Extracellular and Cytoplasmic Domains of Endoglin Interact with the Transforming Growth Factor-β Receptors I and II*

Endoglin is an auxiliary component of the transforming growth factor-β (TGF-β) receptor system, able to associate with the signaling receptor types I (TβRI) and II (TβRII) in the presence of ligand and to modulate the cellular responses to TGF-β1. Endoglin cannot bind ligand on its own but requires the presence of the signaling receptors, supporting a critical role for the interaction between endoglin and TβRI or TβRII. This study shows that full-length endoglin interacts with both TβRI and TβRII, independently of their kinase activation state or the presence of exogenous TGF-β1. Truncated constructs encoding either the extracellular or the cytoplasmic domains of endoglin demonstrated that the association with the signaling receptors occurs through both extracellular and cytoplasmic domains. However, a more specific mapping revealed that the endoglin/TβRI interaction was different from that of endoglin/TβRII. TβRII interacts with the amino acid region 437–558 of the extracellular domain of endoglin, whereas TβRI interacts not only with the region 437–558 but also with the protein region located between amino acid 437 and the N terminus. Both TβRI and TβRII interact with the cytoplasmic domain of endoglin, but TβRI only interacts when the kinase domain is inactive, whereas TβRII remains associated in its active and inactive forms. Upon association, TβRI and TβRII phosphorylate the endoglin cytoplasmic domain, and then TβRI, but not TβRII, kinase dissociates from the complex. Conversely, endoglin expression results in an altered phosphorylation state of TβRII, TβRI, and downstream Smad proteins as well as a modulation of TGF-β signaling, as measured by the reporter gene expression. These results suggest that by interacting through its extracellular and cytoplasmic domains with the signaling receptors, endoglin might affect TGF-β responses.

Endoglin is an auxiliary component of the transforming growth factor-␤ (TGF-␤) receptor system, able to associate with the signaling receptor types I (T␤RI) and II (T␤RII) in the presence of ligand and to modulate the cellular responses to TGF-␤1. Endoglin cannot bind ligand on its own but requires the presence of the signaling receptors, supporting a critical role for the interaction between endoglin and T␤RI or T␤RII. This study shows that full-length endoglin interacts with both T␤RI and T␤RII, independently of their kinase activation state or the presence of exogenous TGF-␤1. Truncated constructs encoding either the extracellular or the cytoplasmic domains of endoglin demonstrated that the association with the signaling receptors occurs through both extracellular and cytoplasmic domains. However, a more specific mapping revealed that the endoglin/T␤RI interaction was different from that of endoglin/T␤RII. T␤RII interacts with the amino acid region 437-558 of the extracellular domain of endoglin, whereas T␤RI interacts not only with the region 437-558 but also with the protein region located between amino acid 437 and the N terminus. Both T␤RI and T␤RII interact with the cytoplasmic domain of endoglin, but T␤RI only interacts when the kinase domain is inactive, whereas T␤RII remains associated in its active and inactive forms. Upon association, T␤RI and T␤RII phosphorylate the endoglin cytoplasmic domain, and then T␤RI, but not T␤RII, kinase dissociates from the complex. Conversely, endoglin expression results in an altered phosphorylation state of T␤RII, T␤RI, and downstream Smad proteins as well as a modulation of TGF-␤ signaling, as measured by the reporter gene expression. These results suggest that by interacting through its extracellular and cytoplasmic domains with the signaling receptors, endoglin might affect TGF-␤ responses.
Members of the transforming growth factor-␤ (TGF-␤) 1 su-perfamily (1, 2) exert their biological effects through binding to a heteromeric complex containing two different transmembrane serine/threonine kinases known as type I and type II signaling receptors (3,4). Upon ligand binding to the type II receptor, the association between type I and type II receptors is induced, leading to phosphorylation and activation of the type I receptor by the constitutively active type II receptor (4,5). Then, activated type I receptor propagates intracellular signal to the nucleus by phosphorylating members of the Smad family of proteins (4,6,7).
The TGF-␤ receptor complex also contains two auxiliary co-receptors named endoglin and betaglycan (8 -10). These are transmembrane proteins with large extracellular domains and serine/threonine-rich cytoplasmic regions without consensus signaling motifs. Endoglin binds TGF-␤1, TGF-␤3, activin-A, BMP-2, and BMP-7 in the presence of the signaling receptor types I and II (8,11,12) and modulates TGF-␤1-dependent cellular responses (11,13,14). It is highly expressed on endothelial cells and several lines of evidence support an important role for endoglin in cardiovascular development and vascular remodeling. Thus, endoglin expression is regulated during heart development in humans and chicken (15,16), and genetic inactivation of endoglin in the mouse shows that embryos homozygous for mutant endoglin die at 10 -10.5 days postcoitum due to vascular and cardiac anomalies (17)(18)(19). Furthermore, genes encoding endoglin and ALK-1 (a type I TGF-␤ receptor) are targets for the autosomal dominant disorder known as hereditary hemorrhagic telangiectasia (20 -22). Despite the data supporting an important role for endoglin in the TGF-␤ system, the molecular basis of endoglin function is still poorly understood.
Several lines of evidence suggest that endoglin is not a receptor per se: (a) in endothelial cells, only a small percentage of endoglin molecules is capable of binding TGF-␤; (b) endoglin does not bind ligand on its own but requires the presence of the signaling receptors (11,12), (c) even in the presence of the signaling receptors, ligand association with endoglin can only be visualized in the presence of a cross-linking agent (12); and (d) cross-linking of endoglin transfectants with radiolabeled TGF-␤ showed the existence of ligand-free endoglin associated with TGF-␤-loaded signaling receptors (11). All of these data suggest that endoglin association with the signaling receptors is critical for endoglin access to ligand and for its modulation of the ligand-induced cellular responses. Here, we have investigated the domains involved in the association between endoglin and the signaling receptors as well as the consequences of this interaction on the phosphorylation status of the proteins involved.

EXPERIMENTAL PROCEDURES
Cell Culture and Transfections-The monkey kidney COS-7 and the rat myoblast L 6 E 9 cell lines were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum in 5% CO 2 at 37°C in a humidified atmosphere. COS-7 cells were transiently transfected with expression vectors encoding wild type and mutant constructs of endoglin, T␤RI, or T␤RII using LipofectAMINE Plus as indicated by the manufacturer (Invitrogen). Functional assays were carried out 48 h after transfection. Treatment of cells with recombinant human TGF-␤1 (Peprotech) was performed at a concentration of 10 ng/ml for 30 min.
