Specific Activation of Smad1 Signaling Pathways by the BMP7 Type I Receptor, ALK2*

BMP7 and activin are members of the transforming growth factor β superfamily. Here we characterize endogenous activin and BMP7 signaling pathways in P19 embryonic carcinoma cells. We show that BMP7 and activin bind to the same type II receptors, ActRII and IIB, but recruit distinct type I receptors into heteromeric receptor complexes. The major BMP7 type I receptor observed was ALK2, while activin bound exclusively to ALK4 (ActRIB). BMP7 and activin elicited distinct biological responses and activated different Smad pathways. BMP7 stimulated phosphorylation of endogenous Smad1 and 5, formation of complexes with Smad4 and induced the promoter for the homeobox gene,Tlx2. In contrast, activin induced phosphorylation of Smad2, association with Smad4, and induction of the activin response element from the Xenopus Mix.2gene. Biochemical analysis revealed that constitutively active ALK2 associated with and phosphorylated Smad1 on the COOH-terminal SSXS motif, and also regulated Smad5 and Smad8 phosphorylation. Activated ALK2 also induced the Tlx2promoter in the absence of BMP7. Furthermore, we show that ALK1 (TSRI), an orphan receptor that is closely related to ALK2 also mediates Smad1 signaling. Thus, ALK1 and ALK2 induce Smad1-dependent pathways and ALK2 functions to mediate BMP7 but not activin signaling.

Recent studies have advanced significantly our understanding of how TGF␤ 1 superfamily members mediate their biological effects. The discovery of TGF␤ receptors and Smad proteins along with recent insights into the mechanism of their activation have allowed us to trace a TGF␤ signal transduction pathway from the cell membrane to the nucleus (reviewed in Refs. 1 and 2). TGF␤ family members initiate signaling at the cell surface by binding and bringing together two different but related serine/threonine kinase receptors, type I and type II.
First, the ligand binds to the type II receptor, which recruits and transphosphorylates the type I receptor on the GS domain, a region in the juxtamembrane domain that is rich in serine and glycine residues (3). A mutation in the GS domain can lead to constitutive activation of the receptor (4) and such an activated receptor mimics the effects of the entire receptor-ligand complex in the absence of growth factor and the type II receptor. Thus, the type I receptor is considered the primary transducer of signals to downstream components of the pathway. So far, seven type I receptors or activin receptor-like kinases (ALK1-7) have been identified in vertebrates (reviewed in Refs. 2 and 5).
Once the type I receptor is activated, it associates with specific receptor-regulated Smad proteins and phosphorylates them on the last two serine residues on the carboxyl-terminal domain (6 -9). Smad proteins are essential components of TGF␤ signaling that link ligand/receptor signals to transcriptional control (10 -16). All members possess two highly conserved MAD homology domains in the amino (MH1) and carboxyl (MH2) terminus that are connected by a proline-rich nonconserved region (reviewed in Refs. 2, 17, and 18). Once receptor-regulated Smad proteins are phosphorylated they can associate with a common partner, Smad4, and translocate to the nucleus. Once there, they may associate with specific DNAbinding proteins in order to generate transcriptional complexes, which subsequently will activate specific target genes. For instance, TGF␤ and activin regulate the Xenopus genes Mix.2 (13), XFKH1 (19), and goosecoid (20), whereas the Xenopus genes Xom (21), Xvent-2 (22), and Msx-1 (23), and the mouse homeobox gene Tlx2 (24) are regulated by BMP signals.
Biochemical and biological studies have established at least two distinct TGF␤ family pathways: one shared by TGF␤ and activin and the other by the BMP2 and 4 subfamily. Characterization of these pathways has shown that a given ligand can induce the formation of heteromeric complexes between different receptors and that a given receptor can recognize different ligands and substrates, suggesting a certain redundancy or cooperativity in signaling. Thus, T␤RI (ALK5) in concert with T␤RII can transduce the TGF␤ responses (25,26), whereas ActRIB (ALK4) acting in union with ActRII or ActRIIB transduce the same signals but in response to activin (26,27). Consistent with these findings, T␤RI and ActRIB, which display 90% identity between their kinase domains, can signal through the same downstream components, Smad2 and 3, to generate the same set of responses (6,28,29). Thus, TGF␤ and activin illustrate a case of convergent signaling. A similar scenario is displayed by the BMP subfamily. In this case, BMP2/4 bind BMPRII in concert with ALK3 or ALK6 (30,31) and activate Smad1 and 5 signaling pathways (7,32,33). The existence of a plausible third pathway shared by activin and BMP7 has been suggested, since both ligands can bind ActRII and IIB and recruit the type I receptor, ALK2. Hence, ALK2 could represent a candidate receptor that mediates common responses for these factors.
