Signal Relay by Retinoic Acid Receptors α and β in the Retinoic Acid-induced Expression of Insulin-like Growth Factor-binding Protein-3 in Breast Cancer Cells*

Neither retinoic acid receptor-β (RARβ) nor insulin-like growth factor-binding protein-3 (IGFBP-3) is expressed in breast cancer cell line MCF-7. The expression of both proteins can be induced in response to all-trans-retinoic acid (atRA). By using an RARα-selective antagonist (Ro 41-5253), we demonstrated that RARβ expression was induced by atRA through an RARα-dependent signaling pathway and that RARβ induction was correlated with IGFBP-3 induction. However, MCF-7 cells transfected with sense RARβ cDNA expressed IGFBP-3 even in the presence of the RARα-selective antagonist Ro 41-5253. On the other hand, antisense RARβ cDNA transfection of MCF-7 cells blocked atRA-induced IGFBP-3 expression, indicating that RARβ is directly involved in the mediation of IGFBP-3 induction by atRA. Induction of IGFBP-3 expression by atRA occurs at the transcriptional level, as measured by nuclear run-on assays. Finally, we showed that atRA-induced IGFBP-3 is functionally active in modulating the growth-promoting effect of IGF-I. These experiments indicate that RARα and RARβ, both individually and together, are important in mammary gland homeostasis and breast cancer development. By linking IGFBP-3 to RARβ, our experiments define the signal intersection between the retinoid and IGF systems in cell growth regulation and explain why loss of RARβ might be critical in breast cancer carcinogenesis/progression.

Retinoids induce growth inhibition and apoptosis in a variety of tumor cells, including breast cancer cells (1). Recently, we proposed a mechanism by which all-trans-retinoic acid (atRA) 1 synergizes with interferon to inhibit the growth of both estrogen receptor-positive and estrogen receptor-negative breast cancer cell lines (2). Here we studied mechanisms by which atRA counteracts the growth-promoting effects of insulin-like growth factors (IGFs) in breast cancer cells, focusing on the involvement by retinoic acid receptors (RARs).
It is known that the molecular actions of retinoids are pri-marily mediated by their nuclear receptors (RAR␣, ␤ and ␥, and the retinoid X receptors (RXRs) ␣, ␤, and ␥), which function as liganded transcription factors (3). These receptors show both spatiotemporal patterns of expression during development and tissue-specific distribution in adults, suggesting that the various receptors play different roles in transducing retinoid signals. Among the RARs, RAR␣ is expressed ubiquitously in adult tissues, RAR␥ is expressed mainly in skin, and RAR␤ is expressed primarily in epithelial cells, including those in mammary tissue (4). Expression of RAR␤ is lost in the majority of breast cancer cell lines; it can be induced by retinoic acid (RA) in estrogen receptor-positive breast cancer cell lines but not in estrogen receptor-negative cancer cell lines (4 -7). The latter are believed to represent more advanced forms of breast carcinoma. Induction of RAR␤ expression correlates well with the growth-inhibitory and apoptotic effects of retinoic acid (8,9), suggesting that loss of RAR␤ expression may be one of the critical events involved in breast carcinogenesis/progression and in responsiveness of breast cancer cells to retinoid chemotherapy. At the same time, there is strong evidence that RAR␣ is the mediator of the growth inhibition of breast cancer cells by retinoids (10,11). In general, RAR␣ expression is lower in estrogen receptor-negative breast cancer cell lines than in estrogen receptor-positive lines; this corresponds to the responsiveness of these cell lines to RA. Taken together, these observations raise the possibility that both RAR␣ and RAR␤ are involved in the physiological action of retinoic acid in breast cancer cells. The insulin-like growth factor system includes IGF-I and IGF-II, their corresponding receptors, six IGF-binding proteins (IGFBPs), and four IGFBP-related proteins (12). IGF-I and IGF-II are thought to be important growth factors for breast cancer. IGF-I and -II receptors and IGFBP-2 and -4 proteins have been found in breast cancer cell lines and in tissue specimens (13). Although IGF-I and -II proteins are not expressed in breast cancer cell lines, they are expressed in breast cancer specimens, possibly by stromal cells (13), suggesting that IGFs, through a paracrine mechanism, promote breast cancer cell growth and underscoring the importance of IGFBPs for their ability to modulate IGF-I actions in the extracellular matrix.
