Interaction and Functional Cooperation of PEBP2/CBF with Smads

Smads are signal transducers for members of the transforming growth factor-β (TGF-β) superfamily. Upon ligand stimulation, receptor-regulated Smads (R-Smads) are phosphorylated by serine/threonine kinase receptors, form complexes with common-partner Smad, and translocate into the nucleus, where they regulate the transcription of target genes together with other transcription factors. Polyomavirus enhancer binding protein 2/core binding factor (PEBP2/CBF) is a transcription factor complex composed of α and β subunits. The α subunits of PEBP2/CBF, which contain the highly conserved Runt domain, play essential roles in hematopoiesis and osteogenesis. Here we show that three mammalian α subunits of PEBP2/CBF form complexes with R-Smads that act in TGF-β/activin pathways as well as those acting in bone morphogenetic protein (BMP) pathways. Among them, PEBP2αC/CBFA3/AML2 forms a complex with Smad3 and stimulates transcription of the germline Ig Cα promoter in a cooperative manner, for which binding of both factors to their specific binding sites is essential. PEBP2 may thus be a nuclear target of TGF-β/BMP signaling.

(Co-Smads), and inhibitory Smads. Smad2 and Smad3 are R-Smads that transmit TGF-␤/activin signals, whereas Smad1, Smad5, and Smad8 act as R-Smads mediating BMP signals. Smad4 is the only Co-Smad identified in mammals. Upon ligand stimulation, R-Smads are phosphorylated by the serine/ threonine kinase receptors, form complexes with Co-Smad, and translocate into the nucleus, where they cooperatively regulate the transcription of target genes with other transcription factors, including Xenopus FAST1 and its mammalian homologues (3)(4)(5) and also the c-Jun/c-Fos complex (6,7). TGF-␤ is a potent growth inhibitor for most cell types, including hematopoietic cells and lymphocytes. In addition, TGF-␤ directs class switching to IgA in splenic B cells (8,9). BMPs play important roles in early embryogenesis and in the induction of bone formation in vivo (10). It is thus important to identify and classify transcription factors that serve as nuclear targets of TGF-␤/ BMP signals and regulate these biological events.
Polyomavirus enhancer binding protein 2/core binding factor (PEBP2/CBF) is a transcription factor complex composed of ␣ and ␤ subunits (11,12). Three mammalian ␣ subunits have been identified, termed PEBP2␣A/CBFA1/AML3 (referred to as ␣A in this report), PEBP2␣B/CBFA2/AML1 (␣B), and PEBP2␣C/CBFA3/AML2 (␣C), whereas only a single ␤ subunit (PEBP2␤/CBFB) with several spliced variants is present in mammals. The ␣ subunits of PEBP2, which contain the highly conserved Runt domain, are responsible for binding to DNA and transcription activity. In contrast, the ␤ subunit does not bind to DNA by itself, but it enhances the DNA binding activity of the ␣ subunits by interacting via the Runt domain. PEBP2/ CBF plays critical roles in growth and differentiation of cells in certain specific tissues, i.e. ␣A in bone formation (13)(14)(15) and ␣B in definitive hematopoiesis (16,17); ␣C appears to be important in class switching to IgA because of its ability to activate the germline Ig C␣ promoter (18). Abnormalities of the PEBP2 genes are linked to human diseases. Mutations in one allele of the human PEBP2␣A/CBFA1 gene cause human cleidocranial dysplasia syndrome (19,20), whereas PEBP2␣B/ AML1 gene is frequently disrupted by chromosomal translocations in several types of human leukemia (11,12).
PEBP2 has been shown to interact with several transcription factors and co-activators and support context-dependent transcription of target genes (21)(22)(23). Because BMPs and ␣A play critical roles in bone formation, and TGF-␤ and ␣C in transcription of germline Ig ␣ transcripts required for IgA class switching, we examined the functional cooperation between the PEBP2␣ subunits and Smads. Our findings suggest that PEBP2␣ subunits and R-Smads cooperate to synergistically activate transcription in both the TGF-␤ and BMP signaling pathways, thereby regulating the function of cells in specific tissues upon activation by TGF-␤-like factors.

