Constitutive negative regulation in the processing of the anti-Müllerian hormone receptor II

The more than 30 cysteine-knot-containing secreted morphogens that form the transforming growth factor-β (TGF-β) superfamily (TGFβ-SF) of ligands play crucial roles in health and disease through the regulation of development, differentiation, immune responses, cell growth arrest and tissue or organ regeneration and maintenance (Heldin et al., 2009; Hogan, 1996; Massagué, 1998; Reddi, 1998; Wakefield and Hill, 2013). The superfamily includes TGFβs, activins, inhibins, bone morphogenetic proteins (BMPs), growth and differentiation factors (GDFs), nodal and the anti-Müllerian hormone (AMH), also known as Müllerian inhibitory substance (Hinck, 2012; Weiss and Attisano, 2013). Receptors of the superfamily are also structurally similar, with a cytosolic serine-threonine kinase domain involved in the transduction of signals, and an extracellular domain (ECD) characterized by a three-finger toxin fold that mediates ligand recognition and binding (Hinck, 2012). The similarity of the transduction of signals by the TGFβ-SF members also extends to the repertoire of downstream mediators employed by the receptors. Thus, all superfamily receptors signal through Smad effectors (Massagué and Gomis, 2006; Weiss and Attisano, 2013). In the case of AMH, the ligand-recruiting receptor (anti-Müllerian hormone type-II receptor, AMHRII) binds to AMH (di Clemente et al., 2010) and activates a type-I receptor – ALK2, ALK3 or ALK6 (also known as ACVR1, BMPR1A and BMPR1B, respectively) (Clarke et al., 2001; Gouédard et al., 2000; Jamin et al., 2002; Orvis et al., 2008; Sèdes et al., 2013; Visser et al., 2001). The activated type-I receptor phosphorylates Smad1, Smad5 or Smad8 (R-Smads) and promotes their interaction with Smad4 and the translocation of the R-Smad–Smad4 complex to the nucleus, where it regulates gene transcription. In addition to sharing the same repertoire of type-I receptors, AMH also shares its R-Smad effectors with BMPs.

Early in the development of male fetuses, during gonad differentiation, AMH is synthesized by Sertoli cells and mediates the regression of Müllerian ducts – the anlagen of the uterus and Fallopian tubes. In accord with its proposed role in male sex differentiation, mutations in the AMH and AMHRII genes result in persistent Müllerian duct syndrome (PMDS), where Müllerian derivatives (e.g. Fallopian tubes, uterus and the upper part of the vagina) persist in virilized genetic males (Belville et al., 2004; di Clemente and Belville, 2006). Postnatally, AMH is also produced by ovarian granulosa cells, has roles in the preservation of the primordial follicle pool and serves as a marker of ovarian function and health (Dewailly et al., 2014). Additionally, a putative use of AMH as a therapy agent in ovarian cancer has been proposed (Meirelles et al., 2012; Wei et al., 2010).

In accordance with the far-reaching consequences of the activation of TGFβ-SF signals on cell health and function, and the strict dependence of such signals on the plasma membrane subpopulation of signaling-competent receptor complexes, TGFβ-SF receptors are regulated at multiple levels. These levels include, but are not limited to, oligomerization, post-translational modifications and trafficking (Chen et al., 2009; Chen, 2009; Di Guglielmo et al., 2003; Ehrlich et al., 2012; Ehrlich et al., 2011; Hartung et al., 2006; Hirschhorn et al., 2012; Kim et al., 2012; Nohe et al., 2002; Partridge et al., 2004; Shapira et al., 2012; Shapira et al., 2014; Yao et al., 2002). Of these regulatory mechanisms, modifications related to the biosynthetic maturation of TGFβ-SF receptors have been relatively less studied. The importance of regulation of such receptors at early stages of folding, maturation and trafficking is exemplified by the molecular fate of receptors carrying genetic disease-related mutations, which induce the intracellular retention of AMHRII and the type-II BMP receptor (BMPRII, also known as BMPR2), and abolish the signaling response to their respective ligands (Belville et al., 2009; Faure et al., 1996; Frump et al., 2013; Li et al., 2010; Sobolewski et al., 2008).

Proteolysis-based ectodomain shedding of cytokine receptors commonly downregulates their signaling through the generation of ligand-binding competent soluble receptors (i.e. ligand traps) and a concomitant reduction in the amounts of receptor at the plasma membrane (Levine, 2008). Within the TGFβ-SF, co-receptors endoglin and betaglycan (also known as TGFBR3) are shed through the activity of different matrix metalloproteases (Hawinkels et al., 2010; Tobar et al., 2014; Velasco-Loyden et al., 2004). In addition, BMPRII and the type-I receptors for BMP and TGF-β undergo stimulated proteolytic cleavage (e.g. with phorbol esters), in accordance with roles for signaling-activated proteases, such as disintegrin and metalloproteinase 17 (ADAM17) (Liu et al., 2009; Mu et al., 2011; Singhatanadgit et al., 2006a; Singhatanadgit et al., 2006b).

