Juvenile hemochromatosis is a severe and rapidly progressing hereditary disorder of iron overload, and it is caused primarily by defects in the gene encoding repulsive guidance molecule c/hemojuvelin (RGMc/HJV), a recently identified protein that undergoes a complicated biosynthetic pathway in muscle and liver, leading to cell membrane-linked single-chain and heterodimeric species, and two secreted single-chain isoforms. RGMc modulates expression of the hepatic iron regulatory factor, hepcidin, potentially through effects on signaling by the bone morphogenetic protein (BMP) family of soluble growth factors. To date, little is known about specific pathogenic defects in disease-causing RGMc/HJV proteins. Here we identify functional abnormalities in three juvenile hemochromatosis-linked mutants. Using a combination of approaches, we first show that BMP-2 could interact in biochemical assays with single-chain RGMc species, and also could bind to cell-associated RGMc. Two mouse RGMc amino acid substitution mutants, D165E and G313V (corresponding to human D172E and G320V), also could bind BMP-2, but less effectively than wild-type RGMc, while G92V (human G99V) could not. In contrast, the membrane-spanning protein, neogenin, a receptor for the related molecule, RGMa, preferentially bound membrane-associated heterodimeric RGMc and was able to interact on cells only with wild-type RGMc and G92V. Our results show that different isoforms of RGMc/HJV may play unique physiological roles through defined interactions with distinct signaling proteins and demonstrate that, in some disease-linked RGMc mutants, these interactions are defective.
- repulsive guidance molecule
- juvenile hemochromatosis
- bone morphogenetic protein
iron is an essential element required for many cellular processes, including oxygen transport, energy metabolism, and respiration (1, 11). Iron homeostasis is tightly regulated, and there are major health consequences linked to both its deficiency and excess (1, 11). Inadequate iron impairs hemoglobin synthesis and red blood cell production, while chronic excess leads to tissue damage (1, 6, 11). In mammals, it is thought that the liver-derived hormone, hepcidin, largely controls systemic iron balance by regulating dietary iron intake and release from intracellular stores (1, 6). Normal iron homeostasis is disrupted in hemochromatosis, a heterogeneous hereditary disorder of iron overload that is characterized by low levels of hepcidin, and organ damage and dysfunction caused by chronic accumulation of iron in tissues (6). Juvenile hemochromatosis is a rare early-onset form of this condition that has been linked to the HFE2 locus, which encodes the protein hemojuvelin (6, 30). To date, over two-dozen mutations have been identified in the hemojuvelin gene, with many of the alterations being point mutations in the coding region that lead to amino acid substitutions in the protein (6, 30). Because mice engineered to lack hemojuvelin have diminished hepcidin expression and increased iron accumulation in tissues (12, 28), the evidence strongly supports a key role for hemojuvelin in normal iron metabolism.
Hemojuvelin is identical to RGMc, a member of the repulsive guidance molecule family of proteins that includes RGMa and RGMb/Dragon. RGMc expression is limited to striated muscle and the liver, while RGMa and RGMb are found predominantly in the central nervous system (16, 29, 30, 37, 38). RGMa, the first member of this family to be characterized, was identified as a developmental guidance cue for temporal retinal axons (26) and appears to function in neuronal patterning, survival, and differentiation (23, 24, 29, 34), while RGMb has been shown to enhance neuronal adhesion (37). Members of the RGM family are glycosylphosphatidylinositol (GPI)-anchored proteins with a relatively high degree of amino acid identity (42–50% in mouse) (29, 30, 38), and they share several structural motifs, including a partial von Willebrand factor type D domain, a predicted internal cleavage site (GlyAsp↓ProHis), and a COOH-terminal GPI-addition motif (29, 30, 38).
Recent studies have indicated that RGMs may modulate the actions of bone morphogenetic proteins (BMPs) 2, 4, and 9 (2, 4, 36, 42). BMPs comprise a subset of the transforming growth factor (TGF)-β superfamily of growth factors and are involved in a wide range of developmental processes (39, 43). BMPs are soluble proteins that bind to specific cell-surface receptors, which function as ligand-stimulated serine-threonine protein kinases, and activate several intracellular mediators, including members of the Smad family of transcriptional activators (39, 43). Several reports have concluded that all three RGM proteins when expressed on the cell membrane can function as putative BMP coreceptors and can enhance BMP-stimulated activation of promoter-reporter genes (2, 4, 36, 46). Cell-associated RGMc has been found to increase BMP-2-mediated induction of hepcidin promoter activity and hepcidin gene expression (2, 3), while soluble RGMc in contrast appears to inhibit hepcidin mRNA accumulation (3, 19, 21).
