␣ linked to GST and various forms of IGFBP-3 proteins. GST-RXR-␣ was able to "pull-down" various forms of natural and recombinant IGFBP-3. As shown in Fig. 1, GST-RXR-␣ bound recombinant IGFBP-3 as well as the NLS mutant IG-FBP-3, which has 2 amino acids mutated in the NLS region of IGFBP-3 (Table I). However, the IGFBP-3-HBD-BP1 mutant, in which 11 of the 18 amino acids in the HBD domain have been substituted to simulate the homologous region in IGFBP-1, which does not contain a heparin-binding domain (Table I), displayed no binding capacity for RXR-␣. This suggests that RXR-␣ binding is specific to a region of IGFBP-3, which is within the HBD domain, but does not involve the NLS domain, which is part of it. In separate experiments, GST-RXR-␣ bound recombinant IGFBP-3 of both the glycosylated (Chinese hamster ovary-derived as well as baculovirus-derived) and nonglycosylated (E. coli-derived) forms, and also pulled-down serum, cell lysate, and conditioned media IGFBP-3 (data not shown). Reconstitution experiments with IGFBP-3 and the Sepharose beads indicated that less than 5% of the IGFBP-3 binds to the beads without RXR-GST being present.
Western ligand dot blot assays confirmed IGFBP-3 binding to RXR-␣ as shown in Fig. 2. Increasing amounts of GST, GST-RXR-␣ protein, and IGF-I were dot-blotted onto nitrocellulose membranes, then incubated with 125 I-IGFBP-3 with or without peptides homologous to the N terminus of the IGFBP-3 protein, or the HBD domain of IGFBP-3. IGFBP-3 potently bound to both RXR-␣ and IGF-I to a similar degree as shown in the blot depicted in Fig. 2A and in the phosphorimaged quantification in Fig. 2B. GST alone demonstrated no binding to  1. GST-RXR-␣ pulls down IGFBP-3. Recombinant E. coliderived 29-kDa IGFBP-3 protein, E. coli-derived 29-kDa IGFBP-3 NLS mutant, and baculovirus-derived 35-kDa IGFBP-3-HBD-BP1 mutant incubated with or without GST-RXR-␣ fusion protein, which was loaded on glutathione-Sepharose 4B beads. Proteins were analyzed by Western immunoblotting using anti-IGFBP-3 antibodies. RXR-␣ bound the wild type and the NLS mutant, but not the HBD-BP-1 mutant. Fig. 2A, there was no appreciable binding of IGFBP-3 to either human or bovine albumin nor to insulin. The N terminus peptide of IGFBP-3 blocked IGF-I binding, but had no effect on RXR-␣ binding (Fig.  2B), whereas the HBD peptide blocked RXR-␣ binding, but had no effect on IGF-I binding to IGFBP-3 (Fig. 2B). This indicates that RXR-␣ binds near the HBD domain of IGFBP-3, while IGF-I binding involves the N-terminal domain. In additional blots (data not shown) insulin, human albumin, and bovine albumin were dot blotted and probed with radiolabeled IG-FBP-3 and showed no binding, indicating that RXR-IGFBP-3 binding is specific.

IGFBP-3 (data not shown). As shown in
In order to further investigate the specificity of RXR-␣ binding to IGFBP-3 in vivo, we performed co-immunoprecipitation experiments. In Fig. 3A, we used anti-IGFBP-3 and anti-RXR-␣ antibodies to immunoprecipitate these proteins and any interacting molecules from HeLa nuclear extracts then precipitated the samples with protein A-agarose. After separating the proteins using SDS-PAGE, we immunoblotted the membranes with anti-human IGFBP-3 antibodies (left) and anti-human RXR-␣ antibodies (right). IGFBP-3 antibodies immunoprecipitated all of the IGFBP-3 in the nuclear extracts samples and 39 Ϯ 11% of the RXR-␣ available in the nuclear extracts. RXR-␣ antibodies successfully precipitated all the RXR-␣ in nuclear extracts and more than 84 Ϯ 17% of the IGFBP-3 in the nuclear extract samples.
In Fig. 3B, 125 I-IGFBP-3 was incubated with HeLa nuclear extracts. Antibodies for IGFBP-3, RXR-␣, or RAR were bound with protein A-agarose and then further incubated with the nuclear extract mixtures. After precipitation, proteins were separated on 12.5% SDS-PAGE and autoradiographed. Both the IGFBP-3 antibody and the RXR-␣ antibody precipitated a complex of 125 I-IGFBP-3 and RXR-␣, whereas an RAR antibody did not precipitate IGFBP-3. This indicates that the IGFBP-3 -RXR-␣ binding is specific and that IGFBP-3 only binds RXR homodimers in HeLa nuclear extracts but does not bind RXR-RAR heterodimers.
