Molecular mechanism of cleavage at R271 during prothrombin activation revealed by cryo-EM

Key Points

• •The structure of mzTΔF1 bound to factors Va and Xa is solved by cryo-EM at 3.8 Å resolution.
• •The study reveals the molecular mechanism of cleavage at R271 during prothrombin activation.

Visual Abstract

Abstract

The conversion of the inactive zymogen prothrombin to the active protease thrombin in the common pathway of the coagulation cascade is the molecular event responsible for the pathophysiology of hemostasis and thrombosis. The conversion entails 2 proteolytic cleavages at R320 and R271 by the prothrombinase complex composed of the enzyme factor Xa (fXa), the cofactor fVa, Ca²⁺, and phospholipids. A recent cryogenic electron microscopy (cryo-EM) structure revealed how cleavage at R320 generates the active intermediate meizothrombin in the first step of the activation pathway. Here we present the 3.8 Å resolution cryo-EM structure of a truncated form of meizothrombin (mzTΔF1) bound to fVa and fXa that reveals how the second cleavage at R271 generates thrombin. The cleavage is brokered by molecular contacts that involve mostly the protease domains of mzTΔF1 and fXa and largely validate the results from biochemical studies. The switch in cleavage site from R320 to R271 involves a significant reorientation rather than conformational transitions of the protease domain of mzTΔF1 that moves the guanidinium group of R271 more than 20 Å into the primary specificity pocket of fXa. The findings complete the cryo-EM structural analysis of prothrombin activation along the meizothrombin pathway and advance our molecular understanding of a reaction critical to the pathophysiology of blood coagulation.

It is underappreciated that many plasma coagulation factors perform double cleavage of their substrates to release a series of products. The molecular mechanisms of these multidomain rearrangements require elucidation, and Stojanovski et al report a new cryogenic electron microscopy (cryo-EM) structure of the prothrombinase complex bound to meizothrombin, a partially activated species of thrombin. This arrangement promotes a secondary proteolytic cleavage to release fully activated thrombin at R271, further elucidating how thrombin is generated during thrombosis.

Introduction

The response to vascular injury triggers a cascade of proteolytic events where inactive zymogens are converted to active proteases.^1,2 Two alternative but interconnected pathways of activation of the coagulation cascade, intrinsic and extrinsic, converge to the common pathway where the enzyme factor Xa (fXa), the cofactor fVa, Ca²⁺, and phospholipids assemble in the prothrombinase complex and catalyze a rapid conversion of prothrombin to thrombin.^{1, 2, 3, 4} Because this cofactor-dependent activation is paradigmatic of analogous reactions of the blood coagulation and complement cascades,^1,5 its study bears broad relevance to basic and clinical sciences. Once thrombin is generated from prothrombin, it goes on to cleave fibrinogen and PAR1 to promote clot formation and platelet aggregation that are at the basis of the pathophysiology of hemostasis and thrombosis.⁶

The biochemistry of prothrombin activation has been studied extensively.^{1,2,7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18} Conversion of prothrombin to thrombin by the prothrombinase complex requires 2 proteolytic cleavages at R271 and R320 along 2 mutually exclusive pathways that are context dependent^1,17 (supplemental Scheme 1, available on the Blood website). On the surface of platelets, cleavage at R271 sheds the auxiliary γ−carboxyglutamate (Gla) domain and 2 kringles to generate prethrombin-2 (pre2), the inactive precursor of thrombin comprising only a protease domain (PD).^19,20 On red blood cells,²¹ synthetic liposomes^{15,22, 23, 24} and microparticles²⁵ activation proceeds through the alternative cleavage at R320 and generates the active intermediate meizothrombin (mzT), which retains the Gla-kringle1-kringle2-PD multidomain assembly of prothrombin but folds the active site in the PD according to the Huber-Bode mechanism²⁶ to enable catalysis. Thrombin is generated either by cleavage of mzT at R271 or by cleavage of pre2 at R320 as an active protease containing only the PD.⁶ When prothrombin activation is studied in vitro, the presence of fVa promotes activation along the mzT pathway,^10,11,13 with or without phospholipids,^10,11,23,27 and directs preferential cleavage at R320 independent of the sequence in the activation domain.¹⁷ On the other hand, absence of fVa promotes cleavage at R271 and activation along the pre2 pathway.^9,11,17 Absence of fVa also results in an additional cleavage at R155 and generates the inactive intermediate prethrombin-1 (pre1)¹¹ containing only kringle2 and the PD.²⁸ Activation of pre1 by subsequent cleavage at R320 generates the active intermediate mzTΔF1 that differs from mzT for the lack of Gla and kringle1 domains²⁹ and from thrombin for the presence of the kringle2 domain.

