Identification of Distinct Inhibin and Transforming Growth Factor β-binding Sites on Betaglycan

Betaglycan is a co-receptor that mediates signaling by transforming growth factor β (TGFβ) superfamily members, including the distinct and often opposed actions of TGFβs and inhibins. Loss of betaglycan expression, or abrogation of betaglycan function, is implicated in several human and animal diseases, although both betaglycan actions and the ligands involved in these disease states remain unclear. Here we identify a domain spanning amino acids 591–700 of the betaglycan extracellular domain as the only inhibin-binding region in betaglycan. This binding site is within the betaglycan ZP domain, but inhibin binding is not integral to the ZP motif of other proteins. We show that the inhibin and TGFβ-binding residues of this domain overlap and identify individual amino acids essential for binding of each ligand. Mutation of Val614 to Tyr abolishes both inhibin and TGFβ binding to this domain. Full-length betaglycan V614Y, and other mutations, retain TGFβ binding activity via a distinct site, but are unable to bind inhibin-A. These betaglycan mutants fail to mediate inhibin antagonism of activin signaling but can present TGFβ to TβRII. Separating the co-receptor actions of betaglycan toward inhibin and TGFβ will allow the clarification of the role of betaglycan in disease states such as renal cell carcinoma and endometrial adenocarcinoma.

Betaglycan is a co-receptor for members of the TGF␤ 3 superfamily, with vital roles in mediating and regulating signaling of diverse superfamily members. The importance of betaglycan is highlighted by loss of expression or mutation of betaglycan in various human diseases. For example, there is a strict correlation between a loss of betaglycan expression in human patients and development of both renal cell carcinoma and endometrial adenocarcinoma (1,2). These results also highlight the difficulty in understanding the role of betaglycan in different cellular contexts, because betaglycan can impinge on signaling by so many TGF␤ superfamily members. In the studies cited above, loss of betaglycan in renal cell carcinoma has been proposed to impair TGF␤-2 signaling, because impaired TGF␤ signaling due to loss of T␤RII expression has also been detected in these carcinomas (1). In contrast, loss of betaglycan in endometrial adenocarcinoma is speculated to involve betaglycan regulation of inhibin action, due to the fact that the inhibin ␣-subunit is also disrupted in endometrial adenocarcinomas (2). However, these conclusions will remain primarily correlative until methods are developed to dissect the various co-receptor roles of betaglycan.
Betaglycan directly binds TGF␤ isoforms, which is functionally vital in the case of TGF␤-2. Betaglycan facilitates TGF␤-2 binding to the TGF␤ type II receptor T␤RII, leading to the formation of a ternary complex containing TGF␤-2, T␤RII, and betaglycan (3). Once this complex has formed, T␤RI (ALK5) then binds to TGF␤-2 and T␤RII, displacing betaglycan. T␤RII then phosphorylates T␤RI, which, in turn, propagates TGF␤ signals into the cell via phosphorylation of intracellular Smad proteins (reviewed in Ref. 4). Expression of betaglycan in cells that normally lack this co-receptor leads to an increase in cellular sensitivity to TGF␤-2 and a concomitant increase in TGF␤-2 binding to T␤RII and T␤RI (5). Because TGF␤-2 does not interact with T␤RII or T␤RI with high affinity in most cells, betaglycan has as essential, although indirect, role in promoting TGF␤-2 signaling.
Betaglycan also directly binds inhibin isoforms and promotes inhibin binding to type II receptors (6), primarily the activin type II receptors ActRII and ActRIIB and the BMP type II receptor BMPRII. However, the functional consequences of inhibin binding to betaglycan are diametrically opposed to the events that occur subsequent to TGF␤ binding. Inhibin-betaglycan-type II receptor complexes are very stable and, because inhibins do not interact with type I receptors, these complexes do not result in Smad phosphorylation and downstream signaling. Instead, this inhibin-betaglycan-type II receptor complex sequesters type II receptors, thereby forestalling their interactions with signaling ligands such as activins or BMPs. Expression of betaglycan in cells that respond poorly to inhibin leads to dramatic increases in inhibin sensitivity (6,7). Because inhibins bind with low affinity to type II receptors in the absence of betaglycan, betaglycan has as essential, direct role in potentiating inhibin antagonism of activin and BMP signaling.
By functioning as a co-receptor for both TGF␤s and inhibins, betaglycan can potentially affect signaling by the majority of TGF␤ superfamily ligands. In addition, betaglycan may inhibit TGF␤ signaling in certain systems, depending on the cell-specific glycosaminoglycan modification of betaglycan (8). In vivo studies demonstrate the functional role of betaglycan as a co-receptor is clearly vital. Mice lacking betaglycan die during embryogenesis with heart and liver defects (9), whereas experimental targeting of betaglycan during development has been shown to disrupt mesenchyme formation in the heart (10) and branching morphogenesis in the lung (11). In the adult, loss of betagly-can is associated with tumorigenesis (1,2), and soluble betaglycan ECD proteins have been used to disrupt tumor progression (12)(13)(14).
Rat betaglycan is an 853-amino acid protein with a large extracellular domain, a single transmembrane domain, and a short intracellular domain lacking obvious signaling motifs (3). The cytoplasmic domain of betaglycan can be removed without grossly altering co-receptor function, although this domain can be bound by ␤-arrestin (15) and the PDZ domain protein GIPC (16). The extracellular domain of betaglycan contains sites for attachment of heparin and chrondoitin sulfate glycosaminoglycan chains as well as sites for N-glycosylation (17), and binds TGF␤ superfamily ligands. Betaglycan has at least two independent binding sites for TGF␤s, which have been roughly delineated within the two halves of the betaglycan ECD (17,18). Individual deletion of either half of the betaglycan ECD does not prevent betaglycan from mediating TGF␤-2-induced phosphorylation of Smad2, suggesting that each half of the betaglycan ECD can independently function to present TGF␤-2 to T␤RII (19).
In contrast to TGF␤s, inhibins may bind only to the membrane proximal half of the betaglycan ECD (19). This portion of the betaglycan ECD contains only one recognizable feature: a ZP domain, a motif found in several extracellular proteins, including endoglin, the ␣and ␤-tectorins of the inner ear, the zona pellucida proteins (ZP1, ZP2, and ZP3), renal uromodulin, and DMBT1 (20). ZP domains are composed of around 260 amino acids, including eight conserved cysteines, which are thought to mediate a related three-dimensional structure to these domains (20). The ZP domains of ZP2, ZP3, and uromodulin are responsible for polymerization of these proteins into filaments (21), although less is known about the role the ZP domain plays in transmembrane proteins, such as betaglycan.
Despite the crucial role of betaglycan as a co-receptor for TGF␤-2 and inhibins, the inhibin and TGF␤-binding sites within betaglycan have not been precisely mapped. The two halves of the betaglycan ECD are not highly related to each other, suggesting their ligandbinding sites may have important differences. Furthermore, the role of ZP domains in binding TGF␤ superfamily ligands has not been investigated. To separate betaglycan co-receptor actions toward TGF␤s and inhibins, there is a need to fully delimitate the sites of betaglycan involved in inhibin and TGF␤ binding, and to develop mechanisms to specifically disrupt either the TGF␤ or inhibin coreceptor actions of betaglycan. Therefore, we have utilized truncation and site-directed mutagenesis studies to delineate the inhibinbinding domain on betaglycan and identify amino acids involved in inhibin and TGF␤ binding. Here we map the inhibin-and TGF␤binding sites within the membrane proximal domain of betaglycan, demonstrate that they overlap, and identify individual amino acids essential for the binding of both ligands. Through binding and functional studies we have generated betaglycan mutants that bind TGF␤s but not inhibins, allowing experimental separation of betaglycan co-receptor actions toward TGF␤s and inhibins.

