Alternatively Spliced Variant of Smad2 Lacking Exon 3

An alternatively spliced variant of Smad2 with a deletion of exon 3 (Smad2Δexon3) is found in various cell types. Here, we studied the function of Smad2Δexon3 and compared it with those of wild-type Smad2 containing exon 3 (Smad2(wt)) and Smad3. When transcriptional activity was measured using the p3TP-lux construct, Smad2Δexon3 was more potent than Smad2(wt), and had activity similar to Smad3. Transcriptional activation of the activin-responsive element (ARE) of Mix.2 gene promoter by Smad2Δexon3 was also similar to that by Smad3, and slightly less potent than that by Smad2(wt). Phosphorylation by the activated transforming growth factor-β type I receptor and heteromer formation with Smad4 occurred to similar extents in Smad2Δexon3, Smad2(wt), and Smad3. However, DNA binding to the activating protein-1 sites of p3TP-lux was observed in Smad2Δexon3 as well as in Smad3, but not in Smad2(wt). In contrast, Smad2(wt), Smad2Δexon3, and Smad3 efficiently formed ARE-binding complexes with Smad4 and FAST1, although Smad2(wt) did not directly bind to ARE. These results suggest that exon 3 of Smad2 interferes with the direct DNA binding of Smad2, and modifies the function of Smad2 in transcription of certain target genes.

Members of the transforming growth factor-␤ (TGF-␤) 1 superfamily transduce signals through two different types of serine/threonine kinase receptors, known as type II and type I receptors (1). In the TGF-␤ receptor system, ligand binds to the TGF-␤ type II receptor (T␤R-II), which has a constitutively active kinase. TGF-␤ type I receptor (T␤R-I) is then recruited into the TGF-␤⅐T␤R-II complex, and phosphorylated mainly at the glycine/serine-rich domain (GS domain), which results in the activation of T␤R-I kinase (2). The T␤R-I kinase transduces intracellular signals by activation of various proteins, including Smad proteins. T␤R-I thus acts as a downstream component of T␤R-II. Mutation in Thr-204 of T␤R-I to aspartic acid (T␤R-I(TD)) results in the constitutive activation of the T␤R-I kinase, which has signaling activity in the absence of TGF-␤ and T␤R-II (3).
Smad proteins have recently been shown to comprise a family of proteins that mediate signals for members of the TGF-␤ superfamily (4 -6). Thus far, eight mammalian Smad proteins have been identified, termed Smad1 through Smad8. Smads are classified into three subgroups based on their structure and function, i.e. pathway-restricted Smads, common mediator Smads, and inhibitory Smads. Pathway-restricted Smads can be further subdivided into those involved in the TGF-␤ and activin signaling pathways, and those activated by the bone morphogenetic protein pathway. Smad2 and Smad3 serve as pathway-restricted Smads for the TGF-␤/activin signaling pathways. Smad1, Smad5, and possibly Smad8/MADH6 are activated by bone morphogenetic protein receptors. Smad4 is a common mediator Smad; thus far only one common mediator Smad has been identified in mammals. Smad6 and Smad7 are inhibitory Smads.
Smads have conserved N-and C-terminal regions known as Mothers against decapentaplegic (Mad) homology domain-1 (MH1) and -2 (MH2), respectively, which are linked by a linker region of variable length and amino acid sequence. The MH2 domain is a functional domain, which has transactivation activity when fused to the Gal4-DNA binding domain (18,19). The MH2 domain also plays important roles in interaction with type I receptors (20), homo-and hetero-oligomerization by Smad proteins (21,22), interaction with a transcription factor, FAST1 (23,24), and association with transcriptional co-activators, p300/CBP (25)(26)(27)(28). The MH1 domain has been shown to physically interact with the MH2 domain, and thereby inhibits the activity of the latter (29). However, it has been shown that the MH1 domain has an intrinsic function in signal transduction, i.e. direct binding to specific DNA sequences. Mad in Drosophila was shown to bind to the quadrant enhancer of vestigial gene (13). Smad3 and Smad4 have also been shown to bind to specific DNA sequences through their MH1 domains (13)(14)(15)(16)(17). * This work was supported in part by Grants-in-aid for Scientific Research from the Ministry of Education, Science, Sports and Culture of Japan, and Special Coordination Funds for Promoting Science and Technology from the Science and Technology Agency of Japan. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Although Smad2 and Smad3 are 91% identical in amino acid sequence, they have certain differences in biological activity. In contrast to Smad3 and Smad4, Smad2 does not directly bind to DNA (12)(13)(14)(15)(16)(17). Binding to a transcriptional regulator, Evi-1, is observed for Smad3 but not for Smad2 (30). A functional difference between Smad2 and Smad3 has also been suggested in the effects of TGF-␤ and activin on the HaCaT keratinocyte cells (31). TGF-␤ and activin inhibit the growth of HaCaT cells, but TGF-␤ is much more potent than activin. TGF-␤ induces the phosphorylation of both Smad2 and Smad3, whereas activin A preferentially activates Smad3 (31).