Structural insight into prothrombin activation has recently come from cryogenic electron microscopy (cryo-EM) studies of the prothrombin-fVa-fXa complex.³⁰ The components assemble in a pyramid with the A2 domain of fVa at the top and the C1-C2 domains of fVa and the Gla domains of prothrombin and fXa at the bottom, aligned on a plane for interaction with a membrane surface. A distinctive feature of the complex, not captured by previous molecular models,^31,32 is the wide spatial separation among the Gla domains of prothrombin and fXa and the C domains of fVa also observed in the presence of nanodiscs.^30,33 Prothrombin activation is driven by protein contacts established independently of the interaction with a membrane surface whose main role is to accelerate assembly of the complex by reducing the effective dimensionality of the system.³⁴ In the cryo-EM structure, prothrombin and fXa face each other against fVa, with fXa positioned along the A2, A3, and C1 domains and prothrombin aligned along the A2, A1, and C2 domains (Figure 1A). Prothrombin assumes the closed form identified independently by single molecular Förster resonance energy transfer (smFRET) measurements,³⁵ rapid kinetics,³⁶ limited proteolysis,³⁷ and X-ray crystallography^38,39 and presents the site of cleavage at R320 to the active site of fXa. In this conformation, the alternative site of cleavage at R271 is >20 Å away^30,33 (Figure 1A) and would require a substantial rearrangement of the PD to bind to the active site of fXa once R320 of prothrombin is cleaved and mzT is generated as a new substrate of prothrombinase. Here we show by cryo-EM analysis how this rearrangement takes place and complete our structural investigation of the mechanism of prothrombin activation along the mzT pathway.

Figure 1.

Structural basis of prothrombin activation. (A) Molecular surface representation of prothrombin bound to fVa and fXa^30,33 in the first step of prothrombin activation along the mzT pathway (cleavage at R320). The ternary complex³³ (supplemental Note 1) is shown in different orientations obtained by 90° rotation as indicated by arrows. The constitutive domains of prothrombin (Gla, K1 = kringle1, K2 = kringle2, PD), fVa (A1, A2, A3, C1, C2), and fXa (Gla, EGF1, EGF2, PD) are labeled and rendered in different shades of purple (prothrombin), green (fVa), and yellow (fXa). Prothrombin engages fVa and fXa only through the PD that binds to the A2 domain of fVa and the PD of fXa. The PD of fXa is in close contact with the A2 domain of fVa. Arrows indicate the sites of prothrombin activation at R271 and R320, both not visible on the surface in the orientations shown. The distance between the sites is >20 Å. (B) Molecular surface representation of mzTΔF1 bound to fVa and fXa depicting the second step of prothrombin activation along the mzT pathway (cleavage at R271), as reported in this study. The ternary complex is shown in different orientations with the constitutive domains of mzTΔF1, fVa, and fXa labeled and rendered in different shades of purple (mzTΔF1), green (fVa), and yellow (fXa). Only the EGF2-PD of fXa could be resolved from the cryo-EM density map. MzTΔF1 engages the PD of fXa through its own PD. The arrow indicates the position of R271 inside the active site of fXa made possible by a significant rotation of the kringle2-PD pair relative to the position observed with bound prothrombin (A).