ALK2, also termed ActRI, SKR1, or Tsk-7L, was initially identified as an activin type I receptor because of its ability to bind activin in concert with ActRII or IIB. However, it was not demonstrated that such ligand-receptor complexes were active and could transduce signals under physiological conditions (34 -36). Later, it was reported that a specific antisera to ALK2 receptor could immunoprecipitate both a BMP7, as well as an activin-receptor complex. This suggested that ALK2 might indeed mediate common responses for activin and BMP7 (37,38) and depicted another example of receptor sharing and convergent signaling. However, the recent studies in early Xenopus embryos using a constitutively active form of the ALK2 receptor has shown that ALK2 generates signals similar to BMPs and that it cannot mimic activin signals in the early embryo (23,39,40).
In order to characterize ALK2 function we investigated the signaling pathway regulated by this receptor. In the present study, we show that BMP7 and activin stimulate different pathways in the BMP and activin responsive cell line, P19. Although, both factors bind ActRII and IIB, BMP7 specifically binds endogenous ALK2 whereas activin binds ALK4. Stimulation of P19 cells with BMP7 leads to phosphorylation of endogenous Smad1 and Smad5 and association with endogenous Smad4 as well as to induction of the promoter of the homeobox gene, Tlx2. In the same cells, activin signals through endogenous Smad2 and Smad4 leading to induction of a specific activin target gene. Thus, the ALK2 receptor appears to signal through Smad1 and 5 since an activated form of ALK2 mimics BMP7 but not activin signals in P19 cells. Moreover, activated ALK2 also induces the phosphorylation of Smad1, 5, and 8, and the association of Smad1 with Smad4 in transfected COS-1 cells. Therefore, we provide evidence that ALK2 functions as a BMP type I receptor and is likely to transduce BMP7 but not activin signals. In addition, we show that the orphan type I receptor ALK1 (TSRI) which is most closely related to the ALK2 receptor (34) also activates Smad1-dependent signaling.

EXPERIMENTAL PROCEDURES
Construction of Expression Vectors-The construction of pCMV5B/ Flag-Smad1, Flag/Smad2, and Flag/Smad2(3SA) has been described previously (6,32,41). The mutant Flag/Smad1(⌬458), where the last 7 residues at COOH-terminal tail were deleted, was generated by a polymerase chain reaction-based strategy and was fully sequenced. The construction of His/Smad1(MH2) domain, for bacterial expression, was performed by subcloning the MH2 domain of Smad1 (amino acids 242-465) into the pRSETB expression vector encoding an amino-terminal hexahistidine tag (Invitrogen). The Smad3, Smad5, and Smad8 constructs were provided by R. Derynck (University of California, San Francisco, CA), J. M. Yingling (Duke University, Durham, NC), and W. Vale (The Salk Institute, San Diego, CA), respectively. The ALK2 and ALK1 receptors were described previously (34) and the constructs ALK2/HA (Q207D) and ALK1/HA (Q201D) were generated by a polymerase chain reaction-based strategy and were fully sequenced. The construction of pTLX2-lux and pA3-lux reporter genes has been described previously (24,42), respectively.
Preparation of Polyclonal Antisera-Bacterial expression constructs for Smad1, 2, and 4 were constructed by subcloning the non-conserved region from Smad1, 2, or 4 into pGEX-4T1. The glutathione S-transferase fusion proteins generated were used as antigens for rabbit polyclonal antisera production. Specific type I receptor antisera obtained as described previously (43) were kindly provided by Dr. Kohei Miyazono (The Cancer Institute, Tokyo, Japan) and Dr. P. ten Dijke (Ludwig Institute for Cancer Research, Uppsala, Sweden) and ActRII and IIB specific antisera were a generous gift of Dr. W. Vale (The Salk Institute, San Diego, CA).
Cell Lines and Transfections-P19 cells were cultured in ␣-minimal essential medium containing 7.5% calf serum and 2.5% fetal bovine serum, and were transiently transfected using the calcium phosphate-DNA precipitation method. MC3T3-E1 derived by Sudo et al. (44) were maintained in ␣-minimal essential medium ϩ 10% fetal bovine serum and were obtained from Dr. J. Aubin (University of Toronto, ON). COS-1 cells were grown in Dulbecco's modified Eagle's medium containing high glucose and 10% fetal bovine serum. COS-1 cells were transiently transfected with the indicated vectors using a DEAE-dextran method as described previously (32).
Immunoprecipitations and Immunoblotting-For endogenous Smads, endogenous type I and type II receptors and anti-HA immunoprecipitations, cell were lysed in lysis buffer (45) subjected to immunoprecipitation with specific Smad antisera, specific type I and II receptors antisera or anti-HA antibody, respectively, followed by adsorption to protein A-Sepharose (Pharmacia). For endogenous Smad immunoprecipitations, cell lysates were precleared with protein A-Sepharose beads and preimmune sera for 30 min. For ␣-Flag or ␣-Myc immunoprecipitations cell lysates were subjected to immunoprecipitation with anti-Flag M2 antibody (IBI, Eastman Kodak) or with 9E10 mouse anti-Myc antibody, respectively, followed by adsorption to protein G-Sepharose (Pharmacia) as described previously (32). Immunoprecipitates were separated by SDS-polyacrylamide gel electrophoresis (SDS-PAGE). For determination of Flag, HA, or Myc-tagged Smad protein levels and endogenous Smad protein levels, aliquots of total cell lysates were separated by SDS-PAGE and assayed by immunoblotting as described previously (32).