EXPERIMENTAL PROCEDURES
Materials-NuPAGE gels, molecular weight markers, and nitrocellulose blotting membranes were purchased from Invitrogen. Recombinant human TGF␤-1 and TGF␤-2 were purchased from Peprotech (Rocky Hill, NJ). Recombinant human activin-A and inhibin-A were produced using an inhibin-␣ and -␤A expressing stable Chinese hamster ovary cell line. Activin-A and inhibin-A were each purified by consecutive heparin affinity, gel filtration, and reverse phase high pressure liquid chromatography steps. Final purity was found to be Ͼ99% as accessed by SDS-PAGE. 125 I-Inhibin-A and 125 I-TGF␤-1 were prepared using a modified chloramine-T method (22). Horseradish peroxidaselinked anti-mouse IgG, chemiluminescent substrate (Supersignal TM ), and TMB-turbo peroxidase (enzyme-linked immunosorbent assay) substrate were obtained from Pierce. Perfectin transfection reagent was purchased from Gene Therapy Systems (San Diego, CA) and SuperFect transfection reagent was purchased from Qiagen (Valencia, CA). Anti-FLAG antibodies (M2) were purchased from Sigma and anti-c-myc antibodies (9E10) were purchased from Santa Cruz Biotechnology (Santa Cruz, CA). High molecular weight polyethyleneimine (PEI) was purchased from Sigma. The betaglycan constructs used in this study were expressed from the pcDNA3 expression vector (Invitrogen).
Mutagenesis of Betaglycan-The internal truncation series of betaglycan constructs were the generous gift of Dr. J. Massague (17). The N-terminal-deleted betaglycan construct ⌬26-575 (which we refer to as BG-(576-853) was a generous gift from Dr. M. O'Connor-McCourt (18). The betaglycan ⌬ZP construct was generated by deleting residues 456-733 of rat betaglycan. The BG/DMBT1 ZP chimera was made by replacing residues 456-733 of rat betaglycan with residues 362-561 of the human DMBT1 cDNA. Most other betaglycan mutants were initially constructed using the rat betaglycan-⌬26-575 cDNA as a template for overlap PCR-based mutagenesis. Standard molecular biology techniques were used to purify and clone mutant constructs into the expression vector pcDNA3. All mutants were fully sequenced before use.
Transfection of Betaglycan and Betaglycan Mutants in HEK293T Cells-HEK293T cells were grown in 5% CO 2 in a 37°C humidified incubator in complete Dulbecco's modified Eagle's medium (DMEM) containing 10% bovine calf serum. Transfection of myc-tagged wildtype or mutant betaglycan constructs was performed using a modified high molecular weight PEI protocol, as previously described (23,24). Briefly, HEK293T were plated at 1 ϫ 10 5 cells per well in 24-well plates coated with poly-D-lysine. After recovery overnight, cells were transfected with a total of 500 ng of DNA per well. DNA was diluted to a final volume of 10 g/ml in serum-free DMEM. PEI (frozen as a 10 mg/ml stock) was diluted to 1 mg/ml in H 2 O and added to the DNA solution to a final concentration of 20 g/ml. The DNA/PEI solution was vortexed and incubated at room temperature for 10 min. During this incubation, media was removed from the plates and 250 l of serum-free DMEM was added to each well. After 10 min, 50 l of the DNA-PEI complexes were added directly to each well and plates were incubated for 4 h at 37°C in 5% CO 2 . After 4 h, wells were aspirated and 300 l of DMEM containing 10% fetal bovine serum was added. Plates were incubated at 37°C in 5% CO 2 to allow cells to recover and express protein for 48 h prior to assay. For co-expression of betaglycan and ActRII, equal amounts of betaglycan and ActRII cDNAs were used and cells were transfected using Perfectin reagent according to the manufacturers instructions.
Detection of Surface-expressed Betaglycan-A cell surface enzyme-linked immunosorbent assay was performed to ensure expression of betaglycan constructs. This method is a modification of a previously described protocol (25). Briefly, 48 h after transfection, intact cells were chilled on ice, washed in cold binding buffer (12.5 mM Hepes, pH 7.4, 140 mM NaCl, and 5 mM KCl) supplemented with 0.1% bovine serum albumin, 5 mM MgSO 4 , 1.5 mM CaCl 2 and blocked with binding buffer containing 3% bovine serum albumin for 1 h. Cells were then incubated with an anti-myc (9E10) antibody (1 g/ml in binding buffer with 1% bovine serum albumin) for 2 h, washed three times in binding buffer, and finally incubated with a 1:2000 dilution of peroxidase-conjugated anti-mouse antibody for 2 h. Cells were washed three times in binding buffer and anti-myc antibody binding was quantitated using TMB-turbo peroxidase sub-strate. Peroxidase activity was quenched using 0.18 N H 2 SO 4 and absorbance was measured at 405 nm. Expression data were analyzed and graphed using the Prism software package (version 2.0 from GraphPad Software (San Diego, CA)).
Inhibin-A and TGF␤-1 Binding to Betaglycan and Betaglycan Mutants-Inhibin and TGF␤ binding was measured in intact cells. 48 h post-transfection, cells were washed in binding buffer and then incubated in binding buffer with 2 ϫ 10 5 cpm/well of 125 I-inhibin-A or 125 I-TGF␤ tracer and increasing concentrations of unlabeled ligand.
Plates were incubated for 2 h with gentle rocking and then wells were washed three times in binding buffer to remove free tracer. Cells were solubilized in 500 l of 1% SDS and 125 I-inhibin-A or 125 I-TGF␤-1 bound in each well were determined using a ␥-counter. Binding data were analyzed and graphed using the Prism program.
Cross-linking of Inhibin-A or TGF␤-1 to Betaglycan and Betaglycan Mutants-For cross-linking, HEK293T cells were plated on six-well plates coated with poly-D-lysine at a density of 5 ϫ 10 5 cells per well. Cells were transfected with 2 g of DNA/well with the indicated constructs using Perfectin according to the manufacturers instructions. 48 h post-transfection, cells were incubated with 5 ϫ 10 5 cpm/well 125 I-inhibin-A or 125 I-TGF␤-1 for 4 h at room temperature in binding buffer with gentle rocking. Cells were washed in binding buffer, resuspended in cross-linking buffer (0.5 mM disuccinylsuberate in HDB), and incubated 30 min on ice. Cross-linking reactions were quenched with TBS (50 mM Tris-HCl, pH 7.5, 150 mM NaCl) and cells were solubilized in TBS containing 1% Nonidet P-40, 0.5% deoxycholate, and 2 mM EDTA and subjected to immunoprecipitation using anti-Myc antibodies as indicated. Immune complexes were analyzed by SDS-PAGE and autoradiography.
Transfection and Luciferase Assays in AtT20 Cells-AtT20 cells were grown in 5% CO 2 in a 37°C humidified incubator in complete DMEM containing 10% bovine calf serum. For transfection, cells were plated at 2 ϫ 10 5 cells/well in 12-well plates and allowed to recover overnight. The next day cells were transfected with SuperFect reagent according to the manufacturer's instructions using 1 g of DNA and 1.2 l of Super-Fect/well. Cells were transfected with the 3TP-Lux reporter plasmid, CMV-␤-galactosidase, and either pcDNA3 vector or rat betaglycan construct. Cells were allowed to recover for 24 h and were then treated with vehicle, 1 nM activin-A, or 1 nM activin-A and 5 nM inhibin-A. After 24 h of treatment, cells were harvested and luciferase and ␤-galactosidase activities were measured. Luciferase activity was normalized to ␤-galactosidase activity. Data were analyzed and graphed using the Prism software package.