In addition to the well established roles of IGFBPs in regulating IGF bioavailability and IGF-I receptor responsiveness to IGF-I, IGFBP-3 has also been recently proposed to function as a negative regulator of growth, independently of the IGF-I receptor (14,15). Supporting its role as a growth inhibitory regulator, IGFBP-3 expression is up-regulated by growth-inhibitory (and apoptosis-inducing) agents, such as retinoic acid (16 -19), vitamin D (20), transforming growth factor-␤ (16,21,22), antiestrogens (23), tumor necrosis factor-␣ (24), and, most compellingly, the tumor suppressor gene p53 (25); IGFBP-3 expression is down-regulated by growth-promoting factors, such as estrogen (26) and epidermal growth factor (27). All of this information clearly indicates that IGFBP-3 is a common downstream effector of many growth regulatory agents. We report here that both RAR␣ and RAR␤, by relaying the atRA signal in MCF-7 cells, are involved in the induction of IGFBP-3, and our experiments suggest that lack of RAR␤ expression in the majority of breast cancer cell lines may result in the failure of IGFBP-3 induction and growth inhibition by retinoids.

EXPERIMENTAL PROCEDURES
Cell Cultures and Retinoids-Cells of the breast carcinoma cell line MCF-7 (American Type Culture Collection, Manassas, VA) were grown in phenol red-free Eagle's minimal essential medium (Sigma) supplemented with 5% charcoal-stripped calf serum (Sigma). Cells from Ͻ15 passages were used for experiments.
atRA was purchased from Sigma. The RAR-specific agonist Ro 13-7410, the RXR-specific agonist Ro 25-7386, and the RAR␣-selective antagonist Ro 41-5253 were generously provided by Hoffmann-La Roche. Retinoids were dissolved in absolute ethanol under lights that were covered with a UV-blocking film (CLHC, Sydlin, Inc., Lancaster, PA). The integrity of atRA was routinely monitored by spectrophotometry.
Preparation of Conditioned Medium-MCF-7 cells were grown as described above for 24 h, washed with phosphate-buffered saline, and then transferred to phenol red-free Eagle's minimal essential medium supplemented with 2 g/ml fibronectin and 2 g/ml transferrin (both from Sigma) for another 24 h before atRA treatment. The conditioned medium was then harvested with the addition of 0.2 mM phenylmethylsulfonyl fluoride and 10 g/ml aprotinin (both from Sigma), dried under speed vacuum, and resuspended for analysis.
Cell Growth Inhibition Assay-MCF-7 cells (4 ϫ 10 3 cells/well) were cultured in the conditioned medium described above in 96-well cell culture plates. Recombinant human IGF-I, recombinant human IG-FBP-3 (both generous gifts of Celtrix, Palo Alto, CA), or medium from atRA-treated cell cultures was added alone or in different combinations to the cell cultures for 2 days. Cells were washed, fixed with 10% trichloroacetic acid for 1 h, and then stained with 1% sulforhodamine B for 1 h. Cells were washed again, and then 100 l of 10 M Tris-HCl, pH 10, was added to release the dye (28). The absorbance was measured at 562 nm.
Immunodepletion-Conditioned medium from atRA-treated or untreated cells was incubated with 2 g/ml of anti-IGFBP-3 antibodies (goat polyclonal antibodies against human IGFBP-3; Santa Cruz Biotechnology Inc., Santa Cruz, CA) or normal goat serum (Santa Cruz Biotechnology) for 2 h. Protein A/Protein G PLUS-Agarose (Santa Cruz Biotechnology) then was added, and the media were rocked at 4°C overnight followed by filter sterilization of the supernatants. Immunoprecipitates were boiled for 3 min in SDS gel loading buffer and were used in a Western ligand blotting.
