Structural investigation of zymogenic and activated forms of human blood coagulation factor VIII: a computational molecular dynamics study

Abstract

Background: Human blood coagulation factor VIII (fVIII) is a large plasma glycoprotein with sequential domain arrangement in the order A1-a1-A2-a2-B-a3-A3-C1-C2. The A1, A2 and A3 domains are interconnected by long linker peptides (a1, a2 and a3) that possess the activation sites. Proteolysis of fVIII zymogen by thrombin or factor Xa results in the generation of the activated form (fVIIIa) which serves as a critical co-factor for factor IXa (fIXa) enzyme in the intrinsic coagulation pathway.

Results: In our efforts to elucidate the structural differences between fVIII and fVIIIa, we developed the solution structural models of both forms, starting from an incomplete 3.7 A X-ray crystal structure of fVIII zymogen, using explicit solvent MD simulations. The full assembly of B-domainless single-chain fVIII was built between the A1-A2 (Ala1-Arg740) and A3-C1-C2 (Ser1669-Tyr2332) domains. The structural dynamics of fVIII and fVIIIa, simulated for over 70 ns of time scale, enabled us to evaluate the integral motions of the multi-domain assembly of the co-factor and the possible coordination pattern of the functionally important calcium and copper ion binding in the protein.

Conclusions: MD simulations predicted that the acidic linker peptide (a1) between the A1 and A2 domains is largely flexible and appears to mask the exposure of putative fIXa enzyme binding loop (Tyr555-Asp569) region in the A2 domain. The simulation of fVIIIa, generated from the zymogen structure, predicted that the linker peptide (a1) undergoes significant conformational reorganization upon activation by relocating completely to the A1-domain. The conformational transition led to the exposure of the Tyr555-Asp569 loop and the surrounding region in the A2 domain. While the proposed linker peptide conformation is predictive in nature and warrants further experimental validation, the observed conformational differences between the zymogen and activated forms may explain and support the large body of experimental data that implicated the critical importance of the cleavage of the peptide bond between the Arg372 and Ser373 residues for the full co-factor activity of fVIII.

Figure 1

The schematic representation of the sequence and domain arrangement of A) the full model of zymogen fVIII; B) the B-domainless fVIII, used in the present study; C) the activated fVIII. The functionally important calcium and copper ions bound to the inter- and intra-domains are also shown.

Figure 2

The solution structural model of human B-domainless fVIII zymogen derived from the MD snapshot that corresponds to 75 ns of MD simulation trajectory. The coordinating calcium (red spheres) and copper (pink spheres) ions, tyrosine sulfation sites (ball-and-stick) are highlighted. The A1-A2 acidic linker may be seen blocking the putative fIXa binding Y555-D569 loop (magenta color) in the A2-domain (orange color). The solvent-exposed proteolytic site (Arg372-Ser373) is also shown. The platelet surface shown in the figure is merely a schematic representation and not modeled in the current study.

Figure 3

The solution structure of activated fVIII derived from the MD snapshot that corresponds to 80 ns of MD simulation trajectory. The conformational reorganization of the A1-A2 acidic linker peptide (orange color), upon the cleavage of the Arg372-Ser373 bond, may be seen from the MD simulations. The exposed fIXa binding loop in the A2-domain is highlighted (magenta color).

Figure 4

The RMS deviations of the backbone atoms of A) the zymogen (left) and B) the activated form (right) of factor VIII as compared to the starting structure.

Figure 5

Backbone superimposed structures of A) zymogenic (left) and B) activated (right) forms of factor VIII, representing the top ten clusters of the MD trajectory. The last 10 ns of the simulation data (70-80 ns) was used for clustering.

Figure 6

Atomic positional fluctuations in A1-A2 and A3-C1-C2 domains of A) active and B) zymogenic forms of fVIII. The data corresponds to 2 ns of MD simulation data extracted from 75 to 77 ns.

Figure 7

Models used for generation of linker peptide between A1 and A2 domains, A) The A1 and A2 domains with missing linker peptide and the distance between the core residues Glu332 (A1 domain) and Pro379 (A2 domain) is shown. B) The top ten structures of the proposed linker peptide derived from clustering the last 30 ns of 100 ns MD trajectory are presented. Converged structures of isolated sequences C) Arg336-Met355 and D) Asn369-Lys380 in the linker peptide (based on 50 ns of MD simulations each) to demonstrate the reproducibility of the helical nature of the sequences is shown. E) The RMS deviations in the backbone (Cα) atoms during the simulations of the linker peptide model with reference to the starting structure are shown.

Figure 8

The structural convergence of the A2-A3 linker peptide generated using a reduced model between A2 and A3 domains. A) The starting conformation, B) the converged structures (based on clustering the last 30 ns of MD trajectory) of the linker peptide and C) the corresponding backbone (Cα atoms) RMS deviations with reference to the starting conformation are shown.

Figure 9

The octahedral calcium coordination network in A) A1-domain; B) at the interface of A2 and A3 domains and C) the distance plot of a solvent calcium ion coordination with one the side-chain carboxylate atoms are shown.

Figure 10

The tetrahedral copper ion coordination network in A) the A1 domain, B) the A3 domain and C) at the interface of A1 and A3 domains are shown.