Smad and AML Proteins Synergistically Confer Transforming Growth Factor b 1 Responsiveness to Human Germ-line IgA Genes*

Transcription of germ-line immunoglobulin heavy chain genes conditions them to participate in isotype switch recombination. Transforming growth factor- b 1 (TGF- b 1) stimulates promoter elements located upstream of the IgA1 and IgA2 switch regions, designated I a 1 and I a 2, and contributes to the development of IgA responses. We demonstrate that intracellular Smad proteins mediate activation of the I a 1 promoter by TGF- b . TGF- b type 1 receptor (ALK-5), activin type IB receptor (ALK-4), and the “orphan” ALK-7 trans-activate the I a 1 promoter, thus raising the possibility that other members of the TGF- b superfamily can also modulate IgA synthesis. Smads physically interact with the AML family of transcription factors and cooperate with them to activate the I a 1 promoter. The I a 1 element provides a canape´ of interspersed high and low affinity sites for Smad and AML factors, some of which are indispensable for TGF- b responsiveness. While AML z Smad complexes are formed in the cytoplasm of DG75 and K562 cells constitutively, only after TGF- b (Amer-sham GST-Smad CNBr-activated Sepharose 4B according the instructions 35 immunoprecipitation, detect

Production of the appropriate antibody isotypes during a humoral immune response is an important prerequisite for the development of protective humoral immunity. Distinct isotypic profiles are shaped by differentially regulating immunoglobulin heavy chain class-switching-associated DNA rearrangements. Such events juxtapose selectively an expressed heavy chain variable region (VDJ) to a new downstream constant region (C H ), deleting the DNA in between containing the earlier expressed IgM and IgD genes (1). Experimental evidence has implicated transforming growth factor-␤1 (TGF-␤1) 1 as a positive regulator of IgA production (2)(3)(4). There are two IgA isotypes in the human system, IgA1 and IgA2, encoded by two different genes, one in each of the two duplication units that compose the human immunoglobulin heavy chain locus (5). Both genes contain, upstream of their switch regions, almost identical TGF-␤1 responsive promoter/enhancer elements designated intron (I) ␣1 and I␣2, respectively (6,7). According to the "accessibility model" for isotype class switching (8,9) transcribed germ line IgA genes become "accessible" to the Ig class switch recombination machinery and thus are preferentially rearranged and expressed. The importance of the I region promoter/enhancer elements for proper guiding of a given immunoglobulin heavy chain locus toward a certain class has been repeatedly illustrated by elegant gene targeting experiments where disruption of I region elements prevented rearrangement of the targeted alleles (10,11).
Genomic segments encompassing approximately 130 base pairs upstream of the human and mouse I␣ transcription initiation sites contain all the information necessary for expression and TGF-␤ responsiveness in transient expression assays (6,7,12). A portion of the above sequences, containing a directly repeated motif (referred to as T␤RE) for TGF-␤ responsive element (12) or DRE for direct repeat element (13) could transfer TGF-␤ responsiveness to an unrelated promoter. We have recently shown that the DRE element contains functional binding sites for transcription factors of the acute myeloid leukemia (AML) family (14), and that AML proteins cooperate with transcription factors of the Ets family to activate the I␣1 promoter. The AML proteins, also known as core-binding factors (CBF) (15), polyoma enhancer binding proteins (PEBPs) (16), or Runt domain transcription factors bind to DNA as ␣/␤ subunit heterodimers (15,17). Three ␣ chains are encoded by the AML1 (CBFA2/PEBP2␣B), AML2 (CBFA3/PEBP2␣C), and AML3 (CBFA1/PEBP2␣A) genes and one ␤ subunit is encoded by the PEBP2␤ (CBF␤) gene. The ␣ subunits contain the DNA binding runt homology domains, as well as sequences necessary for association with the ␤ subunit and trans-activation, whereas the ␤ subunit does not bind DNA but increases the strength of the DNA binding of the AML ␣ subunits (15)(16)(17).
Here we demonstrate that the I␣1 promoter contains an array of interspersed functional Smad and AML-binding sites that mediate in a cooperative manner the TGF-␤ responsiveness of this element. Our findings extend the repertoire of Smad interacting transcription factors and provide the framework for characterizing the signals that regulate the interplay between AMLs and Smads in the context of the I␣1 promoter and other promoter/enhancer elements that contain neighboring AML-and Smad-binding sites.