Intra-chain disulfide bonds stabilize the three-finger toxin fold of the ECDs of TGFβ-SF receptors and contribute to their ability to discriminate amongst different superfamily ligands (Allendorph et al., 2006; Ehrlich et al., 2012; Ehrlich et al., 2011; Greenwald et al., 2003; Hart et al., 2002; Radaev et al., 2010). Accordingly, cysteine substitution mutants in BMPRII cause hereditary pulmonary arterial hypertension (PAH; Li et al., 2010; Sobolewski et al., 2008), and the detrimental role of improper folding of the receptor due to such mutations is underscored by the beneficial effects of chemical chaperones on the membrane localization and ligand binding of mutant BMPRII (Li et al., 2010). The formation and isomerization of disulfide bonds occur co- and post-translationally within the oxidizing environment of the endoplasmic reticulum (ER; Oka and Bulleid, 2013), and constitute a major step in the quality control of secreted proteins (Sitia and Braakman, 2003). Failure to pass such control results in the retention and ER-associated degradation of such proteins (Benyair et al., 2011).

In this study, we characterize the processing and oligomerization of AMHRII and identify its constitutive intracellular cleavage and the formation of higher-order disulfide-bonded AMHRII oligomers upon increase in expression levels. Taken together, our results suggest novel mechanisms for the negative regulation of AMHRII.

Recently, we developed a polyclonal antibody against the ECD of mouse AMHRII (α-mAMHRII, see Materials and Methods). To characterize the expression of endogenous AMHRII, we employed α-mAMHRII and immunoblotted lysates of SMAT-1 mouse Sertoli cells (Belville et al., 2005), which were analyzed under four different conditions: (1) non-reducing, (2) reducing, (3) PNGaseF digestion, (4) EndoH digestion. Digestion of glycans (3 and 4) was performed on reduced samples. Under non-reducing conditions, three bands of apparent molecular mass (M_r) of ∼58 kDa, ∼66–68 kDa and ∼71 kDa were observed (Fig. 1A,B). To assess the folding status of these three apparent mouse AMHRII species, we compared their M_r under non-reducing and reducing conditions, as proteins that contain intra-molecular disulfide bonds and globular folds migrate faster in non-reducing (NR)-SDS-PAGE as compared to reducing (R)-SDS-PAGE. DTT-mediated reduction slowed the migration of the 66–68-kDa band (to ∼69 kDa), whereas no increase in M_r of the 58-kDa or 71-kDa bands was observed, suggesting that only the former protein exhibits disulphide-bond-mediated folding. PGNaseF digestion reduced the M_r of the 58-kDa and 69-kDa bands to ∼55 kDa and 63 kDa, respectively, without affecting the 71-kDa band. These data suggest that the 58-kDa and 69-kDa proteins are N-glycosylated, whereas the 71-kDa protein is not. EndoH digestion reduced the M_r of the 58-kDa protein to 55 kDa, whereas the 69-kDa band was predominantly EndoH resistant. These results suggest that the 58-kDa protein contains mostly immature glycosyl chains (consistent with ER localization), whereas the 69-kDa protein contains mostly mature glycosyl chains (consistent with Golgi or post-Golgi localization). Based on these results, we hypothesized that the 66–68-kDa and the 58-kDa proteins are different forms of AMHRII, whereas the 71-kDa protein is not an AMHRII protein, but is non-specifically recognized by the α-mAMHRII antibody. The latter conclusion is supported by the failure of an anti-AMHRII monoclonal antibody (mAb) to recognize this species (data not shown).

The presence of two forms of AMHRII differing by 8–10 kDa might result from either alternative splicing or post-translational processing of the receptor. To address these possibilities, we transfected SMAT-1 cells with a cDNA construct encoding an N-terminally epitope-tagged human AMHRII (HA–AMHRII). Lysates of transfected and non-transfected cells were immunoblotted with α-mAMHRII and anti-HA (α-HA) antibodies. Under reducing conditions, immunoblotting of lysates of HA–AMHRII-transfected cells with α-mAMHRII, revealed an increased intensity of the 69-kDa and the 58-kDa bands and no change in the intensity of the 71-kDa band (Fig. 2A), confirming our assignment of the 71-kDa band as non-specific. Under non-reducing conditions, immunoblotting of lysates of HA–AMHRII-transfected cells with α-mAMHRII revealed prominent 66–68-kDa and 58-kDa bands, as well as higher M_r forms of AMHRII (Fig. 2A), suggesting that upon overexpression a considerable portion of AMHRII is present in disulfide-bonded oligomers. Notably, immunoblotting of the same cell lysates with α-HA failed to detect the 58-kDa form of AMHRII. In addition, α-HA immunoblotting revealed the 66–69-kDa bands (under reducing and non-reducing conditions), higher M_r forms (under non-reducing conditions) and a ∼70-kDa smear under reducing conditions (shown to be the EndoH-resistant form of overexpressed epitope-tagged HA–AMHRII, see Fig. 4B). Based on these results (Fig. 1; Fig. 2A,B) we hypothesized that the 58-kDa band represents an N-terminally cleaved form of AMHRII (not recognized by α-HA), which is not globularly folded (no alteration in M_r under non-reducing conditions), is partially glycosylated (∼3-kDa reduction in M_r with PNGaseF) and is retained in the ER (fully EndoH sensitive). These characteristics contrast with those of the full-length endogenous AMHRII species, which exhibited globular folding and was mainly EndoH resistant. Moreover, these results suggest that an increase in expression of AMHRII results in the accumulation of disulphide-bond-mediated AMHRII oligomers.

To examine the correlation between AMHRII expression levels and its disulfide-bond-mediated oligomerization, SMAT-1 cells were transiently transfected with HA–AMHRII for different periods of time (0, 4, 6 or 8 h), and cell lysates (separated by NR-SDS-PAGE) were immunoblotted with α-mAMHRII. In untransfected SMAT-1 cells, only a small portion of the endogenous mouse AMHRII was observed in higher-M_r oligomers (Fig. 2C). After 4 h of transfection with HA–AMHRII, a considerable portion of receptor was found in higher-M_r oligomers. Longer transfection times resulted in further increases in the expression of HA–AMHRII, which were reflected mainly in the accumulation of higher-M_r oligomers, suggesting a direct correlation between expression levels and the disulfide-bond-mediated oligomerization of AMHRII (Fig. 2C).