The membrane-spanning protein neogenin was found to be a receptor for RGMa (24, 34, 48). Neogenin along with DCC (deleted in colon cancer) and UNC5H1–4 comprise a family of widely expressed molecules with broad effects in tissue morphogenesis, development, and disease (5, 25, 33). These proteins also are receptors for netrins, a small group of secreted molecules that act as bifunctional neuronal guidance cues, attracting some axons while repelling others (7, 25). Neogenin, but not DCC or UNCs, was shown to mediate the repulsive guidance functions of RGMa and its effects on neuronal survival and differentiation (23, 24, 34). In contrast, the role of neogenin in the biology of RGMc/hemojuvelin is not clear. A single report has shown coimmunoprecipitation of overexpressed RGMc and neogenin in cultured cells (48), with a possible impact on accumulation of soluble RGMc in conditioned culture medium or on cellular uptake of iron (47, 48), but this remains to be substantiated.
The RGMc/hemojuvelin protein undergoes a complex series of biosynthetic and processing steps that result in accumulation of two forms on the cell membrane and two in extracellular fluid (15). Membrane-associated RGMc consists of both a single-chain 50-kDa polypeptide and a disulfide-bonded heterodimer composed of 20-kDa NH2-terminal and 35-kDa COOH-terminal subunits (15, 19, 41, 48), while soluble RGMc consists of 50- and 40-kDa species that are derived from the membrane-bound 50-kDa protein (15). Although the responsible biochemical mechanisms have not been elucidated fully, it appears that 40-kDa soluble RGMc is truncated at its COOH terminus through cleavage by proprotein convertases (20, 40).
To date, little is known about specific pathogenic defects in disease-causing RGMc/hemojuvelin proteins. Here we identify functional abnormalities in three juvenile hemochromatosis-linked mutant RGMc proteins that either impair binding by neogenin or to BMP-2. Using a combination of biochemical and cell-based approaches, we additionally show that neogenin preferentially binds membrane-associated heterodimeric RGMc, while BMP-2 interacts with single-chain RGMc species.
MATERIALS AND METHODS
Antibodies and Reagents
The following antibodies were used: goat anti-neogenin (AF1079) and mouse anti-BMP-2 (MAB3551; R&D Systems, Minneapolis, MN); goat anti-DCC (sc-6535; Santa Cruz Biotechnology, Santa Cruz, CA); goat anti-insulin-like growth factor binding protein-5 (IGFBP-5; sc-6006, Santa Cruz Biotechnology); rabbit anti-pan-cadherin (no. 4068; Cell Signaling Technology, Danvers, MA); mouse anti-T7-tag (no. 69522; Novagen, Madison, WI); Alexa Fluor 680-conjugated goat anti-rabbit IgG, Alexa Fluor 488-conjugated donkey anti-goat IgG, Alexa Fluor 594-conjugated goat anti-rabbit IgG, and Alexa Fluor 488-conjugated goat anti-mouse IgG (all from Molecular Probes, Eugene, OR); IRDye 800-conjugated goat anti-mouse IgG and goat anti-human IgG (Rockland, Gilbertsville, PA); and goat anti-human IgG Cy3 (Jackson ImmunoResearch Labs, West Grove, PA). A rabbit anti-RGMc antibody was generated using His-tagged Escherichia coli-derived mouse RGMc (amino acids 31–350) as antigen, and it was purified by affinity chromatography with mouse RGMc (amino acids 31–350) fused to maltose binding protein (15). The antibody was characterized by showing that the RGMc-maltose binding protein fusion blocked binding to membranes of differentiated muscle cells, which endogenously produce RGMc (15), and by comparison of the pattern of immunoreactive protein bands with muscle cell protein extracts, and with Cos-7 cells expressing recombinant RGMc, which were identical (15). Fetal bovine serum (FBS) was purchased from Hyclone (Logan, UT), Dulbecco's modified Eagle's medium (DMEM) and phosphate-buffered saline (PBS) from Mediatech, (Herndon, VA), and Ultralow IgG FBS and Superscript III first-strand cDNA synthesis system from Invitrogen (Carlsbad, CA). Protease inhibitor, NBT/BCIP tablet, and restriction enzymes were from Roche Applied Sciences (Indianapolis, IN). Okadaic acid and decanoyl-Arg-Val-Lys-Arg-chloromethylketone (RVKR) were purchased from Alexis Biochemicals (San Diego, CA). Sodium orthovanadate, heparin, Type IIS heparin-agarose, and p-nitrophenyl phosphate tablets were from Sigma-Aldrich (St. Louis, MO). Quik-Change site-directed mutagenesis kit was from Stratagene (La Jolla, CA), Advantage2 GC PCR kit from BD Biosciences (Mountain View, CA), and T4 DNA ligase from Fermentas (Hanover, MD). TransIT-LT1 was from Mirus Bio (Madison, WI). Bicinchoninic acid protein assay kit, GelCode Blue stain reagent, EZ-link sulfo-LC-NHS-biotin, and streptavidin agarose were from Pierce Biotechnology (Rockford, IL). Protein A-Sepharose 4B conjugate was purchased from Zymed (South San Francisco, CA). NitroBind nitrocellulose was from GE Water & Process Technologies, and AquaBlock tm/EIA/WIB solution from East Coast Biologicals (North Berwick, ME). All other chemicals were reagent grade and purchased from commercial suppliers.
