Identification of an Inhibin Receptor in Gonadal Tumors from Inhibin α-Subunit Knockout Mice*

Inhibins and activins are dimeric proteins that are functional antagonists and are structurally related to the transforming growth factor-β (TGFβ) family of growth and differentiation factors. Receptors for activin and TGFβ have been identified as dimers of serine-threonine kinase subunits that regulate cytoplasmic proteins known as Smads. Despite major advances in our understanding of activin and TGFβ receptors and signaling pathways, little is known about inhibin receptors or the mechanism by which this molecule provides a functionally antagonistic signal to activin. Studies described in this paper indicate that an independent inhibin receptor exists. Numerous tissues were examined for inhibin-specific binding sites, including the developing embryo, in which the spinal ganglion and trigeminal ganglion-bound iodinated inhibin A. Sex cord stromal tumors, derived from male and female inhibin α-subunit-deficient mice, were also identified as a source of inhibin receptor. Abundant inhibin and few activin binding sites were identified in tumor tissue sections by in situ ligand binding using iodinated recombinant human inhibin A and125I-labeled recombinant human inhibin A. Tumor cell binding was specific for each ligand (competed by excess unlabeled homologous ligand and not competed by heterologous ligand). Based on these results and the relative abundance and homogeneity of tumor tissues versus the embryonic ganglion, tumor tissues were homogenized, membrane proteins were purified, and putative inhibin receptors were isolated using an inhibin affinity column. Four proteins were eluted from the column that bind iodinated inhibin but not iodinated activin. These data suggest that inhibin-specific membrane-associated proteins (receptors) exist.

The inhibin and activin ␤-subunits are 30% homologous to TGF␤ based on alignment of conserved cysteine amino acids. Crystal structures of two TGF␤ family members (TGF␤2 and osteogenic protein-2) have been solved, and the overall structural similarity between these proteins implies that a conserved topology exists between members of the superfamily (8,9). Proteins within the TGF␤ family span multiple species and appear to be central factors in bone formation and morphogenesis of embryos, and the genes that encode them may function as tumor suppressor genes (10).
Activin activity is inhibited by interaction with a bioneutralizing binding protein called follistatin (11). Follistatin binds both inhibin and activin through the ␤-subunit; however, the ability of follistatin to neutralize inhibin activity cannot be determined until a cellular activity is described for inhibin that is not confounded by activin. Follistatin is structurally complex and is highly conserved between species with two conserved amino acid differences between rat and human and 87% homology with the Xenopus follistatin homologue (12,13).
Cellular response to activin is transduced through two single membrane-spanning serine-threonine kinase subunits (3). The ligand binds a type II receptor (70 -75 kDa) that transphosphorylates a type I receptor (50)(51)(52)(53)(54)(55). The holo-receptor complex is then competent to initiate intracellular signaling cascades (3). Two ligand binding type II receptor genes have been identified (type RII and type RIIB receptor subunits), and four alternatively spliced variants of the type IIB receptor have been cloned (14). The isoforms differ by changes in an extracellular proline-rich region immediately preceding the transmembrane domain and in a region between the transmembrane domain and serine-threonine kinase domain. Inhibin is able to bind the type II receptor but not recruit a type I receptor (15)(16)(17). It is therefore likely that the ability of inhibin to inhibit activin action is based, in part, upon this dominant negative interaction with receptor subunits. Receptors for many of the individual ligands have been identified, and they have structural characteristics similar to those described for the activin receptor. Inhibin ␣-subunit is distinct within the TGF␤ superfamily because it is capable of heterodimeric assembly (with activin ␤-subunit) and is not able to homodimerize. The inhibin ␣-subunit is one of four proteins distantly related to the core ligands (activin/TGF␤/bone morphogenic protein). Other ligands with distant homology include Mullerian inhibiting substance (the receptor of which is analogous to the TGF␤ structure), growth differentiation factor-9 (the receptor of which has not been identified), and glial-derived nerve growth factor (the receptor of which includes a glycosylphosphatidyl inositol-anchored binding protein that presents the ligand to a tyrosine kinase receptor) (18 -20).