To assess whether AMHRII cleavage and oligomerization were cell-type specific, we transfected HEK-293T cells with HA–AMHRII–GFP (schematically depicted in Fig. 2G), and analyzed its expression by immunoblotting with anti-GFP antibodies (α-GFP), α-mAMHRII and α-HA (Fig. 2D–F, respectively). α-GFP and α-mAMHRII yielded similar band patterns, revealing three forms of AMHRII protein – a ∼100-kDa species, a ∼90-kDa species and higher-M_r oligomers (under non-reducing conditions). The recognition of all three forms by α-GFP indicates that these forms of AMHRII contain an intact C-terminus. Notably, the ∼90-kDa protein (the expected M_r for the cleaved form of the GFP-fusion protein) was not recognized by α-HA, confirming that this form is missing the N-terminal portion of AMHRII. Taken together, these results show that a significant fraction of AMHRII undergoes cleavage within the ECD, whereas a different fraction undergoes disulfide-bonded oligomerization that is dependent on expression level.

To directly assess the potential of AMHRII-ECD to form disulfide-bonded oligomers, we expressed a chimeric construct of the ECD of AMHRII and the Fc portion of IgG1 (AMHRII–Fc; di Clemente et al., 2010), and analyzed the secreted protein by digestion with endoproteinase LysC, which cleaves after lysines and removes the Fc fragment. Under non-reducing conditions, prior to LysC digestion, multiple high-M_r forms of AMHRII–Fc were observed. After LysC digestion, three bands of 100 kDa, 54 kDa and 42–44 kDa were observed (Fig. 3A). Incubation of the digested proteins with Protein-A–Sepharose removed Fc-containing proteins (100-kDa and 54-kDa bands) resulting in a diffuse single band running around 42–44 kDa (Fig. 3B). We hypothesized that the 42–44-kDa band represented a glycosylated AMHRII-ECD dimer. This was confirmed after reduction, when the 40–42-kDa band collapsed into a band of ∼22 kDa (Fig. 3B). This indicates that at least one interchain disulfide bond forms between ECD monomers within the AMHRII–Fc fusion protein.

Recently, we proposed a model of the structure of the ECD of AMHRII, based on the solved crystal structures of BMPRII, the activin type-II receptor and TβRII (Belville et al., 2009). In this model, the AMHRII-ECD forms five disulfide bridges: Cys24–Cys61, Cys55–Cys79, Cys60–Cys87, Cys92–Cys109 and Cys111–Cys116. To probe for the involvement of specific cysteines or of the predicted cysteine pairs in the formation of the DTT-sensitive higher-M_r oligomers of AMHRII, we mutated individual cysteine residues, or the predicted cysteine pairs, to serine(s), in the context of myc-tagged AMHRII (myc–AMHRII). All cysteine to serine mutants generated comparable levels of disulfide-bonded higher-M_r oligomers (Fig. 3C,D). As such, the homo-oligomerization of myc–AMHRII could not be attributed to any individual cysteine or to any predicted cysteine pair of its ECD. These data fit a scenario in which the formation of such intermolecular disulfide bonds is promiscuous and might involve multiple different cysteines.

To further assess the degree of specificity of the disulfide-bond-mediated oligomerization of AMHRII, we compared the oligomer-forming potential of HA–AMHRII with that of HA–TβRII when expressed in HEK-293T cells and analyzed by R- and NR-SDS-PAGE. In R-SDS-PAGE, HA–TβRII appeared as a faint band of ∼70 kDa and a diffuse band of ∼80–95 kDa (Fig. 4A). Digestion of lysates with EndoH and PGNaseF confirmed that these were the precursor and mature forms of HA–TβRII, respectively (Fig. 4B). In NR-SDS-PAGE, HA–TβRII migrated with a slightly reduced M_r, indicating the presence of intrachain disulfide bonds. Notably, no higher-M_r oligomers of HA–TβRII were observed (Fig. 4A). Similar experiments with HA–BMPRII also failed to reveal significant higher-M_r oligomerization in NR-SDS-PAGE (data not shown). These data contrast with the migration patterns observed with HA–AMHRII, and they suggest that the higher-M_r oligomerization of AMHRII is not a general phenomenon of type-II receptors of the TGFβ-SF. Next, we examined whether the differential oligomerization tendencies of AMHRII and TβRII were also reflected in their intracellular distribution pattern. To visualize the localization of both receptors, we transfected COS7 cells (as their morphology is well suited for confocal microscopy analysis) with myc–AMHRII or myc–TβRII, and co-stained cells for epitope-tagged receptors and endogenous markers of the ER (calnexin) or the Golgi compartment (GM130). Quantitative analysis of the colocalization of the receptors with these markers (Fig. 4C,D) revealed opposite tendencies, as myc–AMHRII colocalized with calnexin (Pearson's correlation coefficient of 0.7±0.02), whereas myc–TβRII prominently colocalized with GM130 (Pearson's correlation coefficient of 0.78±0.01).

