Repulsive Guidance Molecule (RGMa), a DRAGON Homologue, Is a Bone Morphogenetic Protein Co-receptor*

Bone morphogenetic proteins (BMPs) are members of the transforming growth factor β (TGF-β) superfamily of ligands, which regulate many mammalian physiologic and pathophysiologic processes. BMPs exert their effects through type I and type II serine/threonine kinase receptors and the Smad intracellular signaling pathway. Recently, the glycosylphosphatidylinositol (GPI)-anchored protein DRAGON was identified as a co-receptor for BMP signaling. Here, we investigate whether a homologue of DRAGON, repulsive guidance molecule (RGMa), is similarly involved in the BMP signaling pathway. We show that RGMa enhances BMP, but not TGF-β, signals in a ligand-dependent manner in cell culture. The soluble extracellular domain of RGMa fused to human Fc (RGMa.Fc) forms a complex with BMP type I receptors and binds directly and selectively to radiolabeled BMP-2 and BMP-4. RGMa mediates BMP signaling through the classical BMP signaling pathway involving Smad1, 5, and 8, and it up-regulates endogenous inhibitor of differentiation (Id1) protein, an important downstream target of BMP signals. Finally, we demonstrate that BMP signaling occurs in neurons that express RGMa in vivo. These data are consistent with a role for RGMa as a BMP co-receptor.

hemojuvelin also possess an RGD motif, which could be involved in cell attachment (15). RGMa and DRAGON are expressed in a complementary manner in the central nervous system (17)(18)(19)(20), where RGMa mediates repulsive axonal guidance (15,21,22) and neural tube closure (18), while DRAGON contributes to neuronal cell adhesion through homophilic interactions (17). RGMa also binds to the receptor neogenin (21) and functions as a cell survival factor (23). Hemojuvelin is expressed most heavily in the liver, heart, and skeletal muscle, and is mutated in juvenile hemochromatosis, a disorder of iron overload (16 -20, 24).
Here, we investigate whether RGMa is involved in the BMP signaling pathway. Using a reporter assay, we show that transfection of RGMa cDNA into cells enhances BMP, but not TGF-␤, signals in a ligand-dependent fashion. Binding and cross-linking studies in a cell-free system demonstrate that soluble RGMa.Fc fusion protein interacts with the BMP type I receptor ALK6 and binds directly to 125 I-BMP-2 and 125 I-BMP-4, but not other members of the TGF-␤ superfamily. Co-transfection of RGMa cDNA with dominant negative BMP type I receptors or with dominant negative Smad1 inhibits RGMa-mediated BMP signaling, suggesting that RGMa generates BMP signals via the classical BMP pathway. Transfection of RGMa cDNA into cells induces phosphorylation of endogenous Smad1/5/8 and up-regulates endogenous Id1. Finally, immunofluorescence microscopy of adult rat spinal cord sections, reveals that RGMa is expressed in vivo in neurons, which also show nuclear accumulation of p-Smad1/5/8. Taken together, these data suggest that RGMa functions as a BMP co-receptor.

EXPERIMENTAL PROCEDURES
cDNA Subcloning-cDNA encoding murine RGMa was subcloned into the expression vector pCDNA4/HisB (Promega). cDNA encoding the extracellular domain of murine RGMa was amplified by polymerase chain reaction (PCR) and subcloned into the mammalian expression vector pIgplus (R & D Systems, Minneapolis, MN) into the restriction sites BamHI and HindIII in-frame with the Fc portion of human immunoglobulin (IgG) to generate soluble RGMa.Fc fusion protein.
Luciferase Assay-LLC-PK1 cells were transiently transfected with a TGF-␤ responsive firefly luciferase reporter, (CAGA) 12 MLP-Luc (CAGA-Luc, Ref. 25), or a BMP responsive firefly luciferase reporter (BRE-Luc, Ref. 6) (both kindly provided by Peter ten Dijke, Netherlands Cancer Institute), in combination with pRL-TK Renilla luciferase vector (Promega) in a ratio of 10:1 to control for transfection efficiency, with or without co-transfection with RGMa cDNA. Forty-eight hours after transfection, cells were serum-starved in DMEM supplemented with 1% FBS for 6 h and treated with varying amounts of TGF-␤1, BMP-2, BMP-4, or BMP-7 ligands (R & D Systems) for 16 h, in the absence or presence of noggin (R & D Systems) or anti-BMP-2/4 antibody (R & D Systems). Cells were lysed, and luciferase activity was determined with the Dual Reporter Assay (Promega) according to the manufacturer's instructions. Experiments were performed in duplicate or triplicate wells. Relative luciferase units (RLU) were calculated as the ratio of firefly (reporter) and Renilla (transfection control) luciferase values.