An alternative hypothesis is that a separate inhibin receptor or inhibin accessory protein exists that mediates an inhibinspecific signal. Supporting the hypothesis that an independent inhibin receptor exists, inhibin-specific binding sites have been identified on ovarian granulosa cells and testicular Leydig cells (21)(22)(23). The most compelling evidence indicative of an inhibin receptor or cell surface ancillary binding protein is the identification of an inhibin-specific protein complex in a hematopoietic cell line (K562) (16). Taken together, these data indicate that inhibin activity involves an inhibin-specific receptor. Therefore, studies were initiated to isolate the inhibin receptor.
Mice that are genetically deficient in the inhibin ␣-subunit are deficient in inhibin A and inhibin B and overexpress activin A and activin B (24,25). These inhibin knockout mice are normal during embryonic and early postnatal development. However, microscopic focal gonadal tumors of the mixed or incompletely differentiated sex cord stromal (granulosa/Sertoli cell) type develop in mice as early as 4 weeks of age. Eventually, 100% of these mice will have either unilateral or bilateral tumors and die of a cancer cachexia-like syndrome due to the activin secreted from the tumors (26). Upon investigation, these tumors were found to bind inhibin specifically, and data are presented identifying tumor-derived inhibin receptor proteins. Iodination of Ligands-rh-activin A and rh-inhibin A were iodinated by a modified lactoperoxidase method. Briefly, 5 g of ligand was diluted in 0.4 M sodium acetate, pH 5.6, and 0.5 nmol of Na 125 I (0.5 nmol/mCi on calibration date), 0.5 IU of lactoperoxidase, and 0.25 nmol of H 2 O 2 were added sequentially. The ligands were incubated at ambient temperature with intermittent vortexing for 5 min. The reaction was quenched with 450 l of phosphate-buffered saline ϩ 0.05% Tween 20 ϩ 0.5% bovine serum albumin (Intergene, Purchase, NY). A 10-l aliquot of the precolumn fraction was removed for trichloroacetic acid precipitation. Free iodine was removed using Sephadex G-10 column chromatography (PD-10, Pharmacia Biotech Inc.). The specific activity of the ligands was approximately 100 Ci/g (range, 89 -102 Ci/g). To verify that the iodinated inhibin and activin used in these studies were active, ligands were assayed for bioactivity by perifusion of dispersed rat pituitary cells and quantitation of FSH␤ mRNA (27). Both ligands retained full biological activity (data not shown). Moreover, activin was able to bind type RII receptor expressed in a rat pituitary cell line (␣T3), indicating that this ligand would be able to bind its receptor in the eluate if it existed (data not shown).
In Situ Ligand Binding-In situ ligand binding was performed as described previously using rat embryos collected on E13, E15, and E17 or control ovaries and gonadal tumors collected from inhibin ␣-subunit knockout mice (22,28). Briefly, 12-m cryocut tissue sections were incubated for 3 h at room temperature in blocking buffer: Dulbecco's modified Eagle's medium:F-12 (1:1), 20 mM HEPES, 0.05% cytochrome C, 0.3% bovine serum albumin, 0.01 mg/ml phenylmethylsulfonyl fluoride, 0.01% bacitracin, 0.4 g/ml leupeptin. Slides were then incubated at room temperature overnight in the same buffer containing 40 pM 125 I-rh-inhibin A, 40 pM 125 I-rh-activin A or in the presence of 40 nM excess homologous ligand (to define nonspecific background) or heterologous ligand (to define low affinity binding to heterologous ligands). The slides were washed in phosphate-buffered saline (two times for 10 min each); fixed in 3.7% formalin, 2% glutaraldehyde (10 min); rinsed in

FIG. 1. In situ ligand binding autoradiographic images following incubation of rat embryo tissue sections with 125 I-rh-inhibin A.
Rat embryos were analyzed on E13, E15, and E17 of development. Top row, sections were incubated with 40 pM 125 I-rh-inhibin A, and representative sections depicting areas of hybridization are shown. Middle row, sections adjacent to those showing inhibin binding were incubated with 40 pM 125 I-rh-inhibin A plus 40 mM unlabeled inhibin A to demonstrate specificity of the ligand binding. Bottom row, sections adjacent to those showing inhibin binding were incubated with 40 pM 125 I-rh-inhibin A plus 40 nM unlabeled activin A to determine whether the binding was specific for inhibin. Inhibin-specific binding sites were localized to the trigeminal ganglion (TG) and spinal ganglion (SG). The liver (L), spinal cord (SC), dermis (D), and brain (B) binding was nonspecific based on the lack of competition of the labeled ligand with either inhibin or activin. 3-5 embryos (all from the same dam) were examined for each day.
water (four times for 1 s each); and allowed to dry. Dry slides were exposed to autoradiographic film for 1-14 days.
