Reports

Convergence of Transforming Growth Factor-β and Vitamin D Signaling Pathways on SMAD Transcriptional Coactivators

See allHide authors and affiliations

Science  26 Feb 1999:
Vol. 283, Issue 5406, pp. 1317-1321
DOI: 10.1126/science.283.5406.1317

Abstract

Cell proliferation and differentiation are regulated by growth regulatory factors such as transforming growth factor–β (TGF-β) and the liphophilic hormone vitamin D. TGF-β causes activation of SMAD proteins acting as coactivators or transcription factors in the nucleus. Vitamin D controls transcription of target genes through the vitamin D receptor (VDR). Smad3, one of the SMAD proteins downstream in the TGF-β signaling pathway, was found in mammalian cells to act as a coactivator specific for ligand-induced transactivation of VDR by forming a complex with a member of the steroid receptor coactivator–1 protein family in the nucleus. Thus, Smad3 may mediate cross-talk between vitamin D and TGF-β signaling pathways.

Vitamin D receptor (VDR) is a member of the nuclear receptor superfamily, and acts as a ligand-inducible transcriptional factor with coactivators (1, 2) such as the members of the steroid receptor coactivator–1/transcriptional intermediary factor 2 (SRC-1/TIF2) proteinfamily (3) and CREB-binding protein (CBP)/p300 (4). Cooperative actions of the growth regulatory factor TGF-β and vitamin D (5), and the phenotype of VDR knock-out mice (6), indicate that there may be cross-talk between the two signaling pathways. We therefore examined ligand-induced transactivation function of VDR and other nuclear receptors in cells stimulated by TGF-β or bone morphogenetic protein (BMP). VDR expression vectors and chloramphenicol acetyltransferase (CAT) reporter plasmid were transfected into COS-1 cells, and cells were treated with or without 1,25-dihydroxyvitamin D3[1,25(OH)2D3] or TGF-β (7–10). The transactivation function of VDR was significantly enhanced by the treatment with TGF-β (Fig. 1A), but not BMP. The constitutively active forms of either TGF-β type I receptor [TβR-I(TD)] or BMP type I receptor [BMPR-IA(QD) or BMPR-IB(QD)] (11, 12) were transfected into COS-1 cells with various nuclear receptor expression plasmids and CAT reporter plasmids bearing their respective response elements (7–9). TβR-I(TD), but not BMPR-IA(QD) or BMPR-IB(QD), increased the ligand-induced transactivation activity of VDR (Fig. 1A); however, it did not affect the other tested nuclear receptors: estrogen receptor α (ERα), androgen receptor (AR), glucocorticoid receptor (GR), retinoic acid receptor (RAR), and retinoid X receptor (RXR) (9, 13). Transactivation function of VDR was, however, suppressed by transfection with the catalytically inactive TGF-β type I receptor [TβR-I(KR)] (Fig. 1A). Similar results were obtained with other cell lines such as HeLa and HOS (13). The presence of TGF-β or BMP receptors did not affect the expression of VDR protein as estimated by protein immunoblotting (6, 13). Thus, the ligand-induced transactivation function of VDR was enhanced by TGF-β, but not by BMP signals.

Figure 1

Enhanced ligand-induced transactivation function of VDR in cells expressing activated TGF-β receptor. (A) CAT assays were performed with extracts from the COS-1 cells transfected with VDR expression vector (pSG5-VDR) (1 μg) and CAT reporter plasmid bearing a VDRE (7) (DR3T-G-CAT) (3 μg), together with either constitutively active [TβR-I(TD)] or catalytically inactive [TβR-I(KR)] TGF-β type I receptor expression vector (5 μg), constitutively active [BMPR-IA(QD) and BMPR-IB(QD)] or catalytically inactive [BMPR-IA(KR) and BMPR-IB(KR)] BMP type I receptor expression vectors (5 μg), along with Smad1 expression vector (pcDNA3-Smad1) (5 μg) in the presence (+) or absence (–) of 1,25(OH)2D3 (10–9 M) or TGF-β [0.5 (+) or 1 (++) ng/ml]. (B) Increased ligand-induced transactivation function of VDR in cells transfected with Smad3. COS-1 cells were cotransfected with pSG5-VDR (1 μg); either DR3T-G-CAT, OP-G-CAT, or OC-G-CAT (3 μg); either TβR-I(TD) or TbR-I(KR) (5 μg); and Smad2 (pcDNA3-Smad2), Smad3 (pcDNA3-Smad3), or Smad4 (pcDNA3-Smad4) expression vector (5 μg) in the presence or absence of 1,25(OH)2D3 (10–9 M). All values represent averages ± SD of at least three independent experiments.