Cellular Co-localization of RXR-␣ and IGFBP-3-Using fluorescence immunocytochemistry with antibodies for IGFBP-3 and RXR-␣, confocal microscopy revealed that IGFBP-3 and RXR-␣ were present in both the cytoplasm and nucleus of LAPC-4 cells (Fig. 4) and PC-3 cells (data not shown). Moreover, IGFBP-3 and RXR-␣ co-localized to a high degree in both cellular compartments. Furthermore, after the cells were treated with the RXR ligand LG1069 for 24 h, cytoplasmic IGFBP-3 and RXR-␣ both translocated to the nucleus as shown in the lower panel of Fig. 4. Similar results were seen upon treatment with 9-cis-RA (data not shown). Fig. 5 demonstrates the interactions of IGFBP-3 with the DNAtranscription factor complex involving RXR-␣ and the RXR response element (RXRE) in electromobility shift assays. HeLa nuclear extracts were incubated with 32 P-labeled DR-1 RXRE, then separated by 4.5% polyacrylamide gel. Specific binding of RXR-␣ in HeLa nuclear extracts to the DR-1 RXRE was demonstrated in lanes 1-3. As expected, the addition of an RXR-␣ antibody in lane 8 supershifts the complex. The addition of an IGFBP-3 antibody in lane 4 also supershifts the complex, indi-

FIG. 2. IGFBP-3 binds RXR-␣ near the HBD domain.
A, Western ligand dot blot of IGF-I, GST-RXR-␣, insulin, and albumin. 50 and 100 mol of each protein were blotted on membranes, incubated with 125 I-IGFBP-3, and autoradiographed. B, quantitated PhosphorImager data from A and similar experiments on IGF-I and RXR-␣, plotted as bar graphs.
Data is plotted as percent of maximal binding signal in the presence or absence of a 20-mer peptide homologous to the N terminus of the IGFBP-3 protein, which blocked IGF-I binding or an 18-mer peptide homologous to the HBD domain of the IGFBP-3 protein, which blocked RXR-␣ binding. * denotes p Ͻ 0.01. cating that IGFBP-3 is bound to the RXR⅐RXR-␣ complex. Similarly labeled DR-5 RARE binds RAR in HeLa nuclear extracts (lane 5), but this complex did not supershift with an IGFBP-3 antibody, suggesting that IGFBP-3 only forms a complex with the RXR-RXR homodimer, not with RXR-RAR heterodimers. Conducting these experiments in the presence of RXR-specific ligands had no effect on the ability of IGFBP-3 antibodies to supershift the complex, suggesting that RXR-␣ binds IGFBP-3 at a site different than the ligand-binding domain of the RXR-␣ molecule.
Effects of IGFBP-3 on RXR-␣-mediated Signaling-In luciferase-based transcriptional assays, we used the DR1-RXRE and the DR5-RARE reporter systems in COS7 (Fig. 6) and F9 cells (data not shown). In both cases, luciferase signaling was enhanced by co-treatment with the appropriate ligand (SR11235 or RA). However, IGFBP-3 co-transfections potently and dose-dependently inhibited RA signaling via RARE, but enhanced RXR-specific ligand signaling via the RXRE, indicating that IGFBP-3 enhances RXR-RXR homodimer-mediated signaling via the RXRE but blocks RAR-RXR heterodimermediated signaling via the RARE.
Requirement of RXR-␣ for IGFBP-3 Actions-To further study the functional interface of IGFBP-3 and RXR-␣ in the nucleus, we performed viability assays utilizing the F9 embryonic carcinoma cell line and a sister cell line, in which RXR-␣ has been knocked out (Fig. 7). IGFBP-3 treatment dramatically reduced cell viability in the F9 cell line, but IGFBP-3 had no discernible effects in the RXR-␣ knockout line, indicating that RXR-␣ is required for IGFBP-3 induced apoptosis. IGFBP-3 effects were seen to a similar extent in the range of 0.5-2.5 g/ml, which is similar to the concentrations of IGFBP-3 found in biological fluids. This result is particularly dramatic, as we have previously published that N-4-hydroxyphenyl-retinamide can induce apoptosis of the RXR-␣ null cells in a manner equivalent to the WT cells (32).
Synergism between RXR-␣ and IGFBP-3-Using apoptosis enzyme-linked immunosorbent assays as shown in Fig. 8, in the prostate cancer cell lines LAPC-4 (Fig. 8A) and LnCaP (Fig.  8B), treatment with IGFBP-3 (0.5 g/ml) alone, or the addition of RXR-␣ agonist alone (1 M) increased the level of apoptosis when separately added to cells. This is similar to published data (6,33). However, in the presence of both IGFBP3 and the RXR specific ligand, there is an additive enhancement of apoptosis in both models, suggesting that the IGFBP-3 and RXR-␣ signaling pathways are connected. DISCUSSION In addition to its role as an IGF-carrying protein in serum, IGFBP-3 is recognized to modulate the interaction between IGFs and the type I IGF receptor (IGF-R). In this capacity, IGFBP-3 may be inhibitory to growth, by preventing IGF binding to the IGF-R, or growth-potentiating, by presenting IGFs to the IGF-R in a controlled fashion and preventing down-regulation of the IGF-R (34).