Cell Culture and Transfections
The following cell lines were purchased from American Type Culture Collection: Cos-7 (CRL-1651), Hek293 (CRL-11268), Hep3B (HB-8064), and C3H10T1/2 (CCL-226). Cells were grown in DMEM plus 10% FCS at 37°C in humidified air with 5% CO2 and were transfected at ∼70% of confluent density using TransIT-LT1.
Protein Isolation, Purification, and Immunoblotting
Conditions for preparation of whole cell protein lysates, SDS-PAGE, and immunoblotting have been described elsewhere (15). Primary antibodies included the following: RGMc (10 ng/ml), cadherin (1:1,000 dilution), T7 (1:1,000), and BMP-2 (1:1,000). IRDye 800-conjugated goat anti-human IgG (1:10,000) was used to detect neogenin-Fc or DCC-Fc.
Generation of RGMc Mutants, RGMc-Fc, and RGMc-Alkaline Phosphatase Expression Plasmids
Codon substitutions G92V and D165E were engineered into hemagglutinin (HA)-tagged mouse RGMc (15) by site-directed mutagenesis and verified by DNA sequencing. G313V has been described elsewhere(15). To generate RGMc-Fc fusion proteins, codons 1–391 of mouse RGMc or the indicated mutants were subcloned via 5′ EcoRI and 3′ NotI sites into pcDNA3 in frame with the Fc region of human IgG2 (from pFuse-hFc1, Invitrogen). For RGMc-alkaline phosphatase (AP), mouse RGMc (codons 1–391) was subcloned in frame with AP in APtag-2 (Genhunter). For AP-RGMc, mouse RGMc (codons 33–420) was subcloned in frame with AP in APtag-4 (Genhunter). To generate RGMa-AP, mouse RGMa (codons 1–406), cloned by RT-PCR from E14.5 fetal mouse brain, was subcloned in frame with AP in APtag-2. Xenopus noggin [from Jan Christian, Oregon Health & Science University (OHSU)] was cloned in frame with APtag-4 to generate AP-noggin. Human netrin-1-AP (14) was obtained from K. Guan (University of Michigan), and AP-Jeb from Dr. Joseph Weiss (OHSU). APtag-4 was the source for AP alone.
Preparation of Neogenin, Neogenin-Fc, DCC, DCC-Fc, UNC5H2, and UNC5H3 Expression Plasmids
Full-length cDNAs encoding mouse neogenin (from H. Cooper, University of Queensland, Brisbane, Australia), DCC (from H. Cooper), UNC5H2 (from D. Engelkamp, Max Planck Institute, Frankfurt, Germany), and UNC5H3 (from S. Ackerman, Jackson Laboratories, Mount Desert Island, ME) were subcloned into pcDNA3. An in-frame T7 epitope tag was added to neogenin at its COOH terminus by PCR. The extracellular region of mouse neogenin (codons 1–1,133) was subcloned by PCR into pcDNA3 via 5′ HindIII and 3′ EcoRI sites in frame with the Fc region of human IgG2 to generate neogenin-Fc. The extracellular region of mouse DCC (codons 1–1,468) was subcloned via 5′ EcoRI and 3′ NotI sites to produce DCC-Fc. All modifications were confirmed by DNA sequencing.