Expression Vectors-Wild type human endoglin cDNA subcloned into the pCMV5 vector (11) was used to derive the truncated endoglin constructs by PCR amplification. The resulting PCR fragments were cloned into pDisplay (Invitrogen) expression vector, which allows expression of proteins on the cell surface. All endoglin constructs expressed from pDisplay contain the influenza hemagglutinin epitope HA at the NH 2 terminus. The oligonucleotides used to prime the PCR synthesis of endoglin fragments were as follows: TMCT-Endo (amino acids 573-658), GAAGATCTAACATCATCAGCCCTGAC and TCCCCGCGGGGCTATGCCATGCTGCT; EC-Endo (amino acids 26-586), GGGGCCCAGCCGGCCGAAACAGTCCATTGT and TCCCCGCGGGGC-CTTTGCTTGT; ECTM-Endo (amino acids 26-614), GGGGCCCAGCCG-GCCGAAACAGTCCATTGT and TCCCCGCGGTCAGTAGATGTACCA; 558-Endo (amino acids 26-558), GGGGCCCAGCCGGCCGAAACAGTC-CATTGT and TCCCCGCGGCCCGGTCTTGGG; 437-Endo (amino acids 26 -437), GGGGCCCAGCCGGCCGAAACAGTCCATTGT and TC-CCCGCGGTTTCCGCTGTGG; 437/586-Endo (amino acids 437-586), GGGGCCCAGCCGGCCGAAAAAGGTGCACTGC and TCCCCGCGGGC-CTTTGCTTGT. The oligonucleotides were designed to introduce a SfiI restriction site at the 5Ј-end (except for the TMCT-Endo mutant, which contains a BglII restriction site) and a SacII site at the 3Ј-end of the endoglin DNA fragments. The reverse primers for TMCT-Endo and ECTM-Endo include a stop codon. All primer sequences for this study are given from 5Ј to 3Ј. Full-length endoglin was also subcloned into the pDisplay vector to generate an HA epitope-tagged endoglin. The construct was engineered by PCR using forward primer that included a SfiI restriction site (GGGGC-CCAGCCGGCCGAAACAGTCCATTGT) and reverse primer that included a stop codon and a SacII site (TCCCCGCGGGGCTATGCCATGCTGCT).
The pCMV5 expression constructs containing cDNAs for T␤RII/HA (tagged at the COOH terminus with the influenza hemagglutinin (HA) epitope), HA/T␤RII (K277R) (tagged at the NH 2 terminus with the HA epitope), T␤RI/HA, T␤RI (T204D)/HA, and T␤RI (K232R)/HA have been described (23) and were a generous gift from Dr. Liliana Attisano (University of Toronto, Canada).
Antibodies-The endoglin-specific monoclonal antibodies P4A4 and P3D1 have been previously described, and they recognize epitopes contained within the fragments Tyr 227 -Gly 331 and Glu 26 -Gly 230 of human endoglin, respectively (24). For recognition of T␤RII and T␤RI, specific polyclonal antibodies were used (C16 and V22, respectively; Santa Cruz Biotechnology Inc., Santa Cruz, CA). Proteins tagged with the influenza virus HA epitope were detected with the monoclonal antibody 12CA5 (Roche Molecular Biochemicals).
Immunoprecipitation and Western Blot Analyses-Cells were lysed at 4°C for 30 min with lysis solution (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1 mM EDTA, 0.5 mM EGTA, 1% digitonin, 1 mM NaF, 10 mM NaVO 4 , 1 mM Mo 2 VO 4 , and a mixture of protease inhibitors). Aliquots of total cell lysates containing equivalent amounts of total protein were precleared for 4 h with protein G or protein A coupled to Sepharose (Amersham Biosciences) at 4°C. Specific immunoprecipitations of the precleared lysates were carried out in the presence of the appropriate antibody, using protein G-or protein A-Sepharose. After overnight incubation at 4°C, immunoprecipitates were isolated by centrifugation and washed three times with lysis buffer. When required, immunoprecipitates were incubated with alkaline phosphatase (Roche Molecular Biochemicals) at 37°C for 1 h. Total lysates and the precipitated proteins were separated by SDS-PAGE and electrotransferred to a polyvinylidene difluoride membrane for immunodetection with the indicated antibody. Immunodetection was performed with the SuperSignal chemiluminescent substrate (Pierce) according to the manufacturer's instructions.
Flow Cytometry Analysis-COS cells were transiently transfected with the indicated HA-tagged endoglin constructs. After 48 h, cells were incubated with the mouse monoclonal antibody 12CA5 (against the influenza virus HA epitope; Roche Molecular Biochemicals) for 30 min at 4°C. After two washes with phosphate-buffered saline, fluorescein isothiocyanate-labeled F(abЈ)2 rabbit anti-mouse Ig was added, and incubation proceeded for an additional period of 30 min at 4°C. Finally, cells were washed twice with phosphate-buffered saline, and their fluorescence was estimated with an EPICS-CS (Coulter Cientifica), using logarithmic amplifiers. Results are expressed as an expression index calculated as the percentage of marker-positive cells multiplied by their mean fluorescence intensity.
Phosphorylation Assays in Vitro-Immunoprecipitates of T␤RI or T␤RII were washed twice with lysis buffer, washed once with kinase FIG. 1. Endoglin interacts with T␤RI and T␤RII in the absence of exogenous ligand. COS-7 cells were transiently transfected with expression vectors encoding endoglin, T␤RI, or T␤RII, as indicated. Cell lysates were immunoprecipitated (IP) with a mixture of P3D1 and P4A4 anti-endoglin (upper panels) or anti-T␤RI or anti-T␤RII (middle panels) antibodies. Immunoprecipitates were subjected to SDS-PAGE under reducing conditions, followed by Western blot (WB) with the indicated antibodies to allow the detection of co-immunoprecipitated proteins. As a control for transfection, total lysates (TL) were analyzed by Western blot (lower panels). This is a representative experiment of at least three different ones.
buffer (50 mM Tris-HCl, pH 7.5, 5 mM MnCl 2 , 5 mM MgCl 2 , 1 mM CaCl 2 ), and resuspended in 40 l of kinase buffer with 5 M ATP, 1 Ci of [␥-32 P]ATP (3000 Ci/mmol; Amersham Biosciences), and 2 g of recombinant GST, GST-Ecyt, GST-GS/T␤RI, GST-Smad2, GST-Smad3, or GST-Smad4. The reaction was incubated for 30 min at 37°C. Phosphorylation was stopped by the addition of Laemmli sample buffer, and the products were resolved by SDS-PAGE. The incorporation of [ 32 P]phosphate was visualized by autoradiography. Assays for T␤R-II kinase activity were performed as described above but without adding GST proteins.