Rabbit antiserum was prepared against the synthetic peptide "DG-PRPEPRRHRQKLDD" derived from a conserved region in the runt homology domain. The peptides were synthesized with an Applied Biosystems 430 A peptide synthesizer using t-butoxycarbonyl chemistry and were purified by reverse-phase high performance liquid chromatography. The peptides were coupled to keyhole limpet hemocyanin using glutaraldehyde. The coupled peptides were mixed with Freund's adjuvant and used to immunize rabbits.
Plasmid Construction and Transfections-The mouse AML1b cDNA was cloned into the pSP64-poly(A) ϩ KSN vector (a derivative of pSP64poly(A) ϩ (Promega, Madison, WI) where the PstI site was replaced by KpnI, SacII, and NotI sites between the HindIII and SalI sites). The sequence 5Ј-GGTACCACC-3Јwas introduced in front of the start codon of AML1b by PCR with Pfu polymerase and the modified AML1b cDNA was inserted into the KpnI/XhoI sites of pSP64-poly(A) ϩ KSN.
Transient transfections were performed as described previously (7) using 2 g of hCMV-␤-gal plasmid (reference plasmid for normalization), 3 g of reporter plasmid, and 1-6 g of each expression plasmid as indicated. Where necessary, the empty expression vector pBJ9⍀ was added to a total of 10 g of expression plasmids. 10 ϫ 10 6 cells were electroporated followed by incubation in a volume of 10 ml of media. Human TGF-␤1 was added 0.5 h after electroporation where indicated. The cells were harvested 20 h after electroporation.
EMSA and DNase I Footprinting Analysis-EMSAs and footprinting analyses were performed as described previously (13,14).
Production and Purification of Recombinant Proteins-Glutathione S-transferase (GST)-Smad fusion proteins were expressed in Escherichia coli and partially purified by column chromatography according to the instructions of the manufacturer (Amersham Pharmacia Biotech). Briefly, bacteria grown in 1 ϫ Luria broth medium were induced with 0.1 mM isopropyl-␤-D-thiogalactopyranoside. After sonication the GST fusions were isolated using glutathione-Sepharose 4B, washed 3 times, eluted, and dialyzed against phosphate-buffered saline supplemented with 2 mM dithiothreitol and 0.5 nM phenylmethylsulfonyl fluoride. Thereafter, the proteins were purified on heparin-Sepharose columns following the instructions of the manufacturer (Amersham Pharmacia Biotech).
AML1b subcloned in the pSP64-poly(A) ϩ KSN vector was in vitro transcribed and translated using the TnT-coupled reticulocyte lysate system (Promega, Madison, WI) with SP6 polymerase. The proteins were purified by heparin-Sepharose chromatography as described previously (31).
In Vivo AML/Smad Binding Assay-Fifteen hours after transfection, K562 or DG75 cells were lysed in 1 ml of lysis buffer (50 mM Tris-HCl, pH 8, 150 mM NaCl, 1% Nonidet P-40, 0.5% deoxycholic acid, 0.1% SDS, 20 mM sodium pyrophosphate, protease inhibitor mixture (Roche Molecular Biochemicals, Germany)). The cell lysates were precleaned with protein G-Sepharose beads and incubated with Flag M2agarose beads (Aldrich-Sigma) for 2 h at 4°C. After washing the immunoprecipitates with lysis buffer three times, immunoprecipitates and aliquots of cell lysates were separated by SDS-gel electrophoresis using 10% polyacrylamide gels and transferred to a Hybond-C extra membrane (Amersham Pharmacia Biotech). The membrane was then probed with anti-Flag M5 or anti-Myc (9E10 monoclonal antibody; Santa Cruz, CA) antibodies. Primary antibodies were detected with horseradish peroxidase-conjugated goat anti-mouse antibody (Amersham Pharmacia Biotech) and chemiluminescent substrate.
In Vitro AML/Smad Binding Assays-Equal amounts of GST or GST-Smad proteins were coupled to CNBr-activated Sepharose 4B (Amersham Pharmacia Biotech) according to the instructions of the manufacturer. The protein-coupled beads were incubated with 35 Slabeled in vitro translated AML1b protein in binding buffer (20 mM HEPES, pH 8.0, 10% glycerol, 100 mM NaCl, 2 mM dithiothreitol, 0.05% Triton X-100). After incubation for 2 h at 4°C, the beads were washed with binding buffer. Bound proteins were eluted by boiling in SDSpolyacrylamide gel electrophoresis sample buffer, detected by SDSpolyacrylamide gel electrophoresis and autoradiography, and quantitated with a PhosphorImager (Molecular Dynamics, Sunnyvale, CA).

