Calcium-controlled conformational choreography in the N-terminal half of adseverin

Abstract

Adseverin is a member of the calcium-regulated gelsolin superfamily of actin-binding proteins. Here we report the crystal structure of the calcium-free N-terminal half of adseverin (iA1-A3) and the Ca(2+)-bound structure of A3, which reveal structural similarities and differences with gelsolin. Solution small-angle X-ray scattering combined with ensemble optimization revealed a dynamic Ca(2+)-dependent equilibrium between inactive, intermediate and active conformations. Increasing calcium concentrations progressively shift this equilibrium from a main population of inactive conformation to the active form. Molecular dynamics simulations of iA1-A3 provided insights into Ca(2+)-induced destabilization, implicating a critical role for the A2 type II calcium-binding site and the A2A3 linker in the activation process. Finally, mutations that disrupt the A1/A3 interface increase Ca(2+)-independent F-actin severing by A1-A3, albeit at a lower efficiency than observed for gelsolin domains G1-G3. Together, these data address the calcium dependency of A1-A3 activity in relation to the calcium-independent activity of G1-G3.

Figure 1. Crystal structures of the N-terminal half of adseverin in calcium-free conditions (iA1–A3) and of calcium-bound domain A3 (aA3).

(a) The structure of iA1–A3 (residues 6σ349) is shown in cartoon representation and coloured from blue to red (N- to C terminus). (b) Comparison of the structure of iA1–A3 with domains iG1–G3 of gelsolin excised from the structure of full-length gelsolin in calcium-free conditions (PDB 1D0N), showing a similar compact arrangement of the domains that masks the conserved actin-binding interfaces on A1 and A2. G1–G3 is shown in grey cartoon representation. (c) The calcium-bound structure of adseverin domain A3 (aA3, residues 248–349), shown in cartoon representation and coloured from blue to red (N- to C terminus). The calcium ion is shown as a magenta sphere with coordinating residues as sticks. Inset: close up of the calcium-binding site, showing the coordination of the calcium ion by the Glu280 carbonyl, monodentate Glu281 and bidentate Glu305 side chains, as well as three water molecules. The 2F_o-F_c electron density map contoured at 1.5σ is shown as a mesh. (d) Comparison of the structure of aA3 (coloured as blue to red rainbow) and iA1–A3 (in grey). Superimposition highlights the conformational changes in the second β-strand and the long α-helix of A3 in response to calcium binding and dissociation from A1.

Figure 2. SAXS experiments.

(a) SAXS profiles of A1–A3 in the presence of 10 mM EGTA (black spheres), 2 mM EGTA (red spheres) and with increasing calcium concentrations (10, 220, 390 and 490 μM CaCl₂, green, yellow, purple and pink spheres, respectively) measured at protein concentrations of 1, 2, 4 and 8 mg ml⁻¹ (curve series 1, 2, 3 and 4, respectively). Curve series (5) corresponds to a second calcium titration at 4 mg ml⁻¹ that spans a larger range of calcium concentrations (2 mM EGTA, 19 and 49 nM, 175 and 550 μM and 2.3 and 50 mM free Ca²⁺) and was measured on a separate synchrotron visit using a different protein preparation. (b) Changes in radius of gyration as a function of calcium concentration corresponding to the curve series in a: (1) black, (2) red, (3) green, (4) yellow and (5) purple line. (c) Model-free analysis of calcium-induced conformational changes. The data shown in a, curve series (5) is represented as Kratky plots, indicating a transition towards a less globular and more flexible structure.

Figure 3. Proposed mechanism of calcium activation of A1–A3 in solution.