Phosphorylation Assays in Vivo-COS-7 cells were transfected with T␤RI, T␤RII, and endoglin, as indicated. After 48 h, the cells were washed in phosphate-free medium (ICN Biomedicals) and metabolically labeled with 0.5 mCi/ml [ 32 P]orthophosphate for 3 h. Cells were washed with phosphate-buffered saline, lysed with lysis buffer, and immunoprecipitated with the appropriate antibodies. Immunoprecipitates were analyzed on SDS-polyacrylamide gels, and radiolabeled bands were detected by autoradiography and quantified by densitometry using a PhosphorImager 410 and ImageQuant software.
Luciferase Assays-The pARE-lux reporter vector, encoding the ac- Immunoprecipitates were electrophoresed by SDS-PAGE under reducing conditions, followed by Western blot (WB) with anti-HA, anti-T␤RI, or anti-T␤RII antibodies (upper panels). As a control for transfection, recombinant proteins were analyzed in total lysates (TL) by Western blot (lower panels). C, full-length endoglin associates with T␤RI and T␤RII kinases independently of their activation state. COS-7 cells were transiently transfected with expression vectors encoding endoglin, wild type T␤RI kinase, constitutively active T␤RI kinase (T204D), kinase-deficient T␤RI (K232R), wild type T␤RII, and kinase-deficient T␤RII (K277R) as indicated. Total lysates were immunoprecipitated with anti-endoglin antibodies (P3D1 and P4A4), and immunoprecipitates were electrophoresed by SDS-PAGE under reducing conditions, followed by Western blot with anti-T␤RI or anti-T␤RII antibodies (upper panels). As a control for transfection, recombinant proteins were analyzed in total lysates by Western blot (lower panels). Results shown in A and B are representative of four different experiments.
tivin response element of Xenopus mix.2 promoter fused to the luciferase reporter gene, and the TGF-␤ transcriptional co-activator Fast-1 have been previously described (25)(26)(27). L 6 E 9 myoblasts were selected for these studies, because they do not express endoglin or its homologue betaglycan, and they have proved to be a useful model system to analyze the function of endoglin and betaglycan in TGF-␤ responses (11,28). Activity of the TGF-␤ reporter construct was determined by transient transfection of L 6 E 9 myoblasts using Superfect (Qiagen). Briefly, cells in 24-well plates were transfected with the pARE reporter and the endoglin or Fast-1 expression vectors at densities of 5 ϫ 10 4 cells/well. The amount of DNA in each transfection was normalized by using the corresponding insertless expression vector as carrier. After 24 h, cells were washed with Dulbecco's modified Eagle's medium before incubation with 5 ng/ml TGF-␤1 for an additional 24-h period. Relative luciferase units from triplicate samples were determined in a TD20/20 luminometer (Promega, Madison, WI). Each transfection experiment was performed at least four times with different DNA preparations. Correction for transfection efficiency was made by cotransfection with the ␤-galactosidase expression vector pCMV-␤-galactosidase, and the corresponding enzymatic activity was determined using the Galacto-Light kit (Tropix). The mean and S.D. were calculated and experimental results of the promoter constructs were displayed either as arbitrary units of luciferase activity, or as a -fold induction with respect to the corresponding untreated sample.

Endoglin Associates with the TGF-␤ Signaling Receptors in the Absence of Exogenous
Ligand-To study the interaction of endoglin with T␤RI or T␤RII, COS cells were transfected with the corresponding expression vectors in the absence of exogenous ligand, and cell lysates were subjected to specific immunoprecipitation, followed by Western blot analysis. As shown in Fig. 1, endoglin co-immunoprecipitates with either T␤RI or T␤RII as revealed by immunodetection with antibodies to T␤RI and T␤RII, respectively (lanes 3 and 6). Although the transfected COS cells express low levels of signaling receptors, no association between endoglin and the endogenous receptors could be detected (lanes 1 and 4). These results demonstrate that endoglin interacts with T␤RI and with T␤RII in the absence of exogenous TGF-␤ and that the formation of the endoglin-T␤RI or endoglin-T␤RII complexes does not require the presence of T␤RII or T␤RI, respectively.

Formation of Endoglin-T␤RI and Endoglin-T␤RII Complexes Is Not Affected by the Activation State of the Signaling
Receptors-To analyze whether the complex formation of endoglin with T␤RI or T␤RII was affected by the activation state of the signaling receptors, experiments were performed in the presence of exogenous ligand or signaling receptors with varying catalytic activities. First, COS cells were transfected with endoglin and T␤RI or T␤RII and cultured with different concentrations of fetal calf serum, either in the absence or in the presence of 10 ng/ml exogenous TGF-␤1. Fetal calf serum is a minor source of TGF-␤, and preliminary experiments indicated FIG. 4. Analysis of the interaction between the extracellular domain of endoglin and T␤RI and T␤RII. A, the extracellular domain of endoglin associates with T␤RI and T␤RII. COS-7 cells were transiently transfected with expression vectors encoding HA-ECTM-Endo, HA-EC-Endo, T␤RI, or T␤RII, as indicated. Total lysates were immunoprecipitated (IP) with anti-T␤RI or anti-T␤RII antibodies, and immunoprecipitates were electrophoresed by SDS-PAGE under reducing conditions. The presence of endoglin was revealed by Western blot (WB) with anti-HA antibodies (upper panels). As a control for transfection, recombinant proteins were analyzed in total lysates (TL) by Western blot (lower panels). B and C, mapping the extracellular regions of endoglin involved in the association with T␤RI and T␤RII. COS-7 cells were transiently transfected with T␤RI, T␤RII, HA-EC-Endo, HA-437-Endo, HA-558-Endo, or HA-437/586-Endo, as indicated. Total lysates were immunoprecipitated with a mixture of P3D1 and P4A4 antibodies (␣-endoglin), anti-T␤RI, or anti-T␤RII, and immunoprecipitates were electrophoresed by SDS-PAGE under reducing conditions. The presence of T␤RI, T␤RII, or endoglin was revealed by Western blot with anti-T␤RI, anti-T␤RII, or anti-HA antibodies, respectively. As a control for transfection, recombinant proteins were analyzed in total lysates by Western blot. T␤RI coprecipitates with the truncated endoglin mutants EC-Endo, 558-Endo, and 437-Endo, whereas T␤RII only coprecipitates with EC-Endo and 558-Endo (B). T␤RI and T␤RII coprecipitate with the truncated endoglin mutant 437/586-Endo (C). Results shown in A-C are representative of at least four different experiments. that 0.2% is the minimal concentration of fetal calf serum required for cell viability. As shown in Fig. 2A, the amount of T␤RI (left upper panel) or T␤RII (right upper panel) co-immunoprecipitated by anti-endoglin antibodies is similar for all treatments, suggesting that the addition of exogenous ligand does not affect the formation of endoglin-T␤RI or endoglin-T␤RII complexes. Since the active TGF-␤ receptor signaling complex requires both T␤RI and T␤RII, endoglin association was analyzed when both signaling receptors were cotransfected (Fig. 2B). Again, exogenous TGF-␤ did not affect the formation of endoglin complexes with T␤RI or T␤RII. Next, COS cells were transfected with expression vectors for endoglin as well as wild type and different kinase mutant versions of the signaling receptors. As shown in Fig. 2C, the constitutively active T␤RI (T204D) mutant, the constitutively inactive T␤RI (K232R) mutant, and the wild type T␤RI construct were able to associate with endoglin (upper left panel). Similarly, the kinase-inactive T␤RII (K277R), and the wild type T␤RII constructs showed interaction with endoglin (upper right panel). Taken together, these results suggest that endoglin association with T␤RI and T␤RII is independent of the activation state of the signaling receptors.