Methods

Materials

The β form of human fX (residues 1-429, mature numbering) carrying a replacement of the catalytic Ser379 by Ala to prevent (auto)proteolytic activity and a human protein C-4 tag in the activation peptide linker to facilitate purification was expressed, purified, and activated by Russell’s viper venom fX activator (Enzyme Research Laboratory) as described.⁴⁰ Following activation, removal of the activation peptide and of the human protein C-4 tag, fXa was purified as reported.⁴⁰ The construct of human mzTΔF1 carrying the S525A mutation to prevent (auto)proteolytic activity was expressed and purified as pre1 and activated with ecarin (Sigma-Aldrich; 2 units/mg protein) by overnight incubation at ambient temperature in 20mM Tris, 145mM NaCl, 5mM CaCl₂, pH 7.5.⁴¹ The construct was purified by anion exchange chromatography followed by size exclusion on a Superdex-200 column (Cytiva).⁴² Human fVa from plasma was purchased from Prolytix.

Cryo-EM grid preparation and data collection

The ternary complex was formed by incubating fVa (600nM), the mzTΔF1 S525A mutant (1200nM) and the fXa S379A mutant (1200nM) for 30 minutes at room temperature prior to freezing in a buffer composed of 20mM HEPES (N-2-hydroxyethylpiperazine-N′-2-ethanesulfonic acid), 100mM NaCl, 5mM CaCl₂, 40μM DDM (n-dodecyl-B-D-maltoside), pH 7.5. The detergent DDM was added to minimize orientation bias due to interactions with the air-water interface. Samples of the ternary complex were applied on Quantifoil 1.2/1.3 200 mesh grid, blotted, and plunge frozen in liquid ethane using a FEI Vitrobot Mark IV system. After loading the grids onto a Titan Krios G3 cryo–transmission electron microscope (ThermoFisher Scientific) operating at 300 kV and equipped with a Falcon IV (4096 × 4096) direct electron detector, data collection was performed at a pixel size of 0.842 Å, using dose exposures of 47.24, 50.6, and 55.26 e−/Å2 (for 3 data sets collected from the same grid) and a set defocus range of −1 to −2.4 μm. A total of 5179 micrographs were collected, with 28% of the total at 30° tilt to improve the particle angular distribution. Relevant data are summarized in Table 1.

Table 1.

Structural parameters for the cryo-EM structure of the mzTΔF1-fXa-fVa complex

	PDB ID 9YQ8, EMD-73327
Data collection and processing
Nominal magnification	75 000×
Voltage (kV)	300
Electron exposure (e−/Å2)	47.24, 50.6, 55.26
Defocus range (μm)	−1 to −2.4
Pixel size (Å)	0.842
Number of images	5 179
Final particle images	38 200
Symmetry imposed	C1
Resolution unmasked∗ (Å)	4.38
Resolution masked∗ (Å)	3.84
Model composition
Refinement program	Phenix (real space)
Number of amino acid chains	6
Number of protein atoms (non-H)	15 484
Protein residues	1 921
Ligand molecules	None
RMSD
Bond lengths (Å)	0.35
Bond angles (°)	0.58
Validation
All-atom clash score	14
Poor rotamers (%)	0
Ramachandran disallowed (%)	0
Ramachandran allowed (%)	11
Ramachandran favored (%)	89
MolProbity score	2.23

PDB, Protein Data Bank; RMSD, root mean square deviation.

^∗Calculated at the Fourier shell correlation threshold = 0.143.

