Structure of Gremlin-2 in Complex with GDF5 Gives Insight into DAN-Family-Mediated BMP Antagonism

Abstract

The DAN family, including Gremlin-1 and Gremlin-2 (Grem1 and Grem2), represents a large family of secreted BMP (bone morphogenetic protein) antagonists. However, how DAN proteins specifically inhibit BMP signaling has remained elusive. Here, we report the structure of Grem2 bound to GDF5 at 2.9-Å resolution. The structure reveals two Grem2 dimers binding perpendicularly to each GDF5 monomer, resembling an H-like structure. Comparison to the unbound Grem2 structure reveals a dynamic N terminus that undergoes significant transition upon complex formation, leading to simultaneous interaction with the type I and type II receptor motifs on GDF5. Binding studies show that DAN-family members can interact with BMP-type I receptor complexes, whereas Noggin outcompetes the type I receptor for ligand binding. Interestingly, Grem2-GDF5 forms a stable aggregate-like structure in vitro that is not clearly observed for other antagonists, including Noggin and Follistatin. These findings exemplify the structural and functional diversity across the various BMP antagonist families.

Figure 1. Structure of Grem2-GDF5

(A) Ribbon and surface representations of Grem2-GDF5 in a ligandcentric view. Grem2 (green, Chain E; blue, Chain F); GDF5 (pink, Chain A; gold, Chain C). (B) Schematic of the Grem2-GDF5 complex. (C) Grem2 in the apo state (left; PDB:4JPH) or GDF5 bound state (right). (D) Superposition of unbound GDF5 (dark purple/orange, PDB:1WAQ) and Grem2 bound (pink/gold). For Grem2 and GDF5 (underscored), F1, F2 and W correspond to finger 1, finger 2, and the wrist region. Dotted lines show unresolved amino acids 35-44 in Grem2.

Figure 2. Grem2-BMP Binding Interface and Analysis

(A) Grem2-GDF5 complex with BMP-binding interfaces in Grem2 (labeled I through IV) highlighted in green, blue, red, and magenta, respectively. (B-E) Grem2 and GDF5 are colored as in Figure 1. Zoomed in view of the Grem2-GDF5 (B) interface I, (C) interface II, (D) interface III, and (E) interface IV as described in the text. (F) BMP-responsive luciferase reporter assay using a BMP-responsive osteoblast cell-line and showing titration of various Grem2 mutants against 2nM BMP2. Experiments were performed in triplicate. (G-H) Injection (stage 9) of wild type and Grem2 mutants (0.5 μM, 1 μM, and 10 μM) in developing Xenopus embryos. Quantification (G) and representative images (H, top) of BMP-dependent axial development (stage 35). (H, bottom) In situ hybridization (stage 20) for mRNA expression of the BMP-target gene sizzled. Numbers in boxes indicate concentration of the reagent used (top and bottom) as well as the number of embryos scored (bottom).

Figure 3. Differences in Antagonist Inhibition of Ligand-Receptor Interactions

(A-B) Structures of BMP-antagonist and BMP-receptor complexes. (C, D, F, G) SPR experiments injecting mixtures of (C-D) Grem2-BMP2 and (F-G) NBL1-BMP2 onto Alk3-Fc or BMPR2-Fc captured using a protein A chip. Molar ratios 0.1:1, 0.5:1, 1:1, and 2:1 of antagonist:ligand were tested. BMP2 was maintained at a constant 250 nM. (E, H, I) SPR experiments using immobilized Alk3. BMP2 (500 nM) was first injected to form the Alk3-BMP2 complex. Grem2 (E), NBL1 (H), or Noggin (I) were subsequently titrated (0-3 μM) and co-injected following BMP2. (J) SPR performed as in E, H, and I with an additional BMP2 injection to determine if ligand binding could be recovered. (K) ELISA binding assay using immobilized Alk3-Fc treated with labeled BMP2, washed, and then treated with Grem2 (blue) or Noggin (orange) at various concentrations for 1 h. Fluorescence was measured in the supernatant to quantify BMP2 release from Alk3. ** = p-value < 0.05 using student's t-test. All experiments were performed in triplicate. (L) Schematic showing potential DAN-BMP-type I receptor ternary complexes.

Figure 4. Alternating Aggregation of Grem2-BMP

(A) Grem2-GDF5 complex colored as in Figure 1 forms a continuous alternating complex extending beyond the ASU. (B-D) SPR co-injection studies of antagonist and BMP2 over immobilized BMP2. Antagonists were injected followed by alternating injections of BMP2 and antagonist (all at 500 nM). Control experiments show consecutive injections of antagonist. (B) Analysis of Grem2 and NBL1 and (C) Noggin. (D) Direct comparison of Grem2 and Noggin at a slower flow rate with a longer association time. (E) Similar SPR experiment with immobilized Activin A and alternating injections of Follistatin and Activin A. Tic marks and labels show specific protein and time of injection.