Production and Purification of Soluble RGMc, BMP-2, AP Fusion Proteins, RGMc-Fc, Neogenin-Fc, and DCC-Fc
Soluble HA-RGMc was produced by infection of Hep3B cells with Ad-HA-RGMcΔGPI and Ad-tTA (15) at multiplicities of infection (MOI) of 100 and 50, respectively, and medium was collected after a 5-day incubation in DMEM with 2% ultralow IgG FBS. For some experiments, conditioned medium was used directly, and for others, RGMc was first purified by affinity chromatography using antigen-purified RGMc polyclonal antisera coupled to CNBr-activated Sepharose (Amersham-Pharmacia Biosciences, Piscataway, NJ). To generate Ad-BMP-2, the coding region of rat BMP-2 was amplified by RT-PCR, sequence verified, and subcloned into pShuttle-CMV (Qbiogene, Irvine, CA) for adenovirus generation as described previously (45). After infection of C3H10T1/2 cells (MOI 500), BMP-2 was collected from culture medium 5 days later and was enriched by chromatography on heparin agarose (22). For some experiments, conditioned medium was used directly. AP fusion proteins were collected from conditioned medium of transfected Cos-7 cells incubated in DMEM plus 2% FBS (with 2 μg/ml heparin for netrin-AP) for 5 days. Fc-fusion proteins were purified from conditioned culture medium of transfected Cos-7 or Hek293 cells after incubation in DMEM with 2% ultralow IgG FBS by affinity chromatography with protein A-Sepharose, followed by elution with 100 mM glycine pH 3.0 and immediate neutralization with 100 mM Tris-Cl pH 9.5. RGMc-Fc fusion proteins were collected in medium containing RVKR (5 μM). Enrichment was assessed by SDS-PAGE and staining with GelCode Blue, and concentrations were estimated by comparison with bovine serum albumin standards.
Cell-Surface Binding Assays
Soluble AP proteins.
Cos-7 cells on gelatin-coated dishes were transiently transfected with neogenin, DCC, UNC5H2, UNC5H3, or empty vector (pcDNA3), and binding assays were performed by incubating cells with conditioned medium containing equivalent amounts of AP-fusion proteins or AP alone for 90 min at 20°C, followed by six washes with buffer containing Hanks' balanced salt solution, 0.5 mg/ml BSA, 0.1% NaN3, and 20 mM HEPES, pH 7, and incubation at 65°C for 100 min to inactivate endogenous APs. Surface protein binding was assessed by a colorimetric AP activity assay (8).
Cos-7 cells grown on gelatin-coated dishes were transiently transfected with neogenin, DCC, or empty vector and were incubated with conditioned medium containing HA-RGMcΔGPI for 90 min at 20°C, followed by three washes with PBS, fixation with 4% paraformaldehyde in PBS for 15 min at 20°C, and incubation with acetone:methanol (1:1). Proteins were detected by immunocytochemistry using anti-RGMc (30 ng/ml) (15) and either anti-neogenin (1:100 dilution) or anti-DCC (1:500).
Soluble neogenin-Fc and DCC-Fc.
Cos-7 cells grown on gelatin-coated dishes were transfected with wild-type RGMc or amino acid substitution mutants. Purified neogenin-Fc or DCC-Fc (1 μg/ml) was added 24 h later in PBS plus 0.5 mg/ml BSA for 90 min at 20°C, followed by three washes with PBS, fixation with 4% paraformaldehyde in PBS, addition of antibodies, and detection by immunocytochemistry.
Cos-7 cells were grown on gelatin-coated dishes and transfected with wild-type RGMc or amino acid substitution mutants and incubated with conditioned medium containing BMP-2 for 90 min at 20°C, followed by washes, fixation, and detection by immunocytochemistry, using anti-RGMc and anti-BMP-2 (1:250 dilution). All images were captured with a Roper Scientific Cool SNAP Fx camera attached to a Nikon Eclipse TE300 fluorescent microscope using IP Labs Scientific Image Processing software (version 3.9.4r2, Scanalytics).
In Vitro Solution Binding Assays and Ligand Blotting
Neogenin-Fc and DCC-Fc.
Purified neogenin-Fc or DCC-Fc (2 μg) was incubated with 1 μg of purified HA-RGMcΔGPI in PBS with 20 mM HEPES pH 7.3 and 0.1% Tween 20 for 3 h at 20°C, followed by addition of protein A-Sepharose for 2 h at 20°C. Proteins were eluted with 50 μl of 100 mM glycine pH 3.0 and were neutralized with 100 mM Tris-Cl pH 9.5, followed by SDS-PAGE and immunoblotting.
RGMc and BMP-2.