The Extracellular Domain of Endoglin Interacts with T␤RI and T␤RII-To identify the region(s) involved in the association with the signaling receptors, several truncated versions of endoglin were generated (Fig. 3). Construct HA-TMCT-Endo lacks the extracellular domain, HA-EC-Endo lacks the cytoplasmic and transmembrane regions, and HA-ECTM-Endo only lacks the cytoplasmic domain. Additional mutants encoding only part of the extracellular domain (558-Endo, 437-Endo, and 437/586-Endo) were also analyzed. Constructs 437-Endo and 437/586-Endo were generated around the arginine at position 437, because several lines of investigation suggest that this residue might define a protein domain: (a) it is located at a putative protease cleavage site of the protein; (b) artificial constructs truncated at position 437 can be expressed upon transfection of mammalian cells; and (c) its codon is located in the border between exons 9b and 10 (29). In order to facilitate the analysis of the mutants, all constructs contained an epitope tag of HA at the amino terminus. Transfection of these constructs in COS cells confirmed that all mutant versions of endoglin were expressed as evidenced by Western blot analysis (Fig. 3B). Since the wild type endoglin is a disulfide-linked dimer (30), we analyzed whether the truncated forms were also disulfide-linked by subjecting the lysates to electrophoresis under reducing or nonreducing conditions. As shown in Fig. 3B, all of the constructs were expressed in a dimeric form. Interestingly, the smallest construct, TMCT-Endo, contains only one cysteine residue at position 582 of the extracellular domain, suggesting its involvement in the dimerization process. This was further demonstrated by generating the mutant construct TMCT-Endo/C582G, where the cysteine at 582 has been replaced by a glycine. Transfection of this mutant yielded a monomeric form of endoglin, as opposed to the dimer obtained with the TMCT-Endo plasmid (Fig. 3C). Furthermore, similar to the wild type endoglin, all of the truncated forms were also expressed at the cell surface as demonstrated by flow cytometry analysis (Fig. 3D).
To investigate whether the extracellular domain of endoglin was involved in the association with the signaling receptors, HA-ECTM-Endo, T␤RI, and T␤RII were expressed upon transfection in COS cells. Immunoprecipitation analysis with anti-  T␤RI and  T␤RII. COS-7 cells were transiently  transfected with expression vectors encoding HA-TMCT-Endo, wild type T␤RI,  constitutively  active  T␤RI  kinase  (T204D), kinase-deficient T␤RI (K232R), wild type T␤RII, and kinase-deficient T␤RII (K277R), as indicated. T␤RI or T␤RII constructs were immunoprecipitated (IP) from total lysates with anti-T␤RI or anti-T␤RII antibodies, and immunoprecipitates were subjected to SDS-PAGE under reducing conditions. The presence of endoglin was revealed by Western blot (WB) with anti-HA antibodies (␣-HA). As a control for transfection, recombinant proteins were analyzed in total lysates (TL) by Western blot. A, HA-TMCT-Endo coprecipitates with the kinase-deficient form of T␤RI (K232R) but not with the constitutively active T␤RI kinase (T204D). B and C, HA-TMCT-Endo coprecipitates with both wild type (wt) T␤RII and kinase-deficient T␤RII (K277R) but shows a different electrophoretic migration in each case. The addition of alkaline phosphatase to immunoprecipitates (rightmost lane), indicates that the lower electrophoretic migration is due to endoglin phosphorylation. It is worth noting that the intensity of the endoglin band associated with the T␤RII (K277R) is similar to that of the phosphatase-treated sample of T␤RII (wt), suggesting that the lower intensity displayed by the untreated T␤RII (wt) immunoprecipitate is probably due to a higher heterogeneity of the associated endoglin band. Results shown in A-C are representative of at least five different experiments.
T␤RI and anti-T␤RII demonstrated that HA-ECTM-Endo was co-precipitated only when T␤RI or T␤RII were expressed (Fig.  4A), indicating that the extracellular domain of endoglin interacts with both signaling receptors. Parallel studies with HA-EC-Endo construct, which does not contain the transmembrane domain of endoglin, revealed similar levels of truncated protein co-precipitated with anti-T␤RI and anti-T␤RII (Fig. 4A), further confirming the involvement of the extracellular domain and suggesting a nonrelevant role for the transmembrane region in the interaction with the signaling receptors. Next, the interactions of different truncations of the extracellular domain of endoglin with T␤RI and T␤RII were analyzed. Immunoprecipitation studies with anti-endoglin, anti-T␤RI, and anti-T␤RII demonstrated that HA-558-Endo, HA-437-Endo, and HA-437/586-Endo co-precipitated with T␤RI, whereas only HA-558-Endo and HA-437/586-Endo co-precipitated with T␤RII (Fig. 4, B and C). These results suggest that (a) T␤RII interacts with residues 437-558 of the extracellular domain of endoglin, a region proximal to the transmembrane domain and (b) T␤RI interacts not only with residues 437-558 but also with a second region comprised between amino acids 26 and 437 of the extracellular domain of endoglin.
