Repulsive guidance molecules lock growth differentiation factor 5 in an inhibitory complex

Abstract

Repulsive guidance molecules (RGMs) are cell surface proteins that regulate the development and homeostasis of many tissues and organs, including the nervous, skeletal, and immune systems. They control fundamental biological processes, such as migration and differentiation by direct interaction with the Neogenin (NEO1) receptor and function as coreceptors for the bone morphogenetic protein (BMP)/growth differentiation factor (GDF) family. We determined crystal structures of all three human RGM family members in complex with GDF5, as well as the ternary NEO1-RGMB-GDF5 assembly. Surprisingly, we show that all three RGMs inhibit GDF5 signaling, which is in stark contrast to RGM-mediated enhancement of signaling observed for other BMPs, like BMP2. Despite their opposite effect on GDF5 signaling, RGMs occupy the BMP type 1 receptor binding site similar to the observed interactions in RGM-BMP2 complexes. In the NEO1-RGMB-GDF5 complex, RGMB physically bridges NEO1 and GDF5, suggesting cross-talk between the GDF5 and NEO1 signaling pathways. Our crystal structures, combined with structure-guided mutagenesis of RGMs and BMP ligands, binding studies, and cellular assays suggest that RGMs inhibit GDF5 signaling by competing with GDF5 type 1 receptors. While our crystal structure analysis and in vitro binding data initially pointed towards a simple competition mechanism between RGMs and type 1 receptors as a possible basis for RGM-mediated GDF5 inhibition, further experiments utilizing BMP2-mimicking GDF5 variants clearly indicate a more complex mechanism that explains how RGMs can act as a functionality-changing switch for two structurally and biochemically similar signaling molecules.

Keywords: Neogenin; Repulsive guidance molecule; TGFβ/BMP signaling; cell surface receptor; structural biology.

Fig. 1.

Cellular assay for RGM–BMP2/GDF5 signaling and the crystal structure of the RGMB–GDF5 complex at 1.7 Å resolution. (A) Domain organization of human RGMs with the N-terminal domains (RGM_NDs) indicated. (B and C) BMP2 (3 nM) and GDF5 (30 nM) signaling assays in LLC-PK1 cells transfected with RGM and luciferase-reporter vectors. (D and E) BMP2 (3 nM) and GDF5 (30 nM) signaling assays in LLC-PK1 cells similar to B and C but including single amino acid variants of RGMB. Each column in B, D, and E represents an average of data from 32 wells with cells (except D, 16 wells). Bars represent SDs. P values (Student’s t test; two-tailed assuming unequal variance) are shown for selected datasets compared to cells transfected with an empty vector and treated with BMP2/GDF5 (black) or full-length RGMB and treated with BMP2/GDF5 (blue). (F and G) Cartoon representation of the RGMB_ND–GDF5 complex. GDF5 protomers are shown in light and dark blue. RGMB_NDs are shown in light and dark pink. (G) Complex rotated 90° relative to F along the horizontal axis. Disulfide bonds (orange) are labeled with Roman numerals. The N-acetylglucosamine moiety on RGMB Asn120 is shown in gray. (H and I) RGM–GDF5 interfaces (circled in F). Hydrogen bonds between selected atoms (oxygen, red; nitrogen, blue) are shown as dashed gray lines. Water molecules are shown as red spheres in I. Distances (Å) between selected atoms are indicated.

Fig. 2.

The architecture of human RGM–GDF5 complexes is conserved. (A) Superposition of RGMA_ND–GDF5 (orange) and RGMB_ND–GDF5 (pink–blue, two RGMB_ND and GDF5 molecules are shown as ribbons and two in surface representations) complexes calculated with GDF5 as reference. The two views differ by a 90° rotation around a horizontal axis. The “finger 2” region of GDF5 contacting the RGD/RGN motif of RGMs is circled. (B) Superposition of RGMC_ND–GDF5 (cyan) and RGMB_ND–GDF5 (pink–blue) complexes. The GDF5 finger 2 contacting the RGD/RGN motif of RGMs is circled. Distances between corresponding Cα atoms of RGMB and RGMC are indicated, highlighting a relative translation of RGMC (compared to RGMA and RGMB) towards the finger 2 of GDF5. (C–F) Close-up views of the RGM–GDF5 interfaces with molecules colored as in A and B. Hydrogen bonds between selected atoms are shown as dashed lines. Distances (Å) between selected atoms are indicated in C–E.