To directly measure the ability of cleaved and/or disulfide-bond-oligomerized AMHRII to reach the plasma membrane, we employed a surface biotinylation approach. We expressed HA–AMHRII–GFP in HEK-293T cells, surface biotinylated the cells in the cold, and precipitated the receptors from 90% of the cell lysate with α-GFP-conjugated beads. Analysis of the receptor population within whole cells was performed with 10% of the cell lysate. Precipitates and whole-cell lysates were separated by R- and NR-SDS-PAGE and probed with streptavidin–HRP (precipitates, to visualize surface-localized receptors; Fig. 5A) or α-GFP and α-HA immunoblotting (lysates, to visualize the entire population of receptors; Fig. 5B,C). Anti-biotin blotting revealed mainly the 100-kDa band under both reducing and non-reducing conditions, suggesting that full-length monomeric AMHRII is the main form of receptor at the plasma membrane. In contrast, α-GFP probing of the cell lysates clearly revealed all three forms of the receptor (higher-M_r oligomers, 100 kDa and 90 kDa; Fig. 5B), whereas the cleaved 90-kDa form was absent from α-HA immunoblots (Fig. 5C).

To measure the effect of disulfide-bond-mediated AMHRII oligomerization on its function, we compared AMH binding by AMHRII monomers versus oligomers. To this end, we prepared AMHRII monomers by co-expressing (in HEK-293T cells) AMHRII–Fc and the Fc portion of human IgG1. Under these conditions, three proteins are expected to be produced – dimeric AMHRII–Fc, dimeric Fc and a disulfide-linked heterodimer composed of one chain of AMHRII–Fc and one chain of Fc (AMHRII–Fc/Fc), which contains only one AMHRII ECD. Fig. 6A shows a NR-SDS-PAGE analysis of fractions collected by using a Protein-A–Sepharose column (lane 1), and subsequent size-exclusion chromatography column (lane 2). We first compared the abilities of AMHRII–Fc/Fc and AMHRII–Fc to bind a mouse anti-AMHRII monoclonal antibody (mAb-13H8). Following capture of increasing concentrations of the two soluble chimeric constructs (AMHRII–Fc and AMHRII–Fc/Fc) on ELISA plates, we incubated these plates with excess mAb-13H8 and detected bound antibody with anti-mouse Fc antibody (Fig. 6B). In accordance with the fact that AMHRII–Fc has two AMHRII-ECDs, whereas AMHRII–Fc/Fc has only one, AMHRII–Fc bound approximately twice the amount of mAb-13H8 (Fig. 6C). Slight deviations from this ratio (1∶2) might stem from contamination of the AMHRII–Fc/Fc preparation by Fc. Bivalent binding of mAb-13H8 to the two ECDs of AMHRII–Fc might account for its higher apparent affinity compared to AMHRII–Fc/Fc, owing to an avidity effect.

Next, we measured the binding of cleaved AMH to AMHRII–Fc/Fc or AMHRII–Fc, using an ELISA format (Fig. 6D). The level of AMH bound at high concentrations to the chimeric constructs provides an indication of the number of functional ECDs. If AMHRII–Fc contained two functional ECDs, it should bind twice as much AMH as the AMHRII–Fc/Fc, which has one ECD. However, opposite results were obtained, whereby AMHRII–Fc bound much less AMH than AMHRII–Fc/Fc (Fig. 6E), suggesting that only a minor portion of the ECDs are functional in the AMHRII–Fc preparation. AMHRII–Fc/Fc also bound cleaved AMH with a 10–20-fold higher apparent affinity than AMHRII–Fc. Both the lower binding capacity and the lower affinity of AMH for AMHRII–Fc might be due to the ECD interchain disulfide bond(s), which could compromise the ability of AMHRII–Fc to bind AMH. Taken together, the intracellular retention of higher-M_r oligomers of AMHRII and their reduced ability to bind ligand support the notion of their lack of functionality and suggest that oligomerization might serve as a mechanism of negative regulation of AMHRII upon increased expression.

Whereas the covalently bound oligomers of AMHRII were mostly ER-retained (Figs 4, 5), receptors of the TGFβ-SF form oligomeric complexes at the plasma membrane of intact cells (reviewed in Ehrlich et al., 2011) through molecular mechanisms other than intermolecular disulfide bonds. To probe whether AMHRII forms homomeric complexes at the plasma membrane of live cells, we employed computerized immunofluorescence co-patching (IF-coP; Lachmanovich et al., 2003). IF-coP is based on: (1) expression of receptors carrying different epitope tags (or defined epitopes) at their ECDs at the surface of live cells, (2) micro-patching of the receptors with a double layer of primary IgGs of different species (e.g. mouse and rabbit) and anti-species-IgG secondary IgGs conjugated to different fluorophores (e.g. green and red). Receptors residing in the same complex are swept into mutual micropatches, which appear yellow upon overlapping of images acquired with different filters. Here, we co-expressed myc- and HA-tagged AMHRII, TβRII or mucin 1 (MUC1, unrelated control) in COS7 cells and subjected them to IF-coP in the absence or presence of AMH. Green and red patches were identified automatically (as in Lachmanovich et al., 2003) and were defined as overlapping if their intensity peaks were closer than 0.2 µm. Levels of random co-patching (uncorrelated overlap of patches), reflecting the density of patches at the cell surface, were evaluated by overlaying the green image from one region of interest on the red image of an identically sized neighboring region (see Fig. 7C). Randomized values were subtracted to obtain the actual IF-coP percentage (% co-patching). Typical images of an IF-coP experiment are shown in Fig. 7A–C. Averaged data from multiple experiments after subtraction of the contribution of randomized co-patching are given in Fig. 7D. Similarly to previous studies (Rechtman et al., 2009), HA- and myc-tagged TβRII yielded 19.5% co-patching. Owing to statistical considerations, when two identical receptors carrying different epitope tags form dimers, homo-dimerization is higher by a factor of 3/2 than the measured percentage co-patching (discussed in Ehrlich et al., 2012; Ehrlich et al., 2011); thus, ∼29% of TβRII is constitutively homo-dimerized at the cell surface. Both in the absence or presence of AMH, AMHRII presented 38% co-patching in accordance with 57% homo-dimerization. Co-patching of two unrelated receptors, AMHRII and MUC1 yielded negligible levels of hetero-oligomerization, reinforcing the notion of the specificity of the high levels of the interactions between HA–AMHRII and myc–AMHRII.