Smad2 has a region with 30 amino acid residues, which is not found in Smad3 or other Smads in mammals. The Smad2 gene is composed of 11 exons, and this short 30-amino acid region is translated by exon 3. Recently, we found that a Smad2 transcript which lacks exon 3 is present in certain tissues and cells, although the amount of the transcript is about 1/10 of that containing exon 3 (32). Here we studied the function of Smad2 without exon 3 (Smad2⌬exon3) and compared it with those of wild-type Smad2 (Smad2(wt)) and Smad3. In the transcriptional activation assay using the p3TP-lux construct, Smad2⌬exon3 was more potent than Smad2(wt), and had activity almost similar to Smad3. Phosphorylation by T␤R-I and heteromer formation with Smad4 did not differ between Smad2⌬exon3 and Smad2(wt). However, Smad2⌬exon3, but not Smad2(wt), was able to bind to the activating protein (AP)-1 sites from p3TP-lux. In contrast, both Smad2(wt) and Smad2⌬exon3 could form the activin-responsive factor (ARF) with FAST1 and Smad4, and transactivate the Xenopus Mix.2 gene promoter.
Reverse Transcriptase-PCR of Smad2-Poly(A) ϩ RNAs were prepared from various cell lines and reverse-transcribed. The resulting cDNAs were amplified by PCR using primers 2A (5Ј-TTT TCC TAG CGT GGC TTG-3Ј) and 4A (5Ј-TCA GAG AGT TGA GAC ACC AG-3Ј) under conditions as described (32). The PCR products were subjected to the second-round PCR using primers 2B (5Ј-GAA GAG ACT GCT GGG ATG GAA GAA GT-3Ј) and 4B (5Ј-CAA GGC AAT TGA AAA CTG CGA ATA TGC-3Ј). The PCR fragment was run on a 2% agarose gel, and a 330-bp product was cloned using a TA cloning kit (Promega).
Cell Culture and cDNA Transfection-COS7 cells were obtained from American Type Culture Collection. R mutant Mv1Lu cells were provided by J. Massagué. The cells were cultured in Dulbecco's modified Eagle's medium with 10% fetal bovine serum, 100 units/ml penicillin, and 10 g/ml gentamycin. For transient transfection, 60 -80% confluent cells in six-well plates or 10-cm cell culture dishes were transfected using FuGENE6 transfection reagent (Boehringer Mannheim) following the manufacturer's protocol.
Luciferase Assay-R mutant Mv1Lu cells were transiently transfected with p3TP-lux or pAR3-lux (provided by J.L. Wrana) (36) in the presence of various combinations of Smad constructs, T␤R-I(TD), and FAST1. For normalization of transfection efficiency, the Renilla luciferase reporter gene in the pRL-CMV vector (Promega) was co-transfected in each transfection. After transfection, cells were incubated for 36 h, and luciferase activity in the cell lysates was determined with a dual luciferase assay system (Promega) using a luminometer (Lumat LB 9501, EG & G Berthold) according to the manufacturer's recommendations. The luciferase activities of p3TP-lux and pAR3-lux constructs were measured as luminescence of firefly luciferase. As an internal control, Renilla luciferase was measured immediately afterward by quenching the firefly luminescence.