Western Immunoblotting and Western Ligand Blotting-Fifty g of protein from cell lysates or conditioned medium was loaded onto 8 -12% SDS-polyacrylamide gels under nonreducing conditions. After transfer, nitrocellulose blots were incubated with rabbit polyclonal antibodies against human RAR␤ (Santa Cruz Biotechnology). The blots were then incubated with secondary antibodies and developed using an ECL kit (Amersham Pharmacia Biotech). For Western ligand blotting, nitrocellulose blots were initially washed in 3% Nonidet P-40 (Fluka Chemical Corp., Ronkonkoma, NY) for 30 min, followed by blocking in 1% bovine serum albumin (Sigma) for 2 h and 0.1% Tween 20 (Sigma) for 15 min. Blots were then probed with 125 I-labeled recombinant human IGF-II (Bachem California Inc., Torrance, CA) overnight followed by extensive washing with 1% Tween 20 before autoradiography.
Transient Transfection-A luciferase reporter gene construct under the control of a retinoic acid response element (DR5-tk-Luc, provided by Dr. R. M. Evans, Gene Expression Laboratory, Salk Institute for Biological Studies, La Jolla, CA) was used to measure retinoid receptormediated gene activation. Ten g of DR5-tk-Luc was co-transfected into MCF-7 cells with 2 g of ␤-galactosidase expression vector (pCMV␤; CLONTECH, Palo Alto, CA) using Lipofectin reagent (Life Technologies, Inc.). Transfection efficiency was normalized to ␤-galactosidase activity.
Stable Transfection-Plasmid constructs for stable transfection experiments were pRC/CMV-RAR␤ and pRC/CMV-antisense RAR␤ (generous gifts from Dr. X.-K. Zhang, La Jolla Cancer Research Center, La Jolla, CA). MCF-7 cells grown to 50% confluence were washed with serum-free growth medium. Two g of either empty vector or construct was mixed with Lipofectin reagent and added to cells for 5 h. Selection was initiated with 400 g/ml of G418 (Life Technologies, Inc.) on the third day and continued for 17-21 days until drug-resistant colonies emerged. Single colonies were cloned and assayed for the expression of the inserted genes by Northern blotting, and the expression of RAR␤ receptor protein was measured by Western blotting.
Northern Blot Analysis-Total RNA was isolated using TRI Reagent (Sigma). RNA was separated on 1% agarose/1.1 M formaldehyde gels and then transferred and cross-linked to GeneScreen nylon membranes (NEN Life Science Products). Hybridization was carried out using the following probes: T 4 polynucleotide kinase-labeled 40-mer antisense RAR␣, RAR␤, or ␤-actin DNA (Oncogene Research Products, Cambridge, MA) or random primer-labeled IGFBP-3 cDNA (Genentech, Inc., South San Francisco, CA). The results were analyzed with a phosphorimager (Bio-Rad).

IGFBP-3 Expression Is Induced by atRA in a Dose-and
Time-dependent Fashion in MCF-7 Cells-To determine the effects of atRA on the expression of IGFBP-3 in our experimental system, MCF-7 cells were grown in the presence of 0, 10 Ϫ9 , 10 Ϫ8 , 10 Ϫ7 , or 10 Ϫ6 M atRA for 72 h followed by Northern blotting analysis of IGFBP-3 mRNA. As shown in Fig. 1A, MCF-7 cells did not express IGFBP-3 message in the absence of atRA, but as little as 10 Ϫ8 M atRA was effective in inducing the expression of IGFBP-3 mRNA. Higher levels of IGFBP-3 mRNA were detected with increasing concentrations of atRA (Fig. 1A). Fig. 1B shows the temporal effect of 10 Ϫ6 M atRA on the expression of IGFBP-3 message. IGFBP-3 mRNA was detected as early as 24 h after atRA treatment and was maximal at 48 h.