Transfection and Luciferase Assays in HepG2 Cells-HepG2 cells were grown in 5% CO 2 in a 37°C humidified incubator in complete ␣ modification of Eagle's media with 10% bovine calf serum. For transfection, HepG2 cells were plated at 5 ϫ 10 4 cells per well in 24-well plates and allowed to recover overnight. Cells were transfected with the 3TP-Lux reporter plasmid, CMV-␤-galactosidase, and pcDNA3 vector or rat betaglycan construct as shown. Transfections were performed using SuperFect reagent and a total of 500 ng of DNA and 1 l of SuperFect/ well. Cells were allowed to recover for 24 h and then treated with activin-A (500 pM or 1 nM) and various doses of inhibin-A as shown. After 16-20 h of treatment, cells were harvested in solubilization buffer (1% Triton X-100, 25 mM Hepes, pH 7.8, 15 mM MgSO 4 , 5 mM EGTA) and luciferase and ␤-galactosidase activities were measured. Luciferase activity was normalized to ␤-galactosidase activity. Data were analyzed and graphed using the Prism software program and data are presented as percent inhibition.

Identification of a Single, Functional Inhibin-binding Region of
Betaglycan-Betaglycan is able to bind both TGF␤s (3) and inhibins (6). Two distinct TGF␤-binding regions have been localized to the membrane proximal (spanning residues 410-781) (26,27) and membrane distal (spanning residues 1-410) (17) halves of the betaglycan ECD. Inhibin-A, in contrast, binds only to the membrane proximal half of the betaglycan ECD (19). To extend these findings, we tested the ability of additional betaglycan deletion mutants to form complexes with inhibin-A in the presence or absence of ActRII and to mediate inhibin antagonism of activin signaling.
All of the deletion mutants examined were expressed at the cell surface at levels approximating wild-type betaglycan (Fig. 1B). Removal of the cytoplasmic tail (⌬811-853), or mutation of the glycosaminoglycan attachment sites in the betaglycan ECD (S535A/S546A), did not affect inhibin binding (Fig. 1A). Similarly, each of the mutants with deletions through the membrane distal region, except betaglycan-⌬287-409, bound inhibin-A robustly (Fig. 1A). The apparent decrease in inhibin binding observed for the ⌬287-409 mutant has been described previously with respect to TGF␤ binding (19) and may relate to downstream folding effects. As expected, deletion of the membrane proximal region of betaglycan-⌬499-783 completely abrogated inhibin-A binding (Fig. 1A). We next utilized cross-linking assays to examine inhibin complex formation with the betaglycan deletion mutants alone (Fig. 1C), or when co-expressed with ActRII (Fig. 1D). In support of the binding studies ( Fig. 1A), immunoprecipitation of cross-linked complexes with a myc antibody directed against the epitope tag on betaglycan confirmed the importance of the C-terminal, membrane proximal half of the betaglycan ECD for inhibin binding. Deletion of residues 45-199, 200-285, 287-409, 811-853, or the glycosaminoglycan attachment sites did not prevent inhibin cross-linking, although complex formation was substantially lower with the ⌬287-409 mutant (Fig. 1C). In contrast, deletion of residues 499-783 abolished cross-linking. A pattern of weak binding was observed to the betaglycan core protein (ϳ110 kDa for wild-type betaglycan), which lacks glycosaminoglycan chains, and stronger binding to the fully processed betaglycan, which runs as a smear above 200 kDa (Fig. 1C). Consistent with the cooperative nature of inhibin binding to betaglycan and type II receptors, co-expression of ActRII in this system increased the total levels of cross-linked complexes (compare Fig. 1, C and D). ActRII did not, however, enable betaglycan-⌬499-783 to form a cross-linked complex with inhibin-A (Fig.  1D, lane 6). These data support the conclusion that betaglycan-⌬499-783 lacks residues necessary for inhibin binding to betaglycan and formation of the inhibin-ActRII-betaglycan complex.
We previously described the ability of activin to suppress ACTH secretion by AtT20 corticotrope cells and showed that inhibin could not antagonize this activin response (28). However, when AtT20 cells were transfected with betaglycan, the activin actions in these cells become sensitive to inhibin blockade (6). To test which betaglycan mutants could confer inhibin sensitivity, we co-transfected AtT20 cells with the various betaglycan constructs and the activin-responsive 3TP-Lux luciferase reporter and determined the ability of inhibin to block activin signaling. In cells transfected with vector, activin-A induced a 2.5-fold increase in luciferase activity that was unaffected by 5 nM inhibin (Fig.  1E). In contrast, in cells expressing wild-type betaglycan, this dose of inhibin-A was able to antagonize activin signaling (Fig. 1E). Expression of betaglycan constructs that bound inhibin: i.e. betaglycan-⌬44-199, -⌬200-285, -⌬287-409, -⌬811-853, or betaglycan S535/546A, similarly endowed cells with inhibin sensitivity. Of all the mutants tested, only betaglycan-⌬499-783 failed to promote inhibin antagonism of activin signaling (Fig. 1E, sixth set).
The ZP Domain of Betaglycan Is Required for Inhibin Binding-The inhibin-binding region of betaglycan we identified contains a ZP domain (residues 456-733), and we therefore examined whether this domain was necessary and sufficient for inhibin binding. Initially, we removed the entire ZP domain of betaglycan to generate a ⌬ZP betaglycan construct and, subsequently, replaced the betaglycan ZP domain with the ZP domain of DMBT1 (deleted in malignant brain tumor-1) to generate a betaglycan/DMBT1(ZP) chimera. DMBT1 was selected because the DMBT ZP domain has relatively high conservation with the betaglycan ZP domain. The betaglycan construct lacking the entire ZP domain (BG ⌬ZP) and the betaglycan/DMBT1(ZP) chimera each failed to bind inhibin-A directly (Fig. 2, A and C), despite being expressed at the cell surface at levels only somewhat reduced when compared with wild-type betaglycan (Fig. 2B). We also examined the ability of the betaglycan ZP mutants to participate in binding inhibin in complex with ActRII. HEK293T cells were transfected with combinations of FLAGtagged ActRII and wild-type or ZP disrupted betaglycan-myc constructs. Following 125 I-inhibin-A binding, complexes were cross-linked and isolated by immunoprecipitation with a myc antibody. As expected, co-expression of ActRII with betaglycan ⌬ZP resulted in inhibin cross-FIGURE 1. Identification of the juxtamembrane region of the betaglycan ECD as necessary for inhibin binding and betaglycan functional co-receptor action. A, inhibin-A binding to full-length betaglycan or betaglycan deletion constructs. 293T cells were transiently transfected with empty vector or full-length wild-type betaglycan or betaglycan constructs as shown. Cells were treated with 125 I-inhibin-A, washed, and binding was measured as described under "Experimental Procedures." Data are presented as counts per minute (c.p.m.) bound. B, cell surface expression of full-length betaglycan or betaglycan truncation constructs. The same transiently transfected 293T cells as in A were assayed for betaglycan expression using the N-terminal myc epitope tag as described under "Experimental Procedures." C, covalent cross-linking of 125 I-inhibin-A to full-length betaglycan or betaglycan truncation constructs. 293T cells were transiently transfected with empty vector or full-length wild-type betaglycan or betaglycan constructs as shown for 48 h were cross-linked to 125 I-inhibin-A, and betaglycan complexes were immunoprecipitated with anti-myc antibodies, resolved by SDS-PAGE, and visualized by autoradiography as described under "Experimental Procedures." D, cooperative cross-linking of 125 I-inhibin-A to 293T cells transiently transfected with ActRII and full-length betaglycan or betaglycan truncation constructs. 293T cells were transiently transfected with ActRII and wild-type betaglycan or betaglycan truncations as shown for 48 h and cross-linked to 125 I-inhibin-A, and betaglycan-ActRII complexes were immunoprecipitated with anti-myc antibodies, resolved by SDS-PAGE, and visualized by autoradiography as described under "Experimental Procedures." E, functional assay for betaglycan potentiation of inhibin antagonism. AtT20 cells were transiently transfected with an activin responsive luciferase reporter (3TP-Lux), CMV-␤Gal, and either vector or wild-type betaglycan or betaglycan truncation as shown. After recovery, cells were treated with vehicle (white bars), 1 nM activin-A (black bars), or 1 nM activin-A plus 5 nM inhibin-A (gray bars). Luciferase and ␤-galactosidase activities were measured as described under "Experimental Procedures." Luciferase activity is shown in arbitrary units, normalized to ␤-galactosidase activity.