Reverse Transcription Polymerase Chain Reaction (RT-PCR)-LLC-PK1 cells were grown to confluence on 6-cm tissue culture plates. Total RNA was isolated using the RNeasy Mini Kit (Qiagen, Valencia, CA) including DNase digestion with the RNase-Free DNase set (Qiagen) according to the manufacturer's instructions. First strand cDNA synthesis was performed using iScript cDNA Synthesis kit (Bio-Rad) according to the manufacturer's instructions. Transcripts of BMP-2 were amplified using the forward primer 5Ј-CGTGACCAGACTTTTGGACA-C-3Ј and reverse primer 5Ј-GGCATGATTAGTGGAGTTCAG-3Ј. Transcripts of BMP-4 were amplified using the forward primer 5Ј-AG-CAGCCAAACTATGGGCTA-3Ј and reverse primer 5Ј-TGGTTGAGT-TGAGGTGGTCA-3Ј.
Purification and Characterization of RGMa.Fc-HEK 293 cells stably expressing RGMa.Fc were cultured in DMEM supplemented with 5% FBS using 175-cm 2 multifloor flasks (Denville Scientific, Southplainfield, NJ). RGMa.Fc was purified from the media of stably transfected cells via one-step Protein A affinity chromatography using Hi-Trap rProtein A FF columns (Amersham Biosciences) as previously described (26). Purified protein was eluted with 100 mM glycine-HCl, pH 3.2 and neutralized with 0.3 M Tris-HCl, pH 9 as previously described (26).
Purified human RGMa.Fc was subjected to reducing SDS-PAGE using pre-cast NuPAGE Novex 4 -12% Bis-Tris gels (Invitrogen), and gels were stained with Bio-safe Coomassie Blue (Bio-Rad) to determine the purity of RGMa.Fc and quantify protein concentration. For Western blot analysis, gels were electroblotted to polyvinylidene difluoride filter membranes (Bio-Rad). Membranes were blocked with TBS-T (Tris-buffered saline, 0.2% Tween 20) containing 6% milk powder for 1 h and washed three times with TBS-T for 10 min. Membranes were then probed with an affinity-purified rabbit polyclonal anti-mouse RGMa antibody (␣-RGMa) raised against the peptide RM-DEEVVNAVEDRDSQGLYLC (amino acids 296 -316 in the C terminus of mouse RGMa upstream of its hydrophobic tail) 2 (1:2,000) at 4°C overnight, or with anti-Fc antibody (Jackson ImmunoResearch, West Grove, PA) (1:2000) at room temperature for 1 h in blocking solution. Membranes were washed with TBS-T and incubated with donkey antirabbit or anti-goat horseradish peroxidase-linked secondary antibody (1:10,000) (Santa Cruz Biotechnology, Santa Cruz, CA). Antibody binding was detected with chemiluminescence reagent (PerkinElmer Life Sciences, Boston, MA) and exposed to BioMax XAR film (Kodak, Rochester, NY).
Ligand Iodination-2 g of carrier-free human BMP-2 and BMP-4 ligand (R & D Systems) per reaction was iodinated with [ 125 I] by the modified chloramine-T method as previously described (27).
Binding Assay-25 ng of purified RGMa.Fc in 1ϫ Tris-buffered saline/casein-blocking buffer (BioFX, Owings Mills, MD) or buffer alone was incubated with 125 I-BMP-2 or 125 I-BMP-4 in a total volume of 200 l overnight at 4°C, either alone or in the presence of 80 ng of cold BMP-2, BMP-4, BMP-7, or TGF-␤1 for competition assays. For mixing studies, buffer alone, 10 ng of purified RGMa.Fc alone, 10 ng of ALK6.Fc alone (R & D Systems), or 10 ng each of RGMa.Fc and ALK6.Fc together were incubated in 1ϫ Tris-buffered saline/caseinblocking buffer with 125 I-BMP-2 in a total volume of 200 l. The reaction mix was then incubated for 1.5 h at 4°C on protein A-coated plates (Pierce), plates were washed with wash solution (KPL, Gaithersberg, MD), and individual wells were counted with a standard ␥ counter.