Conditioned medium containing BMP-2 (1 ml at ∼500–1,000 ng/ml) and HA-RGMcΔGPI (200 μl) were incubated overnight at 4°C, followed by immunoprecipitation with anti-BMP-2 (10 μg) and protein A-Sepharose. Protein complexes were assessed by SDS-PAGE and immunoblotting. For interactions of BMP-2 with AP fusion proteins, equivalent amounts of AP enzymatic activity was incubated with conditioned medium containing BMP-2 (1 ml) overnight at 4°C, followed by immunoprecipitation with anti-BMP-2 (10 μg) and protein A-Sepharose, and AP assays (8). For pull-down assays, purified RGMc-Fc (500 ng) was mixed with purified recombinant BMP-2 (250 ng) for 3 h at 20°C in PBS with 20 mM HEPES pH 7.3 and 0.1% Tween 20, followed by addition of protein A-Sepharose (10 μl of a 50% slurry) for 2 h at 25°C. After washing and elution, interacting proteins were analyzed by immunoblotting. For ligand blotting, affinity-purified HA-RGMcΔGPI (100 ng) and purified IGFBP-5 [100 ng (27)] were separated by nondenaturing SDS-PAGE, transferred to nitrocellulose, and incubated with purified BMP-2 [50 ng/ml] in PBS plus 0.1% Tween 20 for 16 h at 4°C, followed by detection by immunoblotting.
Cell-Surface Biotin Labeling
Monolayer cultures were incubated with 1 mg/ml EZ-link sulfo-LC NHS-biotin for 30 min at 4°C and were quenched by three washes with 100 mM glycine in PBS. After addition of streptavidin-agarose to protein extracts, labeled proteins were detected after SDS-PAGE by immunoblotting (15). To detect binding of added RGMc, Cos-7 cells expressing neogenin or empty vector were incubated with conditioned medium containing HA-RGMcΔGPI for 3 h at 4°C, followed by addition of EZ-link, pull-down of protein extracts with streptavidin-agarose, SDS-PAGE, and immunoblotting.
Characterization of Biosynthetic Defects in Selected Juvenile Hemochromatosis-Associated RGMc Mutants
More than two-dozen juvenile hemochromatosis-linked mutations have been identified within the human hemojuvelin/RGMc gene, including >20 different amino acid substitutions, which appear to be scattered throughout the protein (6, 30). Because very little is known about these mutant proteins, their pathogenic defects remain to be discovered. We sought to test the hypothesis that some of these mutant proteins have defects in their biosynthesis, processing, or interactions with other molecules. To address this question, three disease-linked amino acid substitutions were engineered into mouse RGMc, comprising alterations at G92V (corresponding to human G99V), within the intramolecular cleavage sequence (D165E, equivalent to human D172E), and at G313V (human G320V), the most prevalent human juvenile hemochromatosis mutation (6) (Fig. 1A). All three mutant proteins were successfully expressed in Cos-7 cells and could be detected in protein extracts, on the cell membrane after surface biotin labeling (Fig. 1B), and in the extracellular fluid (Fig. 1C). Both 50- and 40-kDa species were detected in the culture medium (Fig. 1C), although the larger protein was seen only after incubation of cells with the proprotein convertase inhibitor, RVKR, in agreement with recent reports (20, 41). Only wild-type RGMc and G92V appeared to be efficiently processed into the membrane-associated 20- and 35-kDa heterodimer (Fig. 1B, gray arrows), yet lack of intramolecular processing did not prevent either cell-surface expression of mutant RGMc proteins or their accumulation in extracellular fluid.
BMP-2 Preferentially Binds to 40-kDa RGMc
Recent studies have demonstrated that RGMa, RGMb, and RGMc are able to bind BMP-2, -4, and -9 and have suggested that each RGM may function as a BMP coreceptor (2, 4, 36, 46). We performed a series of in vitro binding experiments with RGMc and recombinant BMP-2 to determine whether these two proteins interacted directly with each other. We first asked whether BMP-2 could coprecipitate an AP-RGMc fusion protein. AP-RGMc, AP alone, or controls AP-noggin [noggin is a high-affinity binding protein for BMPs (39, 43)] or AP-Jeb [Jeb is a secreted Drosophila protein that binds to a tyrosine kinase receptor (17)] were incubated with or without BMP-2, and enzymatic activity was assessed after immunoprecipitation with a BMP-2 antibody. Only AP-RGMc and AP-noggin were recovered in BMP-2 immunoprecipitates, and the AP activities were similar (Fig. 2A). We also examined interactions with soluble purified RGMc by ligand blotting. As shown in Fig. 2B, BMP-2 bound to both 40- and 50-kDa RGMc species but did not bind to another purified secreted protein, IGFBP-5. Since nonreducing gel electrophoresis was used to separate the RGMc proteins before ligand binding, heterodimeric disulfide-linked 20- and 35-kDa RGMc comigrated with the 50-kDa single-chain species (15), and we could not determine whether binding occurred to one or both of these RGMc isoforms. We thus used a solution-based assay to look for preferential interactions of BMP-2 with different RGMc species. BMP-2 was incubated with soluble RGMc [derived from RGMcΔGPI, which lacks the GPI attachment sequence (15)], and associated proteins were detected by immunoblotting after immunoprecipitation with anti-BMP-2 antibody. As depicted in Fig. 2C, there was clear enrichment of 40-kDa RGMc over other species. Thus, BMP-2 shows a binding preference for one RGMc protein isoform.