Transcriptional Response Assay-For BMP or activin-inducible luciferase assays, P19 cells were transiently transfected with the reporter plasmid pTlx2-lux or pA3-lux, respectively, and pCMV␤-gal. Cells were seeded at 20% confluency in 24-well plates and transfected overnight with 0.5-1 g of DNA per well using the calcium phosphate-DNA precipitation method as described previously (32). To induce the luciferase reporter, cells were treated overnight with the appropriate ligands, lysed, and luciferase activity was measured using the luciferase assay system (Promega) in a Berthold Lumat LB 9501 luminometer. ␤-Galactosidase activity was measured as described previously (8) and data were used to determine transfection efficiency.
Phosphate and Metabolic Labeling-For [ 32 P]phosphate labeling, cells were washed and preincubated with phosphate-free medium containing 0.2% dialyzed fetal bovine serum. The cells were then incubated with media containing 1 mCi/ml [ 32 P]PO 4 for 2 h at 37°C as described previously (45). For 35 S-metabolic labeling, cells were washed and then preincubated with methionine-free medium containing 50 Ci/ml [ 35 S]methionine for 2 h at 37°C as described previously (45). Cells were then lysed and subjected to immunoprecipitation with specific antibodies. Labeled proteins were analyzed by SDS-PAGE and autoradiography.
Affinity Labeling-P19 cells or transiently transfected COS-1 cells were affinity labeled using [ 125 I]BMP7, [ 125 I]BMP2, or [ 125 I]activin in media containing 0.2% fetal calf serum at 37°C for 30 min and the receptors were cross-linked to the ligand as described previously (46). Cells were lysed in lysis buffer containing 10% glycerol as described previously (45). In order to study specific ligand/receptor binding, lysates were immunoprecipitated using specific type I and II receptors polyclonal antisera. For determining Smad-receptor complex interactions, lysates were immunoprecipitated using anti-Flag M2 monoclonal antibody as described previously (6). Complexes were analyzed by SDS-PAGE and autoradiography.
In Vitro Kinase Assay-Receptor complexes were immunoprecipitated from transiently transfected COS-1 cell lysates as described above except that immunoprecipitates were washed three times with TNTE buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 0.1% Triton X-100, and 1 mM EDTA) followed by two washes in kinase buffer. Receptor complexes were incubated for 30 min at room temperature in kinase buffer containing His/MH2 domain of Smad1 protein which was expressed in bacteria. Briefly, the His/MH2 domain of Smad1 was expressed in BL21(DE3) cells and purified from isopropyl-1-thio-␤-D-galactopyranoside-induced bacterial cultures. The protein was collected on Ni-NTA agarose beads, washed with TNT buffer (50 mM Tris, pH 7.4, 100 mM NaCl, 0.5% Triton) containing 40 mM imidazole, and eluted with 250 mM imidazole. The eluted fusion protein was dialyzed overnight, concentrated, and 2 g used as a substrate in the kinase assay as described previously (6). Protein phosphorylation was analyzed by SDS-PAGE and autoradiography.

Characterization of BMP and Activin Receptors on P19
Cells-To investigate BMP7 signaling we used P19 cells which respond to both BMPs and activin. To characterize the profile of BMP7 receptors, P19 cells were affinity labeled using [ 125 I]BMP7 and receptor complexes were immunoprecipitated using specific antibodies directed against various type I and type II receptors. BMP7 receptor complexes were detected on P19 cells and these were found to be composed of the previously characterized type II receptors, ActRII and ActRIIB, together with the type I receptor, ALK2. No cross-linking of BMP7 to ALK4 was detected, consistent with previous conclusions that ALK4 is an activin-specific type I receptor (26,27). Since previous studies indicated that ALK2 may bind activin under some conditions (34), we also explored the profile of activin receptors expressed on these cells. Affinity labeling with 125 Iactivin, followed by immunoprecipitation with antibodies to ALK4 or ALK2 revealed the presence of complexes of either ActRII or ActRIIB together with ALK4. However, no interaction of activin with ALK2 was detected, suggesting that in P19 cells ALK2 functions predominantly as a BMP7 receptor (Fig. 1A).
We also examined the profile of BMP2 receptors on these cells. BMP2 bound efficiently to both ALK3 and ALK6, as described previously for these receptors (37,47), and formed heteromeric complexes with both ActRII and ActRIIB. Although earlier studies suggest that ActRII and IIB receptors do not bind BMP2 (48), subsequent work with the Drosophila activin-like type II receptor, Punt, showed that it functions as a BMP2 receptor when coexpressed with the appropriate type I receptor (49). Thus, our studies are consistent with these findings and demonstrate that like PUNT, ActRII and IIB can bind BMP2 and function as type II receptors in the presence of the appropriate type I receptors (Fig. 1B).