EXPERIMENTAL PROCEDURES
Plasmid Construction-FLAG-pcDEF3 and 6Myc-pcDEF3 containing six tandem copies of the Myc-epitope tag were previously described (24,25). The constructions of constitutively active forms of TGF-␤ type I receptor (T␤R-I(TD)) and BMP-type IB receptor (BMPR-IB(QD)), T␤R-II, wild-type (WT) Smads, and Smad3(DE) were reported (24 -26). The constructions of ␣A, ␣B, ␣C, and ␤2 have been described elsewhere (27)(28)(29). 2 Deletion constructs of ␣C were prepared by a polymerase chain reaction-based approach. For construction of the isolated Ig C␣/ TGF-␤-responsive element (T␤RE) promoter reporter construct ((T␤RE) 3 -Lux) and its mutants, three tandemly repeated T␤REs (WT or mutant versions) of the Ig C␣ promoter were fused to the heterologous c-Fos (30) and luciferase reporters. All of the polymerase chain reaction products were sequenced.
Immunoprecipitation and Immunoblotting-COS7 cells were transiently transfected with expression constructs for PEBP2␣ subunits, Smads and constitutively active forms of type I receptors. Cells were then washed, scraped, and solubilized (25). Immunoprecipitation and immunoblotting using an enhanced chemiluminescence detection system (Amersham Pharmacia Biotech) were performed as described (25).
Glutathione S-transferase (GST) Pull-down Assay-A GST pull-down assay was performed as described previously (22). GST-fusion proteins containing the full-length Smad3 or the Mad homology (MH)1 or MH2 domain of Smad3 were expressed and purified as described (32). In vitro transcription and translation of C-terminal deletion constructs of ␣C were done using the TNT system (Promega) in the presence of [ 35 S]methionine. GST-Smad3 (full-length), Smad3 (MH1), Smad3 (MH2), or GST bound to glutathione-Sepharose was mixed with ␣C proteins in 500 l of Tris-buffered saline, pH 7.4, containing 0.5% Nonidet P-40 for 1 h and washed vigorously three times with 1 ml of the same buffer. After boiling in the SDS-sample buffer, they were analyzed by SDS-polyacrylamide gel electrophoresis followed by autoradiography. 2 Y.-W. Zhang and Y. Ito, unpublished data. Cell lysates were immunoprecipitated (IP) with anti-FLAG antibody followed by immunoblotting (Blot) using anti-Myc antibody. ␣A and Smad4 co-immunoprecipitated with R-Smads are indicated. Expression levels of 6Myc-Smad4, 6Myc-␣A, and FLAG-R-Smads are shown. B, COS7 cells were transfected with the indicated combinations of FLAG-Smad1 or -Smad3, 6Myc-␣A, -␣B, or -␣C, and constitutively active forms of type I receptors. Complex formation between PEBP2␣ subunits and Smads was detected by anti-FLAG immunoprecipitation followed by anti-Myc immunoblotting. Expression of 6Myc-PEBP2␣ subunits and FLAG-Smads is indicated.

FIG. 2. Interaction between ␣C and Smad3.
A, COS7 cells were transfected with the indicated combinations of cDNAs encoding FLAGtagged Smad3, 6Myc-Smad4, 6Myc-␣C, and T␤R-I(TD)-HA. Cell lysates were immunoprecipitated with anti-FLAG antibody followed by immunoblotting using anti-Myc antibody. ␣C and Smad4 co-immunoprecipitated (IP) with Smad3 are indicated. Expression levels of Smad4, ␣C, and Smad3 were confirmed. B, regions in ␣C proteins essential for their interaction were determined by GST pull-down assay. The structure of ␣C is shown in the upper panel. GST-Smad3 and GST alone were incubated with a series of C-terminal deletion constructs of ␣C. 10% of [ 35 S]methionine-labeled proteins used for the assay were applied as controls (Input). C, the MH1 and MH2 domains of Smad3 were examined for the interaction with ␣C proteins by GST pull-down assay. GST-Smad3 (MH1), GST-Smad3 (MH2), and GST alone were incubated with C-terminal deletion constructs of ␣C. 25% of [ 35 S]methioninelabeled proteins were applied as controls (Input).