The Cytoplasmic Domain of Endoglin Interacts with T␤RI and T␤RII-When studying the interaction between two different transmembrane proteins, it is important to assess both the involvement of their extracellular domains and that of their cytoplasmic domains. Endoglin is constitutively phosphorylated in Ser/Thr residues (31,32), and both signaling receptors are Ser/Thr kinases (33,34), thus providing a hint for the participation of their cytoplasmic domains in protein association. To study the interaction between endoglin cytoplasmic domain and T␤RI or T␤RII, the HA-TMCT-Endo construct (lacking the extracellular domain) was co-transfected with wild type or kinase mutants of T␤RI and T␤RII (Fig. 5). Immunoprecipitation analysis with anti-T␤RI and anti-T␤RII demonstrated that the cytoplasmic domain of endoglin associates with T␤RI when the kinase domain is inactive (K232R) but not with the constitutively active (T204D) form (Fig. 5A). In the same experiment, a weak band of endoglin was found associated with the wild type T␤RI. This association is probably due to the FIG. 6. Endoglin phosphorylation by T␤RI and T␤RII kinases. A, the cytoplasmic domain of endoglin is phosphorylated by T␤RI and T␤RII kinases in vitro. COS-7 cells were transiently transfected with HA-tagged wild type T␤RI kinase, constitutively active T␤RI kinase (T204D), kinase-deficient T␤RI (K232R), wild type T␤RII, and kinase-deficient T␤RII (K277R), as indicated. Cells were collected 48 h later, and their lysates were immunoprecipitated (IP) with anti-HA antibodies. Immunoprecipitates were incubated with equal amounts of GST or GST-Ecyt proteins in kinase buffer containing [␥-32 P]ATP, and phosphorylation was allowed to proceed. Samples were analyzed by SDS-PAGE followed by autoradiography. The positions of GST and GST-Ecyt are indicated by arrowheads. As a control for transfection, recombinant T␤RI and T␤RII proteins were analyzed in total lysates (TL) by Western blot (WB). B, endoglin is phosphorylated by T␤RI and T␤RII kinases in vivo. COS-7 cells were transiently transfected with fulllength endoglin, wild type T␤RI kinase, kinase-deficient T␤RI (K232R), wild type T␤RII, and kinase-deficient T␤RII (K277R), as indicated. Cells were metabolically labeled 48 h later with [ 32 P]orthophosphate, lysed, and cell lysates were immunoprecipitated with antiendoglin antibodies (P4A4 and P3D1). Immunoprecipitates were subjected to SDS-PAGE under reducing conditions, and radiolabeled bands were detected by autoradiography. As a control for transfection, recombinant endoglin, T␤RI, and T␤RII proteins were analyzed in total lysates by Western blot. C, radiolabeled bands from B were quantified by densitometry. The sample number is the same as in B. Results shown in A-C are representative of at least four different experiments.
T␤RI activation induced by oligomerization of the signaling receptors that occurs upon transfection (35,36). In addition, the cytoplasmic domain of endoglin interacts with both the active (wild type) and inactive (K277R) forms of T␤RII (Fig.  5B). These results demonstrate that the endoglin cytoplasmic domain interacts with T␤RI and with T␤RII.
The Cytoplasmic Domain of Endoglin Is Phosphorylated by T␤RI and T␤RII-A detailed analysis of the bands separated by SDS-PAGE, corresponding to the endoglin cytoplasmic domain, revealed that the electrophoretic mobility of the HA-TMCT-Endo associated with the constitutively active T␤RII (wild type) was lower than that of the HA-TMCT-Endo associated with the kinase-inactive (K277R) form of T␤RII (Fig. 5, B and C; upper panels). Similar electrophoretic differences were observed when analyzing the endoglin cytoplasmic domain in total lysates derived from cells overexpressing either the wild type or the constitutively inactive form of T␤RII, respectively (Fig. 5, B and C; lower panels). It is well known that phosphorylated proteins migrate more slowly that the unphosphorylated forms, suggesting that the electrophoretic differences observed above might be due to different phosphorylation states of endoglin. This interpretation would be compatible with the fact that the migration displayed by HA-TMCT-Endo is dependent on the activation state of the associated kinase. To confirm this hypothesis, co-precipitates of HA-TMCT-Endo and constitutively active T␤RII were treated or not with alkaline phosphatase. As shown in Fig. 5C, upon phosphatase treatment, endoglin associated with wild type T␤RII showed an electrophoretic mobility identical to that of endoglin associated with the constitutively inactive T␤RII, suggesting that endoglin not only associates with T␤RII but is also a substrate for T␤RII.
At variance with T␤RII, no changes in the electrophoretic mobilities among the HA-TMCT-Endo constructs associated

FIG. 7. T␤RII phosphorylation is diminished in the presence of endoglin.
A, COS-7 cells were transiently transfected with HA-endoglin and wild type T␤RII, as indicated. Cells were metabolically labeled with [ 32 P]orthophosphate and lysed, and cell lysates were immunoprecipitated (IP) with anti-T␤RII antibodies. Immunoprecipitates were subjected to SDS-PAGE under reducing conditions, and radiolabeled bands were detected by autoradiography. As a control for transfection, recombinant endoglin and T␤RII proteins were analyzed in total lysates (TL) by Western blot (WB). B, COS-7 cells were transiently transfected with HA-TMCT-Endo, wild type T␤RII, and kinase-deficient T␤RII (K277R), as indicated. Cells were metabolically labeled with [ 32 P]orthophosphate and lysed, and cell lysates were immunoprecipitated with anti-T␤RII antibodies. Immunoprecipitates were subjected to SDS-PAGE under reducing conditions, and radiolabeled bands were detected by autoradiography. As a control for transfection, recombinant endoglin and T␤RII proteins were analyzed in total lysates by Western blot. C, radiolabeled bands from B were quantified by densitometry. The sample number is the same as in B. D, COS-7 cells were transiently transfected with HA-endoglin, wild type T␤RII, and kinase-deficient T␤RII (K277R), as indicated. Cell lysates were immunoprecipitated with anti-T␤RII antibodies, immunoprecipitates were subjected to SDS-PAGE under reducing conditions, and radiolabeled bands were detected by autoradiography. The data show that the autophosphorylation capacity of T␤RII kinase is diminished in the presence of endoglin. As a control for transfection, recombinant endoglin and T␤RII proteins were analyzed in total lysates by Western blot. E, radiolabeled bands from D were quantified by densitometry. The sample number is the same as in D. Results shown in A-E are representative of at least four different experiments.
with the different versions of T␤RI were observed (Fig. 5A). Since the cytoplasmic domain of endoglin associates with the kinase-inactive but not with the constitutively active form (Fig.  5A), it is possible that, if phosphorylated by T␤RI, the endoglin cytoplasmic domain would be detached from the active kinase complex.