BMP7 Induces a BMP2-like Response-To investigate BMP7
responses in P19 cells, we examined regulation of Tlx2, a homeobox gene of the HOX11 class. Previously, we showed that Tlx2 functions during gastrulation in the mouse and is strongly induced by BMP2 (24). Analysis of the Tlx2 gene showed that it contained a BMP-responsive promoter that was induced by BMP2 in P19 cells (24). To characterize BMP7 signaling, pTlx2-Lux was transiently transfected into P19 cells and cultures were incubated in the presence or absence of 10 nM BMP7, 10 nM BMP2, or 2 nM activin prior to analysis of reporter gene activation. Under these conditions, BMP7 induced the Tlx2 promoter 7-fold, similar to the response observed for BMP2 ( Fig. 2, TLX2-Lux). In contrast, activin had no detectable effect on the Tlx2 promoter, similar to previous observations (24).
We also sought to determine whether BMP7 could induce activin-dependent signaling pathways. For this, we used the activin response element (ARE) that was initially characterized in Xenopus and mediates activin-specific induction of the Mix.2 gene. In P19 cells activin treatment of cells transiently transfected with the ARE reporter, pA3-lux, resulted in a 6-fold induction of the reporter gene ( Fig. 2, A3-Lux). In contrast, neither BMP7 nor BMP2 induced this promoter. Together, these results suggest that BMP7 specifically induces a BMPlike signaling pathway in P19 cells.
BMP7 Leads to Specific Phosphorylation of Smad1 and Smad5-MAD-related proteins are critical intracellular mediators of TGF␤ superfamily signaling (reviewed in Refs. 2, 17, 18, 50, and 51). To investigate the intracellular pathways activated by BMP7 in P19 cells, we examined the regulation of endogenous Smads in these cells. For these studies, we generated Smad antisera that are specific for different subclasses of Smads. To characterize these antisera, tagged Smad proteins were expressed in COS-1 cells and total cell lysates were immunoblotted using specific Smad antisera. In Fig. 3, we show that anti-Smad1 recognizes Smad1 and the closely related Smad5, anti-Smad2 recognizes Smad2 and Smad3, and anti-Smad4 is specific for Smad4 protein. Although these antisera displayed some cross-reactivity between closely related Smads, both Smad1 and Smad5, and Smad2 and Smad3 exhibited distinct mobilities when analyzed by SDS-PAGE. This allowed us to distinguish these Smads during our subsequent analysis.
To study the ability of BMPs to regulate endogenous Smads, P19 cells were labeled with [ 32 P]phosphate and treated with 10 nM BMP2, BMP7, or activin for 30 min prior to lysis and immunoprecipitation with Smad1/5, Smad2/3, or Smad4-spe- cific antisera. In control cells, Smad1 and 5, Smad2 and 3, and Smad4 displayed low levels of basal phosphorylation (Fig. 4). However, brief treatment of P19 cells with activin led to an induction of Smad2 and Smad3 phosphorylation with no detectable effect on phosphorylation of Smad1 or 5 (Fig. 4A). Conversely, when cells were treated with BMP7, phosphorylation of Smad1 and 5 was induced to the same degree as was observed for BMP2 with no effect on Smad2 or 3 phosphorylation (Fig. 4B). To further characterize the BMP7 response, we analyzed the time course of Smad1 and 5 phosphorylation. Both BMP7 and BMP2 induced Smad1 and 5 phosphorylation within 15 min of ligand addition, with the highest levels being reached by 30 -60 min (Fig. 4C). Maximal levels of Smad1 and 5 phosphorylation were observed for at least 2 h during which time we observed no changes in the levels of Smad1 or Smad5 protein as determined by Western blotting of whole cell lysates (data not shown). In contrast to the regulation of receptorregulated Smad phosphorylation, we were unable to detect any effect of BMPs or activin on phosphorylation of endogenous Smad4 protein (Fig. 4D).
Since phosphorylation of receptor-regulated Smads results in heteromeric complex formation with the common mediator, Smad4, we investigated whether BMP7 induced association between endogenous Smad1 and Smad4. For these studies, P19 cells were treated with BMP7 for 1 h and then lysates were subjected to immunoprecipitation with the Smad1/5 or Smad2/ 3-specific antisera. The immunoprecipitates were then resolved by SDS-PAGE and immunoblotted with antisera specific to Smad4. In resting P19 cells, no detectable complexes of Smad1 and Smad4 were detected. However, treatment with BMP7 or BMP2 for 1 h resulted in the formation of Smad1-Smad4 heteromeric complexes (Fig. 5, top panel). In contrast to these results, analysis of Smad2-Smad4 complexes revealed some constitutive association in P19 cells. This interaction was significantly enhanced by the addition of activin, but was unchanged in cells stimulated with BMP7 (Fig. 5, bottom panel). Interestingly, in Smad4 immunoprecipitates obtained from [ 32 P]phosphate-labeled cells (Fig. 4D), we observed a phosphorylated protein that coprecipitated with Smad4 and comigrated with Smad2, and thus likely represents endogenous Smad2 associated with Smad4. Since, P19 cells express low levels of activin (52), our observations of constitutive Smad2-Smad4 complexes may thus reflect a low basal level of autocrine signaling by activin.