Image processing, map calculation, and model building

All image processing steps were carried out in cryoSPARC,⁴³ unless stated otherwise. After performing patch-based motion correction, dose weighting and patch-based contrast transfer, initial particles were picked using the blob picker function, extracted with a box size of 400 pixels, and cleaned by 2-dimensional (2D) classification. The cleanest 2D classes were used as templates for particles picking and the process was repeated several times to improve the quality and diversity of 2D classes. Particles from selected 2D classes were used subsequently to create a training data set for the Topaz neural network–based picker⁴⁴ and the resulting model was used to pick particles using Topaz Extract. Following extraction (400-pixel box size) and 2D classification, particles were combined from different picker jobs, duplicates removed, and ∼108 000 particles were used to generate an initial model of the ternary complex by ab initio reconstruction. An initial consensus refinement was performed using homogenous refinement, but significant level of conformational heterogeneity was identified after subjecting the initial consensus map to 3D-variability analysis. To isolate homogenous subset of particles and minimize the level of conformational variability, 3D classification with a total of 2 classes was performed. The resultant 2 volumes were used for subsequent heterogenous refinement, separating the 108 000 particles into 2 classes, with class 0 comprised of 68 000 particles and class 1 comprised of 40 000 particles. Visual inspection of the 2 volumes in University of California San Francisco (UCSF) ChimeraX revealed significant differences in the orientation of mzTΔF1, but in subsequent refinement steps we were only able to generate a map of sufficiently high quality for the particles from class 0. To this end, the 68 000 particles belonging to class 0 were subjected to heterogenous refinement using the 2 initial representative reconstructions as map inputs. Two additional rounds of heterogenous refinement resulted in further improvement of particle homogeneity. The final 38 200 particles were used for homogenous and local refinement to produce a consensus map with an overall resolution of 3.5 Å (Electron Microscopy DataBank [EMD]-73318). Two sets of masked refinement with and without particle subtractions were performed to improve the quality of specific regions in the map. In one masked refinement job (EMD-73320) a softly padded mask was generated against the fXa/mzTΔF1 portion of the map and all 38 200 particles were refined without particle subtraction. In a second masked refinement job (EMD-73319), we initially used a mask of mzTΔF1 to subtract projections of the masked subvolume from particle images. The subtracted particles were then used for masked refinement using a softly padded mask generated against the fVa/fXa part of the map. The 2 maps were sharpened by DEMhancer⁴⁵ and combined in Phenix⁴⁶ to generate the final composite map (EMD-73327). The global resolution of the composite map was 3.84 Å, as calculated by Phenix Mtriage⁴⁷ using gold-standard Fourier shell correlation at a threshold of 0.143. The model of the ternary complex was generated through a combination of rigid body fitting and morphing. Initially, we used UCSF ChimeraX⁴⁸ to rigid body fit into the map the Protein Data Bank structures 3E6P for mzTΔF1,²⁹ 1XKB for fXa,⁴⁹ and 8TN9 for fV lacking the B-domain.⁵⁰ The model was subsequently subjected to multiple rounds of real space refinement in Phenix using the phenix.real_space_refine program.⁵¹ Initially, the model was morphed into the map to improve fitting and in subsequent stages real space refinement implemented reference model and Ramachandran restraints. COOT⁵² was used to manually improve the fitting of main chains and side chains within the experimental density in regions of higher resolution, as well as to address outstanding geometry issues including clashes, rotamer, and Ramachandran outliers. The process was repeated iteratively using both restrained refinement in Phenix (without morphing) and COOT to obtain the final deposited model 9YQ8. Molecular contacts were calculated using PISA.⁵³ The clash score of 14 largely results from >85% of close contacts generated by H atoms not included in our model but added for validation purposes.

smFRET measurements

Labeling with Alexa Fluor 555 and Alexa Fluor 647 probes at the engineered Cys of the prothrombin and mzTΔF1 R271C/Q576C/S525A mutants was carried out as described.³⁵ The construct was engineered to validate the information derived from the cryo-EM structure about the distance between the site of cleavage at R271 and the lower rim of the active site in the PD. Data collected by smFRET were processed as described.¹⁷

Numbering

Deposited sequences are numbered sequentially for the mature forms of fVa, mzTΔF1 and fXa. Residues of the PD of fXa in the text are also numbered according to chymotrypsin^54,55 and labeled with a “c” following the residue number.