BMP-2 Interacts at the Cell Membrane With Selected Juvenile Hemochromatosis-Associated RGMc Proteins
To extend the results shown in Fig. 2, we performed pull-down experiments by incubating BMP-2 with purified chimeric proteins either containing wild-type RGMc or G92V, D165E, or G313V mutants fused in frame to the Fc fragment of human IgG heavy chain. As shown in Fig. 3A, only wild-type RGMc, D165E, and G313V were able to bind BMP-2 under these conditions, with binding to the wild-type protein being strongest (12% of input, compared with 4% or 1% for D165E and G313V, respectively). We also examined the ability of BMP-2 to bind to the membrane of Cos-7 cells expressing wild-type RGMc, RGMcΔGPI, or the three mutants. As seen in Fig. 3B, wild-type RGMc and the mutant proteins were detected on the cell surface after immunocytochemistry of nonpermeabilized cells, whereas RGMcΔGPI was not. BMP-2 colocalized at the membrane with wild-type RGMc and with the D165E and G313V mutants, but not with G92V. Thus, BMP-2 interacts selectively with some disease-linked mutant RGMc proteins and does not bind to others.
RGMc Binds to Neogenin at the Cell Membrane
The membrane-spanning protein neogenin was identified as an RGMa receptor by expression cloning studies using an RGMa-AP fusion protein as ligand and was confirmed by biochemical and functional assays (24, 34). Along with DCC (deleted in colon cancer) and UNC5H1–4, neogenin is a member of a family of six “dependence” receptors that bind netrins and can induce cell death when expressed in the absence of ligands (25, 33). We tested the hypothesis that one of these proteins also could bind RGMc. RGMc-AP, RGMa-AP, netrin-AP, and AP alone were incubated with Cos-7 cells transiently transfected with expression vectors for neogenin, DCC, UNC5H2, or UNC5H3. As seen in Fig. 4A, RGMc-AP (and RGMa-AP) bound to cell membrane-associated neogenin but not to any of the other receptors, while netrin-AP bound to all four proteins as expected (13, 18). None of the AP fusion proteins associated with cells expressing empty vector, and AP alone did not bind under any conditions (Fig. 4A).
We performed additional experiments to address the interaction between RGMc and neogenin. Cos-7 cells expressing either neogenin or DCC were incubated with soluble mouse RGMc. As seen by immunocytochemistry, RGMc became associated with cells only if neogenin was also present (Fig. 4B). We also asked whether soluble neogenin could bind to cell-surface-associated RGMc. Cos-7 cells were transfected with expression plasmids for full-length RGMc or RGMcΔGPI and were incubated with purified chimeric proteins consisting of the extracellular domains of neogenin or DCC fused in frame with the Fc fragment of human IgG heavy chain (neogenin-Fc or DCC-Fc; immunoblot is pictured in Fig. 5B). As shown by immunocytochemistry of nonpermeabilized cells, neogenin-Fc colocalized with membrane-associated RGMc, but DCC-Fc did not (Fig. 4C). As expected, neogenin-Fc did not bind to cells expressing RGMcΔGPI, as this latter protein was not found on the membrane (15) (Fig. 4D). On the basis of the results in Fig. 4, we conclude that cell-associated RGMc binds to neogenin at the cell membrane.
Neogenin Preferentially Binds Heterodimeric RGMc
To determine which RGMc protein species can bind to neogenin, we first incubated Cos-7 cells transfected with either neogenin or empty expression vector with soluble RGMc (RGMcΔGPI) and labeled membrane-associated proteins with cell-impermeable biotin, followed by selection of biotin-containing molecules by binding to streptavidin-agarose. As seen in Fig. 5A, added RGMc was labeled with biotin only in cells expressing neogenin (lane 2), while the endogenously synthesized membrane protein cadherin was labeled in all cells (bottom, lanes 2 and 3). In addition, the 20/35-kDa RGMc heterodimer was preferentially labeled compared with the 50-kDa species, while 40-kDa RGMc was not detected. These results further demonstrate that RGMc binds to cell membrane-associated neogenin.