RESULTS
Smad Proteins Regulate the TGF-␤ Responsiveness of the I␣1 Promoter Elements-Work over the past few years has demonstrated that Smad proteins are major transcription regulators of TGF-␤ responsive promoter/enhancer elements (18,19). However, TGF-␤ family members activate other signaling pathways as well (32,33). In order to clarify whether the TGF-␤ responsiveness of the I␣ promoters is Smad-dependent we co-transfected the I␣1 promoter-driven luciferase reporter construct with expression constructs for different Smad proteins in the Burkitt's lymphoma B cell line DG75. This cell line has been extensively used for analyzing I␣ promoter activity (7,13). Smad3 increased significantly (ϳ2.5-fold induction) the ligand-independent and ligand-dependent I␣1 promoter activity, whereas Smad2 and Smad4 increased it moderately (ϳ1.4fold induction). However, when Smad2, Smad3, and Smad4 were combined, a stronger transactivation was observed with the highest response obtained when all three proteins were co-expressed (ϳ10-fold induction for ligand-independent and ϳ5-fold induction for ligand-dependent transactivation). Smad1 and Smad5 transfected individually, together, or in combination with Smad4, did not influence the I␣1 promoter ( Fig. 1 and data not shown). BMP responsive Smad combinations did not up-regulate the I␣1 promoter even when they were co-transfected with the constitutively active ALK-1, thus excluding the absence of the necessary receptors in the utilized cell lines as the reason for the I␣1 unresponsiveness (data not shown). While the Smad variants that interfere with the TGF-␤/activin signaling pathway, such as Smad7 and dominant negative variants of Smad3 (3⌬c) or Smad4 (4⌬c) (30) suppressed basal and TGF-␤1 induced I␣1 activity, Smad6, if anything, induced a small increase in promoter activity (Fig. 1). In conclusion, Smad3 and Smad4 and to some extend Smad2 are major regulators of TGF-␤ responsiveness of the I␣1 promoter.
Constitutively Activated T␤R-I/ALK-5, ActR-IB/ALK4, and Orphan ALK-7 Can Stimulate the I␣1 Promoter-Different members of the TGF-␤/activin family can activate similar Smads (18 -20). Therefore, we questioned the exclusive involvement of TGF-␤1 in the regulation of I␣1. We tested the effect of TGF-␤ superfamily members to stimulate the I␣1 driven reporter construct. TGF-␤1, TGF-␤2, activin A, but not BMP2 and BMP7, were able to induce a transcriptional response in K562 and DG75 cells (data not shown). Analysis of TGF-␤ superfamily receptors on these cells showed the presence of TGF-␤ and activin receptors but absence of BMP receptors (data not shown). Therefore, we also analyzed the capacity of different constitutively active type I receptors to stimulate the I␣1 driven reported construct. As shown in Fig. 2, expression of constitutively active ALK-1, ActR-I/ALK-2, BMPR-1A/ALK-3, or BMPR-1B/ALK-6, capable of activating Smad1 and Smad5 did not affect the I␣1 promoter activity in DG75 cells. However, expression of constitutively active ActR-1B/ALK-4 or T␤R-I/ ALK-5, involved in activin and TGF-␤ signaling, respectively, and the "orphan" ALK-7 (29), induced the I␣1 reporter in a dose-dependent manner. ALK-4 and ALK-7 in particular could reconstitute almost 80 -90% of the transcription activity obtained by stimulating the cells with TGF-␤1 (Fig. 2). Thus in addition to TGF-␤, activins or even not as yet identified members of the TGF-␤ superfamily, can potentially activate transcription of the germ line IgA genes, by modulating the activity of Smad3 and Smad4, and consequently participate in regulation of IgA humoral responses.