Immunoprecipitation and Immunoblotting-Immunoprecipitation and immunoblotting have been described previously (33). Briefly, COS7 cells were transfected with expression constructs for Smads, T␤R-I(TD), and FAST1. Forty-eight hours after transfection, the cells were solubilized, and the cell lysates were incubated with the anti-Flag M2 (Eastman Kodak Co.) or anti-Myc 9E10 antibodies (Santa Cruz Biotechnology), followed by incubation with protein G-Sepharose beads. The immunocomplexes were then eluted by boiling for 3 min in the SDS sample buffer containing 10 mM dithiothreitol and subjected to SDSpolyacrylamide gel electrophoresis (PAGE). Aliquots of the cell lysates were directly subjected to SDS-PAGE. Proteins were electrotransferred to polyvinylidene difluoride membrane (ProBlott membranes, Applied Biosystems) and immunoblotted with the anti-Flag M2, anti-Myc 9E10, anti-HA 3F10 (Boehringer Mannheim), or anti-phosphoserine antibodies (Zymed Laboratories Inc.) and developed using an enhanced chemiluminescence detection system (Amersham Pharmacia Biotech). For re-blotting, the polyvinylidene difluoride membranes were stripped following the manufacturer's protocol.
Gel-mobility Shift Assay-Gel-mobility shift assay was performed as described previously (14). Briefly, whole cell extracts were prepared from the COS7 cells transfected with Smad, T␤R-I(TD), and FAST1 expression constructs. Cells were lysed in a lysis buffer containing 50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 0.5% Nonidet P-40, 50 mM NaF, 1 mM NaPO 4 , 1 mM dithiothreitol, and protease inhibitors. Glutathione S-transferase (GST) fusion proteins were prepared as described (16). For supershift analysis, anti-Flag and/or anti-Myc antibodies (1-2 l each) or antisera against Smad2 or Smad3 (2 l each, provided by P. ten Dijke) (33) were added to the whole cell lysates or GST-fused Smad proteins. The probe containing AP-1 sites (77 bp) was created by digestion of the p3TP-lux with NdeI and SphI, and that containing ARE was prepared by digestion of the pAR3-lux with Acc I and BamHI. The probes were then subjected to [␣-32 P]dCTP Klenow labeling. Whole cell lysates (3 l, containing 3 g of protein) or GST fusion proteins (150 ng each) were added with a premix solution (13.4 l) containing 1 g of poly(dI-dC) and 1 l of the probe labeled to an activity of 2.0 ϫ 10 4 cpm/l (37). The final concentration of NaCl in the samples was adjusted to 110 mM by hypotonic and hypertonic lysis buffers. Complexes were then resolved on a 4% polyacrylamide gel and analyzed by autoradiography.

Construction of Smad2⌬exon3
Plasmid-Smad2 has a 30amino acid region in the middle of the MH1 domain (amino acid 79 -108), which is not found in other Smads in mammals (Fig.  1A). mRNAs were prepared from various human cell lines, i.e. the HaCaT keratinocyte cell line, HEL human erythroleukemia cell line, U937 human monocytic leukemia cell line, and Molt-4 human T cell leukemia cell line. PCR was performed using primers (2A and 4A) corresponding to the sequences in exons 2 and 4, and the PCR products were subjected to second-round PCR using primers 2B and 4B. Similar to previous results obtained with mRNAs from placenta and HaCaT keratinocytes (32), we observed two bands in HaCaT and Molt-4 cells, i.e. major bands of 420 bp and faint bands of 330 bp (Fig. 1B). The 330-bp bands were only very weakly detected in the other two cell lines. By DNA sequencing, we confirmed that the 330-bp products correspond to Smad2 lacking exon 3. By inserting the 330-bp product into Smad2(wt) at PvuII-RcaI sites, we prepared a construct for Smad2⌬exon3.
The Smad constructs were transfected into COS7 cells, and the expression of proteins was analyzed by immunoblotting. N-terminally Flag-tagged Smad2⌬exon3 was efficiently expressed in COS7 cells, with protein levels similar to those with Smad2(wt) and Smad3 (data not shown). The size of Smad2⌬exon3 is smaller than that of Smad2(wt), and larger than that of Smad3, consistent with their estimated molecular weights.