The members of the SMAD protein family (Smad1 through Smad8) are signal transducers of the TGF-β–BMP superfamily (14). Smad1 and Smad5 transduce signals for BMPs (15, 16), and signals for TGF-β are mediated by Smad2 and Smad3 (17). Smad4 acts as a common partner for these pathway-specific SMAD proteins (18). When TGF-β or BMP receptors are activated by the binding of cognate ligands, pathway-specific SMADs are phosphorylated by the type I receptor serine-threonine kinases. Phosphorylated SMADs form stable complexes with Smad4, and these complexes translocate into the nucleus where they activate transcription as coactivators or DNA-binding transcription factors (14, 1719), though the overexpressed Smad3 and Smad4 are predominantly localized to the nucleus (20). We therefore investigated whether SMAD proteins could enhance the transactivation function of VDR. Neither Smad2 nor Smad4 stimulated the transactivation of VDR (Fig. 1B). However, expression of equivalent amounts of Smad3 did enhance the ligand-induced transactivation function of VDR, and Smad4 and TβR-I(TD) only slightly increased the action of Smad3 (Fig. 1B). When smaller amounts of Smad3 were expressed, TβR-I(TD) caused an additional increase in VDR transactivation beyond that induced by Smad3 (13). Moreover, a dominant-negative mutant of Smad3 that inhibits phosphorylation of wild-type Smad3 eliminated the enhancement of VDR transactivation function by wild-type Smad3 (13). The enhanced transactivation function of VDR by Smad3 was also observed in vitamin D response elements (VDREs) derived from mouse osteopontin (OP) or human osteocalcin (OC) gene promoters (Fig. 1C) (8). Smad1, a signal transducer of BMPs, had no effect on the ligand-induced transactivation of VDR (Fig. 1A). Thus, only Smad3 appears to enhance the ligand-induced transactivation function of VDR.

To test whether SMADs might serve as coactivators of VDR, we examined whether Smad3 physically interacts with VDR in vivo. Although no interaction of VDR with Smad2 was detected in the mammalian two-hybrid system (21), which detects protein-protein interactions in vivo, the ligand-dependent interaction of VDR with Smad3 was evident when compared to that of VDR with RXR (Fig. 2A). Interaction of Smad3 with VDR was also tested by coimmunoprecipitation (22) of Smad3 and VDR from COS-1 cells transfected with full-length VDR [VDR(1-424)] and FLAG-tagged Smad3. VDR was detected in anti-FLAG immunoprecipitate by protein immunoblotting with antibody to VDR. The presence of 1,25(OH)2D3 enhanced complex formation of Smad3 and VDR and heterodimerization with RXR (Fig. 2B). These findings indicate that VDR and Smad3 interact in vivo in a ligand-dependent manner.

Figure 2

Ligand-dependent interaction of Smad3 with VDR in vivo. (A) Interactions of SMADs with VDR were examined in the mammalian two-hybrid system in the presence (solid columns) or absence (open columns) of 1,25(OH)2D3 (10–9 M). CAT assays were done as described (Fig. 1). COS-1 cells were transfected with CAT reporter plasmid (3 μg) bearing the GAL4-binding element (17M5-CAT), expression vector containing the DEF domain of VDR fused to GAL4 DNA-binding domain [GAL4-VDR(DEF)] (5 μg), and the expression vector [5 (+) or 10 (++) μg] bearing Smad2, Smad3, ER, or RXR fused to VP16 activation domain (VP16-Smad2, VP16-Smad3, VP16- ER, or VP16-RXR). (B) Coimmunoprecipitation of Smad3 and VDR. Smad3-VDR interaction was analyzed by immunoprecipitation with antibody to FLAG followed by immunoblotting using antibody to VDR. COS-1 cells were cotransfected with pSG5-VDR, pcDNA3-FLAG-Smad3, pcDNA3-FLAG-ER, or pcDNA3-FLAG-RXR (10 μg) in the presence or absence of 1,25(OH)2D3 (10–9 M). Cells were lysed in TNE buffer, immunoprecipitated with monoclonal antibody to FLAG, and interacting proteins were detected by immunoblotting with antibody to VDR (6). Expression of FLAG-Smad3, FLAG-ER, and FLAG-RXR is shown by immunoblotting with antibody to FLAG.