Isolation of Membrane Proteins-20 -25 grams of gonadal tumor tissue isolated from adult male and female inhibin ␣-subunit knockout mice were mechanically homogenized in 0.15 M NaCl plus a mixture of protease inhibitors (aprotinin (10 g/ml, serine protease inhibitor); EDTA-Na 2 (0.5 mg/ml, metalloprotease inhibitor); leupeptin (0.5 g/ml, serine and cysteine protease inhibitor), and pepstatin (0.7 g/ml, aspartate protease inhibitor)). Insoluble material was removed by centrifugation at 10,000 ϫ g. The resultant cleared extract was centrifuged at 100,000 ϫ g, and the membrane pellet was resuspended in 85 mM Tris, pH 7.8, 0.1% octyl ␤-glucoside, 30 mM NaCl, and protease inhibitors. The suspension was centrifuged at 100,000 ϫ g, and insoluble membranes were discarded. Four separate isolations of protein were performed.
Construction of an Inhibin Affinity Column and Affinity Purification-An inhibin affinity column was prepared by coupling 5 mg of rh-inhibin A to AffiGel 10 (Bio-Rad) following the manufacturer's protocol. Coupling efficiency was Ͼ98%. Soluble membrane proteins were passed in series though a blank AffiGel-10 column (bed volume, 5 ml) and then over the rh-inhibin A-coupled column (bed volume, 5 ml). The eluate was reprocessed twice through the inhibin column. The column was washed extensively with 25 mM Tris, pH 7.5, 150 mM NaCl, 0.1% octyl ␤-glucoside. After a stable baseline was reached, the columnbound proteins were eluted with 100 mM glycine, pH 3.0, 0.1% octyl ␤-glucoside and immediately neutralized with 3 M Tris, pH 8.5. The column fractions were analyzed by chemical cross-linking and mobility on SDS-PAGE (26). Briefly, 100,000 cpm of biologically active iodinated inhibin or activin was added to equivalent aliquots of column fractions and incubated at room temperature for 1 h. Disuccinimidyl suberate, dissolved in Me 2 SO (Pierce), was added to a final concentration of 500 M and allowed to react with the bound complexes for 1 h at room temperature. The products were analyzed by 12% SDS-PAGE. The gels were dried and subjected to autoradiography.

RESULTS
The successful identification of activin/TGF␤/BMP/MIS receptor subunits has proceeded based on degenerate oligonu-cleotide priming from the conserved serine-threonine kinase domain of the activin receptor subunit. Studies in our laboratory were initiated to identify a subunit receptor that bound inhibin and not activin using degenerate oligonucleotides against the conserved serine-threonine kinase domain in pituitary and inhibin ␣-subunit tumor tissue. No independent receptor isoform was identified using this approach (data not shown). Second, an expression library generated from rat ovaries was generated and screened for inhibin-binding proteins. No inhibin-specific binding proteins were identified using this methodology (data not shown).
Ligand binding studies in ovary (22), testis (23), and the developing embryo ( Fig. 1) indicate that separate inhibin and activin binding sites exist in specific cellular groups. In the embryo, inhibin-specific binding sites are detected in the trigeminal ganglion and spinal ganglion. Because activin receptors and follistatin are also present near or coincident with the inhibin binding sites in ovary, testis, and the embryo, no attempt was made to purify the inhibin receptor from these source tissues (22,23,28). These studies indicate, however, that several tissues or cellular groups have inhibin binding sites that are distinct from activin binding sites.