Localization of the Smad Responsive Motifs in the I␣1 Promoter-To identify the sequences mediating Smad responsiveness in the I␣1 promoter, deletion mutants were constructed and their responsiveness to TGF-␤1 in the presence or absence of expression plasmids for Smad proteins was studied in the B cell line DG75. As shown in Fig. 3, deletion of sequences between Ϫ142 and Ϫ99, encompassing the DRE element reduced dramatically the capacity of the promoter to respond to TGF-␤1 in the presence or absence of exogenous Smad proteins. The small remaining Smad responsiveness is probably due to additional Smad-binding sites located downstream (see below, Fig.  4).
To define more specifically the Smad-interacting sequences of the I␣1 promoter, we produced recombinant Smad proteins in E. coli and used the highly purified proteins in EMSA and DNase I footprinting analyses. EMSA demonstrated that Smad1, Smad3, and Smad4 were able to bind to a doublestranded oligonucleotide spanning the DRE element (Ϫ131/ Ϫ96). Smad4 bound with the highest affinity since almost 10 times more Smad1 and 4 times more Smad3 was required to produce the same level of retarded probe. Smad2 did not bind even at 100-fold higher concentration (data not shown). Footprinting analyses with recombinant Smad1, Smad3, and Smad4 proteins demonstrated the presence of an array of protected segments within the I␣1 region (Fig. 4, A and B). Simi- Constitutively active type I receptors (mutations: ALK1-Q201D, ALK2-Q207D, ALK3-Q233D, ALK4-T206D, ALK5-T204D, ALK5-T204D, ALK6-Q204D, and ALK7-T194D) as well as wild-type receptors were co-transfected with the I␣1 reported construct (Ϫ142/ϩ79) in DG75 cells and their capacity to stimulate the I␣1 promoter was assayed 20 h later. Relative fold activation is calculated considering the activity of the I␣1 reported construct without and ligand or receptor stimulation as 1.
larly to the EMSA analyses, Smad4 generated footprints at lower protein concentrations than Smad3, suggesting a higher affinity for the target DNA, and Smad1 bound less efficiently than Smad3. Smad2 did not footprint even at the highest concentration used, thus serving as a negative control. Interestingly, the Smads differed from each other not only regarding their affinity for DNA, but also in the way by which they interacted with it. At optimal concentrations Smad1, Smad3, and Smad4 produced footprints around three areas (Ϫ127/ Ϫ100, Ϫ95/Ϫ65, and Ϫ51/Ϫ20). However, the footprints were similar, but not identical (Fig. 4, A and B, and footprints for upper strand not shown). The footprints of Smad3 and Smad4 within the Ϫ127/Ϫ67 area were identical, but Smad4 footprinted a larger stretch in the Ϫ51/Ϫ20 area. Smad1 protected a limited stretch within the Ϫ127/Ϫ100 segment and only the lower strand was weakly protected around the Ϫ51/Ϫ20 area. Furthermore, while all three induced DNase I hypersensitivity at position Ϫ88(G) in the upper strand (which lies within a putative CRE/ATF site), all three Smad molecules induced additional unique hypersensitive sites. Smad4, for example, was the only one that induced a hypersensitive site on the upper strand next to the high affinity AML site, position Ϫ111(C), and another on the upper strand at position Ϫ29(G). The recombinant Smads initially used in our studies were synthesized as GST fusion proteins. Identical results were obtained using recombinant Smad3 and Smad4 proteins produced in baculovirus-infected insect cells (data not shown).
Furthermore, phosphorylated Smad3 produced by co-expressing the activated T␤R-I/ALK-5 kinase domain in insect cells gave identical footprints as the non-phosphorylated and the GST-Smad3 proteins, however, it required only one-fifth protein concentration. Collectively, the footprinting analyses reveals differences in the way the various Smads interact with DNA, not only regarding the affinity of interaction, but also regarding the effect their binding has on the DNase I accessibility of the DNA, especially on neighboring sequences.
Alignment of all the sequences protected by Smad proteins demonstrated the presence of the CAGAC(C/A) motif or its complementary (T/G)GTCTG in all of them (Fig. 4B). Recent structural and molecular studies have demonstrated that indeed the AGAC sequence that is present within the above motifs constitutes the core of Smad-binding sites (34 -38).