P19 cells express ALK3 and ALK6 receptors. ALK6 receptors have been shown to bind BMP7 in some cells (37), while binding of BMP7 to ALK3 is poor and has only been observed in one cell type (37). In addition to BMP7, ALK6 also binds other ligands such as BMP2, BMP4, and GDF5 and forms heteromeric complexes with the type II receptors, BMPRII, ActRII, and ActRIIB. Therefore, it is plausible that ALK6 might mediate some BMP2 and BMP7 signaling in P19 cells. Thus, to confirm that BMP7 binding to ALK2 can induce activation of Smad1, we also investigated BMP7 regulation of Smad1 and 2 in the mouse calvarial cell line, MC3T3-E1. These cells are particularly suitable because the only major BMP7 type I receptor expressed on this cell line is ALK2 (37). Analysis of ALK2, 3, and 6 expression by reverse transcriptase-polymerase chain reaction confirmed that our MC3T3-E1 clone expressed only ALK2 and ALK3 but not ALK6 (data not shown). Similar to our observations in P19 cells, BMP7 strongly induced phosphorylation of Smad1 in MC3T3-E1 cells, but had no detectable effect on Smad2 (Fig. 6A). Importantly, activin was unable to mediate induction of Smad1 phosphorylation in these cells (Fig.  6A). Together, our results suggest that BMP7 can induce specific activation of Smad1-dependent pathways through the type I receptor, ALK2. Furthermore, since activin was unable to mediate Smad1 or Smad5 activation in either P19 or MC3T3-E1 cells, our data suggest that ALK2 is not a functional receptor for activin.
ALK2 Activates Smad1 and Induces BMP-specific Signals-To directly assess the regulation and activation of Smads by ALK2 we substituted the glutamine residue at position 207 with an aspartate residue. Previous studies have shown that this type of mutation constitutively activates the kinase domain of type I Ser/Thr kinase receptors (4). Expression of constitutively active ALK2 together with wild type Smad1 resulted in a strong induction of Smad1 phosphorylation (Fig.  6B), and association of Smad1 with Smad4 (data not shown). This regulation was specific, since Smad2 phosphorylation was unaffected by the constitutively active ALK2. Since phosphorylation of receptor-regulated Smads occurs on the COOHterminal SSXS motif (6 -9), we sought to determine whether ALK2 might similarly regulate Smad1. In contrast to wild type Smad1, activated ALK2 was unable to induce phosphorylation of Smad1(⌬458). Since this mutation removes the COOH-terminal serines, these results suggest that the SSXS motif of Smad1 is the site for ALK2-dependent phosphorylation of the protein. We also tested whether activated ALK2 could induce phosphorylation of Smad5 and Smad8. Consistent with our observations in P19 cells and previous reports, both Smad5 and Smad8 were efficiently phosphorylated by the activated receptor (Fig. 6B, lower panels).
To confirm that ALK2 activates Smad1-dependent signaling pathways, we examined the regulation of pTLX2-Lux and pA3-Lux by constitutively active ALK2 in P19 cells. As was observed for BMP7 ligand, expression of activated ALK2 in P19 cells induced the Tlx2 promoter to a level comparable to that observed for the BMP2 type I receptor, ALK6. Furthermore, activated ALK2 was unable to induce the ARE reporter gene. These results were contrasted by those obtained with the activin type I receptor ALK4, which mediated induction of the ARE but not Tlx2. Thus, the type I Ser/Thr kinase receptor, ALK2 specifically mediates activation of Smad1-dependent signaling, but not Smad2 (Fig. 6C).
Smad1 Is a Substrate for ALK2-To determine whether Smad1 directly interacts with BMP7 receptor complexes, we expressed Smad1 or Smad1(⌬458) together with ActRIIB and either wild type or kinase-deficient ALK2. BMP7 receptor complexes were then affinity labeled with [ 125 I]BMP7 and complexes coprecipitating with Smad1 analyzed by SDS-PAGE and autoradiography. Similar to our previous observations on the TGF␤ receptor (6), no association between wild type Smad1 and wild type BMP7 receptors could be detected. However, analysis of complexes composed of kinase-deficient type I receptor showed coprecipitation of BMP7-labeled receptors with Smad1. Furthermore, introduction of a phosphorylation site mutation into Smad1 allowed efficient coprecipitation of wild type BMP7 receptors with Smad1. These data are consistent with the notion that Smad1 transiently associates with activated ALK2 and that phosphorylation leads to dissociation of the Smad1 substrate from the receptor. In none of these experiments were we able to detect associations between BMP7 receptors and either wild type Smad2 or the phosphorylation site mutant, Smad2(3SA) (Fig. 7A).