DSS Cross-linking in Solution-100 l of 125 I-BMP-2 or 125 I-BMP-4 (400,000 cpm) were incubated overnight at 4°C with an equal volume of 20 mM HEPES (pH 7.8), 0.1% bovine serum albumin, and protease inhibitors (Roche Diagnostics, Mannheim, Germany) alone, or containing 25 ng of RGMa.Fc or ALK5.Fc (R & D Systems), in the absence or presence of 80 ng of cold BMP-2 or BMP-4. This mixture was incubated in the absence or presence of 2.5 mM disuccinimidyl suberate (DSS, Sigma) in dimethyl sulfoxide for 2 h on ice, followed by quenching of DSS activity with 40 mM Tris (pH 7.5) for 15 min. The mixture was then centrifuged and the supernatant incubated with protein A-Sepharose beads (Amersham Biosciences) at 4°C for 2 h to precipitate hot BMP-2 or -4 bound to RGMa.Fc. Beads were washed with phosphate-buffered saline (PBS) and protein eluted by non-reducing Laemmli sample buffer (Bio-Rad). Eluted protein was separated by SDS-PAGE and analyzed by autoradiography.
For receptor cross-linking studies, 200 ng of RGMa.Fc, 200 ng of ALK6.Fc, and/or 100 ng of BMP-2 were incubated in 100 l of 20 mM HEPES (pH 7.8), 0.1% bovine serum albumin, and protease inhibitors at 4°C overnight. The mixtures were then cross-linked with DSS, incubated with protein A beads, and eluted with non-reducing Laemmli sample buffer as described above. The protein complex was then separated by non-reducing SDS-PAGE, electroblotted to polyvinylidene difluoride membranes, and analyzed by Western blot using RGMa antibody (1:2000) as described above for RGMa.Fc.
Northern Blot-Adult rat total RNA was separated on a 1.5% formaldehyde agarose gel and blotted onto GeneScreen Plus membrane (PerkinElmer Life Sciences) as previously described (28) Immunohistochemistry-Freshly dissected adult rat lumbar spinal cord was embedded in OCT (Sakura, Tokyo, Japan), frozen on dry ice, cut by cryostat in 16-m sections, and stored at Ϫ80°C. Spinal sections were fixed in 4% paraformaldehyde, washed three times in PBS, and incubated for 1 h at room temperature in blocking buffer (1% bovine serum albumin, 0.5% Triton X in PBS). Fixed sections were incubated overnight at 4°C in blocking buffer with rabbit polyclonal ␣-RGMa (1:500) or rabbit polyclonal ␣-p-Smad1/5/8 (1:100), in combination with mouse monoclonal anti-neuron-specific nuclear protein antibody (␣-NeuN, 1:1000) (Chemicon, Temecula, CA) to visualize neuronal cell bodies. Sections were then washed three times in PBS and incubated with cyanin 3 (Cy3)-conjugated antirabbit and fluorescein isothiocyanate (FITC)-conjugated anti-mouse secondary antibodies (1:200 each, Jackson ImmunoResearch) for 1 h at room temperature. Finally, sections were washed three times in PBS and visualized by fluorescence microscopy. Transfected cells were then incubated with or without BMP-2 or TGF-␤1 for 16 h followed by measurement of luciferase activity. In the absence of RGMa, stimulation with BMP-2 or TGF-␤1 increased the relative luciferase activity for their respective reporters compared with unstimulated cells (Fig. 1, A and B, compare bars 2 to 1). Co-transfection with RGMa similarly increased BRE luciferase activity even in the absence of exogenous BMP ligand (Fig. 1A, bar 3). The RGMa-mediated BMP signaling was dose-dependent ( Fig. 1C, white bars), reaching a peak at about 200 ng of cDNA per transfection. RGMa also augmented signaling produced by exogenous BMP-2 ( Fig. 1C, black bars). In contrast, co-transfection with RGMa (up to 1 g) did not increase the TGF-␤ responsive CAGA luciferase activity above baseline (Fig. 1B, bar 3). Similar results were seen in another cell line (HepG2 cells, data not shown). Taken together, these results demonstrate that like DRAGON, RGMa mediates BMP, but not TGF-␤, signaling.