To further analyze the phosphorylation of endoglin by the signaling receptors, in vitro and in vivo studies were carried out. First, immunoprecipitates of T␤RI, T␤RII, or their mutant forms were incubated with a GST fusion protein containing the cytoplasmic domain of endoglin (GST-Ecyt) in the presence of [␥-32 P]ATP to allow kinase activity. As shown in Fig. 6A, the active forms of T␤RI (T204D) and T␤RII (wild type) are able to phosphorylate the endoglin cytoplasmic domain. The wild type T␤RI also displays endoglin phosphorylation activity despite the absence of TGF-␤ activation. This is probably due to the T␤RI activation induced by oligomerization of the signaling receptors that occurs upon transfection (35,36). By contrast, no significant endoglin phosphorylation was observed in the presence of inactive signaling receptors (T␤RI/K232R and T␤RII/ K277R), excluding the possible contribution of endogenous signaling receptors due to heteromeric associations with the recombinant constructs. As a negative control for specificity in all samples, no phosphorylation could be observed when using GST alone as a substrate. Next, we analyzed endoglin phosphorylation in vivo. Cells transfected with full-length endoglin and signaling receptors were metabolically labeled with [ 32 P]orthophosphate, and the phosphorylation state of endoglin was analyzed by immunoprecipitation followed by SDS-PAGE analysis (Fig. 6B). In mock-transfected cells, a basal level of endoglin phosphorylation was observed (lane 2), probably due to the endogenous signaling receptors. Upon transfection with the wild type T␤RI (lane 6) and T␤RII (lane 3), a marked increase in endoglin phosphorylation levels could be detected. Also, the constitutively kinase-active T␤RI (T204D) yielded similar levels of endoglin phosphorylation as the wild type form (data not shown). By contrast, expression of the kinase-inactive forms of T␤RI (lane 7) and T␤RII (lane 4) yielded background signals, supporting the specificity of the wild type signaling receptor-induced endoglin phosphorylation. In addition, experiments performed with the truncated form TMCT-Endo, lacking the extracellular domain of endoglin, resulted in a similar phosphorylation pattern by the wild type and mutant forms of the signaling receptors (data not shown), indicating that phosphorylation of endoglin occurs at the serine/threonine-rich cytoplasmic domain.
Endoglin Expression Modulates the Phosphorylation State of T␤RII and T␤RI-The phosphorylation state of T␤RII and T␤RI plays a critical role in TGF-␤-induced cell signaling (35,36). Since endoglin not only associates with these signaling receptors but is also a substrate for their kinase activity, we analyzed whether endoglin expression affected the phosphorylation state of T␤RII and T␤RI. Thus, in vivo and in vitro studies were carried out. First, we investigated the effect of endoglin expression on T␤RII phosphorylation after metabolic labeling with [ 32 P]orthophosphate. Immunoprecipitation of transfected T␤RII from endoglin-deficient cells resulted in a broad band (between 70 and 90 kDa) of the autophosphorylated kinase (Fig. 7A), as previously described (35). Similar experiments with cells cotransfected with full-length endoglin yielded a weaker T␤RII band overlapping with a strong signal at 90 kDa of phosphorylated endoglin (Fig. 7A). Since the overlapping between T␤RII and endoglin bands interfered with the estimation of T␤RII phosphorylation levels, we decided to carry FIG. 8. T␤RI phosphorylation is enhanced in the presence of endoglin. A, phosphorylation of the juxtamembrane and GS regions of T␤RI in vitro. COS-7 cells were transiently transfected with HA-endoglin and wild type T␤RII, as indicated. Cell lysates were immunoprecipitated (IP) with anti-T␤RII antibodies, and immunoprecipitates were incubated with GST or GST-GS/T␤RI in the presence of [␥-32 P]ATP. Phosphorylated substrates were subjected to SDS-PAGE under reducing conditions, and radiolabeled bands were detected by autoradiography. The data show that the phosphorylation of the juxtamembrane/GS regions of T␤RI is increased in the presence of endoglin. As a control for transfection, recombinant endoglin and T␤RII proteins were analyzed in total lysates (TL) by Western blot (WB). B, radiolabeled bands from A were quantified by densitometry. The sample number is the same as in A. These results are representative of at least four different experiments. C, phosphorylation of T␤RI in vivo. COS-7 cells were transiently transfected with HA-endoglin and wild type T␤RI, as indicated. Cells were metabolically labeled with [ 32 P]orthophosphate, lysed, and cell lysates were immunoprecipitated with anti-T␤RI antibodies. Immunoprecipitates were subjected to SDS-PAGE under reducing conditions, and radiolabeled bands were detected by autoradiography. As a control for transfection, recombinant endoglin and T␤RI proteins were analyzed in total lysates by Western blot. D, radiolabeled bands from C were quantified by densitometry. The sample number is the same as in C. These results are representative of at least four different experiments. out experiments using a smaller endoglin construct, HA-TMCT-Endo, which contains only the transmembrane and cytoplasmic domains. Thus, cotransfection of HA-TMCT-Endo resulted in a marked decrease of T␤RII phosphorylation levels (Fig. 7, B and C; compare lane 3 with lane 1). As expected, no significant phosphorylation was obtained in the presence of the kinase-inactive mutant of T␤RII (lane 4). Since T␤RII is constitutively autophosphorylated (35,37), we analyzed whether the decrease of T␤RII phosphorylation levels observed in vivo was due to a decreased autokinase activity of T␤RII. Thus, anti-T␤RII immunoprecipitates from cell lysates, containing or not containing full-length endoglin, were incubated with [␥-32 P]ATP, and the radiolabeled T␤RII was analyzed by SDS-PAGE (Fig. 7, D and E). As expected, the T␤RII present in endoglin-deficient cells (lane 2) showed a clear autophosphorylating activity as compared with a negative control (lane 1). Interestingly, co-expression of endoglin resulted in a markedly reduced level of T␤RII phosphorylation (lane 3), suggesting that endoglin inhibits the autophosphorylation of T␤RII. As expected, the kinase-inactive mutant showed background levels of phosphorylation either in the presence (lane 4) or in the absence of endoglin (not shown). These results suggest that endoglin expression inhibits the phosphorylation of T␤RII.
The phosphorylation state of specific Ser/Thr residues in the T␤RII is able to regulate its kinase activity (2)(3)(4). Since the juxtamembrane and GS regions of T␤RI are direct substrates for the T␤RII kinase, we investigated whether the endoglin-dependent modulation of the T␤RII phosphorylation affected the T␤RI phosphorylation. Anti-T␤RII immunoprecipitates from cell lysates, with or without endoglin, were incubated with a recombinant protein containing the juxtamembrane and GS regions of T␤RI (GST-GS/T␤RI) in the presence of [␥-32 P]ATP, and the radiolabeled substrate was analyzed by SDS-PAGE (Fig. 8, A and B). In the absence of endoglin, T␤RII induced the specific phosphorylation of GST-GS/T␤RI (lane 2), as compared with a negative control (lane 1). However, upon endoglin expression, a markedly enhanced phosphorylation of the GST-GS/T␤RI protein was observed (lane 4). These results suggest that endoglin expression leads to an increased phosphorylation of the T␤RI within the juxtamembrane and GS regions. This observation was confirmed by analyzing the effect of endoglin expression on T␤RI phosphorylation after metabolic labeling with [ 32 P]orthophosphate. As depicted in Fig. 8, C and D, the basal phosphorylation levels of T␤RI (lane 3) were clearly increased when endoglin was cotransfected (lane 4).
Overall, these results indicate that endoglin modulates the phosphorylation levels of both signaling receptors.