The high levels of co-patching of AMHRII were unique when compared to other TGFβ-SF type-II receptors (Gilboa et al., 2000; Nohe et al., 2002; Rechtman et al., 2009). We were interested in understanding whether such uniqueness was also reflected in the distribution and mobility of plasma-membrane-localized AMHRII. To characterize the membrane distributions of AMHRII and TβRII, COS7 cells expressing either myc–AMHRII or myc–TβRII were sequentially labeled in the cold with mouse anti-myc antigen-binding fragment (Fab′) followed by Alexa-Fluor-546-labeled Fab' fragment of goat anti-mouse-IgG. Such a labeling protocol allows for the visualization of the unperturbed distribution of receptors, as it does not induce clustering. Labeled cells were fixed, and the ventral membrane of multiple randomly selected cells was imaged by fluorescence microscopy. Whereas TβRII appeared to be uniformly distributed at the plasma membrane, AMHRII was distributed into numerous diffraction-limited fluorescent clusters (Fig. 8A). To obtain a quantitative assessment of the aggregation of AMHRII and TβRII, we calculated the coefficient of variation (standard deviation divided by mean) of the fluorescence signal in selected regions of interest (∼10×10-µm squares). The rationale of such measurement resides in the increased variance of fluorescence signal (spots of high and low signal intensity, see inset of 8A) in clustered staining patterns as opposed to a lesser variance that is characteristic of uniform staining. The coefficient of variation of the membrane staining of AMHRII was significantly higher than that of TβRII (Fig. 8A), confirming the different distribution of the receptors and the clustering tendency of AMHRII.

Fluorescence recovery after photobleaching (FRAP) measurements allow for characterization of the lateral mobility of transmembrane receptors. To probe whether AMHRII and TβRII differed in lateral mobility, we transfected and labeled cells as in the experiments that measured receptor clustering (Fig. 8A), and conducted FRAP experiments with a Gaussian-spot confocal-FRAP setup (see Materials and Methods). The mobility parameters R_f (mobile fraction) and D (diffusion coefficient) of TβRII (Fig. 8B,C) were similar to previously published data (Rechtman et al., 2009). Notably, both the R_f and the D of AMHRII were significantly lower than those of TβRII (P<10⁻⁷), pointing to the restricted mobility of AMHRII, as compared to TβRII. Taken together, the IF-coP, clustering analysis and FRAP experiments show that a high percentage of AMHRII resides in homo-oligomers, distributes in clusters and presents limited lateral mobility, all consistent with a different organization of AMHRII at the plasma membrane when compared to TβRII or BMPRII (Ehrlich et al., 2011; Marom et al., 2011; Rechtman et al., 2009).

In this study, we have used molecular and cell biology experimental approaches, together with quantitative fluorescence microscopy and FRAP experiments, to investigate the expression, processing and oligomerization of AMHRII. Our main findings are that: (1) AMHRII undergoes constitutive cleavage, generating a form of the receptor that lacks a major portion of its ECD and is retained in the ER; (2) increased AMHRII expression results in the accumulation of disulfide-bonded AMHRII oligomers, which are defective in ligand binding and are retained in the ER; and (3) at the plasma membrane, AMHRII forms homomeric complexes prior to ligand addition, has restricted lateral mobility and presents a clustered distribution pattern. These features distinguish AMHRII from other type-II receptors of the TGFβ-SF (e.g. TβRII and BMPRII) and suggest unique modes of regulation with potential implications for its interactions with ligand, type-I receptors and downstream signaling mediators.

Our conclusions that AMHRII is cleaved within its ECD and that the cleaved receptors are ER retained and non-functional are supported by the following evidence: (1) immunoblotting with antibodies recognizing mouse AMHRII-ECD (α-mAMHRII; in case of endogenous or tagged receptors) or GFP (in case of HA–AMHRII–GFP) identified two forms of AMHRII that differed by ∼8–10 kDa (depending on conditions), whereas antibodies against N-terminal tags in AMHRII (HA or myc) identified only the longer of the two forms of the receptor (Figs 1, 2); (2) the shorter form of AMHRII was fully EndoH sensitive, and presented identical migration in R- and NR-SDS-PAGE, suggesting a lack of mature glycosyl chains and of intramolecular disulfide bonds and globular folding (Fig. 1); and (3) the shorter form of AMHRII was not detected at the cell surface by surface biotinylation (Fig. 5). In retrospect, our interpretation of the identity of the cleaved form of AMHRII can be expanded to the lower-M_r form of wild-type AMHRII reported in Faure et al. (Faure et al., 1996), which exhibited the same characteristics as we report here – constitutive generation (cleavage), EndoH sensitivity and failure to reach the plasma membrane. Notably, this lower-M_r form (Faure et al., 1996) behaved similarly (same M_r, processing and intracellular retention) to an AMHRII mutant devoid of the sequence encoded by the second exon of the AMHRII gene (amino acids 18–78 of AMHRII). Importantly, the absence of such sequence was associated with lack of AMHRII function, as the mutant receptor was isolated from a male patient with persistent Müllerian derivatives (Faure et al., 1996). Accordingly, a similarly alternatively spliced AMHRII that is naturally expressed in rabbits does not bind ligand (di Clemente et al., 1994). In contrast to the constitutive cleavage of AMHRII, Alk5 (also known as TGFBR1; Liu et al., 2009; Mu et al., 2011), Alk3, Alk6 and BMPRII (Abdulhussein et al., 2008; Abduljabbar et al., 2012) were shown to undergo cleavage following phorbol ester stimulation, demonstrating that signaling by TGFβ-SF receptors can also be dampened by inducible cleavage events.