Based on our in situ ligand binding studies, we hypothesized that gonadal tumor tissues arising from the genetic deletion of the inhibin ␣-subunit in mice would be a potential source of inhibin receptor. Gonadal tumors (both ovarian and testicular) were collected from adult inhibin ␣-subunit-deficient mice and embedded immediately on dry ice. Tumor tissues were processed to examine inhibin or activin binding sites using in situ ligand binding (Fig. 2). The tumors had a mixed sex cord stromal phenotype, and all cells bound labeled ligands (Fig. 2, B ( 125 I-inhibin A) and F ( 125 I-activin A)). Addition of 1000-fold molar excess unlabeled inhibin A competed with the 125 I-inhibin A (Fig. 2C), and excess activin A competed with 125 Iactivin A (Fig. 2G). Inhibin and activin did not cross-compete , which is identical to that shown previously in the rat. High levels of 125 I-rh-inhibin A binding to all cells of the gonadal tumors was noted (B). The 125 I-rh-inhibin A binding was competed by unlabeled rh-inhibin A (C) and was not competed by unlabeled rhactivin A (D). In contrast, the control ovarian tissue sections had easily detectable 125 I-rh-activin A binding to follicle structure (E) in a pattern identical to that demonstrated previously in the rat. The tumor tissue had low 125 I-rh-activin A binding (F), which was competed by activin (G) but not inhibin (H). This low level activin binding may represent interaction of the 125 I-rh-activin A probe with follistatin or activin type RII receptor subunit.

FIG. 3. Analysis of ovarian and testicular tumor homogenates for inhibin binding (receptor) proteins.
Ovarian and testicular tumor homogenates were prepared, and membrane proteins were solubilized. Homogenates were incubated with 125 I-rh-inhibin A, crosslinked, and analyzed by nonreducing SDS-PAGE. Specificity of the 125 I-rh-inhibin A binding was determined by competition experiments in which 100-fold unlabeled inhibin was co-incubated with 125 I-rhinhibin A and protein homogenates. Both ovarian and testicular tumors contained proteins that bound 125 I-rh-inhibin A, and the ability of 125 I-rh-inhibin A to bind these proteins was competed for partially (ovarian) or entirely (testicular) by unlabeled inhibin A, suggesting that the proteins cross-linked to the labeled ligands did so in a specific manner (Precolumn fractions). Ovarian or testicular homogenates were passed over an inhibin affinity column, and proteins eluted from the column were analyzed by cross-linking study (Postcolumn fractions). Proteins of similar apparent molecular weights were identified in ovarian and testicular homogenates, and these proteins corresponded to the proteins that were specifically competed by unlabeled inhibin A. The arrow indicates the position of 125 I-rh-inhibin A. The lettered bands are proteins that correspond to proteins isolated in all purifications (for example, see Fig. 4, proteins a, b, and c).
with heterologous labeled ligand, indicating that the tissues have distinct inhibin and activin binding sites (Fig. 2, D and H). Sections of ovaries obtained from wild-type littermate animals had low level inhibin binding to antral granulosa cells ( Fig. 2A) and higher levels of activin A binding to ovarian follicles (Fig.  2E). The pattern and intensity of inhibin and activin binding in control mouse ovaries is identical to the binding pattern of the two ligands previously described in normal rat ovary (22).
Because the tumor tissues appeared to be enriched for an inhibin binding moiety and had little activin binding (which we predicted would represent binding to follistatin, a cytoplasmic protein), ovarian and testicular tumor tissues were homogenized, and membrane proteins were isolated from multiple tumors. Protein was incubated with iodinated inhibin, crosslinked with disuccinimidyl suberate, and analyzed under nonreducing conditions on denaturing SDS-PAGE (Fig. 3). In addition, iodinated inhibin formed complexes with proteins in both ovarian and testicular tumor extracts. Incubation of membrane extracts with 100-fold excess unlabeled rh-inhibin A reduced the inhibin binding in ovarian tumor extracts and competed efficiently for the binding in testicular tumor homogenates. Ovarian and testicular tumor homogenates were passed over an inhibin affinity column, resulting in an enrichment of proteins that bind inhibin. Ovarian and testicular tumor homogenates were combined from this experiment forward.
To examine whether the proteins that could be isolated would bind specifically to inhibin, solubilized membrane protein from ovarian and testicular tumors was passed over the inhibin affinity column twice, the column was washed, and the protein that bound the immobilized inhibin was eluted using a low pH buffer. Fractions representing the eluted protein peak were neutralized with Tris, and aliquots were incubated with iodinated inhibin and iodinated activin. The proteins were cross-linked with disuccinimidyl suberate and analyzed by SDS-PAGE (Fig. 4). When incubated with fractions representing the protein peak eluted from the inhibin affinity column, 125 I-rh-inhibin A was specifically shifted upward in the chromatogram, whereas 125 I-rh-activin A was not (Fig. 4, A and B,  lanes 5-9). Four specific 125 I-rh-inhibin A complexes corre-sponding to sizes of 130, 116, 86, and 72 kDa were identified (Fig. 4A, lane 7, a, b, c, and d, respectively). These proteins were identified in four independent experiments. However, attempts to microsequence the proteins were unsuccessful due to blocked N termini or lack of sufficient material. Internal sequence analysis was attempted in the cases of blocked N termini; however, no sequence data were generated due to the small amount of protein recovered in the purifications.