RGMa Mediates BMP, but Not TGF-␤, Signaling-Because
RGMa-mediated BMP Signaling Is Ligand-dependent-The ability of RGMa to generate BMP signals even in the absence of exogenous BMP ligand raises the question of whether RGMa acts in a ligand-independent manner, or whether it augments signaling by endogenous BMP ligands. To investigate this question, we examined whether RGMa-mediated signaling was inhibited by noggin, a soluble BMP inhibitor that binds and sequesters BMP ligands, barring access to membrane receptors (4,29). The effects of noggin on exogenous BMP-2 and TGF-␤1 stimulation were used as positive and negative controls respectively. Results were confirmed using a neutralizing antibody against BMP-2 and BMP-4 (␣-BMP-2/4). In the absence of noggin, co-transfection with RGMa cDNA increased BRE luciferase activity 8-fold above baseline ( Fig. 2A, compare bar 3 to 1). Similarly, exogenous BMP-2 increased BRE luciferase activity 15-fold over baseline ( Fig. 2A, compare bar 5 to 1). This stimulation by either RGMa transfection or exogenous BMP-2 was blocked by noggin ( Fig. 2A, bars 4 and 6). Noggin also decreased basal BMP signaling in cells neither transfected with RGMA nor stimulated with exogenous BMP-2 ( Fig. 2A, compare bar 2 to 1). In contrast, noggin did not affect TGF-␤1 induced CAGA luciferase activity (data not shown). Similar results were seen with a neutralizing antibody against BMP-2 and BMP-4 (Fig. 2B). These data suggest that RGMa generates BMP signals in a ligand-dependent manner, presumably via endogenously expressed BMP ligands. This hypothesis is supported by the fact that mRNA for both BMP-2 and BMP-4 were detected in these cells by RT-PCR (Fig. 2C).
Production and Characterization of Soluble RGMa.Fc Fusion Protein-To further characterize the role of RGMa in the BMP signaling pathway, we produced RGMa.Fc fusion protein by fusing the extracellular domain of RGMa to the Fc portion of human IgG. We also generated an affinity-purified rabbit polyclonal antibody raised against a C-terminal peptide sequence of RGMa upstream of its GPI anchor (␣-RGMa). Purified RG-Ma.Fc was analyzed by reducing SDS-PAGE followed by Western blot using anti-human Fc antibody (␣-Fc) and ␣-RGMa. Both antibodies recognized two bands of ϳ60 and 75 kDa not seen in mock-transfected cells, confirming the presence of both the RGMa and Fc domains, and validating the RGMa antibody (Fig. 3). These bands were not seen when the RGMa antibody was preincubated with competing antigenic peptide (data not shown). The larger band is consistent with the predicted size of the full-length RGMa.Fc fusion protein, and the smaller band is consistent with the predicted size of RGMa.Fc fusion protein, which has been proteolytically cleaved as previously described for both the chick (15) and mouse homologues (18) of RGMa.
RGMa.Fc Binds Selectively to BMP-2 and BMP-4, but Not BMP-7 or TGF-␤1-Next, we tested whether RGMa directly interacted with BMP ligands using soluble RGMa.Fc fusion protein in a cell-free binding system. RGMa.Fc was incubated overnight with 125 I-BMP-2 with or without excess cold BMP-2, -4, -7 or TGF-␤1, followed by incubation on protein A-coated plates and determination of radioactivity. RGMa.Fc bound to 125 I-BMP-2 (Fig. 4A, compare bar 2 to 1), and this binding was competitively inhibited by excess cold BMP-2 or BMP-4, but not by BMP-7 or TGF-␤1. Similar findings were seen with 125 I-BMP-4 (data not shown).