atRA Activates IGFBP-3 Gene Transcription, and RAR, Rather Than RXR, Mediates This Process-We next wished to determine whether the retinoic acid-induced expression of IG-FBP-3 in MCF-7 cells was mediated by RAR or RXR and whether atRA directly activates the transcription of the IG-FBP-3 gene. The second point was of interest because it is known that retinoids can regulate gene expression posttranscriptionally (29,30). For these experiments, MCF-7 cells were incubated for 48 h with 10 Ϫ6 M of either atRA, the RAR-specific agonist Ro 13-7410, or the RXR-specific agonist Ro 25-7386. Nuclei were isolated, and nuclear run-on assays were performed. As indicated in Fig. 2A, both atRA and the RARspecific agonist Ro 13-7410 activated IGFBP-3 gene transcription, but the gene was not transcribed in cells treated with vehicle only or with the RXR-specific agonist Ro 25-7386. The ␤-actin gene was transcribed normally under all of these experimental conditions. These results indicate that 1) RAR but not RXR is involved in transducing the atRA signal to induce IGFBP-3 expression, and 2) atRA and Ro 13-7410 directly activate IGFBP-3 gene transcription. IGFBP-3 mRNA was measured in parallel experiments following treatment of MCF-7 cells with the various retinoids for 72 h (Fig. 2B). IGFBP-3 mRNA was only present in cells treated with atRA and Ro 13-7410, the RAR-specific agonist.
To verify the ability of the synthetic retinoids, Ro 13-7410 and Ro 25-7386, to activate retinoid receptors in our experimental system, a luciferase reporter gene under the control of a DR5 element, the canonical retinoic acid response element activated by RARs, was introduced into MCF-7 cells. Luciferase activity was measured 72 h later in the presence of 10 Ϫ6 M of atRA, Ro 13-7410, or Ro 25-7386. As documented in Fig. 2C, Ro 13-7410, the RAR-specific agonist, activated the expression of luciferase gene at a level similar to that of atRA, but Ro 25-7386, the RXR-specific agonist, was not effective in activating the expression of luciferase gene. These results validate the use of the synthetic retinoids in our experimental system. RAR␤ Expression Is Induced by atRA in an RAR␣-dependent Pathway, and RAR␤ Relays the atRA Signal That Leads to the Induction of IGFBP-3 Expression in MCF-7 Cells-It has been shown that the transcription of the RAR␤ gene is induced rapidly after retinoid treatment, peaking by 6 h, and that it is independent of new protein synthesis (31,32). Furthermore, the level of RAR␣ expression in breast cancer cell lines appears to be correlated with the induced levels of RAR␤ expression (5,8,9). Thus, it is reasonable to postulate that RAR␤ is induced in MCF-7 cells by atRA through a signaling pathway mediated by RAR␣. To test this hypothesis, MCF-7 cells were grown in the presence or absence of 10 Ϫ6 M atRA for 72 h. Total RNA was extracted, and 30 g was used to measure mRNAs for RAR␣ and RAR␤ by Northern blotting. As documented in Fig. 3, the levels of RAR␣ expression in MCF-7 cells were similar in the presence or absence of atRA, whereas RAR␤ expression was detectable only after atRA treatment. These results indicate that RAR␣ mediates the atRA-induced expression of RAR␤.