To confirm that the proteins were receptors and did not include follistatin, two experiments were conducted. First, io- Western blot analysis using a follistatin-specific monoclonal antibody (B). By ligand blot, using 125 I-activin A as probe, no follistatin was detected in the solubilized membrane preparation or in the recovered column fractions. An antibody raised against rh-follistatin (provided by Dr. Patrick Sluss, Massachusetts General Hospital) detected a high molecular weight protein in the membrane preparation but not in a column fraction that contained inhibin binding activity. The protein detected in the solubilized membrane preparation did not bind iodinated activin (compare panel A, lane 1, and panel B, lane 2) nor is it present in the inhibin affinity column eluate. dinated activin was used as a probe in a ligand binding blot to detect follistatin in the solubilized membrane protein fraction and in fractions isolated from the inhibin affinity column (Fig.  5A). Iodinated activin was used in this experiment because iodinated inhibin does not bind follistatin in this format (data not shown). Iodinated activin binds rh-follistatin in the positive control lane, yet it does not detect free follistatin in the total membrane protein fraction or in the column eluate fractions. This suggests that the homogenate and eluate are essentially devoid of free follistatin and indicates that the membrane preparation is likely free of cytoplasmic proteins. To determine whether membrane-bound proteoglycan-associated follistatin was present in the membrane or column fraction, an antibody against follistatin was used in an immunoblot analysis of isolated protein (Fig. 5B). The antibody detected a high molecular weight protein in the membrane fraction but not in the column eluate. These data suggest that follistatin is membrane-associated but is not in a form capable of binding activin and indicate that the proteins eluted from the inhibin affinity column do not include follistatin.

DISCUSSION
Inhibin is an gonad-derived dimeric protein hormone. The principle biological activity of inhibin is to suppress pituitary FSH secretion in a classic endocrine fashion (2,29,30). It may also participate in ovarian follicle development and oocyte maturation (31)(32)(33). Closely related to inhibin (␣-␤) is activin (␤-␤). Activin stimulates pituitary FSH synthesis and secretion in a dose-dependent manner and causes follicle atresia; however, it is synergistic with inhibin in stimulating oocyte maturation (2,21,(31)(32)(33). In addition to its role in ovarian and pituitary function, activin is known to regulate erythrodifferentiation (34), promote neuronal survival (35), and regulate mesoderm development in Xenopus and mouse embryos (36,37). Activin regulates these functions through a family of receptor kinases (38 -40). One subunit, the type RII(B) receptor, binds the ligand and then transphosphorylates a second type RI receptor. The ligand-RII-RI complex is a functional serine threonine kinase, the functional targets of which are the members of the Smad cytoplasmic protein family (41).
Although specific activin receptor subunits have been identified and cloned, efforts to isolate an inhibin receptor have not been successful. Numerous studies were done to identify an inhibin receptor. The inability to identify an inhibin receptor using oligonucleotides directed against conserved regions of known activin receptors suggests that the inhibin receptor may differ from the activin receptor subunits. Numerous in situ ligand binding studies were done using a wide variety of tissues to localize inhibin binding sites that could be indicative of novel inhibin function and a source of potential inhibin receptor. In the ovary, inhibin-specific binding sites are associated with the granulosa cell (22). The ovary is capable of producing inhibins, activins, and follistatins in response to pituitary gonadotropins and to local growth regulatory factors (42). Inhibin A and inhibin B are released from the ovary and regulate pituitary FSH in a traditional endocrine feedback manner (43,44). In addition to endocrine regulations, the follicular granulosa cell, theca, cell and ooctye are able to respond to inhibin and to activin, and these effects can be modulated by the binding protein follistatin. Inhibin stimulates granulosa cell proliferation, theca cell androgen production, and ooctye maturation (21, 32-35, 45, 46). Each of these effects may be coordinated through an independent inhibin receptor. In the testis, inhibinspecific binding sites are present on interstitial cells that are also positive for 3␤-hydroxysteroid dehydrogenase, suggesting that these cells are the steroidally active Leydig cells (23).