Our conclusion that increased expression of AMHRII specifically results in its negative regulation through disulfide-bond-mediated oligomerization and ER retention is supported by: (1) the direct correlation between the levels of expression of AMHRII (endogenous versus exogenous expression) and the portion of receptor that is present as higher-M_r oligomers in NR-SDS-PAGE (Fig. 2C). Notably, generation of higher-M_r oligomers of AMHRII, which did not depend on cell type and was not affected by N- or C-terminal tagging of the receptor, was unique to AMHRII when compared to TβRII (Fig. 4A) or BMPRII (not shown). (2) The high EndoH sensitivity (Fig. 4B) and preferential colocalization with calnexin (Fig. 4C) of exogenously expressed AMHRII, both indicative of retention in the ER. In contrast, exogenously expressed TβRII was mostly EndoH resistant (Fig. 4B) and showed preferential colocalization with the Golgi marker GM130 (Fig. 4D). Additionally, a high portion of the full-length endogenous mouse AMHRII was EndoH resistant (Fig. 1), indicating an inverse correlation between AMHRII expression levels and its Golgi processing. (3) The marked differences observed in the ratio of monomers and oligomers in the population of receptors on the cell surface versus the population in the total cell lysate (Fig. 5).

Former studies by us and others have detected ligand-independent and ligand-induced homomeric complexes of TGF-β and BMP receptors (reviewed in Ehrlich et al., 2012; Ehrlich et al., 2011). After confirming that the vast majority of plasma-membrane-localized AMHRII molecules were not covalently oligomerized (Fig. 5), we assessed the formation of AMHRII complexes at the plasma membrane of live cells with IF-coP. Following the statistical corrections necessary for calculation of the percentage of homodimers, the high levels of constitutive AMHRII homomeric complexes (57%) were unique when compared with those of TβRII (29%; Fig. 7) or BMPRII (Ehrlich et al., 2012; Ehrlich et al., 2011; Marom et al., 2011). Importantly, these data are also in accordance with the notion that receptors of the TGFβ-SF are present as heterogeneous populations of monomers and oligomers at the plasma membrane, enabling flexibility in the cellular responses that might be elicited by different receptor complexes (Ehrlich et al., 2012; Ehrlich et al., 2011; Marom et al., 2011).

The notion of a differential regulation of the membrane distribution and membrane dynamics of AMHRII was further supported by FRAP measurements, which revealed significant differences in the mobile fractions and in the diffusion coefficients of AMHRII and TβRII, reflecting slower diffusion and immobilization of a significant portion of AMHRII on the experimental timescale employed here (minutes, Fig. 8B,C). These data raise the possibility that the high levels of homo-oligomerization and the restricted lateral mobility of AMHRII are inter-related. However, according to the well-accepted Saffman–Delbrück theory (Saffman and Delbrück, 1975), transmembrane proteins should show only a weak logarithmic dependence of diffusion on the radius occupied by their transmembrane domains, suggesting that oligomerization of AMHRII should not significantly affect its mobility. This notion has been recently challenged (Gambin et al., 2006); and newer models stipulate that both the diffusion of large complexes or the mobility of receptors within crowded membrane domains might deviate from the Saffman–Delbrück model (Guigas and Weiss, 2006; Ramadurai et al., 2009). In this context, it is notable that AMHRII (but not TβRII; Fig. 8A) presented a clustered distribution when labeled with a double layer of monovalent Fab′ antibody fragments. Such clustering is consistent with either the aggregation of plasma-membrane-localized AMHRII or the existence of AMHRII-rich membrane subcompartments. Ligand activation of downstream signaling pathways has been shown to be influenced by the localization of TGFβ receptors to cholesterol-rich domains (Shapira et al., 2014) and by the degree of lateral mobility or confinement of BMP receptors into membrane microdomains (Guzman et al., 2012). Thus, there are likely to be significant consequences of the AMHRII localization pattern that we have elucidated in this study on its signaling output.

HEK293T and COS7 cells (ATCC) and mouse immature Sertoli SMAT-1 cells (Dutertre et al., 1997) were maintained in DMEM (Dulbecco's modified Eagle's medium; Gibco®) supplemented with 10% fetal calf serum (FCS), 25 µg/ml penicillin, 40 µg/ml streptomycin, and 5 mM glutamine (all from Biological Industries, Beit HaEmek, Israel). Transfections were with calcium phosphate (HEK293T), Lipofectamine 2000 (Invitrogen; SMAT-1) or JetPRIME® (Polyplus; COS7).