Smad2⌬exon3 Has Transcriptional Activity Different from That of Smad2(wt)-Next, we studied the transcriptional activity of Smad2⌬exon3 and compared it with those of Smad2(wt) and Smad3 using the p3TP-lux promoter reporter construct, which contains the promoter region of plasminogen activator inhibitor (PAI)-1 and three tandemly linked AP-1 sites (38) (Fig. 2A). The R mutant Mv1Lu cells, which lack functional T␤R-I, were used to determine the transcriptional activity of Smads. In the presence of small amounts of T␤R-I(TD) plasmid, a slight increase in the transcriptional activity on p3TP-lux was observed. Smad3 efficiently induced transcription even in the absence of exogenous Smad4, while Smad2(wt) was less efficient in transcriptional activation than Smad3. Co-expression of Smad4 led to higher transcriptional activity of Smad2(wt), although it was less than that of Smad3. Interestingly, Smad2⌬exon3 had transcriptional activity similar to that of Smad3 in the absence or presence of Smad4, which was more potent than that of Smad2(wt).
The transcriptional activity of Smad2⌬exon3 was also tested in the reporter gene pAR3-lux containing the ARE of Mix.2 promoter in Xenopus (36) in the presence or absence of the specific DNA-binding protein, Xenopus FAST1 (10). In contrast to the results obtained using p3TP-lux, Smad2 was slightly more potent than Smad3 in inducing luciferase activity of pAR3-lux (Fig. 2B). The luciferase induction by Smad2⌬exon3 was similar to that by Smad3, and slightly less effective than that by Smad2.
Phosphorylation by T␤R-I and Complex Formation with Smad4 -In order to understand mechanism of the higher transcriptional activity in p3TP-lux of Smad2⌬exon3 than of Smad2(wt), we examined the phosphorylation of Smad2 and Smad3 by T␤R-I. Flag-tagged Smad constructs were transfected into COS7 cells together with various amounts of T␤R-I(TD), and immunoprecipitated with anti-Flag antibody, followed by immunoblotting using anti-phosphoserine antibodies (Fig. 3A). Although Smad2⌬exon3 and Smad3 were more potent than Smad2(wt) in inducing transcriptional activity on p3TP-lux, there was no significant difference between the levels of phosphorylation in Smad2⌬exon3, Smad2(wt), and Smad3.
Heteromer formation with Smad4 was also studied using Smad2⌬exon3, Smad2(wt), and Smad3 (Fig. 3B). Weak interaction with Smad4 was observed in Smad2⌬exon3 and Smad2(wt) even without T␤R-I(TD), and was correlated with the weak phosphorylation of the Smad2 proteins. Strong heteromer formation was induced in all Smad2 and Smad3 constructs by T␤R-I(TD), and we found no differences in between Smad2⌬exon3, Smad2(wt), and Smad3.
Smad2⌬exon3 but Not Smad2(wt) Binds to the AP-1 Sites of p3TP-lux-We then studied the formation of DNA-binding complexes by Smad2⌬exon3 at the AP-1 sites of p3TP-lux. It was previously shown that Smad3 and Smad4, but not Smad2, bind to probe prepared from p3TP-lux DNA (13). Whole cell extracts were prepared from COS7 cells transfected with the indicated Smad constructs and T␤R-I(TD), and subjected to gel-mobility shift analysis (Fig. 4A). None of the Smad proteins formed DNA-binding complexes in the absence of T␤R-I(TD) (data not shown). In the presence of T␤R-I(TD), Smad3 and Smad2⌬exon3, but not Smad2(wt), formed DNA-binding complexes. The complexes supershifted in the presence of the anti-Flag antibody. Smad4 alone did not form a DNA-binding complex in the presence of T␤R-I(TD), but it did participate in DNA-binding complexes in the presence of Smad2⌬exon3 or Smad3, but not of Smad2(wt). The bands shifted in the presence of anti-Myc antibody, although they were weak compared with those obtained with the anti-Flag antibody. Moreover, addition of both anti-Myc and anti-Flag antibodies led to shift with a slower mobility, indicating that the DNA-binding complexes contained Smad4. These results indicate that by deletion of exon 3, Smad2 acquired the ability to participate in DNA-binding complexes.