In experiments described in this paper, we identified inhibin binding sites in the trigeminal ganglion and spinal ganglion of the developing rat embryo. The ability of inhibin to act on these sites requires that the ligand be present. Inhibin ␣-subunit mRNA is expressed in the somites of the embryonic rat on E12, in the dorsal root ganglion from E12 to E20 (47), and in the somites of the 10.5 day mouse embryo (48); immunoreactive ␣-subunit protein is localized in the somites of chick embryos (49). Moreover, both inhibin ␣and ␤-subunit mRNAs are detected in different stages of the early mouse embryo and in embryonic stem and embryonic carcinoma cells that model early murine fetal development (50,51). Likewise, human preembryos have been shown to secrete immunoreactive ␣-subunit protein (52). Therefore, the developing embryo likely produces inhibin in restricted cellular sites, and this inhibin may have effects specifically on cellular loci such as the trigeminal ganglion and spinal ganglia, where inhibin-specific binding sites have been localized. Further analysis will be required to delineate the specific effect(s) of inhibin (and activin) on these cellular sites. A powerful model system to delineate the physiological relevance of inhibin and activins is the genetic deletion of subunit genes through homologous recombination (the knockout mouse model). A series of animals deficient in the ␣or ␤-subunits, in the receptors for activin, and in the activin/inhibin-binding protein, follistatin, have been generated (24,(53)(54)(55)(56)(57). Animals deficient in the inhibin ␣-subunit develop gonadal tumors (24 -26), activin type RII receptor is down-regulated (57), and, as shown herein, the tumors bound iodinated inhibin A preferentially. Upon identification of the ovarian tumors as a tissue source of a potential inhibin receptor, we initiated studies to purify the protein using classical affinity chromatography methods. Four inhibin-binding proteins were partially purified. The receptor proteins were identified by cross-linking labeled ligand to putative receptor and examining the retardation of the complex on SDS-PAGE. Our inability to sequence the N terminus of the proteins isolated by affinity purification means that we can only speculate on the relationship between the proteins that bind iodinated inhibin. The two smaller proteins (complexes of 86 and 72 kDa) may be proteolytic cleavage products of the larger proteins. A broad spectrum mixture of protease inhibitors was used throughout the purification procedure; however, degradation of the larger proteins is a possibility. Alternatively, the larger molecular weight shifts may represent complexes of the smaller proteins with the iodinated inhibin. A third possibility is that several classes of inhibin receptor proteins exist in the tumor tissues. For example, inhibin may bind a yet-unidentified type II receptor, may bind and activate a type RI receptor, may have a structurally dissimilar receptor, or may use an adapter protein, such as type RIII receptor, to act in concert (or competition) with activin type RII or RI receptors. The existence of an inhibin-specific receptor and competition with the activin receptor are not mutually exclusive conclusions. Indeed, an inhibin-specific protein band has been identified in the context of activin type RII and RI receptors in human erythroid precursor cells (16). Whether the protein identified in the K562 cell system is similar to one of the proteins identified in this study remains to be clarified. Indeed, the complete elucidation of the functional relationship between inhibin and activin receptors awaits the cloning of the inhibin receptor.
The fact that the inhibin receptor can be clearly identified in the knockout mouse tumor tissue is significant. It is known that one of the phenotypes of these tumors is low expression of type RII receptor, likely due to down-regulation by activin (24 -25, 57). Similarly, the inhibin receptor may be up-regulated by persistent activin or by the lack of negative feedback of inhibin. Clearly, additional studies are necessary to determine what role, if any, activin and inhibin have in tumorigenesis and whether an inhibin receptor is expressed in human epithelial ovarian tumors.
In summary, inhibin binding moieties were found to be abundant in gonadal tumors that arise from the genetic elimination of the ␣-subunit of inhibin. Four inhibin-binding proteins were purified by inhibin affinity chromatography from these tumors. Finally, the proteins eluted from the inhibin column were found to be distinct receptor proteins and not follistatin. Further work will be required to clone and characterize the inhibin receptor-generated proteins identified by affinity column purification; however, the results presented in this study represent an important first step toward the elucidation of the inhibin receptor and signal transduction system.