We used an in vitro binding assay to extend and to quantify these observations. RGMcΔGPI was purified by immuno-affinity chromatography, the purified proteins were incubated with purified neogenin-Fc or DCC-Fc, and protein complexes were pulled down with protein A-Sepharose. As depicted in Fig. 5B, DCC-Fc and neogenin-Fc each bound equivalently to protein A (lanes 2 and 3, top), but RGMc associated only with neogenin-Fc (bottom). The 20/35-kDa heterodimer showed greater binding to neogenin-Fc than did the 50-kDa species, while 40-kDa RGMc did not bind (compare lane 3 to lane 1, bottom). Quantification of results from six independent experiments showed a fourfold enrichment of heterodimeric RGMc compared with the 50-kDa species, and negligible binding of 40-kDa RGMc (Fig. 5C). On the basis of these observations, we conclude that neogenin preferentially binds 20/35-kDa heterodimeric RGMc.
Lack of Binding by Neogenin to Selected Juvenile Hemochromatosis-Associated RGMc Mutants
We next used immunocytochemistry of nonpermeabilized cells to address potential interactions between the juvenile hemochromatosis-associated RGMc amino acid substitution mutants pictured in Fig. 1 and soluble neogenin (neogenin-Fc). As illustrated in Figs. 3B and 6, all three mutant RGMc proteins were detected on the membrane of transiently transfected Cos-7 cells, yet as shown in Fig. 6, neogenin-Fc bound only to cells expressing wild-type and G92V RGMc. Thus, on the basis of these data and the results depicted in Fig. 5, neogenin appears to bind cell-surface-associated RGMc molecules that are processed into the 20/35-kDa heterodimeric species, whereas, in contrast, BMP-2 can bind to single-chain RGMc proteins (Fig. 2 and 3), which are found both on the membrane and in the extracellular fluid (15).
Genetic evidence in both humans and mice demonstrates that RGMc plays a critical role in iron homeostasis, with loss of the protein causing the rapidly progressing iron overload disorder, juvenile hemochromatosis (12, 28, 30). Despite this overall significance of RGMc in iron metabolism, its mechanisms of action have not been elucidated. One complicating feature in discerning the specific biological actions of RGMc is the existence of multiple protein species that result from a complex series of biosynthetic and processing steps. Mature RGMc consists of at least four protein isoforms, including two cell-membrane-associated species, a 50-kDa single-chain protein, and a disulfide-linked heterodimer composed of 20- and 35-kDa subunits, and two single-chain soluble proteins of 50- and 40-kDa (15). The mechanisms controlling the production and turnover of each of these species are beginning to be characterized, but their individual contributions to the regulation of iron homeostasis are not yet defined. Here we show that three naturally occurring RGMc mutants that are associated with juvenile hemochromatosis either undergo abnormal biosynthesis and processing or do not interact with potential effector molecules, BMP-2 and neogenin. In conjunction with these observations, we have used a series of biochemical and cell-based assays to define preferential binding between distinct RGMc protein species and both BMP-2 and neogenin.
We show that there is a direct interaction between BMP-2 and RGMc. We find that BMP-2 binds to RGMc as effectively as to noggin, a known high-affinity BMP-interacting molecule (39, 43), and detect a preference of BMP-2 for single-chain RGMc, particularly the 40-kDa species, which is derived from cleavage of 50-kDa RGMc by proprotein convertases (20, 40). Our results thus extend the observations of Babitt et al. (2), who used protein cross-linking and competition binding assay to establish an association between I125-labeled BMP-2 and an RGMc-IgG Fc fusion protein, and those of Halbrooks et al. (9), who showed using surface plasmon resonance that soluble RGMc could bind BMP-2 with nanomolar affinity (9). We also demonstrate an interaction of BMP-2 on the cell surface with GPI-linked RGMc, and with two RGMc amino acid substitution mutants, D165E and G313V, that are expressed on the membrane only as single-chain molecules, thus showing that the binding between these two proteins can take place under physiologically relevant conditions. Surprisingly, BMP-2 does not bind to cells expressing the G92V mutant, even though it appears to be processed similarly to wild-type RGMc and can be detected on the membrane as both a single-chain 50-kDa isoform and as a 20/35-kDa heterodimer. The latter results suggest either that glycine 92 in RGMc may be critical for binding of BMP-2 or, alternatively, that the valine substitution results in an altered conformation of RGMc that precludes this interaction.