Smad Proteins Co-operate with AML1 to Stimulate the I␣1 Promoter-We have previously demonstrated that AML proteins and their corresponding binding sites in the I␣1 promoter are involved in its TGF-␤1 responsiveness (14). In light of the findings that Smad proteins integrate physically and functionally with other transcription factors (20,30,39,40) we investigated the interplay between Smad and AML proteins. For these experiments we utilized the K562 erythroleukemia cell line which expresses very low levels of AML proteins and displays very low TGF-␤ responsiveness as measured by the I␣1 reporter construct (maximally 2-fold induction). As shown in Fig. 5, overexpression of AML1b (the largest splice forms of the mouse AML1) or a combination of Smad3 and Smad4, increased dramatically the TGF-␤1 responsiveness of K562 cells. When AML and Smad proteins were co-expressed, a synergistic effect on the I␣1 activity was observed. The synergistic effect was clearer in the ligand independent responses (open bars, Fig. 5). Better synergistic effect where obtained in AML1b/Smad3/Smad4-transfected TGF-␤1-stimulated cells when the fetal calf serum in the culture medium was reduced from 5 to 1% instead (data not shown).
The synergistic action of AML and Smad sites on the activity of the I␣1 promoter was illustrated more dramatically when the highest affinity AML and the closely located high affinity Smad site were mutated (Fig. 5). EMSA analyses using wildtype and mutated oligos demonstrated that mutations that reduced binding of recombinant AML proteins to the AML D site did not abrogate binding of recombinant Smad proteins in the nearby SMAD B site and vice versa (Fig. 5, inset). However, mutation of either the AML D or the SMAD B sites abrogated the positive effect that overexpression of AML1b and/or Smad3/ Smad4 had on the promoter activity, in the presence or absence of TGF-␤ stimulation. Mutation of both these sites affected the I␣ promoter even more. In addition to the above two sites, we mutated the SMAD A , SMAD D , AML U , and CRE/ATF sites individually or in combinations. In all cases the mutations decreased the I␣1 promoter activity, indicating that all these sites even those with low affinity for the corresponding transcription factors are necessary for optimal I␣1 function (data not shown).
Receptor-dependent Formation of Novel DNA⅐Protein Complexes Containing AML, Smad3, and Smad4 -To detect receptor dependent complex formation in EMSAs, we co-transfected DG75 cells with different combinations of epitope-tagged AML1b, Smad3, Smad4 and constitutively active ActR-IB/ ALK-4 receptor (Fig. 6). The constitutively active ALK-4 stimulates the same R-Smads as the constitutively active T␤RI/ ALK-5 and as shown in Fig. 2 it can trans-activate the I␣1 promoter. Similarly transfected cells were also stimulated with TGF-␤1 (20 h after transfection for 30 min) instead of the constitutively active receptor. Nuclear extracts were prepared and analyzed in EMSA using a probe that contained the AML U , SMAD A , AML D , and SMAD B sites (Fig. 6). Overexpression of tagged Smad3, Smad4, and AML1 resulted in the appearance of a weak, novel, lower mobility band. However, when the cells were stimulated by TGF-␤1 or co-transfected with the constitutively active type I receptor, this novel band increased significantly. The novel complex was supershifted/inhibited with antibodies against the AML1b protein (anti-HA), Smad3 (anti-Myc), or Smad4 (anti-Flag), indicating that all these proteins can bind to the I␣1 DRE sequences in a receptor-dependent manner. Thus, activation of the receptor induces the appearance of novel complexes that contain Smad3, Smad4, and AML proteins.

Smad and AML Proteins Can Physically Interact in Vivo and in
Vitro-Recent studies have demonstrated that Smads regulate transcription through their ability to bind to DNA directly and/or indirectly and to induce transcriptional responses through cooperativity with other transcription factors (20, 30, 39 -41). To investigate the molecular basis of AML/Smad synergy we expressed epitope-tagged AML1b and Smad3 in K562 and DG75 cells in the presence or absence of constitutively active ALK-4 (Fig. 7). AML1b and Smad3 were co-immunoprecipitated in both cells analyzed. The AML1b/Smad3 interaction was constitutive in both cell lines tested. Co-transfection of a PEBP2␤ expression vector did not improve the interaction, probably due to the already existing adequate endogenous levels in DG75 and K562 (14). Fractionation of cytoplasmic from nuclear extracts showed the majority of complexes to be in the cytoplasm (data not shown).
To map the domains of Smad3 and AML1b involved in this protein-protein interaction, different epitope-tagged derivatives of either molecule (Fig. 8A) were expressed in DG75 cells and their interaction was investigated by co-immunoprecipitation. As shown in Fig. 8C, fragments containing the C-terminal MH2 domain of Smad3 exhibited the best AML1b binding activity.