As supporting evidence for the interaction between RGMa.Fc and BMP-2 and -4, chemical cross-linking experiments were performed using DSS in a cell-free system. 125 I-BMP-4 was cross-linked with RGMa.Fc in the presence of DSS (Fig. 4B, bar 4), and this cross-linking was inhibited by excess cold BMP-4 ( Fig. 4B, bar 5). As negative controls, no band was seen in the absence of DSS (Fig. 4B, bars 1 and 2) or when buffer alone (Fig. 4B, bar 3) or ALK5.Fc (a TGF-␤ type I receptor, Fig. 4B, bar 6) was used in place of RGMa.Fc. Similar results were seen for cross-linking with 125 I-BMP-2 (data not shown). Taken together, these data suggest that RGMa.Fc binds directly and selectively to BMP-2 and BMP-4, but not BMP-7 or TGF-␤1. RGMa Mediates BMP signaling via BMP type I receptors-We next investigated whether RGMa acts via the classical BMP signaling pathway through BMP receptors. Dominant negative mutants of BMP type I receptors ALK3 (ALK3 DN) and ALK6 (ALK6 DN), which are deficient in kinase activity and therefore unable to phosphorylate Smad1/5/8, have been previously described (30,31). We therefore examined the effects of co-transfection with dominant negative ALK3 and ALK6 mutants on RGMa-mediated BMP signaling. The effect of these mutants on exogenous BMP was also studied as a control. Transfection with RGMa or incubation of cells with exogenous BMP-2 increased BRE luciferase activity 6 -10-fold above baseline (Fig. 5A, compare bar 2 to 1 and 6 to 5). This stimulation by either RGMa or exogenous BMP-2 was blocked by co-transfection with dominant negative ALK3 (Fig. 5A, bars 3 and 7) or dominant negative ALK6 (Fig. 5A,  bars 4 and 8).
To determine whether RGMa interacted directly with BMP type I receptors, purified RGMa.Fc, ALK6.Fc, and/or BMP-2 were incubated in solution either individually or in various combinations, in the presence of the cross-linker DSS. Complexes were pulled-down with protein A beads and analyzed by non-reducing SDS-PAGE, followed by Western blot with A, 400,000 counts 125 I-BMP-2 was incubated overnight alone (Background) or in combination with 25 ng of RGMa.Fc, in the absence (Total binding) or presence of excess unlabeled BMP-2, -4, -7, or TGF-␤1 as indicated, followed by incubation on protein A-coated plates and determination of radioactivity using a standard ␥ counter. B, buffer alone (C), 25 ng RGMa.Fc, or TGF-␤ type I receptor ALK5.Fc were incubated with 400,000 counts of 125 I-BMP-4 with or without excess cold BMP-4 overnight at 4°C. This mixture was then incubated in the absence (ϪDSS) or presence of 2.5 mM DSS in Me 2 SO (ϩDSS) as indicated for 2 h at 4°C. After quenching of DSS activity, the mixture was incubated with protein A beads at 4°C for 2 h, and the eluted protein complex analyzed by non-reducing SDS-PAGE, followed by autoradiography.
␣-RGMa. As previously shown, RGMa.Fc formed a complex in solution with BMP-2, demonstrated by a more slowly migrating band under non-reducing conditions compared with RGMa.Fc alone (Fig. 5B, compare arrowhead in lane 4 to lane 3). RG-Ma.Fc also formed a complex with ALK6.Fc, even in the absence of BMP-2 (Fig. 5B, compare arrowhead in lane 5 to lane 3). In the presence of BMP-2, an even larger shift was seen, suggesting the possible formation of a complex containing RG-Ma.Fc, ALK6.Fc, and BMP-2 (Fig. 5B, lane 6, arrowhead). No bands were seen for ALK6.Fc or BMP-2 in the absence of RGMa.Fc (Fig. 5B, lanes 1 and 2).