To confirm that Smad1 is a substrate of the receptor complex we isolated ALK2 complexed with different type II receptors and subjected the heteromeric complexes to an in vitro kinase assay using bacterially expressed Smad1 MH2 domain as the substrate (6,7). As shown in Fig. 7B, none of the type II receptors alone were able to significantly phosphorylate Smad1(MH2). However, Smad1 was efficiently phosphorylated by the BMP2 receptor complex, ActRIIB/ALK6, and the BMP7 receptor complex BMPRII/ALK2. Interestingly, we detected only minimal phosphorylation of Smad1 by ALK2 containing complexes that were comprised of the type II receptors ActRII or ActRIIB. It is unclear what the basis for this difference is, however, the kinase activity of ALK2 may be relatively unstable in vitro and may be best stabilized through association with BMPRII. Consistent with this possibility, we detected no autophosphorylation of ALK2 in any of the ActRII and IIB immunoprecipitates or when ALK2 was directly immunoprecipitated from cell lysates. A previous study similarly failed to observe any in vitro kinase activity for purified ALK2 (7). It is currently unclear whether there is any biological relevance to the differential effect of type II receptors on the in vitro kinase activity of ALK2. Together our results show that ALK2 is a functional type I receptor for BMP7 but not activin and further demonstrate that this type I receptor mediates BMP-specific signaling by directly associating with and phosphorylating Smad1.
The Orphan ALK1 Receptor Also Activates Smad1-dependent Signaling-Comparison of the kinase domains of type I receptors indicates that ALK2 is more closely related to ALK1 (80% identity) than it is to ALK3 and ALK6. ALK1 (TSRI) was initially identified as a TGF␤ receptor due its ability to bind TGF␤ when overexpressed with T␤RII (34). However, no TGF␤ mediated signals have been characterized and its endogenous ligand has not been identified. Since ALK1 is most closely related to ALK2, and given that ALK2 activates Smad1 signaling, we examined whether Smad1 is also a target for ALK1. To investigate this, we constructed a constitutively active version of ALK1 by replacing the glutamine at position 201 with an aspartate. The capacity of wild type and activated forms of ALK1 to activate Smad1 was then assessed in transiently transfected COS-1 cells. Expression of constitutively active ALK1 together with wild type Smad1 or Smad2 resulted in a strong induction of Smad1 phosphorylation, but not Smad2, similar to that observed for ALK2 (Fig. 8A), whereas T␤RI specifically induced Smad2 phosphorylation (Fig. 8A). We also found that like ALK2, ALK1 also induced phosphorylation of Smad5 (data not shown). To examine whether ALK1 might also activate Smad1-dependent signaling pathways, we examined the regulation of pTlx2-Lux and pA3-Lux by the constitutively active ALK1 in P19 cells. Expression of activated ALK1 in P19 cells induced the Tlx2 promoter to a level similar to that observed for ALK2 (Fig. 8B), whereas activated ALK1 was unable to induce the activin-responsive reporter gene, pA3-lux (data not shown). Thus, the orphan receptor ALK1 like ALK2, also appears to function to regulate BMP signaling. DISCUSSION The characterization of BMP and activin signaling has shown that these factors induce formation of specific heteromeric combinations of type I and type II Ser/Thr kinase receptors. For example, BMPs bind the type II receptor BMPRII in concert with the type I receptors, ALK3 or ALK6, and activin promotes the association of the type II receptors, ActRII or IIB, with the type I receptor, ALK4 (ActRIB). In contrast to these specific type I receptors, ALK2 has been observed to bind activin in concert with ActRII and ActRIIB when overexpressed in COS cells (34), as well as mediating binding of BMP7 in cell lines that express endogenous ALK2 (37). Therefore, it has been proposed that ALK2 is a shared receptor for activin and BMP7 and that it might be responsible for mediating common responses to these factors (38). Thus, the goal of the present study was to characterize the signaling pathways activated by ALK2 and to investigate the potential function of this receptor in mediating responses to BMP7 and activin. Using P19 cells, we show that BMP7 and activin bind to the common type II receptors, ActRII and IIB, but recruit different type I receptors into the ligand-receptor complex and activate distinct Smad signaling pathways. Thus, BMP7 signaling through the ALK2 receptor can activate Smad1 and 5 and induce the BMP-responsive homeobox gene, Tlx2, whereas activin, through ALK4 activates Smad2 to induce the ARE promoter. Furthermore, we show that ALK2 associates with Smad1, mediates phosphorylation of Smad1 on its COOH-terminal SSXS motif and that activated forms of the receptor can mimic BMP-dependent signaling in P19 cells. ALK2 also induces phosphorylation of Smad5 and Smad8 and we show that the orphan ALK1 receptor, which is most closely related to ALK2, also activates Smad1. Thus, ALK2 can mediate BMP7dependent signaling and does not appear to encode a functional receptor for activin.