(a) Model for A1–A3 transition between the inactive and active (actin-binding competent) states. The release of the A1/A3 latch from the inactive form (I) leads to a first intermediate (Int1), followed by the loss of the A1A2 interface (Int2). Finally, the calcium-stabilized interface between A2 and A3 is formed, resulting in the active conformation (a). (b) Fitted SAXS profiles. For clarity, only a subset of the data shown in Fig. 2a is represented. Experimental data are shown as black spheres and theoretical SAXS profiles for optimized ensembles of 50 models are drawn as red lines. (1) 490 μM CaCl₂, 1 mg ml⁻¹ of A1–A3; (2) 490 μM CaCl₂, 2 mg ml⁻¹ of A1–A3; (3) 490 μM CaCl₂, 4 mg ml⁻¹ of A1–A3; (4) 49 nM, 175 and 550 μM, and 2.3 mM CaCl₂, 4 mg ml⁻¹ of A1–A3; (5) 490 μM CaCl₂, 8 mg ml⁻¹ of A1–A3; (6) 2 mM EGTA, 8 mg ml⁻¹ of A1–A3 mg ml⁻¹. (c) Relative populations of inactive (I, black), intermediate (Int1, red and Int2, green) and active (A, blue) conformations of A1–A3, plotted as a function of free calcium concentration (x axis, bottom) and ratio of calcium ion over protein (x axis, top). All data measured at 4 mg ml⁻¹ were used (Fig. 2 A, curve series 3 and 5). Error bars represent the s.d. of ensemble optimization results obtained from 10 independent ensembles of 6,000 models (see Methods section). (d–f) Two-dimensional (2D) histogram representations of the distribution of A1–A2 and A2–A3 interdomain distances (taken between their centres of mass) within the optimized ensembles, using data measured in the presence of 2 mM EGTA (d), 10 μM (e) and 550 μM free Ca²⁺ (f). The 2D histogram was calculated using a binning of 15 × 15 and coloured from red to blue (0–50 models).

Figure 4. MDS of A1–A3 in the inactive state.

(a,b) r.m.s.d. of Cα atoms from the starting structure in the presence of 50 mM CaCl₂ (a) or NaCl (b). The results from two independent simulations are shown as black and red lines, respectively. (c) Root mean square fluctuations (r.m.s.f.) of Cα atoms along the protein sequence. The averages of the two independent simulations are shown as black and red lines for systems simulated in the absence and presence of calcium, respectively. (d) Analysis of calcium binding to inactive iA1–A3 by MDS. Calcium ions displaying the lowest time-averaged distance to protein atoms were extracted from eight snapshots taken at 100 ns of interval and are represented as magenta spheres. Four overlaid snapshots of the protein are shown in two views rotated by 90° in cartoon representation with the protein backbone coloured from blue to red (N- to C terminus). The protein backbone from four additional snapshots, for which the calcium ions are shown, have been hidden for clarity. Close ups of the calcium-binding sites of interest are shown as insets.

Figure 5. Effect of calcium on A2A3 linker dynamics as observed in MDS.

In each panel, the crystal structure of iA1–A3 is shown in grey cartoons for comparison with the A2A3 linker represented using a larger cartoon loop radius. (a) Linker dynamics in the absence of calcium. Ten overlaid snapshots of the protein taken at 10 ns of interval (10–100 ns) are shown in cartoon representation with the protein backbone coloured from blue to red (N- to C terminus). (b–d) Linker dynamics in the presence of calcium, illustrating the increased flexibility and deviation from the crystal structure. Ten overlaid snapshots of the protein taken at 10 ns of interval were extracted from one of the two 1 μs MDS of iA1–A3 in the presence of calcium and are shown as rainbow-coloured cartoons (0–90, 400–490 and 900–990 ns in b–d respectively). In each case, bound calcium ions are represented in magenta spheres for one of the snapshots, with coordinating side chains as sticks (100 ns in b, 500 ns in c and 1,000 ns in d).

Figure 6. Calcium-independent actin-severing activity of A1–A3 is increased by mutations that disrupt the A1/A3 interface.

(a) Details of the A1/A3 interface. The mutated residues Met310 and Glu314 are shown as grey sticks. (b) Similar view of the G1/G3 interface. (c,d) Actin depolymerization assays. A total of 6 μM of each protein (actin control, black; A1–A3wt, red; E314S mutant, green; M310D mutant, yellow; and G1–G3, purple) was added to 6 μM of F-actin in the presence of 1 mM EGTA (c) or 0.5 mM CaCl₂ buffer (d).

Figure 7. Role of the A2A3 linker in activation and comparison with gelsolin.

(a) Conformation of the A2A3 linker in the inactive structure (grey cartoons) and active model (rainbow-coloured cartoons) of A1–A3. The crystal structure of iA1–A3 was superimposed onto a homology model of the gelsolin-like active conformation through structural alignment of A2. (b,c) Calcium coordination at the type II binding site of domain 2 in the active A1–A3 model (b) and in the crystal structure of actin-bound G1–G3 (PDB 3FFK) (c), highlighting sequence divergence in this region with a more acidic linker in adseverin.