BMP ligands exhibit high affinity for BMP type I receptors and low affinity for BMP type II receptors (1). We therefore tested whether the combination of RGMa and BMP type I receptors increased binding of BMP ligands compared with BMP type I receptors alone. Purified RGMa.Fc and ALK6.Fc were incubated overnight in solution with 125 I-BMP-2, followed by incubation on protein A-coated plates and determination of radioactivity. RGMa.Fc and ALK6.Fc alone each significantly bound 125 I-BMP-2 (Fig. 5C, compare bars 2 and 3 to 1). As a negative control, BMP type II receptor was unable to bind (data not shown). The combination of RGMa.Fc and ALK6.Fc increased binding to 125 I-BMP-2 compared with ALK6.Fc alone (Fig. 5C, compare bar 4 to 3).
As supporting evidence that RGMa mediates BMP signaling through the Smad signaling pathway, we examined the effect of RGMa expression on phosphorylation of endogenous Smad1/ 5/8. LLC-PK1 cells were transiently transfected with RGMa cDNA and cell lysates were assayed for p-Smad1/5/8 by Western blot (Fig. 6B, ␣-p-Smad1/5/8). Blots were stripped and reprobed for total Smad1 as a loading control (Fig. 6B, ␣-Smad1). Results were compared with mock-transfected cells as a negative control, and cells stimulated for 2 h with 50 ng/ml exogenous BMP-2 as a positive control. Expression of p-Smad1/ 5/8 relative to Smad 1 was quantitated using IPLab Spectrum software (Fig. 6C). Consistent with our prior data supporting the notion of endogenous BMP signaling in these cells, mocktransfected cells (without RGMa or exogenous BMP-2 stimulation) did have some basal level of p-Smad1/5/8 (Fig. 6B, left two  lanes). Transfection with RGMa cDNA increased p-Smad1/5/8 levels compared with mock-transfected cells (Fig. 6B, compare middle three lanes to left two lanes; Fig. 6C, compare bar 2 to 1). As a positive control, stimulation with exogenous BMP-2 also increased p-Smad1/5/8 levels (Fig. 6B, right two lanes; Fig. 6C,  bar 3).
RGMa Is Widely Expressed-Previous studies have focused on detailing the expression pattern of endogenous RGMa in the central nervous system and during development (17)(18)(19)(20). To begin to elucidate the role of RGMa in vivo, we performed Northern blot analysis of endogenous RGMa expression in a variety of adult rat tissues. RGMa message is widely expressed in many of the tissues tested, including heart, brain, lung, liver, skin, kidney, and testis (Fig. 7). Two distinct bands were seen in some tissues, possibly representing alternative transcription initiation or alternative splicing.
BMP Signaling Occurs in Neurons of the Adult Spinal Cord Which Express RGMa-Next, we explored whether RGMa expression in vivo correlates with its hypothesized role as a co-receptor for BMP signaling. RGMa mRNA has been previ-ously documented to be widely expressed in a complementary fashion to DRAGON in the central nervous system (17)(18)(19)(20), including ventral horn neurons of the spinal cord (17). We therefore determined whether RGMa protein was expressed in ventral horn neurons, and we examined whether these neurons also showed evidence of BMP signaling, i.e. nuclear accumulation of p-Smad1/5/8. Adult rat spinal cord sections were analyzed by immunofluorescence microscopy with ␣-RGMa, ␣-p-Smad1/5/8, and/or anti-neuron-specific nuclear protein antibody (␣-NeuN) to visualize neuronal cell bodies (17). RGMa staining colocalized with NeuN staining in ventral horn motor neurons (Fig. 8, panels A-C). Ventral horn motor neurons were also positive for nuclear p-Smad1/5/8 (Fig. 8, panels D-F), suggesting that there is basal signal transduction via the BMP pathway in these cells. Thus, endogenous RGMa is expressed in ventral horn motor neurons, which also generate BMP signals, consistent with a role for RGMa as a BMP co-receptor in vivo.

DISCUSSION
BMPs are members of the TGF-␤ superfamily of ligands, which play a pleitropic role in vertebrate development and adult tissues (2)(3)(4). These functions require tight spatiotemporal regulation and specific activation via receptor complexes of particular intracellular signaling pathways. In order to generate specificity and to finely tune these signals, regulation occurs at multiple levels extracellularly, at the membrane surface, and intracellularly (1,4,36).