The discovery of IGF-independent modulation of cell growth and death by IGFBP-3 provided indirect evidence for the presence of specific IGFBP-3-binding proteins, which may be cell surface-associated, cytosolic, or nuclear in location. Evidence for specific IGFBP-3 receptors also comes from experiments characterizing IGFBP-3 as a growth-inhibitory factor in murine knockout cells lacking the IGF-R (35). Oh et al. (13) have demonstrated specific binding of IGFBP-3 to uncharacterized cell surface proteins of 20, 26, and 50 kDa in the estrogen receptor negative breast cancer cell line Hs578T by affinity cross-linking and immunoprecipitation. We have similarly identified proteins with IGFBP-3 binding ability in prostate whole cell lysates and plasma membranes (6). IGFBP-3 has also been demonstrated to bind a molecularly uncharacterized 400-kDa protein, which can also bind transforming growth factor-␤ (37). Furthermore, the reports showing that IGFBP-3 is localized to the nucleus suggest that a nuclear receptor for IGFBP-3 may also exist or that IGFBP-3 may bind nuclear proteins and modulate their actions.
In this report, we describe for the first time the identification of RXR-␣ as an IGFBP-3-binding protein/receptor. After cloning RXR-␣ in a two-hybrid screen, we have demonstrated specific RXR-␣ to IGFBP-3 binding through several in vitro methods, including GST-pull down, co-immunoprecipitation, Western blot techniques, and confocal microscopy. These studies demonstrated not only that RXR-␣ associates with IGFBP-3 in nuclei, but that this binding occurs at a specific region of IGFBP-3, near the NLS/HBD domain.
We have further demonstrated physiologically significant ramifications of RXR-␣/IGFBP-3 interactions on the modulation of cell proliferation and apoptosis in several mammalian cellular systems. RXR-␣ agonists and IGFBP-3 are both growth inhibitory in many cancer cells (38), and the co-incubation of these molecules in the LAPC-4 model resulted in an additive effect on apoptosis. This phenomenon is consistent with IG-FBP-3 binding RXR dimers, and with the ability of IGFBP-3 to enhance RXRE-mediated signaling. This observation suggests that rexinoids may be more effective as anti-cancer agents in the presence of high IGFBP-3 levels. This is compatible with reports that show that high IGFBP-3 levels in serum protect from the risk of colon (39), breast (40), and prostate cancers (41).
The effects of IGFBP-3 on cell growth and apoptosis appear to require an intact RXR-signaling pathway, as RXR knockout cells were unresponsive to IGFBP-3-induced apoptosis. In mice, the targeted disruption of the RXR-␣ gene is embryonic lethal (42), however, cell-specific effects of RXR-␣ signaling has been unraveled with the use of the F9 cell system in which the essential role of RXR-␣ in mediating retinoid signaling has been established (43). The cell-regulatory effects of RXR-related transcriptional systems are very diverse. RXR-␣ is required for the actions of multiple natural ligands including thyroid hormone, vitamin D, and retinoic acid. RXR-␣ is also critical for the effects of several classes of novel drugs such as peroxisome proliferator activating receptor agonists, and synthetic retinoids, which are being developed as cancer therapies as well as agents for the treatment of diabetes and osteoporosis. The relationship between RXR-␣ and the IGF-IGFBP-3 axis is poorly understood. Retinoids enhance the expression of IGFBPs, including IGFBP-3, but rexinoids do not (44). Since we observed that IGFBP-3 blocks RA-mediated RAR signaling, this raises the possibility that a negative feedback loop which involves IGFBP-3 limits the extent of retinoid signaling through induction of IGFBP-3 which then blocks further RARmediated transcription. A further component of this loop may be related to the observation that IGF-I induces RAR-␤ expression (36) (which would also be blocked by IGFBP-3).
The discovery of RXR-␣ as an IGFBP-3 interacting protein adds a further level of complexity to the modulation of cellular growth. In addition to the modulation of IGF activity at the cell membrane and direct interactions with its own specific receptors, IGFBP-3 may also affect cell growth through several RXR-␣-dependent mechanisms. Namely, IGFBP-3 could enhance rexinoid action but block signaling mediated by ligands of other RXR-␣ partners. This represents a new paradigm in our understanding of the actions of peptide growth factors. As such, the IGFBP-3/RXR-␣ interaction represents an interface of two previously unrelated signaling pathways and opens new directions in studying cross-talk between growth factors and nuclear receptor ligands in cancer and other diseases. Using apoptosis enzyme-linked immunosorbent assays, in the prostate cancer cell lines LAPC-4 (A) and LnCaP (B), treatment with IGFBP-3 (0.5 g/ml) alone, or the addition of the RXR-␣ agonist LG1069, alone (1 M) increased the level of apoptosis, relative to serumfree (SF) media, when separately added to cells. However, in the presence of both IGFBP3 and the RXR specific ligand, there is an additive enhancement of apoptosis in both models. * denotes p Ͻ 0.05 relative to serum free. ** denotes p Ͻ 0.01 relative to either treatment alone.