Luciferase Assay-A20.3 B lymphocytes were transfected with the germline Ig C␣ promoter (18) together with the expression constructs for ␣C, Smads, and T␤R-I(TD). P19 murine embryonal carcinoma cells were transfected with WT or mutant versions of (T␤RE) 3 -Lux together with ␣C, Smads, and T␤R-I(TD). Firefly and Renilla luciferase activities were assayed with the dual luciferase assay system (Promega) using Lumat LB 9507 (EG&G Berthold). Firefly luciferase activity was normalized with respect to the Renilla luciferase activity.
Electrophoretic Mobility Shift Assay (EMSA)-EMSA was performed as described (27) with minor modifications. Briefly, COS7 cells were transfected with a mixture of expression plasmids encoding T␤R-I, T␤R-II, Smads, ␣C, and ␤2. Whole-cell extracts were prepared, mixed in vitro in combinations as indicated if necessary, and used for EMSA with a 32 P-labeled T␤RE.

Interaction of ␣ Subunits of PEBP2 with R-Smads in
Vivo-We first tested complex formation between ␣A and R-Smads activated by BMPs. ␣A interacted with Smad1 and Smad5, which were activated by, BMPR-IB(QD), a constitutively active form of BMPR-IB (Fig. 1A). Smad4 was co-immunoprecipitated with R-Smads activated by the receptors. Importantly, ␣A also interacted with Smad2 and Smad3 activated by T␤R-I(TD), an active T␤R-I.
We therefore examined whether the other PEBP2␣ subunits associate with different R-Smads. Smad3 activated by T␤R-I(TD) formed complexes not only with ␣A but also with ␣B and ␣C (Fig. 1B). Smad1 activated by BMPR-IB(QD) also formed complexes with ␣A, ␣B, and ␣C. Other R-Smads, i.e. Smad2 activated by T␤R-I(TD) and Smad5 activated by BMPR-IB(QD), also interacted with all three ␣ subunits (data not shown). ␣B and ␣C formed complexes with Smad1 and Smad3 equally well, whereas ␣A associated more strongly with Smad1 and Smad5 than with Smad3 (Fig. 1, A and B). These results indicate that all three mammalian PEBP2␣ subunits can form complexes with R-Smads.
Interaction of ␣C with Smad3-Because ␣C is predominantly induced by TGF-␤ in B lymphocytes and is critical for the induction of the promoter for germline Ig C␣ transcripts upon TGF-␤ stimulation (18), complex formation between ␣C and Smad3 was studied in detail. The ␣C/Smad3 complex was observed in the presence and absence of T␤R-I(TD) (Fig. 2A), and Smad4 interacted with Smad3 upon stimulation by T␤R-I(TD). The mode of interaction between ␣C and Smad3 was studied by GST pull-down assays using deletion constructs of these proteins. When a series of C-terminally truncated constructs of ␣C was examined, deletion of a C-terminal region (␣C (1-283; see Fig. 2B)) corresponding to a part of the transcriptional activation domain (AD) identified in ␣B (27) resulted in a reduction of association with GST-Smad3, and interaction became undetectable by deletion of transcriptional AD (␣C (1-234)) (Fig. 2B). Smads have highly conserved MH1 and MH2 domains in their N-and C-terminal regions, respectively (1,2). A GST pull-down assay revealed that the MH2 domain bound to ␣C (Fig. 2C). In addition, the MH1 domain weakly interacted with ␣C, but the exact location in ␣C where MH1 interacts could not be determined unequivocally because of the weakness of the interaction.