Effect of Endoglin on TGF-␤ Signaling-Since endoglin associates with the signaling receptors and modulates their phosphorylation status, we examined whether endoglin expression affected Smad phosphorylation and TGF-␤ gene responses. In the TGF-␤ signaling pathway, phosphorylation of Smad2 and Smad3 by T␤RI is a necessary step for downstream signal transduction. Anti-T␤RI immunoprecipitates from cell lysates, with or without endoglin, were incubated with Smad2, Smad3, or Smad4 recombinant proteins in the presence of [␥-32 P]ATP, and radiolabeled Smad proteins were analyzed by SDS-PAGE (Fig. 9, A and B). In the absence of endoglin, specific phosphorylation of Smad2 and Smad3 could be observed, whereas the presence of endoglin induced an increased phosphorylation of Smad2 but not Smad3 (Fig. 9B). As expected, no phosphorylation was found in Smad4, which lacks the T␤RI target sequence within the MH2 domain. Since these results suggest that endoglin expression leads to an increased phosphorylation of Smad2, we investigated the reporter expression of the mix.2 gene promoter, whose activation by TGF-␤/activin involves the formation of a transcriptional complex composed of Smad2/ Smad4 and Fast-1 (26). L 6 E 9 cells were transfected with the pARE-lux reporter vector in the absence or in the presence of endoglin (Fig. 9, C and D). Endoglin was able to consistently increase the TGF-␤ responsiveness in a dose-dependent fashion, as analyzed by both absolute activity (Fig. 9C) and -fold induction with respect to untreated samples (Fig. 9D). Interestingly, the basal activity of the pARE reporter vector was inhibited with increasing concentrations of endoglin, consistent with the reported role of endoglin in the absence of exogenous TGF-␤ (38). Taken together, these results suggest that endoglin is able to increase the Smad2-mediated TGF-␤ responses of the mix.2 gene promoter. DISCUSSION Several laboratories have previously shown that endoglin forms heteromeric complexes with several members of the TGF-␤ signaling receptor family in the presence of ligand (8,11,12,13,32,39). Here, we show that full-length endoglin physically interacts with T␤RI and T␤RII, and this interaction is not modified by the presence of exogenous ligand or by the activation state of the signaling kinases. This finding is compatible with the existence of ligand-free endoglin forming a complex with ligand-bound signaling receptors (11) as well as with the requirement of signaling receptors for ligand binding to endoglin (11,12). Also, the direct interaction of endoglin with T␤RI and T␤RII is supported by the existence of ligand-independent complexes of endoglin and ALK-1 in endothelial cells (40,41). We have observed that the formation of the complex endoglin-T␤RI or endoglin-T␤RII does not require the presence of T␤RII or T␤RI, respectively. This finding agrees with the fact that endoglin is co-immunoprecipitated with T␤RII⅐TGF-␤, ActRII-activin-A, or ActRII-BMP-7 complexes, despite the absence of the corresponding type I receptor in the cell (12). Taken together, these results provide a solid molecular basis for ligand binding to endoglin. Thus, endoglin could bind first to the signaling receptors, and this interaction would then allow ligand access to endoglin.
Generation of a panel of endoglin constructs (Fig. 3) has facilitated a more detailed analysis of the association between endoglin and the signaling receptors. Similar to the wild type endoglin, the mutant forms were expressed as disulfide-linked dimers. Interestingly, mutation of Cys 582 abrogated the formation of disulfide linkages in the smallest construct (Fig. 3C), indicating its direct involvement in interchain disulfide bonds. Other cysteine residues contained within the endoglin fragment Cys 330 -Cys 412 have been previously proposed as candidates in endoglin disulfide formation (29,42), but Cys 582 is the first one to be specifically identified.
Analysis on the physical association between endoglin mutants and the signaling receptors has revealed that both the extracellular and cytoplasmic domains participate in this association. Thus, several interaction sites can participate in the formation and stabilization of the endoglin complexes, similarly to the contact sites mapped within the T␤RI⅐T␤RII (34,43) or betaglycan-T␤RII (44) complexes. Despite the structural homology between T␤RI and T␤RII, they show two major differences in their association with endoglin. First, the interaction of T␤RII with the extracellular domain of endoglin maps within the transmembrane proximal region 437-558, whereas T␤RI interacts not only with fragment 437-558, but also with the N-terminal region(s) located above residue 437. Second, the interaction of T␤RI with the cytoplasmic domain of endoglin can only be detected when its kinase domain is inactive; by contrast, both active and inactive forms of T␤RII can associate with the cytoplasmic domain of endoglin. Upon interaction with the signaling receptors, the cytoplasmic domain of endoglin is phosphorylated by T␤RI or T␤RII. This finding fully agrees with the fact that endoglin is constitutively phosphorylated in Ser/Thr residues (31,32) and forms heteromeric complexes with the signaling receptors displaying Ser/Thr kinase activity (33,34). Once phosphorylated by the respective kinase, the cytoplasmic domain of endoglin remains bound to the T␤RII, whereas it is released from T␤RI. Since the kinase activity of T␤RI is only induced in the presence of ligand, these results suggest that the cytoplasmic domain of endoglin dissociates from T␤RI upon exposure to members of the TGF-␤ family. This is not surprising, since in many enzymatic systems the kinase/substrate association is usually transient. In fact, a similar transient interaction has been described for T␤RI-mediated phosphorylation of R-Smads (45). In contrast, the association of T␤RII with the cytoplasmic domain of endoglin seems to be stable even after phosphorylation has occurred. Similarly, phosphorylation of T␤RI by T␤RII kinase does not affect the T␤RI⅐T␤RII association (43). Interestingly, at variance with our endoglin data, the interaction of the T␤RII with the cytoplasmic domain of betaglycan does not occur with a kinaseinactive version of the receptor (44). Since the T␤RII is constitutively active within the cell (35), our results suggests that the constitutive phosphorylation of endoglin is mainly due to this kinase. Nevertheless, although we find that T␤RI or T␤RII is able to phosphorylate endoglin, we cannot exclude the possibility that some of the 19 Ser/Thr residues within the cytoplasmic domain are targeted by additional Ser/Thr kinases. In this regard, putative consensus motifs for protein kinase C and casein kinase have been identified within the cytoplasmic sequence (31) (data not shown).
The high degree of homology between the cytoplasmic domains of endoglin and betaglycan suggests a conserved function/regulation for these domains. Supporting this view, Blobe et al. (44) have described that the cytoplasmic domains of betaglycan and T␤RII interact specifically, and this interaction results in the phosphorylation of the betaglycan cytoplasmic domain, in agreement with the results shown here for endoglin. Both endoglin and betaglycan contain PDZ class I binding motifs (Ser/Thr-X-Ala/Val) at their homologous cytoplasmic regions. Recently, GIPC was identified as a PDZ domain-containing protein able to interact with betaglycan cytoplasmic domain and regulate TGF-␤ signaling (46), although the possible association of GIPC with the endoglin cytoplasmic region remains to be established.