Similarly, different truncations of the AML1b molecule were tested for their capacity to interact with full-length Smad3. A number of functional domains have been recently defined in the AML molecule (42). Based on these mapping data a number of AML1b deletion variants were produced (Fig. 8A) and tested for interaction with full-length Smad3 (Fig. 8B). A complex pattern of interaction was observed since all the AML1b derivatives were able to interact with Smad3, albeit with different efficiency. More specifically, N-terminal deletions up to amino acid residue 411 did not affect binding. Deleting further amino acids 371-411 reduced significantly binding, without com- pletely eliminating it. Deleting even further down to amino acid 291 did not alter the binding, however, a deletion reaching down to amino acid 243 restored binding activity to the level even higher than the intact AML1b molecule. AML variants 6 and 9 containing either the N-terminal or C-terminal half of AML1b were able to interact with full-length or the MH2 variant of Smad3. No interaction was observed with the Mad homology domain-1 portion of Smad3 (Fig. 8C). To make sure that AML1b and Smad3 interact directly with each other and not via an intermediate adapter molecule, we used 35 S-labeled in vitro translated and purified AML1b and E. coli expressed purified GST-Smad fusion protein immobilized on Sepharose beads. As shown in Fig. 9, approximately 50 -80% of the in vitro translated AML1b was retained on beads coupled with recombinant Smad1, Smad2, Smad3, and Smad4, whereas only 10% was bound to GST beads. Smad1 and Smad3 bound better than Smad2 and Smad4. Thus, purified Smads and AML1b can physically associate with each other in the absence of other cytoplasmic proteins.
To make sure that transfection with the expression vectors did not result in the appearance of unphysiological levels of AML1b protein and also to investigate whether TGF-␤ signaling could alter the levels of endogenous AML proteins, cells were harvested at different time points after transfection/ TGF-␤1 stimulation and analyzed by Western blotting with an anti-AML antibody. This antibody was raised against a peptide that is conserved in all AML family members and thus can detect all of them. As shown in Fig. 10, exogenous AML1b was detected 1 h and reached maximal levels 7 h after transfection. The levels of exogenous AML1b protein even when they reached maximum were comparable with the levels of other endogenous AML variants. Since AML1b and Smad3 could be co-precipitated even when 10 times less expression vectors were transfected (data not shown), we conclude that the observed interaction is not due to unphysiologically high levels of expression. It has recently been shown that TGF-␤1 stimulates AML2 mRNA synthesis in a mouse cell line (46). As shown in Fig. 10, overexpression of AML1b, Smad3/4, and TGF-␤1 stimulation, if anything, resulted in reduction of some of the endogenous AML variants. This makes unlikely that the cooperation studied herein is due to an induction of AML gene expression by TGF-␤ signaling.
Collectively, the in vivo and in vitro binding data indicate that parts of the AML1b molecule spanning the RDH region and the area between amino acids 371 and 411 participate in AML1b/Smad3 interaction, however, all these interactions target the MH2 domain of the Smad3 molecule. At least in the two cell lines analyzed, this interaction is constitutive and is not modulated by TGF-␤1 signaling.

The TGF-␤ Responsiveness of the I␣1 Promoter Is Regulated by Direct Interaction of Smads with Specific Sequences within
It-TGF-␤ has been implicated as a major regulator of IgA synthesis. One of the targets of TGF-␤ mediated signaling are the I␣ promoters, which are located upstream of the S␣ regions and function as a regulator of the accessibility of the C␣ germ line loci to class switch recombinases. To clarify the molecular mechanism(s) that skew immunoglobulin production toward IgA, we have characterized the molecular components that participate in the TGF-␤1-regulated transcription through I␣1, Nuclear extracts were prepared 20 h after transfection. TGF-␤1 stimulation was done for 30 min before the termination of the culture and the preparation of the nuclear extracts. EMSA was performed with a probe containing the AML U , Smad A , AML D , and Smad B -binding sites (Ϫ131/Ϫ96). Antibodies specific for the Myc, Flag, and HA epitopes recognizing the epitope-tagged Smad3, Smad4, and AML1b, respectively, were used to supershift the TGF-␤1 or ALK4-induced complex. No nuclear extract was added in the first lane. Complex A represents binding of endogenous AML proteins, B represents the novel TGF-␤1 or caALK-4 receptor induced complexes, C and D represent the anti-Smad3 and anti-Smad4 antibody produced supershifts. The anti-hemagglutinin antibody inhibited complex formation without producing a significant supershift.