We next examined whether Smad2⌬exon3⌬MH2 can directly bind to the AP-1 sites of p3TP-lux. Smad3 lacking the MH2 domain was previously shown to directly bind to DNA (12, 14 -17). GST-fused Smad2(wt)⌬MH2, Smad3⌬MH2, and Smad2⌬exon3⌬MH2 lacking MH2 domains were used for gel shift analysis (Fig. 4B). Similar to the results obtained with whole cell extracts transfected with Smads and T␤R-I(TD), both Smad2⌬exon3 and Smad3 were able to bind the AP-1 sites, but Smad2(wt) did not. The bands shifted in the presence of corresponding Smad antisera, although the shift bands ( lane  15 and 18, Fig. 4B) were weak compared with that obtained with the anti-Flag antibody (lane 19), probably because of the lower affinities of the Smad antisera.
Smad2(wt), Smad3, and Smad2⌬exon3 Participate in ARF-Smad2 associates with Smad4 and FAST1 to form ARF, which binds to ARE of the Mix.2 gene promoter (23,24). Smad2(wt) Cell lysates were immunoprecipitated with the anti-Flag antibody, followed by immunoblotting using anti-phosphoserine (anti-P-serine) antibody. Aliquots of the cell lysates were directly subjected to SDS-PAGE and immunoblotted with the anti-Flag or anti-HA antibody to detect the expression of Smad proteins and T␤R-I(TD), respectively. B, Complex formation of Smad2, Smad2⌬exon3, and Smad3 with Smad4. COS7 cells were transfected with Smad2/3 and Smad4 constructs with or without T␤R-I(TD). Cell lysates were immunoprecipitated with the anti-Myc antibody followed by immunoblotting with the anti-Flag antibody. In order to detect the phosphorylation of Smad2/3, aliquots of the cell lysates were subjected to immunoprecipitation using the anti-Flag antibody followed by immunoblotting using the anti-phosphoserine antibodies, as described in A. Expression of Flag-Smad2/3 was detected after stripping the membrane and immunoblotting using the anti-Flag antibody. Expression of T␤R-I(TD) was detected by immunoblotting using the anti-HA antibody.
FIG. 4. Smad2⌬exon3 and Smad3 bind to the AP-1 sites derived from p3TP-lux. A, Flag-Smad2(wt), Flag-Smad2⌬exon3, or Flag-Smad3 and Myc-Smad4 were transfected into COS7 cells in the presence of T␤R-I(TD)-HA. Whole cell lysates were prepared, and binding to the 32 P-labeled probe containing the AP-1 sites was analyzed by gelmobility shift assay. For supershift analysis, whole cell lysates were incubated with the anti-Flag (F) and/or anti-Myc (M) antibodies, and subjected to gel-shift assay. B, binding of purified GST-Smad proteins to DNA. GST-fused Smad2(wt)⌬MH2, Flag-Smad2⌬exon3⌬MH2, and Smad3⌬MH2 lacking MH2 domains were prepared, and gel shifts were performed using the probe containing the AP-1 sites. For supershift analysis, the antiserum to Smad2 (2) or Smad3 (3), or anti-Flag (F) antibody was used. For competition (competitor; C) of the DNA binding, excess amounts of cold probe was added.
was slightly more potent in inducing transcriptional activity on pAR3-lux than Smad3 or Smad2⌬exon3; however, complex formation with Xenopus FAST1 was observed similarly in between Smad2(wt), Smad2⌬exon3, and Smad3 in the presence and absence of T␤R-I(TD) (Fig. 5A). We, therefore, examined whether there are any differences in the DNA-binding abilities of the ARF complexes containing different Smads. ARF containing Smad2(wt), Smad4, and FAST1 efficiently bound ARE in response to TGF-␤ receptor activation, similar to those containing Smad3 or Smad2⌬exon3 (Fig. 5B). Addition of anti-Flag, anti-Myc, or anti-HA antibodies led to shifts of the bands, indicating that the DNA-binding ARF complexes contained Smad2, Smad3, or Smad2⌬exon3, together with Smad4 and FAST1. FAST1 was shown to interact with Smad2 through the MH2 domain (23,24). These findings indicate that the presence of exon 3 of Smad2 does not interfere with the formation of DNA-binding ARF complex containing FAST1.