Endoglin has been shown to modulate cellular responses to TGF-␤1 (11,13,14). However, the molecular mechanism of this modulation is not known. The data shown here suggest that endoglin might regulate TGF-␤ signaling by its direct interaction with the signaling receptors. Recently, betaglycan has been reported to inhibit TGF-␤ signaling by preventing the association between T␤RI and T␤RII, an interference mediated by the glycosaminoglycan modification of betaglycan (47). Unlike betaglycan, endoglin does not seem to affect the formation of T␤RI⅐T␤RII complexes (12). This, and the absence of glycosaminoglycan recognition motifs in the endoglin molecule, suggests a functional mechanism for endoglin distinct from that reported for betaglycan (47). Endoglin could regulate either ligand access to the signaling receptors through its extracellular interaction or the kinase activity of the signaling receptors through its cytoplasmic association. Supporting this hypothesis, we find that endoglin inhibits phosphorylation levels of the T␤RII in vivo and in vitro. Previous reports have demonstrated that several phosphorylation sites on T␤RII, including residues Ser 213 and Ser 409 , are both necessary for downstream signaling (37). However, phosphorylation of Ser 416 of T␤RII can lead to inhibition of receptor-mediated signaling (34,37). Thus, by altering the phosphorylation pattern of T␤RII, endoglin might regulate its kinase activity. Although the exact mechanism of endoglin action on T␤RII phosphorylation is not known, several possibilities can be proposed: (a) since endoglin is a substrate for T␤RII, it could compete for the autophosphorylation site(s) on the T␤RII kinase; (b) the autophosphorylation site(s) on the T␤RII could be masked by the direct association of endoglin with the T␤RII; (c) endoglin interaction with the T␤RII might induce a conformational change that affects the activity/specificity of the kinase; and (d) when associated with T␤RII, endoglin could recruit phosphatases able to dephosphorylate T␤RII. On the other hand, endoglin was found to affect not only the phosphorylation status of T␤RII but also that of T␤RI. Interestingly, endoglin clearly increased the phosphorylation levels of T␤RI (Fig. 8). This phosphorylation occurs, at least, within the juxtamembrane and GS regions (amino acids 146 -207) of T␤RI, where previous reports have described several amino acid residues as targets for serine/threonine phosphorylation (3,4). Thus, ligand-induced phosphorylation of the serines and threonines in the GS region ( 185 Thr-Thr-Ser-Gly-Ser-Gly-Ser-Gly 192 ) of T␤RI by T␤RII is required for signaling (35,48,49). T␤RII is also able to phosphorylate Ser 165 , located in the juxtamembrane region of T␤RI, and mutation of this Ser 165 results in potentiation of TGF-␤Ϫmediated growth inhi-FIG. 10. Schematic representation of a possible model of endoglin interaction with T␤RI and T␤RII kinases. Endoglin associates with T␤RII through at least one interaction site at region 437-558 of the extracellular domain and one interaction site at the cytoplasmic domain. Upon association, endoglin is phosphorylated by the T␤RII kinase, and the constitutive phosphorylation levels of T␤RII are diminished. Endoglin association with T␤RII results in an increased phosphorylation of T␤RI, which, in turn, phosphorylates Smad2. On the other hand, endoglin associates with T␤RI through at least two interaction sites at regions 437-558 and 26 -437 of the extracellular domain and at least one interaction site at the cytoplasmic domain. Upon association, endoglin is phosphorylated by the T␤RI kinase, and then the endoglin cytoplasmic domain is released from the T␤RI counterpart. Endoglin interactions with T␤RI and T␤RII also affect downstream signaling events, including the reporter gene expression of mix.2. Cysteine 582 at the juxtamembrane region of the extracellular domain forms disulfide linkages between endoglin monomers (see yellow circle). Phosphorylated residues are indicated by red circles. The heteromeric associations of T␤RI⅐T␤RII-endoglin have been omitted for simplification. bition, whereas the TGF-␤-induced apoptosis is reduced (48). Furthermore, mutation of Ser 172 or Thr 176 , within the same juxtamembrane region of T␤RI, selectively impairs TGF-␤-mediated growth inhibition but does not modify the TGF-␤-induced PAI-1 or fibronectin synthesis (50). Taken together, these data underscore the relevance of the phosphorylated residues on the signaling receptors as well as the specific TGF-␤ gene response analyzed. Therefore, it will be interesting to determine those residues whose phosphorylation status is modified by the presence of endoglin.
The presence of endoglin also appears to affect signaling downstream of the T␤RI⅐T␤RII complex. This is evidenced by the endoglin-dependent increased phosphorylation of Smad2 and stimulation of the mix.2 gene promoter-derived reporter responsiveness to TGF-␤ (Fig. 9). By contrast, previous studies have demonstrated an endoglin inhibition of the PAI-1 gene promoter responsiveness to TGF-␤ (11). This differential gene regulation suggests the involvement of gene-specific transcriptional components. This is in agreement with the fact that PAI-1 and mix.2 promoters display distinct transcriptional requirements. Thus, Smad4 is essential for the transcriptional activation of the mix.2 promoter in response to TGF-␤, whereas TGF-␤-induced activation of the human PAI-1 promoter can occur, circumventing the Smad4 requirement by increasing other signaling components (51). Together, these findings suggest that endoglin might differentially regulate the TGF-␤-dependent transcriptional machinery, namely the Smad family of proteins, which act in concert with their corresponding coactivators and corepressors (4,6,7). Consistent with this hypothesis, endoglin was found to increase Smad2 phosphorylation (Fig. 9, A and B); also, the endoglin homologue betaglycan displayed a modulatory effect on Smad2/3 phosphorylation (47).
Based on the data presented above, a model can be proposed to illustrate the role of endoglin in the TGF-␤ receptor complex (Fig. 10). Endoglin associates with T␤RII through its extracellular (at least one interaction site at region 437-558) and cytoplasmic domains, is phosphorylated by the T␤RII kinase, and diminishes the constitutive phosphorylation levels of T␤RII. Endoglin association with T␤RII results in an increased phosphorylation of T␤RI, which in turn phosphorylates Smad2. On the other hand, endoglin associates with T␤RI through its extracellular (at least two interactions sites at regions 437-558 and 26 -437) and cytoplasmic domains; upon activation of the T␤RI, endoglin is phosphorylated by the T␤RI kinase, and then the endoglin cytoplasmic domain is released from the T␤RI counterpart. All of these events result in the modification of specific downstream TGF-␤ gene responses.