For the BMP pathway, most regulatory mechanisms identified to date are inhibitory. Soluble BMP antagonists such as noggin, chordin, chordin-like, follistatin, FSRP, DAN, cerebrus, and gremlin bind BMPs in the extracellular space and mask receptor binding interfaces for BMP type I and type II receptors (4,29). At the membrane surface, BAMBI (BMP and activin receptor membrane-bound inhibitor), which is structurally related to type I receptors in the extracellular domain but lacks the intracellular serine/threonine kinase domain, inhibits BMP signals by stably associating with type II receptors and pre- venting formation of the active receptor complexes (37). Inhibin, in concert with its co-receptor betaglycan (also known as the TGF-␤ type III receptor), competes with BMPs for access to BMP type II receptors (38). Inside the cell, inhibitory Smads (Smad 6 and Smad 7) inhibit signaling by either interacting with phosphorylated type I receptors to prevent activation of receptor-activated Smads (39 -41), or through competition to prevent formation of the receptor-activated Smad/Co-Smad complex (42). Other intracellular molecules Smurf1 and Smurf2 (Smad ubiquitination regulatory factors), selectively target-activated type I receptors and Smad proteins for degradation (43)(44)(45).
For other TGF-␤ superfamily members, accessory or co-receptors also play an important regulatory role to promote or inhibit ligand binding (1, 10 -13). Recently, we identified DRAGON (RGMb) as the first known co-receptor for BMP signaling (14). We therefore investigated whether another RGM family member, RGMa, was similarly involved in the BMP signaling pathway. Here, we have demonstrated that RGMa is a BMP co-receptor, which enhances cellular responses to BMP, but not TGF-␤.
Although transfection of RGMa into LLC-PK1 cells enhanced BMP signal transduction without exogenously added ligand, our results suggest that this is a ligand-dependent process. Ligand dependence is supported by the fact that RGMa-mediated BMP signaling was inhibited by noggin, a soluble BMP inhibitor which binds and sequesters BMP ligands, preventing access to BMP receptors (4. 29). However, this does not pinpoint the endogenous ligand(s) responsible for RGMa-mediated BMP signaling in these cells, because noggin has been shown to bind and antagonize several BMPs, including BMP-2, BMP-4, and BMP-7, as well as some other TGF-␤ superfamily members, including growth and differentiation factor 5 (4,29,46,47). Our findings that a neutralizing antibody to BMP-2 and BMP-4 also inhibited RGMamediated BMP signaling suggest that the major endogenous ligand(s) in these cells may be BMP-2 and/or BMP-4. Indeed, RT-PCR confirmed that these cells do endogenously express both BMP-2 and BMP-4. However, this does not preclude the possibility that some portion of the signaling is related to other BMP ligands. Indeed, the manufacturer of this neutralizing antibody does report some minimal cross-reactivity to other BMP ligands. Additionally, this antibody did not completely inhibit RGMamediated BMP signaling to baseline levels.
Further evidence for the role of RGMa as a co-receptor for BMP-2 and/or BMP-4 is provided by binding and cross-linking studies of purified RGMa.Fc in solution. These assays allow the determination of the binding properties of single types of receptors and combinations of receptors in isolation, avoiding the presence of any confounding co-expressed accessory proteins that may also be present at the cell surface (26). In these studies, RGMa.Fc bound directly and specifically to 125 I-BMP-2 and 125 I-BMP-4, and this binding was competitively inhibited by excess cold BMP-2 and BMP-4, but not BMP-7 or TGF-␤1. RGMa.Fc also formed a complex with the BMP type I receptor ALK6.Fc, and the presence of RGMa.Fc in combination with ALK6.Fc increased binding to 125 I-BMP-2 compared with ALK6.Fc alone. No further increase in binding was seen with the addition of the BMP type II receptor.Fc (data not shown). Although this increased binding was additive, not synergistic, the fact that RGMa.Fc formed a complex with ALK6.Fc raises the possibility that on the cell surface, RGMa may associate with BMP type I receptors, thereby increasing overall binding of BMP ligands to the receptor complex and enhancing BMP signal transduction. However, the detailed mechanism by which RGMa enhances BMP signaling remains to be fully elucidated.