linking only to ActRII (Fig. 2C, lane 5). Unexpectedly, co-expression of ActRII with the betaglycan/DMBT1(ZP) chimera revealed inhibin cross-linking to both ActRII and a band consistent with the betaglycan/ DMBT1(ZP) chimera protein (Fig. 2C, lane 6). Higher molecular weight forms of this complex would not be expected, as the glycosaminoglycan attachment sites of betaglycan are within the betaglycan ZP domain, and thus absent from this chimera. This result implies that the betaglycan/DMBT1(ZP) chimera, while unable to bind inhibin directly, is able to participate in an ActRII-inhibin complex.
Our model of betaglycan co-receptor action hypothesizes that betaglycan promotes a stable complex between inhibin and ActRII, forestalling ActRII interaction with type I receptors. Viewed from this framework, the crucial question is not whether the betaglycan/DMBT1(ZP) chimera is present in the complex, but whether it can promote the interaction of inhibin with ActRII. HepG2 cells are an endogenous activin and inhibin responsive system that express all relevant receptors including betaglycan. Inhibin-A antagonizes responses to both activins and BMPs in this system with low potency, however, transfection of these cells with betaglycan increases inhibin potency dramatically (7). The extent to which transfected betaglycan mediated inhibin antagonism of activin in HepG2 cells was similar to that previously observed in AtT20 cells. To test for the co-receptor action of the betaglycan/ DMBT1(ZP) chimera, HepG2 cells were transfected with an activin responsive promoter (3TP-Lux) and either empty vector or betaglycan constructs. Cells were treated with activin-A and increasing doses of inhibin-A, and the potency of inhibin antagonism of activin signaling was measured. As shown in Fig. 2D, HepG2 cells transfected with vector exhibited an inhibin-A potency with an IC 50 in the low nanomolar range. Transfection with wild-type betaglycan led to a dramatic shift of over 2 orders of magnitude in inhibin-A potency, with an IC 50 in the low-picomolar range (Fig. 2D). This shift exemplifies the co-receptor function of betaglycan in these cells. HepG2 cells transfected with the betaglycan ⌬ZP construct or the betaglycan/DMBT1(ZP) chimera had inhibin potency similar to vector-transfected cells (Fig. 2D). Thus, despite participating in an inhibin complex with ActRII, the betaglycan/ DMBT1(ZP) chimera does not function as an inhibin co-receptor because it does not potentiate inhibin antagonism of activin signaling. These findings are consistent with the idea that residues within the ZP domain of betaglycan are required for betaglycan to function as an inhibin co-receptor. However, inhibin binding is not integral to the conserved ZP domain motif, as evidenced by the inability of the betaglycan/DMBT1(ZP) chimera to bind inhibin directly.
Identification of the Minimal Inhibin-binding site of Betaglycan-The ZP domain of betaglycan spans residues 456 -733. We next examined if the entire ZP domain of betaglycan was required to bind inhibin. First, we tested a construct previously used to examine TGF␤ binding (27). In this construct, betaglycan residues 1-575 have been truncated, removing nearly half of the ZP domain.
As shown in Fig. 3A, betaglycan-(576-853) binds robustly to inhibin-A. Inhibin binding is retained with the removal of an additional 5 or 15 residues (betaglycan-(591-853)), however, truncation beyond residues 591 (even to residues 596) abolishes inhibin binding ( Fig. 3A and data not shown). Thus, the N-terminal half of the ZP domain is dispensable for inhibin binding. Importantly, this boundary at residue FIGURE 2. An intact ZP domain is required for betaglycan to bind inhibin and potentiate inhibin antagonism of activin. A, inhibin-A binding to wild-type betaglycan or betaglycan mutants with impaired ZP domains. 293T cells were transiently transfected with empty vector (pcDNA3), full-length betaglycan, or betaglycan constructs with ZP domain deleted (BG ⌬ZP) or a betaglycan chimera containing the ZP domain from DMBT1 (BG/DMBT-ZP). Cells were treated with 125 I-inhibin-A, washed, and binding was measured as described under "Experimental Procedures." Data are presented as counts per minute (c.p.m.) bound. B, the same transiently transfected 293T cells as in A were assayed for betaglycan expression using the N-terminal myc epitope tag as described under "Experimental Procedures." C, covalent cross-linking of inhibin-A to wild-type betaglycan or betaglycan mutants with impaired ZP domains in the presence or absence of ActRII. 293T cells transiently transfected for 48 h with betaglycan mutants and ActRII as indicated were cross-linked to 125 I-inhibin-A and betaglycan-ActRII complexes were immunoprecipitated with anti-myc antibodies against the epitope tag on betaglycan, resolved by SDS-PAGE, and visualized by autoradiography as described under "Experimental Procedures." D, functional assay for potentiation of inhibin antagonism in HepG2 cells by betaglycan ZP domain mutants. HepG2 cells were transiently transfected with 3TP-Lux, CMV-␤Gal, and either vector or betaglycan mutants as shown. Cells were treated with vehicle or 1 nM activin-A in the presence of increasing doses of inhibin-A as shown and luciferase and ␤-galactosidase activities were measured as described under "Experimental Procedures." Luciferase activity is shown in arbitrary units, normalized to ␤-galactosidase activity. 591 is identical to the boundary found for TGF␤-1 binding (27). Interestingly, this lack of binding appears to be due to a failure of these betaglycan variants to be expressed on the cell surface (Fig. 3B). We next confirmed that the betaglycan truncation mutants that bound inhibin could also potentiate inhibin antagonism in the HepG2 system. HepG2 cells were transfected with 3TP-Lux and various betaglycan constructs and then treated with activin-A and increasing doses of inhibin-A. Consistent with the binding and expression data, betaglycan constructs (576-853), (581-853), and (591-853) shifted inhibin potency from the subnanomolar to the low-picomolar range. This is similar to the potencies observed with full-length betaglycan (compare Fig. 2D with Fig.  3C). Betaglycan constructs (601-853), (611-853), and (621-853) did not alter inhibin sensitivity from that observed in vector-transfected cells.
Taken together, these results highlight the importance of betaglycan residues 591-853 for inhibin binding. Deletion of residues distal to this domain have little effect on inhibin-A binding (Fig. 3A) or on betaglycan function as an inhibin co-receptor (Fig. 3C). These results do not, how-ever, conclusively demonstrate that residues immediately C-terminal to this junction are necessary for inhibin binding. Rather, the loss of detectable cell surface expression of constructs bearing further truncations suggest that these residues are probably required for correct folding or expression of this betaglycan region. Loss of both binding and betaglycan co-receptor functions are most likely a corollary to this expression effect.