BMP and Activin Receptors-P19 cells are an embryonic carcinoma cell line that are ideal for investigating the function of activin and BMP signaling pathways. These cells are BMP and activin-responsive and express endogenous BMP type I receptors, such as ALK3 and ALK6 (53) as well as ALK2 (36). These cells also express activin receptors that have not previously been characterized (52). In the present study, we show that BMP2, in addition to BMP7 and activin, bind endogenous ActRII and IIB in P19 cells. In addition to these type II receptors, P19 cells also express BMPRII (53) that interacts with BMP2 and 7 but not activin (31,54). Thus it is likely that several type II receptors can function in BMP signaling (Fig.  9A). It is interesting to note that previous studies in Drosophila which demonstrated that the type II receptor PUNT, which is highly related to ActRII and binds activin with high affinity, only binds BMP2 when the appropriate type I receptor is coexpressed (49). In COS-1 cells, we have observed a similar behavior for BMP2 binding to the mammalian activin type II receptors, 2 and previous studies on GDF5 have shown that this ligand binds ActRII or IIB in union with ALK6 (55). All of this suggests that ActRII and IIB have broad specificity and function as type II Ser/Thr kinase receptors for numerous TGF␤ family members. This is supported by functional studies in Xenopus that show that a truncated ActRII or IIB receptor blocks both activin and BMP signaling in vivo (38, 56 -58) whereas such cross-reactivity does not occur for a truncated BMPRII, which only binds BMPs (59). It is interesting that the 2 J. L. Wrana, unpublished data. broad specificity of activin type II receptors is conserved in Drosophila, raising the possibility that overlapping specificity may be of functional relevance during development, perhaps by allowing different ligands to regulate each other by competing for common receptor subunits.
Although BMPs and activin may share the same type II receptors, each ligand-type II receptor complex recruits a distinct type I receptor (summarized in Fig. 9A). Our analysis in P19 cells indicates that the major activin type I receptor is ALK4 (ActRIB), consistent with previous reports for this receptor (27) and that BMP2 binds specifically to ALK3 and ALK6. Moreover, BMP7 was the only ligand of these three that significantly bound to endogenous ALK2 receptors. These latter findings are consistent with previous studies that showed that ALK2 was the predominant BMP7 receptor in a variety of cell lines (37). Our ability to simultaneously assay both BMP and activin signaling in the same cells also allowed these binding studies to be extended functionally. Thus activin induced the ARE from the Xenopus Mix.2 gene, while none of the BMPs we assayed were able to regulate this promoter. The ARE, which is bound by the forkhead domain proteins, FAST1 in Xenopus or FAST2 in mouse (13,60), is specifically targetted by complexes of Smad2/Smad4 (60,61). Our results are thus consistent with these observations, since we also showed that activin induced Smad2 activation in P19 cells. In contrast, BMP2 and BMP7 strongly induced the Tlx2 promoter, while activin had no effect. Consequently, although all of these receptor complexes have ActRII and IIB as a common subunit, the specificity observed in the biological response to different ligands supports the idea that the identity of the type I receptor is what defines the signaling capacity of the receptor complex ( Ref. 1 and Fig. 9A).
ALK2 Mediates BMP7 but Not Activin Signaling-Considerable work in Xenopus has defined distinct biological responses to BMPs and activin/TGF␤. For example, both factors are implicated in the induction of mesoderm, but BMPs act as mesoderm ventralizers, whereas activin induces dorsal mesoderm (62,63). In addition, other studies have shown that the expression of dominant negative versions of BMP7 and BMP2/4 induce neuroectoderm in animal caps, whereas dominant negative versions of activin do not mimic this phenotype (64). Similarly, both BMP2 and 7 have inhibitory effects on neurogenesis in Xenopus (65). In early mammalian development this distinction between BMP and activin signaling pathways has not been defined. P19 cells have been used as a model for studying mesodermal and neural differentiation in the mouse. BMP2/4 and activin induce mesoderm markers (66) and inhibit retinoic acid-induced neural differentiation in these cells (67,68). However, while BMPs induce apoptosis in combination with retinoic acid (67,69), activin functions as a mitogen of undifferentiated P19 cells (70). However, the mechanisms underlying these complex biological responses are poorly defined.
In the present study we have investigated BMP7 and show that there is no overlap between BMP7 and activin signaling pathways. This specificity is manifested by the activation of distinct receptor complexes that recognize different downstream Smad substrates. Thus BMP7 activates Smad1, 5, and likely Smad8 and induces BMP-like responses, while activin signaling through ActRIB activates Smad2 and 3 pathways. Notably, we never observed activation of Smad1 by activin. This was highlighted in our analysis of MC3T3-E1 cells, which express ALK2 and respond well to BMP7, but showed no measurable response to activin. Thus, despite previous observations that ALK2 can bind activin under conditions of overexpression, we could find no evidence that endogenous ALK2 receptors mediate activin signaling. We conclude that ALK2 is a BMP7 receptor and is not a functional activin receptor.