Fig. 3.

SPR-based equilibrium binding experiments between RGMs and GDF5. (A–H) SPR-based equilibrium binding experiments showing direct interactions between RGMs and GDF5. The extracellular domain of RGMB (RGMB_ECD) bound to GDF5 with a similar affinity (K_d 8.8 μM, A) as RGMB_ND (K_d 2.7 μM, B), highlighting RGMB_ND as the major site mediating RGMB–GDF5 interactions. Mutations of RGM residues at the RGM–GDF5 interface weaken interactions (E–H). SPR sensorgrams and corresponding isotherms are shown. B_max, maximum response at saturating concentration of analyte (RGM); RU, response units.

Fig. 4.

RGMs compete with the type 1 (but not type 2) receptor binding site on BMP2 and GDF5. (A) Superposition of BMPR1B–GDF5 (GDF5 and BMPR1B are shown in surface and cartoon representations, respectively; PDB ID code 3EVS) and RGMB_ND–GDF5 (only RGMB is shown). (B) Superposition of BMPR1A–BMP2–ActR2b (shown in surface representation, except BMPR1A shown as a cartoon; PDB ID code 2H62) and RGMB_ND–BMP2 (only RGMB is shown; PDB ID code 4UHZ). (C and D) Superposition of RGMB_ND–GDF5 and RGMC_ND–GDF5 (C), and RGMB_ND–GDF5 and RGMA_ND–GDF5 (D) complexes. (E and F) Superposition of RGMB_ND–BMP2 (PDB ID code 4UI0) and RGMC_ND–BMP2 (PDB ID code 4UI1) (E), and RGMB_ND–BMP2 and RGMA_ND–BMP2 (F) complexes. Distances (Å) between selected Cα atoms (shown as spheres) are indicated with dashed lines. (G and H) Superposition of BMPR1B–GDF5 (PDB ID code 3EVS) and RGMB_ND–GDF5 reveals regions of GDF5 that contact BMPR1B (but not RGMB) (G) and regions of GDF5 that contact RGMB (but not BMPR1B) (H). (I) SEC-MALS analysis of the RGMB_ND–GDF5–ActR2b complex. The experimental molecular mass of the RGMB_ND:GDF5:ActR2b complex is 64.2 kDa, corresponding to a 2:2:2 arrangement (theoretical molecular mass: 71.2 kDa). Traces of absorbance at 280 nm are shown as continuous black lines. The Inset shows a close-up of the peak with indicated molecular weight values (with associated statistical uncertainties, calculated using the Astra software from Wyatt Technologies).

Fig. 5.

Structure of the NEO1–RGMB–GDF5 complex. (A and B) Ribbon representation of the NEO1–RGMB–GDF5 complex. RGMB and GDF5 are colored as in Fig. 1, NEO1 is in yellow. The RGMB_ND–RGMB_CD linker and C termini of RGMB are shown as dotted lines. View in A (parallel to cell membrane) and B (view from the top) differ by a 90° rotation around a horizontal axis. N and C termini are marked. (C) The NEO1–RGMB–BMP2 complex (17) shown in similar orientations as the NEO1–RGMB–GDF5 assembly in A and B. (D and E) GDF5 (30 nM) signaling is inhibited by RGMs in the presence or absence of NEO1 (lacking the cytoplasmic domain, NEO1_ΔC), D; full-length NEO1, E in LLC-PK1 cells. Each column represents an average of data from 16 (D) or 31 to 32 (E) wells with cells. Bars represent SDs. P values (Student’s t test; two-tailed assuming unequal variance) are shown for selected data sets compared to cells transfected with an empty vector and treated with GDF5.

Fig. 6.

Control of BMP2 and GDF5 signaling by RGMs. Dimeric BMP2 and GDF5 ligands assemble a complex comprising type 1 and type 2 receptors (BMPR1A and ActR2b, respectively) to activate downstream signaling. Type 1 receptors and RGMs occupy a highly overlapping epitope on BMP2 and GDF5, However, while membrane-anchored RGMs enhance BMP2 signaling, GDF5 signaling is inhibited by RGMs. Ternary signaling complexes were modeled based on BMPR1A–BMP2–ActR2b (PDB ID code 2H62), BMPR1A–GDF5 (PDB ID code 3QB4) and RGMB–GDF5 (present study). The C-terminal domain of RGM is not shown for simplicity.