Expression and Inhibin-A Binding of Betaglycan/DMBT1 Chimeras-The inability of the betaglycan/DMBT1(ZP) chimera to bind inhibin suggested that residues within the ZP domain that differ between betaglycan and DMBT1 are key mediators of inhibin binding. To avoid the effects on expression that we encountered in our truncation studies, and still address the importance of these residues for inhibin binding, we developed a scanning mutagenesis strategy based on betaglycan homology with DMBT1. We engineered more precise chimeras where we selected ϳ20 amino acid stretches within the betaglycan ZP domain and exchanged all residues that differed between betaglycan and DMBT1 with the corresponding residues from DMBT1. In all we generated six such chimeras: BG/DMBT1: 576-595, 596-614, 615-634, 635-654, 655-673, and 674-692. Each of these chimeras, except BG/DMBT1:615-634 and BG/DMBT1:635-654, was well expressed at the cell surface (Fig. 4B) suggesting that the insertion of DMBT1 residues within this region of betaglycan was less disruptive to protein folding than truncation of these sequences. Despite comparable expression, BG/DMBT1:576-595 bound inhibin at levels significantly lower than those observed with wild-type betaglycan-(576-853), whereas BG/DMBT1:596-614, BG/DMBT1:655-673, and BG/DMBT1:674-692 were unable to bind inhibin-A at all in this assay (Fig. 4A).
This series of betaglycan/DMBT1 chimeras were also tested for their ability to functionally potentiate inhibin action in HepG2 cells. As expected, wild-type betaglycan-(576-853) and BG/DMBT1:576-595 potentiated inhibin antagonism of activin signaling (Fig. 4C), increasing inhibin potency ϳ100-fold in these cells. In line with their inability to bind inhibin, most of the other BG/DMBT1 chimeras failed to function as inhibin co-receptors (Fig. 4C). Surprisingly, BG/DMBT1:674-692, which did not bind detectable levels of inhibin-A, increased inhibin potency ϳ10-fold in this assay system (Fig. 4C). To reconcile this discrepancy, we would speculate that this BG/DMBT1 chimera retains the residual capacity to bind low levels of inhibin-A. In transfected HepG2 cells the expression of BG/DMBT1:674-692 is clearly superphysiological, so even a low level of co-receptor action could lead to a measurable shift in inhibin potency. Alternately, this construct may retain some ability to bind inhibin cooperatively with ActRII.
Overall, these findings confirm that the inhibin-binding site of betaglycan begins between residues 576 and 595. This region can be further resolved to residues 591-595 as inhibin binding was not affected by deletion of residues 26-591 (Fig. 3A). Replacement of residues 591-595, within the context of the BG/DMBT1:576-595 chimera, decreases but does not abolish inhibin binding (Fig. 4A), but removal of these residues disrupts expression (Fig. 3B). This suggests that these residues are required for proper folding of an independent domain in the betaglycan ECD but are not directly essential for inhibin binding. By contrast, changing betaglycan residues 596-614 to the corresponding DMBT1 sequence results in a chimera that is expressed but does not bind inhibin or function as an inhibin co-receptor. Similar results are observed with other betaglycan/DMBT1 chimeras, suggesting that determinants for inhibin binding span at least the 97 residues covered by these chimeras (residues 596-692), although we cannot rule out the involvement of more N-terminal residues as well. The same transiently transfected 293T cells as in A were assayed for betaglycan expression using the N-terminal myc epitope tag as described under "Experimental Procedures." C, functional assay of betaglycan truncations for potentiation of inhibin antagonism in HepG2 cells. HepG2 cells were transiently transfected with 3TP-Lux, CMV-␤Gal, and either vector or betaglycan truncations as shown. Approximately 24 h later, cells were treated with vehicle or 1 nM activin-A in the presence of increasing doses of inhibin-A as shown and luciferase and ␤-galactosidase activities were measured as described under "Experimental Procedures." Luciferase activity is shown in arbitrary units, normalized to ␤-galactosidase activity.

Scanning Mutagenesis of the Inhibin-binding Site of Betaglycan Identifies Critical Functional Residues-
The large BG/DMBT1 chimeras, in which ϳ20 amino acids stretches were exchanged, proved to have disrupted inhibin binding and co-receptor function. Refining this strategy, we generated smaller variants, including point mutants, where shorter stretches of betaglycan were replaced with the corresponding regions of DMBT1. Each smaller chimera or point mutant was expressed in HEK293T cells and assayed in parallel for both cell surface expression and inhibin binding (Table 1). Of the smaller chimeras generated, BG/DMBT1:608-614, BG/DMBT1:615-620, and BG/DMBT1:635-638 mimicked their larger, parent chimeras and failed to bind inhibin-A. Individual point mutants were generated for each of the residues altered in these BG/DMBT1 chimeras. All 13 point mutants tested were expressed at the cell surface at levels approximating betaglycan-(576-853) ( Table 1). Of these mutants, 9 had moderate to severe effects on inhibin binding. Of particular note were mutants V614Y, A615D, and H619D, which bound inhibin poorly or not at all and S608V, which bound inhibin-A more robustly than betaglycan-(576-853). Five other mutations (G610S, E616L, V620L, F635L, and I637V) resulted in apparent decreases of 50-65% in inhibin-A binding to betaglycan. These findings are consistent with our having identified a binding interface within the ZP domain of betaglycan that is surface exposed and directly interacts with inhibins.
The inhibin-binding site we have identified is located within the reported membrane proximal half of the betaglycan ECD that binds TGF␤s. Betaglycan-(576-853) bound inhibin-A and TGF␤-2 with similar affinity, suggesting this site was equally capable of binding inhibins or TGF␤s (data not shown). Therefore, we investigated whether the mutations that we had identified that affected inhibin binding would also affect TGF␤ binding. HEK293T cells were transfected with the indicated betaglycan mutants and parallel binding experiments were performed with 125 I-inhibin-A and 125 I-TGF␤-1 (Fig. 5). Like inhibin-A, TGF␤-1 was unable to bind to the BG/DMBT1:615-620 chimera or to the V614Y point mutant (Fig. 5, A and B). Somewhat surprisingly, the BG/DMBT1:608-614 chimera, which incorporates the V614Y point mutation, retained some ability to bind to TGF␤ despite being unable to bind inhibin-A (Fig. 5, A and B). Other differences were that the E616L mutation was more disruptive for TGF␤-1 binding than for inhibin-A binding (Fig. 5, A and B), whereas the H619D and V620L mutations (Fig.  5, A and B) had little effect on TGF␤-1 binding but were moderately disruptive for inhibin-A binding. In addition, the S608V mutation that promoted inhibin binding to betaglycan had no effect on TGF␤-1 binding (Fig. 5, A and B). Covalent cross-linking of 125 I-inhibin-A and 125 I-TGF␤-1 to HEK293T cells expressing the betaglycan mutants, followed by immunoprecipitation and SDS-PAGE supported the binding results (Fig. 5, C and D). These results confirm that inhibin and TGF␤ interact FIGURE 4. Amino acids spanning region 595-695 are involved in inhibin binding to betaglycan and betaglycan co-receptor action toward inhibin. A, inhibin-A binding to betaglycan/DMBT1 chimeras. 293T cells were transiently transfected with wild-type betaglycan or the indicated betaglycan/DMBT1 chimera. Cells were treated with 125 Iinhibin-A, washed, and binding was measured as described under "Experimental Procedures." B, cell surface expression of betaglycan/DMBT1 chimeras. The same transiently transfected 293T cells as in A were assayed for betaglycan expression using the N-terminal myc epitope tag as described under "Experimental Procedures." C, functional assay in HepG2 cells of betaglycan/DMBT1 chimeras for potentiation of inhibin antagonism. HepG2 cells were transiently transfected with 3TP-Lux, CMV-␤Gal, and either vector or betaglycan/DMBT1 chimeras as shown. Cells were treated with vehicle or 1 nM activin-A in the presence of increasing doses of inhibin-A as shown and luciferase and ␤-galactosidase activities were measured as described under "Experimental Procedures." Luciferase activity is shown in arbitrary units, normalized to ␤-galactosidase activity.