Mutations and epitope tagging were performed by overlapping PCR. Primer sequences are in supplementary material Table S1. Human AMHRII cDNA (in pCMV6, OriGene) was corrected for two mutations (Q208; Y338N). The AMHRII–Fc fusion was described previously (di Clemente et al., 2010). HA (YPYDVPDYA) or myc (EQKLISEEDL) epitopes were inserted after P18 of AMHRII. Tagged AMHRII constructs were cloned into BamHI and NotI sites of pCDNA3.1b (Invitrogen) or into the EcoRI and SacII sites of pEGFP-N1 (Clontech). Myc-tagged AMHRII Cys-to-Ser mutants were at positions 24, 55, 60, 61, 79, 87, 92, 109, 111 or 116. Myc- and HA-tagged TβRII pCDNA1 expression vectors were a gift from Prof. Yoav Henis (Tel Aviv University, Israel).

The following reagents were used in this study: cleaved AMH (di Clemente et al., 2010), DL-dithiothreitol (DTT), N-ethylmaleimide (NEM; Sigma-Aldrich); EZ-Link® Sulfo-NHS-LC-Biotin (Thermo scientific); streptavidin–HRP (Invitrogen); endoglucosidase H (EndoH) and peptide-N-Glycosidase F (PNGaseF; New England BioLabs).

The following antibodies were used: anti-myc IgG and Fab′ (9E10 hybridoma; from Prof. Yoav Henis; Tel Aviv University, Israel); mouse monoclonal anti-HA (western blotting; Sigma Aldrich); rabbit polyclonal anti-GFP (western blotting; MBL Co. Ltd); anti-GM130 (immunofluorescence; 1∶250; Sigma Aldrich); rabbit polyclonal anti-HA.11 (IF-coP; 1∶350, Covance); anti-calnexin (immunofluorescence; 1∶200; Santa Cruz Biotechnology); mouse anti-C-terminal AMH mAb 22A2 (di Clemente et al., 2010) and mouse mAb anti-human AMHRII ECD 13H8 (Soazik Jamin, Chrystèle Racine, Nassim Arouche et al., unpublished). Other antibodies were as follows: secondary Alexa-Fluor-488- and Alexa-Fluor-555-conjugated goat anti-mouse (GαM)- and goat anti-rabbit (GαR)-IgG (1∶200; Invitrogen-Molecular Probes); fluorescent Alexa-Fluor-546-GαM-F(ab′)2, converted to monovalent Fab′ as described previously (Henis et al., 1994); normal goat γ-globulin (NGG; 200 µg/ml); HRP-conjugated GαM and GαR secondary antibodies (both 1∶12500 for western blotting); and goat anti-human-Fc antibody (Jackson ImmunoResearch).

6×His-tagged mouse AMHR-II ECD was expressed in Escherichia coli and purified on nickel-nitrilotriacetic acid resin (Qiagen) under native conditions. Serum from immunized rabbits (3 months) was purified by affinity with the fusion protein (O'Quick Pure System kit; Sterogene, Isnes, Belgium).

Transiently transfected cells were analyzed at 24 h post-transfection. Equal numbers of cells were lysed in 150 mM NaCl, 10 mM HEPES pH 7.4, 0.5% Igepal CA-630, 1% Triton X-100, protease inhibitors (Roche) and phosphatase inhibitors (Sigma-Aldrich). The lysis buffer of non-reduced samples was supplemented with 25 mM NEM. Sample buffer (10 mM Tris-HCl pH 8, 10% glycerol, 2% SDS) was supplemented with 20 mM DTT for reduced samples or an equivalent volume of water for non-reduced samples. 100 mM NEM was added to non-reduced samples before boiling and to reduced samples after boiling. PNGaseF and EndoH digestions were performed according to the manufacturer's instructions and involved reduction of samples. Immunoblots are representative of at least three independent experiments.

AMHRII–Fc (487 µg) was incubated with 0.65 µg/ml endoproteinase LysC (1 h, 37°C and 30 min at room temperature). Leupeptin (40 µM) was added to stop the digestion, and the Fc was removed with Protein-A–Sepharose (45 min at room temperature). To reduce the AMHRII-ECD protein, AMHRII-ECD was incubated with 0.1 mM TCEP (60 min, 37°C) followed by addition of 1 mM NEM and 10 mM Tris-HCl, pH 7.4.

At 24 h post-transfection, coverslips were washed twice with cold PBS (4°C), fixed [4% paraformaldehyde (PFA); 20 min; room temperature], blocked and permeabilized (3% BSA, 0.1% Triton X-100 in PBS; 1 h), stained with primary antibodies (1% BSA, 0.1% Triton X-100 in PBS; 1 h; staining solution) and secondary antibodies (1∶200 dilution in staining solution supplemented with 1 µg/ml DAPI stain, Sigma Aldrich; 30 min). Mounting was with fluorescence mounting medium (Golden Bridge).

Cells were imaged with a spinning-disk confocal microscope setup: Zeiss 100×, NA 1.45; Yokogawa CSU-22; Zeiss fully automated inverted 200M; solid state lasers (473, 561 and 660 nm); piezo-controlled Z-stage all under the command of Slidebook^TM. Images were acquired with an HQ2 CCD camera (Photometrics). Typically, a 1×1 binning was employed, yielding a pixel size of 0.065 µm. For analysis of fluorescence signal distribution, single planes of the ventral membrane of cells expressing myc–AMHRII or myc–TβRII, labeled with 9E10 mouse anti-myc Fab′ and goat anti-mouse-IgG Alexa-Fluor-546–Fab′ (4°C), and fixed with 4% PFA were acquired by epifluorescence microscopy (100×, 1×1 binning, 0.065 µm /pixel). Regions of interest of 10×10 µm of flat continuous membrane were selected with Slidebook^TM. Pearson's coefficient correlation, mean intensity and standard deviation of the fluorescence signal were calculated with Slidebook^TM. Pseudo-coloring of images with similar signal averages was performed with Slidebook^TM.