Convergent Signaling to Smads 1, 5, and 8 by BMPs-The specific regulation of Smad1 signaling by ALK2 is analogous to the previously described role for Smad1 as a downstream target of BMP2 signaling through ALK3 and ALK6 (1,7). Indeed, in our current studies, the time courses and magnitudes of Smad1 activation by BMP2 or 7 were indistinguishable. These results suggest that there is considerable convergence in BMP signaling. Thus, multiple BMPs functioning through a limited subset of cell surface receptors may all activate a common downstream signaling pathway to elicit shared biological responses (Fig. 9A). Apart from Smad1, two other receptor-regulated Smads, Smad5 and 8, are likely to participate in this pathway. Smad5 was previously characterized in the BMP pathway (71), while Smad8 was recently isolated and shown to be regulated by constitutively activated ALK2 (39). Thus, ALK1, ALK2, ALK3, and ALK6 all likely activate this common set of Smads and appear to possess considerable overlap in function. Interestingly, BMP7 can also bind ALK6 in some cells and in one cell line has been observed to bind ALK3. This raises the possibility that in addition to utilizing ALK2, BMP7 can activate Smad1, 5, or 8 through other type I receptors. This convergence could suggest a remarkable degree of overlap in ligand function during development. While a certain amount of redundancy is likely to occur, the restricted expression patterns of ligands during development may ensure that many of them fulfill precise roles at specific places and times to control patterning and tissue remodeling. Moreover, cell type specific responses to BMPs may be controlled by the regulated expression of distinct Smad1 nuclear targets. Thus while Smad1, 5, and 8 activation may be a common pathway for a large number of BMP ligands, the complex nature of the biological response to BMPs may be dictated by an array of nuclear targets.
The apparent convergence of BMP signaling pathways onto a common set of receptor-regulated Smads is consistent with studies in Xenopus that have assessed the biological response to enforced expression of constitutively activated versions of the ALKs. Most notably, Armes and Smith (40) investigating the role of ALK2 in Xenopus, observed that constitutively active versions of this receptor mimicked the effects of BMP, but not activin signaling. These observations are consistent with our demonstration that BMP2 and BMP7, signaling through distinct type I receptors, activate a shared downstream signaling pathway to elicit common biological responses.
In the mouse embryo, we recently showed that BMP2 induces the homeobox gene Tlx2 during gastrulation and it was suggested that this gene might also be regulated by other members of the BMP family such as BMP4 (24). In this report, we show that BMP7 and ALK2, in addition to BMP2 and its receptors ALK3 and ALK6, are able to induce Tlx2. Whether BMP7 also regulates Tlx2 during gastrulation remains to be evaluated. BMP7 is expressed in the early embryo at the time of Tlx2 expression (72,73), as is ALK2 (53), so a role for these factors in controlling Tlx2 expression is plausible. It will be interesting to test this possibility through the future genetic analysis of BMP7 and ALK2-deficient mice.
Receptor Determinants of Smad Targetting-Since the identity of the type I receptor in the ligand-receptor complex determines the nature of the signal, the study of the structural requirements in the kinase domain that determine downstream signaling specificity has received considerable attention. Recently, a structural analysis of type I receptor kinase domains has led to the identification of a 9-amino acid region, termed the L45 loop, which seems to be critical in specifying the signal. Thus, replacing the L45 loop in ALK2 with that from T␤RI, conferred on the chimeric type I receptor the capacity to signal TGF␤-like responses (74). This loop is predicted to be located on an exposed region of the kinase domain between subdomains IV and V and thus could represent a Smad-docking site.
Since ALK1, ALK2, ALK3, and ALK6 all induce activation of Smad1, we were interested in identifying any similarities in their L45 loops that might define a Smad1-specific site. Interestingly, alignment of the L45 loops from all the known type I receptors from mammals and Drosophila led to the identification of three major groups, with one Drosophila receptor represented in each group (Fig. 9B). Thus, ALK4, ALK5, ALK7, and Drosophila ATR-I possessed the same L45 loop, while similar groupings were identified for ALK3, ALK6, and Drosophila TKV; and ALK1, ALK2, and Drosophila SAX (Fig. 9B). However, comparison of L45 sequences between these three groups revealed a low level of identity and we could not define any particular sequence motif that might specify Smad1 interaction with the receptor. Indeed, to our surprise the L45 loop from ALK3/ALK6 was more similar to T␤RI-like receptors than it was to the ALK1/ALK2 loop. Thus, it is unclear how the ALK1 and ALK2 kinase domains share the Smad1 and Smad5 substrates with ALK3 and ALK6, and whether the L45 loop plays a direct role. Nevertheless, given that the L45 sequence is highly conserved within the groups and that each contains a homologous receptor from Drosophila, it is intriguing to speculate that a conserved function for these receptor subclasses may be defined by the L45 loop.