TABLE 1 Mutations within the betaglycan (BG) ZP-domain affect cell surface expression and inhibin-A binding
Inhibin-A binding and cell surface expression of betaglycan-(576 -853) mutation constructs. 293T cells were transiently transfected with empty vector or betaglycan-(576 -853) mutations as indicated. Cells were assayed for inhibin binding and cell surface expression in parallel. For binding assays, transfected cells were treated with 125 I-inhibin-A, washed, and binding was measured as described under "Experimental Procedures." Cell surface expression of betaglycan mutants was assayed using the N-terminal myc epitope tag as described under "Experimental Procedures."

Construct
Cell surface expression Inhibin-A binding with the same binding site in the membrane proximal region of betaglycan (19), although there are differences in the contribution of individual amino acids to binding these two ligands. The selectivity of these mutants between inhibin and TGF␤ strongly suggests to us that we are altering surface residues involved in ligand interactions, and not inducing global changes in domain folding and structure. Binding assays were performed in the presence of both betaglycan variants and ActRII to investigate if these constructs could engage in

Identification of residues within the inhibin-binding site of betaglycan that are required for inhibin and TGF␤ binding and inhibin co-receptor function.
A, inhibin-A binding to betaglycan-(576-853) or betaglycan-(576-853) mutations. 293T cells were transiently transfected with the indicated betaglycan constructs. Cells were treated with 125 I-inhibin-A, washed, and binding was measured as described under "Experimental Procedures." B, TGF␤-1 binding to betaglycan-(576-853), or betaglycan-(576-853) mutations. 125 I-TGF␤-1 binding to the same betaglycan-transfected cells as in A were performed as described under "Experimental Procedures." C, cross-linking of 125 I-inhibin-A to betaglycan mutants. 293T cells transiently transfected with the indicated betaglycan-(576-853) mutants were cross-linked to 125 I-inhibin-A, and betaglycan complexes were immunoprecipitated with anti-myc antibodies, resolved by SDS-PAGE, and visualized by autoradiography as described under "Experimental Procedures." D, cross-linking of 125 I-TGF␤-1 to betaglycan mutants. 125 I-TGF␤-1 cross-linking to the same cells as in C was performed as described under "Experimental Procedures." E, cooperative inhibin-A binding to cells co-expressing betaglycan mutants and ActRII. 293T cells were transiently transfected with combinations of betaglycan mutants and ActRII as shown. 48 h later, cells were treated with 125 I-inhibin-A, washed, and 125 I-inhibin-A binding was measured as described under "Experimental Procedures." F, functional assay for betaglycan potentiation of inhibin antagonism in HepG2 cells. HepG2 cells were transiently transfected with 3TP-Lux, CMV-␤Gal, and either vector or betaglycan mutants as shown. Cells were treated with either vehicle or 1 nM activin-A in the presence of increasing doses of inhibin-A as shown and luciferase and ␤-galactosidase activities were measured as described under "Experimental Procedures." Luciferase activity is shown in arbitrary units, normalized to ␤-galactosidase activity. cooperative binding of inhibin. As shown in Fig. 5E, co-expression of ActRII and betaglycan-(576-853) did lead to an increase in total inhibin binding when compared with either ActRII or betaglycan-(576-853) alone, as previously reported (6,7). Betaglycan S608V behaved essentially like wild-type betaglycan-(576-853) in this assay (Fig. 5E). Other mutants showed a similar level of binding to inhibin alone, or in the presence of ActRII, suggesting that mutations affected both inhibin binding and cooperation with ActRII.
Three of the betaglycan point mutants with altered inhibin binding, S608V, V614Y, and A615D, were subsequently examined for their ability to function as inhibin co-receptors in HepG2 cells transfected with the activin reporter 3TP-luciferase. As observed previously, transfection of these cells with betaglycan-(576-853) increased inhibin potency more than 100-fold (Fig. 5F). In binding assays, betaglycan-(576-853): S608V appeared to bind inhibin more robustly than wild-type betaglycan-(576-853), but in this functional co-receptor assay these two betaglycan variants were indistinguishable (Fig. 5F). The ability of betaglycan-(576-853) mutants V614Y and A615D to potentiate inhibin activity, in contrast, did reflect their relative abilities to bind inhibin. Mutant V614Y does not bind inhibin, despite levels of cell surface expression similar to wild-type (Table 1), and was unable to increase inhibin potency above vector-transfected cells (Fig. 5F). Mutant A615D, however, consistently exhibited a very low level of inhibin binding and, when transfected into HepG2 cells increased inhibin potency 10-fold (Fig. 5F), somewhat less than wild-type betaglycan.
Overall, these results reveal that betaglycan-(576-853):V614Y does not bind inhibin-A or TGF␤-1 and does not function as an inhibin co-receptor. Larger betaglycan chimeras BG/DMBT1:608-614 and BG/DMBT1:615-620 also failed to alter inhibin potency in the HepG2 assay (data not shown), indicating that they too were impaired in their ability to function as inhibin co-receptors.
Characterization of Full-length Betaglycan with Mutations That Impair Inhibin Binding and Co-receptor Function-The three most disruptive betaglycan mutations, both for inhibin binding and co-receptor function, were BG/DMBT1:608-614, V614Y, and BG/DMBT1:615-620. These mutations were incorporated into full-length betaglycan to assess their effect on the co-receptor action of betaglycan toward both inhibins and TGF␤s. Inhibin-A bound to HEK293T cells transfected with wildtype betaglycan, but not to cells transfected with the full-length betaglycan mutants: BG/DMBT1:608-614, V614Y, or BG/DMBT1:615-620 (Fig. 6A). This important finding emphasizes that there is only one inhibin-binding site in betaglycan. Because TGF␤s can bind to both the membrane distal and membrane proximal halves of the betaglycan ECD, full-length betaglycan with mutations in the membrane proximal inhibin (and TGF␤)-binding site should still retain an intact TGF␤binding site in the membrane distal half of the ECD. As expected, each full-length variant retained detectable TGF␤-1 binding (Fig. 6B). TGF␤-1 binding to BG/DMBT1:608-614 was only slightly decreased compared with wild-type betaglycan (Fig. 6B), whereas betaglycan V614Y bound TGF␤-1 at about half the level of wild-type betaglycan (Fig. 6B). Full-length betaglycan BG/DMBT1:615-620 bound the least TGF␤-1 of the mutants tested (Fig. 6B). The reasons for these differences may involve residual binding in the inhibin-binding site or possibly interactions between the TGF␤-binding domains. Consistent with our previous truncation and chimera binding studies, inhibin-A was unable to form cross-linked complexes in cells transfected with any of the three full-length betaglycan mutants (Fig. 6C, right side), despite the fact that each of these mutants retained the ability to cross-link to TGF␤-1.
We also examined inhibin binding and cross-linking to the full-length betaglycan mutants in the presence of ActRII (Fig. 6, D and E). In cells expressing both wild-type betaglycan and ActRII, cross-linking identified inhibin complexed to ActRII (80 kDa), the betaglycan core protein (110 kDa), and the higher molecular mass forms of betaglycan (Ͼ200 kDa) (Fig. 6E, lane 3). As seen previously, full-length BG/DMBT1:608-614 does not cross-link to inhibin-A when expressed alone (Fig. 6E, lane  4). However, when co-expressed with ActRII, BG/DMBT1:608-614 forms a complex with inhibin-A (Fig. 6E, lane 5), reminiscent of the BG/DMBT1(ZP) chimera. Inhibin-A complexes are observed with both betaglycan and the betaglycan core protein. In contrast, full-length V614Y and BG/DMBT1:615-620 do not cross-link to inhibin-A, even when expressed with ActRII (Fig. 6E, lanes 7 and 9). Betaglycan core protein complexes are not observed, and there is no increase in the intensity of the inhibin-A-ActRII complex. These cross-linking results correspond closely to the binding results presented in Fig. 6D, and suggest that these betaglycan variants have lost the ability to independently bind inhibins. However, full-length BG/DMBT1:608-614 is still able to participate in complexes containing inhibins and type II receptors.