We further investigated the ability of different GST-fused Smad proteins to directly bind to DNA containing the ARE sequence in the absence of FAST1. Similar to binding to the AP-1 sites (Fig. 4A), Smad3⌬MH2 and Smad2⌬exon3⌬MH2 recognized the ARE sequence, but Smad2(wt)⌬MH2 did not (Fig. 5C). The bands shifted in the presence of Smad antisera (lanes 6 and 8). Shift of the bands was more remarkable with the anti-Flag antibody (lane 9), probably because of the higher affinity of the anti-Flag antibody and possible induction of oligomerization. These data indicate that in addition to indirectly binding to DNA as ARF complexes, Smad3 and Smad2⌬exon3, but not Smad2(wt), have abilities to directly bind to ARE . DISCUSSION Smad proteins have MH1 and MH2 domains, which are linked by a linker region. Inhibitory Smads have MH2 domains, but have divergent MH1-like regions. The MH2 domain plays important roles in various functions of Smads, i.e. interaction with type I receptors, homo-and hetero-oligomer formation, association with DNA-binding proteins, and interaction with transcriptional coactivators, e.g. p300/CBP (4 -6, 25-28). Smad2, which lacks the MH1 domain, is constitutively located in the nucleus and activates target genes (39). The MH2 domain interacts with type I receptors, and the L3 loop, a 17amino acid region in the MH2 domain, plays a critical role in this interaction (20). In addition, ␣-helix 2 of the MH2 domain has recently been shown to determine the binding specificity to DNA-binding proteins such as FAST1 (40).
The MH1 domain plays an important role as a repressor of the function of the MH2 domain. In addition, the MH1 domain has intrinsic activity, i.e. binding to specific DNA sequences. 5. Smad2(wt), Smad2⌬exon3, as well as Smad3 participate in the ARF complexes. A, complex formation of Smad2(wt), Smad2⌬exon3, and Smad3 with Xenopus FAST1. COS7 cells were transfected with Smad constructs with or without T␤R-I(TD) and Xenopus HA-FAST1. Cell lysates were immunoprecipitated with the anti-Flag antibody followed by immunoblotting with the anti-HA antibody. Expression of Flag-Smad2/3 was detected after stripping the membrane and immunoblotting using the anti-Flag antibody. For the detection of the phosphorylated Smad2/3, the membrane was subjected to immunoblotting using the anti-phosphoserine antibodies. Expression of T␤R-I(TD) and FAST1 was detected by immunoblotting using the anti-HA antibody. B, Flag-Smad2(wt), Flag-Smad2⌬exon3, or Flag-Smad3 and Myc-Smad4 were transfected into COS7 cells in the presence of T␤R-I(TD) and HA-tagged Xenopus FAST1. Whole cell lysates were prepared, and binding to the 32 P-labeled probe containing ARE from the Mix. Drosophila Mad, an ortholog of mammalian Smad1 and 5, directly binds to the quadrant enhancer of vestigial gene (13). Smad3 and Smad4, but not Smad2, have been shown to bind to the AP-1 sites of p3TP-lux (14,41). More recently, Smad3 and Smad4 have been shown to bind to specific DNA sequences, and luciferase reporter constructs containing multiple copies of these specific DNA sequences have been shown to be activated by Smad3 and Smad4 (15)(16)(17).
Smad2 and Smad3 are structurally very similar, and serve as pathway-restricted Smads in the TGF-␤ and activin signaling pathways. However, functional differences between Smad2 and Smad3 have been suggested. In addition to the difference between them in DNA-binding ability, Smad3, but not Smad2, binds to a transcriptional regulator Evi-1, which consequently suppresses the activity of Smad3 (30). In the HaCaT keratinocyte cell line, TGF-␤ induces phosphorylation of both Smad2 and Smad3, while activin A preferentially activates Smad3 (31). Since TGF-␤ and activin have different activities in the growth inhibition and differentiation of this cell type, 2 the difference in phosphorylation between Smad2 and Smad3 may be, at least in part, responsible for their distinct biological effects.