FIG. 7.
Interaction of AML1b with Smad3 in vivo. DG75 and K562 cells were transfected with the indicated combinations of epitopetagged AML1b and Smad3 with or without constitutively active ALK-4 expression vector (caALK4). The physical interaction between AML1b and Smad3 was analyzed as described under "Experimental Procedures." the regulatory module of the human IgA1 locus. Our previous studies and the results presented in the current report demonstrate that the I␣1 promoter contains an array of interspersed binding sites for AML and Smad proteins whose cooperative activity underlines I␣1 germ line transcription. Mutation of AML and Smad sites, individually or in combinations, demonstrated that those of high affinity are indispensable for TGF-␤ responsiveness, whereas those of low affinity contribute to it, however, only when the high affinity sites are intact. The presence of multiple AML and Smad sites with different degrees of importance in the I␣1 promoter, and possibly in other Smad-regulated promoters, might allow the promoter to function as a "rheostat" responding gradually to signals of varying strength instead of responding in an all-or-none manner. This possibility is compatible with the finding that synthetic TGF-␤ responsive promoters respond better when they have several copies of Smad-binding sites. In an in vivo situation the varying parameter could be the strength of ligand induced signaling, or it could be the relative abundance of different Smads in the responding cells.
Crystallographic studies have demonstrated that Smad3 binds DNA via an 11-amino acid ␤ hairpin in the Mad homology domain-1 domain that contacts the AGAC motif (38). Smad2 cannot bind DNA despite its very high sequence similarity to Smad3. This is probably due to a sequence insert in Smad2, immediately upstream of the DNA binding ␤-hairpin, which could potentially hinder Smad2-DNA interaction (38,43). Interestingly, our footprinting experiments demonstrate that while Smad3 and Smad4, and less efficiently Smad1, bind to the same core sequence (CAGAC(C/A)), the way they impose themselves on the DNA helix differs. They protect to different extent the two strands (Fig. 3, A and B) and more importantly, the different Smads induce the appearance of different DNase I-hypersensitive sites on the DNA template indicating that they influence to different extents the access of neighboring DNA sequences to DNase I. If different Smads hinder to different extent the accessibility of neighboring transcription factor-binding sites, then the outcome of TGF-␤ signaling will depend on which Smads will eventually bind to a given site. The cooperative function of multiple Smad-binding sites and the differential influence that various Smads exert on neighboring binding sites upon binding, could allow the Smad signaling system to transfer a whole spectrum of quantitatively and qualitatively different instructions on a target promoter/ enhancer and be important in the context of the function of TGF-␤ family members as morphogens during development (44).
Recent findings including those of the present study have demonstrated that recombinant Smad proteins can bind directly to certain DNA sequences that contain the AGAC core motif. Truncation of the MH2 domain in Droshophila Mad protein increases the capacity of the Mad homology domain-1-Linker portion to bind DNA (45). To make sure that the binding activity in our preparations was not due to degradation products mimicking the MH2 truncation we have carefully purified full-length proteins. Removal of smaller size contaminants increased the specific activity of the preparation instead of reducing it, suggesting that the full-length receptor unmodified Smads can bind DNA (data not shown).
Smads Cooperate with AML Proteins to Stimulate the I␣1 Promoter-Our studies extend the list of transcription factors that integrate functionally with the Smad proteins by adding the AML family of transcriptional regulators. Smad3 and AML1b physically interact in vitro and in vivo. Their interaction is constitutive and at least in the two cell lines used is not affected by receptor signaling, and appears to be of lower affinity than the interaction of T␤R-I-phosphorylated Smad3 with Smad4. However, it is only after stimulation either by TGF-␤1 or the constitutively active type I receptors that higher order complexes containing AML1b, Smad3, and Smad4 are detected in nuclear extracts by EMSA (Fig. 6). It appears thus that despite the fact that these molecules form complexes constitutively, the effect of their interplay on I␣1 transcription can potentially be modulated at other levels, such as nuclear availability.