We assessed interaction of a glutathione S-transferase (GST)–VDR fusion protein (23) with in vitro–translated Smad3 protein (24). The proteins interacted directly in the presence of 1,25(OH)2D3. A series of truncated Smad3 proteins showed that the NH2-terminal Mad homology 1 (MH1) region of Smad3 is required for this interaction (Fig. 3A). The MH1 region was also indispensable for the interaction of Smad3 with VDR in the mammalian two-hybrid system with the truncated Smad3 proteins fused to VP16 and GAL4-VDR(DEF) (Fig. 3A). The MH1 region was also required for immunoprecipitation of VDR with Smad3 proteins (Fig. 3B). These results indicated that the MH1 region mediates the interaction of Smad3 with VDR. Smad3 mutants [S3(41-435), S3(68-435), S3(147-435), and S3(238-435)], lacking the VDR interaction domain, did not enhance the ligand-induced transactivation function of VDR, even when the intrinsic transactivation domain (MH2 domain) of Smad3 remained intact (Fig. 3A). In fact, overexpression of such Smad3 mutants suppressed ligand-induced transactivation function of VDR (Fig. 3A).

Figure 3

Smad3 domain required for interaction with VDR. (A) Interaction domain of Smad3 for VDR. VDR was tested for the interaction with the indicated portions of Smad3 in a GST–pull down assay and a mammalian two-hybrid assay (–, no interaction; +, interaction; no symbol, not determined). The effects of Smad3 mutants on transactivation function of VDR were estimated by CAT assays, using the extracts from COS-1 cells expressing VDR and the truncated Smad3 mutants in the presence of 1,25(OH)2D3; the fold activations by Smad3 proteins are in the right panel. (B) Coimmunoprecipitation of Smad3 deletion mutants and VDR. Interactions between Smad3 deletion mutants [S3(21-435) and S3(41-435)] and VDR were analyzed by immunoprecipitation with anti-FLAG followed by immunoblotting using antibody to VDR (Fig. 2). COS-1 cells were transfected with pSG5-VDR, pcDNA3-FLAG-Smad3, pcDNA3-FLAG-S3(21-435), or pcDNA3-FLAG-S3(41-434) (10 μg) in the presence or absence of 1,25(OH)2D3(10–9 M). Expression of FLAG-Smad3 and its mutants is shown by immunoblotting with antibody to FLAG.

A series of truncated VDR proteins demonstrated that the middle region of the ligand-binding domain (E domain) of VDR is required for the interaction with Smad3 (Fig. 4A). The COOH-terminal end of VDR is essential for the ligand-induced transactivation function of the VDR ligand-binding domain (AF-2) and directly interacts with the nuclear coactivators in a ligand-dependent way (25). The region exhibited no interaction, but rather seemed to have an inhibitory effect on the Smad3 interaction (Fig. 4A). The in vitro binding of Smad3 and VDR was ligand-independent, and the binding was weak relative to that of the heterodimerization of VDR with RXR (Fig. 4A).