Smad2 differs from Smad3 in having a 30-amino acid region in the middle of the MH1 domain. We have shown in the previous report (32) and in the present study that mRNA lacking exon 3 is present in various tissues and cells. Amounts of the Smad2 transcript lacking exon 3 appear to differ between cell types. It would be of great interest to determine whether the mRNA encoding Smad2⌬exon3 is preferentially induced under certain physiological and pathological conditions.
In the present study, we have shown that Smad2 lacking exon 3 has higher transcriptional activity on p3TP-lux than does Smad2(wt). The MH1 domain represses the function of the MH2 domain. Mutation of Arg-133 in Smad2 resulted in increase of the affinity between MH1 and MH2 domains, and this mutant was less active than the wild-type Smad2 (29). Similar results were reported for Smad4 with mutation at Arg-100 (29). In the present study, however, deletion of exon 3 of Smad2 did not lead to increase in phosphorylation by T␤R-I, or heteromer formation with Smad4, suggesting that exon 3 is not directly involved in the repressor activity of the MH1 domain.
Smad2⌬exon3 is able to bind to DNA containing the AP-1 sites, a finding not observed for Smad2(wt). These findings strongly suggest that the direct binding of Smad2/3 to DNA is a crucial step in the transcriptional activation of the p3TP-lux reporter. In whole cell extracts, DNA-binding complex formation was observed in both the presence and absence of transfected Smad4. Direct DNA binding of Smad3 and Smad2⌬-exon3 was detected when GST-Smad proteins lacking MH2 domains were used. These findings together with those of previous reports (12)(13)(14)(15)(16)(17) suggest that the interaction may occur through the MH1 domain, when repression by the MH2 domain is released. Since Smad2⌬exon3 has higher transcriptional activity than Smad2(wt), direct DNA binding of Smads may positively regulate their transcriptional activity on p3TP-lux promoter. However, since Smad2(wt) can, to some extents, induce transcriptional activation of p3TP-lux, Smad proteins can partially activate target genes through a mechanism which does not require the direct DNA binding of Smads.
In contrast to the possible role of DNA binding of Smad2/3 in the transcriptional activation of p3TP, transcription of the Mix.2 gene appeared to be less strongly affected by direct DNA binding of Smad2/3. The Mix.2 promoter contains a sequence similar to the Smad-binding element (15,17), to which Smad3, and possibly Smad2⌬exon3, may bind. Smad2(wt), which failed to directly bind the Mix.2 promoter, was efficient in inducing transcriptional activation of pAR3-lux. Since the binding of Smad2 to FAST1 occurs through the MH2 domain, differences in the structure of the MH1 domain may have less important effects on the transcription of the Mix.2 gene. Interestingly, Smad2 activated the goosecoid promoter together with Smad4 and mouse FAST2, a winged-helix transcription factor related to Xenopus FAST1, whereas Smad3 strongly suppressed transactivation of this promoter (12). Smad3 and Smad4, but not Smad2, directly bound to the GC-rich sequences of the goosecoid promoter. These findings suggest that direct DNA binding of Smads thorough the MH1 domain as well as indirect binding via other DNA-binding factors, e.g. FAST1 and FAST2, may cooperatively regulate the transcription of target genes. Direct DNA binding of Smads may play important roles in transcription of certain target genes, such as PAI-1, while it is less critical or acts negatively for other genes, i.e. Mix.2 and goosecoid.
Expression profile of Smad3 appears to be limited compared with that of Smad2 (32). In certain cell types that lack the expression of Smad3, Smad2⌬exon3 may function as a Smad3like molecule. The region like exon 3 is observed only in Smad2 but not in other Smads in mammals. Exon 3 of Smad2 may play a crucial role in modulating the function of Smad2 by interfering with the direct DNA binding to target genes.