One goal of this study was to test the hypothesis that inhibin and TGF␤ co-receptor activities of betaglycan are separable. To conclusively determine whether we had generated a full-length betaglycan variant that would not function as an inhibin co-receptor, we performed functional assays in HepG2 cells. Full-length betaglycan increased inhibin potency by more than 100-fold in these cells (Fig. 6F ), as previously observed. Full-length BG/DMBT1:608-614 was clearly functionally impaired compared with wild-type betaglycan, but still shifted inhibin potency by at least 10-fold in this assay (Fig. 6F ). This activity would not be expected based only on direct measures of inhibin binding, but correspond well to the level of cooperative interaction observed with ActRII. Full-length BG/DMBT1:615-620 was entirely inactive as an inhibin co-receptor in this system (Fig. 6F ). These betaglycan mutants still bind TGF␤ through the N-terminal domain and should mediate TGF␤-2 phosphorylation of Smad proteins as previously shown (19). We conclude that we have identified betaglycan mutants that are still capable of binding TGF␤ isoforms, but which do not bind inhibins and have lost the ability to function as inhibin co-receptors.

DISCUSSION
Betaglycan is an important regulator of the access of the TGF␤ superfamily of ligands to their type I and type II receptors (29)(30)(31). Betaglycan is of particular interest because it can serve to either promote or inhibit the signaling of most superfamily members in acting as a co-receptor for TGF␤s and/or inhibins. In cells exposed to TGF␤s, betaglycan enhances TGF␤ signaling through the Smad 2/3 pathway. In cells exposed to inhibins and activins, in contrast, betaglycan can help to inhibit activin signaling through the Smad 2/3 pathway. In addition, betaglycan may also help to inhibit signaling through the Smad 1/5/8 pathway in cells exposed to inhibins and BMPs (6,7). In some tissues exposed to multiple ligands, such as bone (32,33) and ovary (34), betaglycan could contribute a balance of positive and negative inputs. Overall, the pleiotropic nature of betaglycan actions makes it difficult to interpret the role of this co-receptor in normal and disease physiology.
Through truncation, point mutation, and chimera studies we have identified a single inhibin-binding site in betaglycan, and demonstrated that this site resides within the conserved ZP domain motif in the membrane proximal region of the betaglycan ECD ( Figs. 1 and 2), in keeping with the findings of Esparza-Lopez et al. (19). Removal of the inhibinbinding site or replacement of the ZP domain with the homologous region of DMBT1 abolishes inhibin-A binding. Using further deletion FIGURE 6. Generation of full-length betaglycan mutants that bind TGF␤ but do not bind inhibin or function as inhibin co-receptors. A, inhibin-A binding to full-length wild-type betaglycan, or betaglycan mutants. 293T cells were transiently transfected with the indicated betaglycan constructs. Cells were treated with 125 I-inhibin-A, washed, and binding was measured as described under "Experimental Procedures." B, TGF␤-1 binding to full-length wild-type betaglycan, or betaglycan mutants. 125 I-TGF␤-1 binding to the same betaglycantransfected cells as in A was performed as described under "Experimental Procedures." C, cross-linking of 125 I-inhibin-A and 125 I-TGF␤-1 to full-length betaglycan mutants. 293T cells transiently transfected with the wild-type betaglycan or the indicated betaglycan full-length mutants were cross-linked to 125 I-inhibin-A or 125 I-TGF␤-1 as shown, and betaglycan complexes were immunoprecipitated with anti-myc antibodies, resolved by SDS-PAGE, and visualized by autoradiography as described under "Experimental Procedures." D, cooperative inhibin-A binding to cells co-expressing full-length betaglycan mutants and ActRII. 293T cells were transiently transfected with combinations of betaglycan mutants and ActRII as shown. Cells were treated with 125 I-inhibin-A, washed, and 125 I-inhibin-A binding was measured as described under "Experimental Procedures." E, complex formation between ActRII and full-length betaglycan mutants. 293T cells transiently transfected with full-length betaglycan mutants and ActRII as indicated were cross-linked to 125 I-inhibin-A and betaglycan-ActRII complexes were immunoprecipitated with anti-myc antibodies against the epitope tag on betaglycan, resolved by SDS-PAGE, and visualized by autoradiography as described under "Experimental Procedures." F, functional assay for betaglycan potentiation of inhibin antagonism in HepG2 cells. HepG2 cells were transiently transfected with 3TP-Lux, CMV-␤Gal, and either vector or full-length betaglycan mutants as shown. Cells were treated with either vehicle or 315 pM activin-A in the presence of increasing doses of inhibin-A as shown and luciferase and ␤-galactosidase activities were measured as described under "Experimental Procedures." Luciferase activity is shown in arbitrary units, normalized to ␤-galactosidase activity. studies and receptor chimeras we have mapped the N-terminal border of the inhibin-binding site to amino acids 591-595 of the betaglycan ECD, demonstrating that approximately three quarters of the betaglycan ECD and half of the betaglycan ZP domain are dispensable for inhibin binding and co-receptor function (Fig. 3). Our mapping of this border for the inhibin-binding site of betaglycan corresponds exactly with the previously identified border of the TGF␤-binding region (27), although we further document that the inactivity of further N-terminal truncations is due to improper expression of the resulting truncated receptors.
We utilized scanning mutagenesis to examine residues within this inhibin-binding site for their contribution to binding inhibins and TGF␤s. Replacing residues of betaglycan with the corresponding residues from the closely related ZP domain of DMBT1 allowed us to probe the extent of the inhibin-binding site of betaglycan and eventually identify individual amino acid changes that altered inhibin and TGF␤ binding.
We generated six ϳ20 amino acid chimeras, commencing at residue 576 and covering the remainder of the conserved region between the betaglycan and DMBT1 ZP domains. Of these six chimeras, only BG/DMBT1:576-595 bound inhibin-A and functioned as an inhibin co-receptor ( Fig. 4 and Table 1). The remaining five chimeras were functionally inactive, highlighting the importance of betaglycan residues throughout the region of 596-692 for inhibin binding. These constructs demonstrate that the inhibin-binding site of betaglycan extends at least throughout the remaining portion of the betaglycan ZP domain, although we cannot exclude that additional residues that lie beyond the ZP domain may also participate in inhibin binding or co-receptor action. In general it is thought that evolutionary conserved domains (such as the ZP domain) are independent structural and functional units, so it might be surprising if functional mapping of the inhibinbinding site of betaglycan did not correspond to the evolutionary conserved border of the domain. However, it should be noted that our results clearly demonstrate that inhibin binding activity resides within the C-terminal half of the betaglycan ZP domain, whereas the well conserved N-terminal half of the domain is completely dispensable both for folding of the remaining ZP region, protein expression, and function with regard to inhibin binding.