These experiments led us to ask whether the signal leading to the induction of IGFBP-3 expression was mediated by RAR␣, or if the induced RAR␤ mediates IGFBP-3 induction. In order to answer this question, MCF-7 cells were cultured for 72 h in the presence of 10 Ϫ7 M atRA plus 0, 10 Ϫ8 , 10 Ϫ7 , or 10 Ϫ6 M of Ro 41-5253, an RAR␣-selective antagonist. A lower concentration of atRA was used because we wanted to minimize the cytotoxicity of retinoids that is observed at high concentrations. After incubation, 30 g of total RNA was used to assay RAR␤ mRNA. As documented in Fig. 4A, 1 molar excess of Ro 41-5253 blocked the induction of RAR␤ expression. With decreasing concentrations of Ro 41-5253, RAR␤ expression increased, indicating that the process is mediated by RAR␣. IGFBP-3 expression was measured in MCF-7 cells grown for 72 h in the presence of the same combinations of retinoids by Northern blotting (Fig. 4B). Paralleling the diminished expression of RAR␤ in the presence of 10 Ϫ6 M of Ro 41-5253, IGFBP-3 expression was also abolished, indicating that retinoid-induced IGFBP-3 expression is correlated with RAR␤ expression.
In order to further document the direct involvement of RAR␤ in the atRA-induced expression of IGFBP-3, RAR␤ sense and antisense cDNA constructs were introduced into MCF-7 cells via expression vectors. Positive colonies were identified, cloned, and tested for RAR␤ expression by Western immunoblotting (Fig. 5). Three clones with average levels of expression of each sense (Fig. 5A, ␤3, ␤5, and ␤6) and antisense (Fig. 5B, As-␤4, As-␤6, and As-␤9) RAR␤ were used for experiments similar to those described above. As exemplified by the results shown for ␤5 (Fig. 6A), the RAR␣-selective antagonist Ro 41-5253 was unable to block atRA-induced IGFBP-3 expression in the three clones of RAR␤ sense transfectants, indicating that RAR␣ is not directly involved in this process. In contrast, in RAR␤ antisense transfectants, the induction of IGFBP-3 expression by atRA was totally blocked (Fig. 6B), indicating that RAR␤ is directly involved in IGFBP-3 gene activation.

IGFBP-3 Is a Downstream Effector of RAR␤ in the Inhibition of Breast Cancer Cell Growth by atRA-
The IGF growth factor system is believed to be actively involved in the growth of breast cancer (14). IGFBP-3 is a secreted protein that has been thought to primarily regulate the biological activities of IGFs extracellularly. In order to investigate the functional integrity of atRA-induced IGFBP-3 in modifying the actions of IGF-I, we first assayed IGFBP-3 secretion by MCF-7 cells after induction by atRA. For this purpose, MCF-7 cells were grown in conditioned medium for 6 days in the presence or absence of 10 Ϫ6 M atRA. Conditioned medium was harvested at 2, 4, and 6 days, concentrated, and analyzed for IGFBP-3 secretion by Western ligand blotting. As shown in Fig. 7A, IGFBP-3 protein was secreted into the conditioned medium at a measurable level on day 2 of atRA treatment; higher amounts of IGFBP-3 were secreted on days 4 and 6. In addition to IGFBP-3, IGFBP-2 and IGFBP-4 were also secreted into the conditioned medium, in both the presence and absence of atRA (Fig. 7A).
We next tested the responsiveness of MCF-7 cells to the IGF system. Exogenous IGF-I or IGFBP-3 was added alone or in different combinations to MCF-7 cells for 4 days, and cell growth was measured by sulforhodamine staining. As documented in Fig. 7B, MCF-7 cells were sensitive to the mitogenic effects of IGF-I, and recombinant human IGFBP-3 (rhIG-FBP-3) inhibited such activity in a dose-dependent manner (0 -10 nM).