Transcriptional Activation of the Germline Ig C␣ Promoter-We next studied the functional consequence of R-Smad/ PEBP2␣ interaction using the mouse Ig C␣ promoter. The promoter for mouse germline Ig C␣ transcripts has been shown to contain a TGF-␤-responsive element, T␤RE (34), in which two PEBP2␣ binding sites have recently been identified (18). The human germline Ig C␣ promoter was also shown to contain PEBP2␣ binding sites in its T␤RE (35). In addition, two potential Smad binding motifs (36 -38) are found in the T␤RE (Fig.  3A). Moreover, an additional PEBP2␣ binding site and one Smad binding motif are observed between the T␤RE and the transcription initiation site. To determine the functional importance of these binding motifs, nucleotide mutations were introduced into the promoter, and a transcriptional response assay was performed using A20.3 B lymphocytes. As previously reported (18), TGF-␤ activates the promoter, which is further enhanced by the presence of ␣C. Mutations in the Smad binding motifs in the T␤RE (T␤RE-mS) and those in the PEBP2␣ binding sites (T␤RE-mP) result in dramatic decreases in transcriptional activity (Fig. 3B). A complete loss of response was observed in the mSP mutant with mutations in all PEBP2␣ and Smad binding motifs, indicating that both of these binding motifs are essential for transcriptional activation.
A dominant negative form of Smad3, Smad3(DE), which prevents the activation of both Smad2 and Smad3 by T␤R-I(TD) (26), inhibited the transcription induced by T␤R-I(TD) and ␣C (Fig. 3C). This finding suggests that transcription may be induced by the endogenous R-Smads activated by T␤R-I(TD). Moreover, co-transfection of Smad3 with ␣C strongly induced transcription from the Ig C␣ promoter (Fig. 3D) but not from the Ig C␣ promoter containing mutations in the T␤RE, as shown in Fig. 3A (data not shown). Interestingly, Smad2 did not significantly induce the transcription, probably because Smad2 is unable to bind to the Smad binding motifs (36 -39).
Transcriptional Activation through T␤RE by ␣C and Smad3/4 -To further study the roles of ␣C and Smad3/4 in activating transcription, three tandemly repeated T␤REs (WT or mutant versions) of the Ig C␣ promoter were fused to the heterologous c-Fos promoter, and transcriptional activity was determined using transfected P19 embryonal carcinoma cells, which have very low levels of endogenous PEBP2␣ activity (32,33). Similar to the results obtained with the natural Ig C␣ promoter using A20.3 B lymphocytes, transcriptional activity of (T␤RE-WT) 3 -Lux was mildly induced by Smad3 and -4, whereas the addition of Smad3/4 and ␣C in cells activated by T␤R-I(TD) greatly induced transcription (Fig. 4A). In contrast, mutant versions of (T␤RE) 3 -Lux, i.e. (T␤RE-mP) 3 -Lux and (T␤RE-mS) 3 -Lux, which have mutations in the two PEBP2 binding sites and two Smad binding motifs, respectively, did not respond to T␤R-I(TD), Smad3/4, or ␣C, indicating that both  2 g), and FLAG-Smad3 (0.8 g) or that containing 6Myc-␣C (0.2 g) and ␤2 (0.2 g), or the cells were mock-transfected with an empty plasmid. Whole-cell extracts were mixed in vitro in combinations as indicated and used for EMSA with a 32 P-labeled T␤RE. The total amount of extract was kept constant using the mock extract. of these binding motifs are essential for transcriptional activation by the Smad3/␣C complex.
To determine the domain(s) in ␣C critical in the transcriptional activation in concert with Smads, a series of C-terminal deletions of ␣C was tested for transcription activity. ␣C mutants containing the transcriptional AD increased transcriptional activation in the presence of T␤R-I(TD) and Smad3/4; however, deletion of one-half of the AD resulted in a significant decrease in transcriptional response; complete loss of response was obtained with the mutants lacking the entire AD (Fig. 4B). This result indicates that the physical interaction between Smad3 and ␣C may be critical for the transcriptional activation through T␤RE (see Fig. 2B).