Within this inhibin-binding site we have identified individual residues required for proper folding and binding to inhibin-A and TGF␤-1, as well as residues required for co-receptor function for these ligands at this site (Table 1 and Fig. 5). This site binds inhibin-A and TGF␤-2 with approximately equal affinity. Our results indicate that, in general, amino acids in the inhibin-binding site of betaglycan also participate in TGF␤ binding. This is consistent with the concept that these amino acids constitute an exposed surface on the betaglycan ECD that is available for ligand interaction. To us, the apparent selectivity of many of these mutations between inhibins and TGF␤s further supports this concept, and argues against general effects on domain folding or structure. Within the inhibin-binding site of betaglycan, mutation of amino acids Ser 608 , Val 614 , Ala 615 , His 619 , and Val 620 have pronounced effects on inhibin-A binding. Of these residues, only Val 614 and Ala 615 are similarly critical for TGF␤-1 binding. Betaglycan mutant S608V, which bound inhibin-A more robustly than wild-type betaglycan-(576-853), had no effect on TGF␤-1 binding. Similarly, altering residues His 619 and Val 620 , whereas disruptive for inhibin-A binding to betaglycan, appeared to have minor effects on TGF␤-1 binding. The opposite selectivity was observed with mutation of Glu 616 , which significantly decreased TGF␤-1 binding without altering inhibin-A binding.
These discrepancies in ligand binding to the inhibin (and TGF␤)binding site are understandable when viewed as alteration of residues on the surface of a ligand interaction interface, especially given the low degree of sequence homology between inhibin and TGF␤ subunits. An understanding of how two widely divergent members of the TGF␤ superfamily can interact with the same protein surface on betaglycan will likely require structural studies, as well as mutagenesis of residues in the two ligands, once the residues of the ligands involved in the interaction have been identified.
When altered in the context of full-length betaglycan, these mutations affect both inhibin and TGF␤ binding. Incorporation of each of the three most disruptive mutations for inhibin binding (BG/DMBT1:608-614, BG/DMBT1:615-620 and betaglycan V614Y) into full-length betaglycan generated receptors unable to bind inhibin-A. Incorporating the V614Y mutation into betaglycan, or introducing the BG/DMBT1:615-620 changes, results in full-length variants that do not function as coreceptors for inhibins (Fig. 6). These mutants also have disrupted TGF␤ binding activity at the inhibin-binding site, but retained TGF␤ binding activity through the membrane distal half of the betaglycan ECD (19).
Interestingly, a region of possible homology to the inhibin-binding site we have characterized can be identified in the membrane distal half of the betaglycan ECD. Between residues Ser 142 and Leu 163 of betaglycan, 10 of the 13 residues we identified as being involved in inhibin and TGF␤ binding are repeated as in Sequence 1. Furthermore, Val 614 and Val 620 are conservatively exchanged for Leu 146 and Leu 163 . It is intriguing to speculate that these residues within the membrane distal half of the betaglycan ECD may support TGF␤ binding but not inhibin binding. Future analysis of the membrane distal TGF␤-binding site of betaglycan might begin by examining the importance of residues Ser 142 to Leu 163 .
Our working model of betaglycan co-receptor action hypothesizes that betaglycan binds inhibin and presents inhibin to type II receptors. Betaglycan does not interact with type II receptors directly, but inhibin (once bound by betaglycan) has increased interaction with type II receptors. Betaglycan, effectively, increases inhibin affinity for type II receptors, resulting in increased potency of inhibin as an antagonist. The data we present here complicate this model. We have identified several betaglycan mutants that do not directly bind inhibin in standard binding assays, but do interact with inhibin in cross-linking assays when coexpressed with ActRII. One possibility is that these betaglycan variants retain some low affinity for inhibin, and binding of inhibin is a cooperative interaction involving both betaglycan and type II receptors, which still do not directly interact. A second possibility is that in addition to binding inhibin, betaglycan and ActRII might interact with each other directly, perhaps only in the presence of inhibin. Thus, betaglycan mutants would participate in the inhibin-ActRII complex based on an interaction with ActRII that was not impaired by the mutation or replacement of the inhibin-binding site. It should be noted that these two possibilities are not necessarily mutually exclusive.
Careful examination of the data suggests a clear trend toward a more severe loss of functionality displayed in the direct binding assay and a less severe loss of functionality seen in co-receptor assays. This can be seen in the data with the small BG/DMBT1 chimeras and point mutants presented in Table 1 and Fig. 5. Similarly, full-length betaglycan constructs BG/DMBT1:608-614, BG/DMBT1:615-620, and betaglycan V614Y did not bind inhibin when expressed alone, but BG/DMBT1: 608-614 did weakly cross-link to inhibin when co-expressed with ActRII and displays moderate co-receptor action (Fig. 6). Betaglycan variants such as A615D have some cooperative ability to bind inhibin in the presence of ActRII and this accounts for their partial co-receptor functions. Other constructs, such as the BG/DMBT1 ZP domain chimera, were functionally inert despite the fact that these chimeras could be cross-linked to inhibin in the presence of ActRII. We speculate that this chimera can interact with ActRII and through this interaction can participate in a complex containing inhibin. However, it does not have the ability to cooperatively bind inhibin and therefore is inactive as a co-receptor.
One main aim of this study was to test if the dual role of betaglycan as a co-receptor for inhibin and TGF␤ isoforms could be separated. Data in Fig. 6 clearly demonstrates that full-length betaglycan V614Y and BG/DMBT1:615-620 variants have been converted into TGF␤-specific co-receptors.
An ability to separate the TGF␤ and inhibin co-receptor actions of betaglycan would be useful for studying inhibin and TGF␤ biology in various genetic systems. Mouse embryos lacking betaglycan die starting at E13.5 with heart and liver defects (9). Related developmental abnormalities are also observed if betaglycan action is disrupted with immunoneutralizing antibodies during mesenchyme formation and migration in the atrioventricular cushion in the heart (10), or when betaglycan expression is abrogated using antisense oligonucleotides during lung branching morphogenesis (11). These embryonic developmental effects are probably due to a disruption of TGF␤ signaling, particularly TGF␤-2. Loss of TGF␤ isoforms during development can result in similar effects. Inhibin null mice (with a disrupted ␣-subunit) are apparently normal during embryonic development, although they develop gonadal tumors and other fatal abnormalities after birth (35,36). Therefore, mice engineered with a betaglycan isoform such as full-length betaglycan V614Y or BG/DMBT1:615-620 that can mediate TGF␤-2 signaling, but not inhibin antagonism, would allow genetic exploration of the full role of betaglycan in inhibin biology.
Additionally, soluble forms of the betaglycan ECD have been expressed and utilized as a treatment for disrupting tumor progression (12)(13)(14). Indeed, systemic administration of a soluble betaglycan ECD protein suppresses tumor growth in both prostate and breast cancer models (13,14). However, administration of these betaglycan ECD proteins would also block the activity of inhibins, a significant drawback given that inhibin may be a tumor suppressor in some of these same systems (37). A betaglycan that specifically targets TGF␤ would be preferable in at least some cases. The use of a betaglycan ECD protein engineered to lack inhibin binding but retain TGF␤ binding (i.e. betaglycan V614Y or BG/DMBT1:615-620) would be ideal in these cases.
Conversely, it would also be beneficial to develop betaglycan variants that function as inhibin co-receptors but do not mediate TGF␤ effects. Based on the mutants presented here S608V, which increases inhibin binding but has no effect on TGF␤ binding and E616L, which selectively disrupts TGF␤ binding, could form the precursors for such reagents. However, these mutations would have to be incorporated with other betaglycan alterations (as yet undeveloped) that fully disrupt both of the TGF␤-binding sites in betaglycan before truly inhibin-specific betaglycan variants could be produced.
In conclusion, we have identified and disrupted the inhibin-binding site in the betaglycan ECD, which can bind both inhibins and TGF␤s. By mutating this site, we have, for the first time, dissociated the co-receptor functions of betaglycan for inhibins and TGF␤s. Utilizing these betaglycan mutants, we are now in a position to differentially disrupt inhibin and TGF␤ actions, thereby defining the roles these ligands play in development, and in diseases such as cancer.