Figure 4

Ligand-dependent formation of a VDR-Smad3 complex enhanced by SRC-1. (A) Interaction domain of VDR for Smad3. The indicated portions of VDR or mutated VDRs were tested for interaction with full-length Smad3 in a GST–pull down assay and a mammalian two-hybrid assay. (B) The in vitro ligand-dependent interaction of VDR with Smad3 from the nuclear extracts. Nuclear extracts were prepared from COS-1 cells expressing FLAG-Smad3, FLAG-ER, or FLAG-RXR with or without SRC-1 in the presence or absence of 1,25(OH)2D3. FLAG-Smad3, FLAG-ER, or FLAG-RXR in the cell extracts were precipitated by GST-VDR. Interacting proteins were detected by immunoblotting with monoclonal antibody to FLAG. (C) Effects of SRC-1 on the interaction between Smad3 and VDR or mutated VDRs. Interactions of Smad3-VDR or Smad3-VDR mutants were analyzed by coimmunoprecipitation. COS-1 cells were cotransfected with pSG5-VDR, pSG5-VDR(1-357), pSG5- VDR(L417S), pSG5-VDR(E420Q), pcDNA3-FLAG-Smad3, pcDNA3-SRC-1, or pcDNA3-SRC-1m123 (10 μg) in the presence or absence of 1,25(OH)2D3(10–9 M). Cells were lysed in TNE buffer, immunoprecipitated with monoclonal antibody to FLAG, and interacting proteins were detected by immunoblotting with antibody to VDR (6). Expression of Smad3 is shown by immunoblotting with the antibody to FLAG. (D) Effect of SRC-1 on Smad3-enhanced transactivation function of VDR. CAT assays were done as described (Fig. 1). COS-1 cells were transfected with pSG5-VDR, pcDNA3-Smad3, or pcDNA3-SRC-1 in the presence or absence of 1,25(OH)2D3. (E) Ligand-dependent interaction of Smad3 with RXR/VDR heterocomplex in vivo. Mammalian three-hybrid system estimated by CAT assays were done as described (Fig. 1). COS-1 cells were transfected with GAL4-RXR, pSG5-VDR, or VP16-Smad3 in the presence or absence of 1,25(OH)2D3.

Because the truncated Smad3 mutants lacking interaction with VDR suppressed the ligand-induced transactivation of VDR, it appears that the MH2 domain of Smad3 and VDR may competitively recruit the same factors. Therefore, we used nuclear extracts of cells (26) overexpressing FLAG-Smad3 to test for interaction with GST-VDR. FLAG-Smad3 in the nuclear extracts showed ligand-dependency in the interaction with VDR in vitro (Fig. 4B), as seen in vivo, and the MH1 region was also required for this in vitro ligand dependency (13). Thus, an unidentified nuclear component may stabilize the ligand-dependent complex formation of VDR with Smad3. Such components might include coactivators for VDR such as the members of the SRC-1/TIF2 protein family, which directly interact with the minimal activation domain (AD) of AF-2 in the COOH-terminal end of the E domain in a ligand-dependent manner (3, 25, 27). To test this possibility, we chose SRC-1, because this coactivator binds to the AF-2 AD of VDR (25,28). Overexpression of SRC-1 enhanced the ligand-dependent interaction between Smad3 and VDR (Fig. 4B). The SRC-1–stabilized interaction of VDR with Smad3 was further confirmed by coimmunoprecipitation experiments (Fig. 4C) and the mammalian two-hybrid system (13). When VDR was either truncated [VDR(1-357)] or mutated [VDR(L417S) and VDR(E420Q)] in the AF-2 AD such that they did not interact with SRC-1 (25) but still interacted with Smad3 in vitro, the ligand-dependent interaction of VDR with Smad3 was abolished (Fig. 4C) [mutation Leu417→ Ser417 indicated as L417S (29)]. Conversely, overexpression of a SRC-1 mutant protein (SRC-1m123), which has point mutations in all of the three LXXLL motifs (29) and does not interact with VDR, inhibited the ligand-dependent interaction between Smad3 and VDR (Fig. 4C). Similar results were obtained in these assays when the TIF2 (28) were used instead of SRC-1 (13). SRC-1 augmented the Smad3-enhanced transactivation function of VDR (Fig. 4D). Thus, the ligand-dependent interaction of VDR with Smad3 apparently requires at least a member of the SRC-1/TIF2 protein family. We examined whether Smad3 binds SRC-1 directly in vitro. A ligand-dependent interaction between GST-VDR and SRC-1 was observed (13). However, Smad3 showed no interaction with SRC-1 or TIF2 (13). Finally, we confirmed the ligand-dependent interaction between Smad3 and RXR/VDR heterocomplex in vivo. Although no interaction of Smad3 with RXR was detected in the mammalian three-hybrid system, the ligand-dependent interaction of Smad3 with RXR/VDR heterocomplex was observed (Fig. 4E).

Our results established a molecular basis for cross-talk between TGF-β and vitamin D signaling pathways. The cooperative actions of vitamin D and TGF-β can be synergistic or antagonistic in a tissue-specific manner. Because SMAD proteins are differentially expressed in target tissues for TGF-β, the tissue-specific amounts of endogenous SMAD proteins may contribute to the cooperative actions.

  • * To whom correspondence should be addressed. E-mail: uskato{at}hongo.ecc.u-tokyo.ac.jp

REFERENCES AND NOTES

View Abstract

Navigate This Article