While the binding and cross-linking experiments were performed using RGMa.Fc and BMP receptor.Fc fusion proteins in solution, support for the role of GPI-anchored, cell-surface RGMa in the BMP signaling pathway is provided by our findings that RGMa-mediated BMP signaling was inhibited by dominant negative BMP type I receptors and by dominant negative Smad1, using a BMP-responsive luciferase reporter assay in cell culture. Additionally, transfection of RGMa into LLC-PK1 cells increased phosphorylation of endogenous Smad1/5/8, and increased expression of endogenous Id1 protein, an important target gene of BMP signaling in many tissues (5)(6)(7)(8)(9). The physiologic role of endogenous RGMa as a BMP co-receptor in vivo is supported by our finding that RGMa was expressed in spinal cord neurons, which also showed nuclear accumulation of p-Smad1/5/8, indicative of BMP signal transduction in these cells. Future studies will be needed to determine the functional significance of BMP signaling in these cells. Future studies will also be needed to determine whether the role of RGMa as a BMP co-receptor is independent of, or related to, its other known functions in mediating axonal repulsive guidance (15,22) and neural tube closure (18). Interestingly, RGMa also acts as a cell survival factor by binding to the receptor neogenin and inhibiting the neogenin pro-apoptotic activity (21,23). Whether RGMa binding to neogenin is affected by BMPs, or conversely, whether RGMa binding to BMPs is affected by neogenin, remains unknown.
RGMa and DRAGON are members of the RGM family of proteins, which also includes the juvenile hemochromatosis gene HJV. RGM family members are highly conserved across vertebrates and invertebrates and share significant sequence homology as well as similar structural features (15)(16)(17)(18)(19)(20). Preliminary data suggest that HJV also mediates BMP signaling, 3 indicating that these family members all share the ability to act as co-receptors to enhance BMP signals.
Our results do not reveal any differences between RGMa and DRAGON in regard to their function as BMP co-receptors. Both bind to BMP-2 and BMP-4, but not BMP-7 or other members of the TGF-␤ superfamily. Both signal via the BMP type I receptors ALK3 and ALK6 and Smad1. Thus, it remains unclear what differentiates these family members with regard to their role as BMP co-receptors, and why different family members have evolved. Further studies will be needed to determine the specificity and affinity of RGM family members for interactions with the full range of BMP ligands and receptors. Whether additional co-receptors exist for other BMP ligands remains unknown. Interestingly, a secreted protein, Kielin/chordin-like protein, has recently been described as a paracrine enhancer of BMP-7 signaling (48). Thus, multiple enhancing regulatory mechanisms may exist for BMP signaling to complement the inhibitory regulatory mechanisms that have been described (1,4,36).
The role of RGM family members may be to differentially increase the sensitivity of cells in which they are expressed to low levels of BMP ligands, thus contributing to the tight spatiotemporal regulation of BMP signal transduction. Northern blot analysis of adult rat tissues revealed that endogenous RGMa is expressed in a variety of organs, including heart, brain, lung, liver, skin, kidney, and testis. Although prior studies have reported a more limited distribution, they were largely focused on expression in the central nervous system and during development (17)(18)(19)(20). Additionally, the one previously published Northern blot of RGMa expression in a variety of tissues was limited by underexposure (18), also a possible explanation for their finding only the larger of the two bands seen in our study. Recent work has also demonstrated a broader tissue distribution for DRAGON, including many tissues throughout the reproductive axis (49). We have also found DRAGON expression in the adult rat heart, liver, and kidney by Northern blot (data not shown). While HJV expression has been described predominantly in the liver, cardiac muscle, and skeletal muscle (16 -20), one recent study suggested that HJV is also expressed in the adult mouse brain, lung, spleen, kidney, testis, blood, stomach, and intestine (24). Thus, RGM family member expression appears to overlap in a variety of tissues. However, it is possible that the distribution within those tissues is different. For example, in the central nervous system, where RGMa and DRAGON expression have been best characterized, they are predominantly expressed in non-overlapping areas (17)(18)(19)(20).
Our results help to define the first family of proteins which function as BMP co-receptors. We hypothesize that RGM family members increase the sensitivity of cells in which they are expressed to BMP stimulation, thereby allowing these cells to respond earlier or more robustly to a low level of BMP ligand. RGM family members thus represent an important addition to the complex array of regulatory molecules which help to generate specificity and tightly coordinate cellular responses to BMP ligands.