To investigate the biological activity of atRA-induced endogenous IGFBP-3, MCF-7 cells were maintained in conditioned medium for 4 days in the presence or absence of 10 Ϫ6 M atRA, and the conditioned medium was collected. IGFBP-3 protein was immunodepleted in half of the conditioned medium from atRA-treated cells. The medium was then filter-sterilized and added to MCF-7 cells for 2 days in the presence of 1 nM IGF-I. Cell growth was measured by sulforhodamine staining and expressed as percentage of absorbance relative to control MCF-7 cell cultures treated with control medium supplemented with 1 nM IGF-I (100%) or 1 nM IGF-I plus 10 nM rhIGFBP-3 (0%) (Fig. 7C). Similar to the results described in Fig. 7B, the conditioned medium from atRA-treated MCF-7 cells was able to block the growth promotion of MCF-7 cells by IGF-I (Fig. 7C, CM/RA/BP3). When IGFBP-3 was depleted (Fig. 7C, CM/RA), the medium was no longer effective in blocking the growth promotion by IGF-I, whereas the normal goat serum-treated control (Fig. 7C, CM/RA/BP3/S) did not remove the growth inhibition effect, suggesting that atRAinduced inhibition of IGF-I-stimulated cell growth is mediated rather specifically by IGFBP-3, not IGFBP-2 or IGFBP-4, because IGFBP-3-depleted medium (CM/RA) was unable to counteract IGF-I even when IGFBP-2 and IGFBP-4 were still present (Fig. 7D, lane 2).
As shown in the Western blots in Fig. 7D, conditioned medium from the 4-day atRA-treated MCF-7 cells contained the IGFBPs 3, 2, and 4 (Fig. 7D, lane 1). After immunodepletion, only IGFBP-2 and IGFBP-4 were present (Fig. 7D, lane 2); when the immunoprecipitate was examined, only IGFBP-3 was found (Fig. 7D, lane 3). These experiments clearly demonstrate that atRA-induced IGFBP-3 is able to function as a downstream effector of RAR␤ to block the growth promotion by IGF-I in MCF-7 breast cancer cells. A semiquantitative Western blot analysis utilizing rhIGFBP-3 as a standard indicated that the concentration of IGFBP-3 in atRA-treated conditioned medium was ϳ3 nM (data not shown). This result is consistent with a partial block of IGF-I action by recombinant IGFBP-3, which resulted in ϳ50% inhibition at 2 nM concentration (Fig.  7B).

DISCUSSION
The tissue-specific distribution of retinoic acid receptors in adults and the spatiotemporal patterns of expression during development indicate that these receptors may play different roles. Yet the coexistence of two or three retinoic acid receptor subtypes in a specific tissue also suggests that some type of compensation/coordination may exist among retinoic acid receptors in transducing retinoid signals. Such a compensation/ coordination of RARs in breast cancer cells was demonstrated in our experiments that showed that RAR␤ expression can be induced in MCF-7 cells by atRA via RAR␣ mediation. The levels of RAR␣ expression were similar in the presence or absence of atRA in MCF-7 cells, but RAR␤ expression, which was undetectable in the absence of atRA, was strongly induced by atRA. When the RAR␣-selective antagonist Ro 41-5253 was used, the induction of RAR␤ expression by atRA was blocked, indicating that RAR␤ induction is dependent on RAR␣.
Both RAR␣ and RAR␤ have been implicated in tumor development. For example, it is well documented that acute promyelocytic leukemia is caused by a reciprocal chromosome 15:17 translocation in which the t(15:17) breakpoint occurs in the RAR␣ gene (33). An involvement of RAR␤ in cancer development was originally suggested by the finding that RAR␤ is integrated by the hepatitis B virus in human hepatoma (34). Moreover, defective RAR␤ expression is believed to be an early event in epithelial carcinogenesis (35). Recently, it has been observed that RAR␤ expression is lost in many epithelial tumors and tumor cell lines, including breast cancer and breast cancer cells (5,6,(35)(36)(37)(38)(39). Furthermore, transgenic mice carrying antisense RAR␤2 develop carcinoma 14 -18 months after birth (40), strongly supporting a role of RAR␤ in tumorigenesis. By demonstrating an RAR␣-dependent RAR␤ induction, our experiments further stress the importance of RAR␤, which eventually becomes noninducible with progression of the tumor, as in estrogen receptor-negative breast cancer cells. The results of these experiments also allow us to clarify why both RAR␣ and RAR␤ have been implicated in retinoid-induced growth inhibition of breast cancer cells.