DNA Binding of the ␣C, ␤2, and Smad3 Complex-The formation of DNA-binding complexes containing ␣C and Smad3 on the germline C␣ T␤RE DNA was studied by EMSA. The ␤ subunit of PEBP2 (␤2 isoform) was included in this assay to enhance the DNA binding of PEBP2. Smad3 activated by T␤R-I(TD) and ␣C independently formed DNA-binding complexes, which could be detected as slowly migrating complexes in EMSA (Fig. 5A, lanes 2 and 3, and B, lanes 3 and 4). In the presence of activated Smad3 and ␣C/␤2, a more slowly migrating complex was formed both in vitro and in vivo (Fig. 5A, lanes  4 and 13, and B, lane 5). These complexes were super-shifted in the presence of corresponding antibodies to the epitope tags or an antibody to the ␤ subunit, indicating that ␣C/␤2 and Smad3 can concomitantly bind to DNA as a multimeric Smad3/␣C/␤2 complex.
Mutations in the Smad binding motifs S1 or S2 resulted in the decrease or loss, respectively, of Smad3 and Smad3/␣C/␤2 bindings, but the binding of ␣C/␤2 still remained (Fig. 5B). When the PEBP2␣ sites were mutated, a mutation in P2, but not in P1, disrupted the bindings of ␣C/␤2 and Smad3/␣C/␤2, but binding of Smad3 was still detected. The Smad binding motifs and PEBP2␣ binding sites thus appear to be specific and sufficient for the binding of corresponding proteins, but both are required for the binding of the Smad3/␣C/␤2 complex to the T␤RE and for activation of the promoter by ␣C and Smad3. DISCUSSION The findings shown in the present study revealed that PEBP2␣ subunits and R-Smads specific for both TGF-␤ and BMP signaling pathways form complexes together with Smad4 and that the complex formation appears to be critical for efficient transcriptional activation of target genes, including the germline Ig C␣ promoter. Our findings suggest that PEBP2 may function as a nuclear target of TGF-␤/BMP signaling pathways and that the biological effects of TGF-␤/BMP may be regulated by cooperation between Smads and PEBP2.
Smads have been reported to interact with various DNAbinding proteins as well as the transcriptional coactivator p300/CBP and co-repressor TGIF (40,41). Because members of the TGF-␤ superfamily have pleiotropic functions, interaction with various transcription factors may be required for Smads to exhibit specific effects in certain cell types. Many of these interacting partners, including c-Jun and the vitamin D receptor (6,42), preferentially interact with Smad3, but Xenopus FAST1 and murine FAST2 have been shown to associate with Smad2 as well (3,4,39). Recently, a homeodomain transcription factor Hoxc-8 has been shown to bind to Smad1 (43). PEBP2 is conspicuous compared with these factors, because all three mammalian ␣ subunits of PEBP2 interact with all R-Smads tested in the present study. Recently, SIP1 has been shown to interact with all R-Smads; in contrast to the PEBP2 ␣ subunits, however, SIP1 is a transcriptional repressor, and interaction with R-Smads may lead to relief of repression of target genes by SIP1 (44).
Smad3 interacts with ␣C mainly through the MH2 domain, whereas the MH1 domain binds weakly to ␣C. Analysis by C-terminal deletion of ␣C revealed that the C-terminal region, including the transcriptional AD of ␣C, is required for efficient interaction with the MH2 domain of Smad3.
PEBP2 is a context-dependent transcription factor, requiring interacting partners for transcriptional activation, including Ets-1 (21,22). In the germline Ig C␣ promoter, both PEBP2 and Smad binding sites are essential for transcriptional activation. In contrast, FAST1 binds to the Mix.2 gene promoter with high affinity, and therefore direct binding of Smads to DNA may be less important than in the Ig C␣ promoter (39). Thus, in certain other promoters to which PEBP2 binds with a high affinity together with other transcription factors, direct DNA binding of Smads may not be critical for cooperative transcriptional activation by PEBP2 and Smads.
Our present study revealed that PEBP2␣ subunits interact with R-Smads activated by TGF-␤/activin, as well as with those activated by BMPs, and that functional cooperation between ␣C and Smad3 is required for transcription driven by the germline C␣ promoter. Germline Ig ␣ transcripts are required for IgA class switching (45). Because members of the TGF-␤ superfamily exhibit a wide variety of biological effects, it will be very important to examine whether PEBP2 is involved in these biological events as a nuclear target of Smads.