At 24 h post-transfection with HA–AMHRII–GFP or pEGFP-N1 empty vector, HEK293T cells were washed with PBS (4°C, three times) and borate buffer (10 mM boric acid, 150 mM NaCl, pH 8.0) and incubated with 0.5 mg/ml membrane-impermeable EZ-Link Sulfo-NHS-LC-Biotin (4°C; 45 min; Thermo Scientific) in borate buffer. Cells were then washed (5 min, three times) with ice-cold PBS supplemented with 50 mM glycine and subjected to precipitation with GFP-Trap® A (antibodies-online.com) according to the manufacturer's instructions. 90% of cell lysates was subjected to immunoprecipitation. Immunoprecipitates and 10% of cell lysates were separated by R- or NR-SDS-PAGE and immunoblotted with HRP–streptavidin (1∶2500, Invitrogen; for immunoprecipitates), and α-GFP or α-HA antibodies (for lysates).

HEK293 cells were co-transfected with AMHRII–Fc fusion protein (in pRLC010-1 expression vector) and an expression vector (pEAG1423) encoding the signal sequence, hinge, CH2 and CH3 domains of human IgG1. Conditioned medium was collected every 2 days for 12 days. Recovery of Fc-containing proteins was with Protein-A–Sepharose columns (overnight by gravity); washes were performed with PBS (six times) and 25 mM sodium phosphate pH 5.5, 100 mM NaCl (six times). Fc-containing proteins were eluted (25 mM sodium phosphate pH 2.8, 100 mM NaCl) and neutralized (0.5 M sodium phosphate, pH 8.6).

ELISA plates (Nunc Maxisorp) were coated with a goat anti-human-Fc antibody (in 50 mM sodium bicarbonate, pH 9.6; 4°C, overnight) and washed five times with water. Subsequently, all incubations were followed by five washes with PBS/0.05% Tween-20. Following blocking (1% BSA, 1% goat serum in PBS), plates were incubated with receptor fusion proteins (1 h) followed by incubation with 1 µg/ml cleaved AMH (2 h). Detection was with 1 µg/ml mouse anti-C-terminal AMH mAb 22A2 (1 h), GαM–HRP (1∶3000, 1 h) and 3,3′,5,5′-tatramethylbenzidine (TMB). Absorbance was read at 450 nm. Comparisons of the abilities of dimeric AMHRII–Fc and monomeric AMHRII–Fc/Fc to bind anti-AMHRII mAb 13H8 were as described above except for the addition of 1 µg/ml mouse α-AMHRII mAb 13H8 (1 h) after the incubation with the receptor fusion proteins.

COS7 cells were co-transfected with differently tagged (HA- or myc-tagged) AMHRII and TβRII, or mucin 1 (MUC1; Horn et al., 2009). At 24 h post-transfection, cells expressing pairs of receptors carrying different epitope tags (e.g. HA–AMHRII together with myc–AMHRII or myc–TβRII) or distinct epitopes (MUC1) were washed twice and incubated (30 min, 37°C) with serum-free medium. After washing twice with cold Hank's balanced salt solution (HBSS; Biological Industries, Beit HaEmek, Israel) supplemented with 20 mM HEPES (pH 7.4) and 2% BSA (HBSS/HEPES/BSA; used for all subsequent incubations and washes), cells were blocked with 200 µg/ml NGG (30 min, 4°C). Primary labeling was with rabbit anti-HA together with mouse anti-myc or mouse anti-MUC1 (20 µg/ml each, 30 min, 4°C), followed by secondary labeling and patching with Alexa-Fluor-555–GαR-IgG and Alexa-Fluor-488–GαM-IgG (20 µg/ml each, 30 min, 4°C). Cells were washed and fixed in methanol (5 min, −20°C) and 1∶1 methanol∶acetone (5 min, −20°C). In experiments where ligand (10 µg/ml cleaved AMH) was present, it was added at 4°C for 30 min with NGG and was included at the same concentration during all subsequent incubations.

COS7 cells were transfected with myc–AMHRII or myc–TβRII. After 48 h, cells were washed with HBSS/HEPES/BSA, blocked with 200 µg/ml NGG (30 min, in HBSS/HEPES/BSA), and labeled with monovalent anti-myc Fab′ followed by Alexa-Fluor-546-conjugated GαM-Fab′ (50 µg/ml each, 45 min), all at 4°C. After washes, coverslips were mounted over a chamber containing HBSS/HEPES/BSA. FRAP measurements were performed either at 15°C, to minimize internalization during the measurement, or at 37°C to study receptor mobility at physiological temperature. Indistinguishable results were obtained at both temperatures, and results at 15°C are shown. An argon ion laser beam (Innova 70, Coherent) was focused through a florescence microscope (AxioImage.D1, Carl Zeiss MicroImaging) to a spot with a Gaussian radius (ω) of 0.77±0.03 µm (63×/1.4 NA oil-immersion objective). After a brief measurement at monitoring intensity (528.7 nm, 1 mW), a 5-mW pulse (20 ms) bleached 60–75% of the fluorescence in the illuminated region, and fluorescence recovery was followed by the monitoring beam. The lateral diffusion coefficient (D) and the mobile fraction (R_f) were extracted from the FRAP curves by nonlinear regression analysis, fitting to a lateral diffusion process (Henis et al., 2006).