RAR␤ has been suspected as a tumor suppressor for a long time, and loss of RAR␤ expression has been thought to be a critical event in the development of breast cancer. We suggest that RAR␤ functions as a tumor suppressor by regulating the expression of other critical cell growth regulatory factors. Our experiments show that the regulation (induction) of IGFBP-3 in MCF-7 cells is mediated by RAR␤, because blocking RAR␤ expression by an RAR␣ antagonist, Ro 41-5253, also blocked the expression of IGFBP-3. When MCF-7 cells were transfected with the sense cDNA of RAR␤, IGFBP-3 was expressed, even in the presence of Ro 41-5253. At the same time, when the antisense cDNA of RAR␤ was transfected into MCF-7 cells, those cells were no longer able to respond to atRA by expressing IGFBP-3.
Using nuclear run-on assays, we showed that atRA directly activates IGFBP-3 gene transcription, supporting the recent finding that a major consensus sequence for retinoic acid is present in the promoter region of the IGFBP-3 gene (41). Whereas MCF-7 cells synthesize and secrete IGFBP-2 and IGFBP-4 into conditioned medium, the application of atRA not only induces the messenger for IGFBP-3, but also results in the appearance of secreted protein in the conditioned medium. IGFBP-3 secretion seems to occur while secretion of IGFBP-2 decreases and secretion of IGFBP-4 increases.
Among their diverse biological activities, IGFBPs are able to negatively modulate the actions of IGF by binding IGFs and preventing them from binding to the type 1 receptor (12, 42). Here, we demonstrated that the application of IGF-I stimulates cell growth in MCF-7 cells and that the application of exogenous rhIGFBP-3 can totally reverse this action. When we tested the biological activity of IGFBP-containing conditioned media in their ability to inhibit the IGF-I-stimulated growth of MCF-7 cells, we expected that changes in all of the IGFBPs (that is, IGFBP-3 induced by atRA along with an increase in IGFBP-4 and a decrease in IGFBP-2) would contribute to the growth inhibition effect. However, immunodepletion of IG-FBP-3 from the conditioned medium removed all growth inhibitory activity, suggesting an IGFBP-3-specific growth inhibitory mechanism. The significance of atRA-induced changes in IGFBP-2 and IGFBP-4 and the reason why these changes do not help counteract the IGF-I stimulation effect are not clear. Although IGFBP-3 induction by retinoids has been consistently observed, inconsistencies exist about retinoid-induced changes in IGFBPs 2 and 4 (17)(18)(19). In addition, as mentioned earlier, the biological activities of IGFBPs are not limited to negative effects on IGFs. Thus, changes in IGFBP-2 and IGFBP-4 may be germane to other, as yet unidentified mechanisms. In fact, the co-presence of IGFBPs 2, 3, and 4 in the cell culture medium may well indicate that these proteins possess different functions rather than simply representing functional redundancy. Another explanation is that IGFBP-3 may act through an IGF-independent pathway. IGF-independent actions of IG-FBP-3 have been reported (14,15,43), and underlying mechanisms are being pursued vigorously. Of particular interest is the recent observation that IGFBP-3 can be translocated into the cell nucleus (44 -46). Both exogenous (47) and endogenous IGFBP-3 2 have been shown to be translocated into the nucleus of breast cancer cells. Given the extremely selective nature of nuclear protein localization, it is reasonable to speculate that IGFBP-3 exerts profound biological activity in the nucleus.
In summary, our experiments show that both RAR␣ and RAR␤ are involved in the growth inhibitory activity of retinoids by mediating the induction of IGFBP-3 expression. By linking IGFBP-3 to RAR␤, our experiments have pinpointed an intersection between retinoid and IGF signals. This information also expands knowledge of the downstream effectors of RAR␤ and explains how RAR␤ might act as a tumor suppressor.