BMPs are members of the TGF-β superfamily of growth factors. The biological effects of these proteins are initiated by binding to a complex of type I and type II receptors, leading to stimulation of receptor-mediated serine-threonine protein kinase activity (39). It has been proposed that RGMc functions as a coreceptor for BMP-2, 4, and 9 and through this interaction enhances BMP-mediated activation of hepcidin gene expression in the liver (2, 3). In this way, RGMc would promote hepcidin production and thereby modulate iron absorption and overall iron homeostasis. At present, the data supporting this putative coreceptor function are indirect. When overexpressed, cell-associated RGMc has been shown to increase several effects of BMPs, including hepcidin mRNA accumulation in hepatocytes and liver cell lines (2, 21), and BMP-dependent promoter-reporter gene expression (2). In contrast, RGMc has not been found to alter binding of BMPs to receptors, to stimulate receptor phosphorylation or kinase activity, or to enhance signaling through downstream effector molecules. Smads comprise a family of signaling proteins that function as inducible transcription factors for TGF-β family members (39, 43). Smads 1, 5, and 8 act downstream of BMPs and become rapidly phosphorylated on serine residues by activated BMP receptors (43). Receptor-phosphorylated Smads bind to a common cofactor, Smad 4, and the heterodimer translocates into the nucleus, where it binds to DNA response elements in chromatin on target genes and subsequently activates transcription (39, 43). There is as yet no evidence showing that RGMc enhances the extent or duration of BMP-directed Smad phosphorylation, and no demonstration of an increase in accumulation of BMP-activated Smads in the nucleus by RGMc.
The transmembrane protein neogenin is a receptor for RGMa and has been shown to mediate the effects of RGMa on axon guidance and neuronal survival (10, 23, 34). Neogenin is additionally a receptor for netrins, which also mediate cellular guidance in both neuronal and nonneuronal cell types (7, 35, 44) and facilitate cell-cell interactions in a variety of tissues (7, 44). We now find that RGMc also binds to cell-associated neogenin, but not to the related protein DCC or other members of the neogenin-DCC family, and we demonstrate that neogenin preferentially interacts with the RGMc heterodimer, to a lesser extent with the 50-kDa single-chain molecule, but not with the 40-kDa isoform. Our results, which extend observations of Zhang et al. (48), raise the question of whether neogenin mediates any biological actions of RGMc, including those related to systemic iron metabolism. Zhang et al. have argued that neogenin may play a dual role, on the one hand enhancing RGMc-mediated cellular iron uptake (48), and on the other limiting the biological effects of RGMc by promoting its release from the cell membrane into the extracellular space (47), although it should be noted that their results will need to confirmed by others. Additionally, since RGMc is a GPI-linked cell membrane-associated protein, perhaps its potential functional interplay with neogenin could be assessed in this context. Several other GPI-linked proteins participate in complicated signaling pathways. For example, glial cell-derived neurotrophic factor and related proteins neurturin, persephin, and artemin bind to specific GPI-anchored α-subunits as a prerequisite for activating the tyrosine kinase receptor, RET (32). To date, however, there is no evidence that RGMc and neogenin together comprise a receptor for another ligand. Alternatively, GPI-linked ephrin-A and the transmembrane EphA receptor participate in bidirectional signaling (31). Perhaps neogenin functions in the role of EphA and RGMc as ephrin-A in liver and muscle cells. Further study, including in vivo experiments, will be needed to determine what role neogenin plays in the biology of RGMc and, indirectly, in modulating systemic iron balance.
In summary, our observations provide a framework for considering that different isoforms of RGMc may play distinct physiological roles. Neogenin binds preferentially to heterodimeric cell-associated RGMc, while BMP-2 interacts primarily with the 40-kDa single-chain soluble species. However, both neogenin and BMP-2 appear able to bind membrane-linked single-chain 50-kDa RGMc, and thus they potentially may compete with one another. Alternatively, these three proteins may cooperatively form a multimolecular complex at the cell surface. Clearly, new approaches will be required to discover the critical mechanisms by which RGMc regulates iron homeostasis. A diagram illustrating some of these hypotheses is presented in Fig. 7.
The studies presented in this article were supported by National Institutes of Health Grants R01-DK-42748 (to P. Rotwein) and F32-DK-076348 (to R. Kuns-Hashimoto).
We thank the following colleagues for gifts of reagents: S. Ackerman, H. Cooper, J. Christian, D. Engelkamp, K. Guan, and J. Weiss.
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Copyright © 2008 the American Physiological Society