The AML family includes the AML1, AML2, and AML3 genes. All of them give rise to a collection of polypeptides by alternative splicing of exons and possibly also alternative use of start codons for translation. Shi and Stavnezer (46) have recently presented evidence supporting the role of AML proteins for the regulation of the mouse I␣ promoter. Since AML2 was the only member of the family that was up-regulated by TGF-␤1 stimulation, they concluded that AML2 is the critical regulator of I␣ transcription. Considering that AML1b is the most efficient I␣ transactivator among the AML isoforms (12,14), the quite rapid induction of germ line transcription (47)(48)(49), the fact that AML2 up-regulation occurs at very late stages of TGF-␤ signaling (46) and also the fact that in the DG75 and K562 cells we have analyzed in the present report TGF-␤1 signaling does not induce AML2 up-regulation (Fig. 10), it is conceivable to either envision a dynamic process, during which different isoforms of AML sequentially participate in the AML/ Smad interplay or to consider that the apparent differences are cell line specific. Furthermore, it is possible to expect that different AML variants will be more important in responses to different activating stimuli and in the context of different promoters. Furthermore, considering the quite significant divergence of AML proteins at their C termini, it is possible that different domains of the AML and Smad molecules will differentially contribute in physical interaction of various AML/ Smad combinations.
Is TGF-␤1 the Exclusive Regulator of I␣ Promoter Activity?-Our finding that constitutively activated ALK-4, ALK-5, and ALK-7 can stimulate the I␣1 promoter (Fig. 2) immediately raises the possibility that I␣ transcription can be achieved not only by TGF-␤1 but also by other members of the TGF-␤ superfamily that can interact with the above receptors. Indeed, TGF-␤2 and activin A can also stimulate the I␣1 reporter construct in DG75 cells (data not shown). TGF-␤1 and TGF-␤2 were more potent than activin A, however, this could be due to the high levels of TGF-␤ type III receptors and the low levels of activin receptors these cells express (data not shown). We conclude that other members of the TGF-␤/activin family are potential regulators of IgA synthesis.
Consequences of the AML/Smad Interplay for Other Biological Systems-AML transcription factors bind to DNA sites in the regulatory regions of a number of hematopoiesis specific genes and play key roles in normal and pathological blood cell processes. Inactivation of either the AML1 gene or its associated factor PEBP2␤ results in death early in fetal development (day 12-14) and a complete block in fetal liver hematopoiesis (15,50). Furthermore, the AML1/PEBP2␤ genes are the most FIG. 10. Western blot analysis of AML family members. DG75 cells were transfected with AML1b-Smad3-Smad4 expression vectors or the control pBJ9⍀ plasmid. Transfected cells were cultured in the presence or absence of TGF-␤1 and cell lysates were prepared at the indicated time points. The anti-AML antibody was raised against a conserved amino acid stretch in the runt homology domain and thus detects all the AML proteins and their splice variants. Using commercially available monoclonal antibodies the lower molecular weight major species was recognized as AML1 and the two upper bands as AML2 (data not shown). A number of minor bands were recognized by antibodies against AML1, AML2, and AML3 and probably represent mixtures of different splice variants with similar electrophoretic mobility. frequent targets of translocation in acute human leukemia (51). Smad proteins, as nuclear TGF-␤ effectors, regulate a large range of biological processes. Furthermore, Smads play important roles during processes leading to malignant transformation. Smad4 was originally identified as a putative tumor suppressor gene, frequently mutated in pancreatic carcinomas and occasionally in other types of tumors (18,21).
In vivo and in vitro studies have suggested that among the extracellular signals that can induce the osteoblastic phenotype are the BMPs (52). Recent studies have demonstrated that a splice form of CBFA1/AML3/PEBP2␣A also plays a critical role in osteoblast differentiation (53)(54)(55). One possible mechanism by which BMPs could mediate regulation of bone development is via the induction of CBFA1 synthesis. AML1b interacts physically not only with the TGF-␤/activin Smads but also with the BMP-activated Smad1 (Fig. 9). It is possible that the AML/Smad interplay involves many or all the members of the two families, and that depending on the cell type and stimuli, different combinations of AML and Smad proteins could be recruited to regulate transcription of selected genes.
Considering the wide range of biological phenomena that AML and Smad proteins regulate and the potential oncogenic character of AML1 and tumor suppressor activity of some Smads, the observed functional interplay between these transcription factors has implications not only for class switching to IgA but also for other biological processes such as regulation of bone ossification as well as normal and pathological hematopoietic cell behavior.