Results

The recent cryo-EM structure of the prothrombin-fVa-fXa complex^30,33 elucidated the first step of prothrombin activation along the mzT pathway where cleavage at R320 generates mzT. In the second step, mzT becomes a new substrate for prothrombinase and presents R271 to the active site of fXa.^1,7 Structural considerations anticipate that moving R271 into the active site of fXa would require movement of fXa relative to fVa and/or a considerable downward shift of the PD of mzT (supplemental Figure 1).

The mzTΔF1-fVa-fXa complex

Several attempts to solve the structure of the mzT-fVa-fXa complex failed to resolve the Gla and kringle1 domains of mzT and therefore, prevented assessment of its overall architecture in terms of the closed and open conformations detected by smFRET measurements.^35,39,56 The high flexibility of the Gla domain has long posed a challenge to the X-ray structural analysis of coagulation factors like prothrombin,⁵⁷ mzT,^29,58 fXa,^49,59 and activated protein C (APC).^60,61 None of the X-ray structures of fXa currently deposited in the Protein Data Bank contains the Gla domain,^49,59 mzT was solved as mzTΔF1,^29,58 that is, devoid of Gla and kringle1 domains, and the only available structure of APC was also solved without the Gla domain.^60,61 Disorder in this domain was recently reported even for the cryo-EM structure of APC bound to fVa.⁶² Therefore, we used mzTΔF1 instead of mzT as a substrate of prothrombinase and hoped that removal of the Gla and kringle1 domains would improve resolution of the complex. Support to this strategy also came from the observation that the Gla and kringle1 domains of prothrombin make no contacts with fVa and fXa in the cryo-EM structure of the prothrombin-fVa-fXa complex.³⁰

The cryo-EM structure of mzTΔF1 bound to the fVa-fXa complex was solved at 3.8 Å resolution (Figures 1 and 2; Table 1). The PD and EGF2 domain of fXa align against the A2 and A3 domains of fVa, whereas mzTΔF1 assumes a tilted orientation with its PD facing the PD of fXa and kringle2 positioned near the A1-A2 domain interface of fVa (Figure 1., Figure 2., Figure 3.). Resolution ranged from ∼3 Å for residues in the core of fVa and at the protein-protein interfaces to ∼6 Å for most of kringle2 of mzTΔF1 and the EGF1 domain of fXa (supplemental Figure 2). The final model included the entire assembly of fVa and mzTΔF1 but only the EGF2-PD domains of fXa because the Gla-EGF1 domains lacked sufficient density to be resolved with confidence, unlike what was seen in recent cryo-EM structures^30,33 (supplemental Figure 2). Independent 3D variability analysis revealed significant conformational heterogeneity in regions of the map containing the EGF1 and Gla domains of fXa (supplemental Video 1). A similar conformational variability was seen for portions of the map around mzTΔF1, consistent with transient dissociation from the active site of fXa (supplemental Video 2). The conformational heterogeneity in these regions of the map explains the lower resolution that characterizes parts of the envelopes of the light chain of fXa and mzTΔF1.

Figure 2.

Cryo-EM structure of the mzTΔF1-fXa-fVa complex. (A) Cryo-EM density of the composite map viewed from top, bottom, and sides with embedded atomic models of the mzTΔF1-fXa-fVa complex shown as a cartoon. The domains of fXa (EGF2 and fXa-PD), fVa (A1-A2-A3-C1-C2), and mzTΔF1 (kringle2 [K2] and mzT-PD) are denoted in different shades of yellow (fXa), green (fVa), and purple (mzTΔF1). (B) Representative 2D classes of the ternary complex. (C) Gold-standard Fourier shell correlation (GSFSC) for the composite map calculated with (orange) and without (black) masking using the Phenix mtriage program.

Figure 3.

Molecular view of the fXa-fVa interface. Cartoon (fXa, yellow) and spheres (fVa, green) representation of the atomic model depicting structural elements important for stabilizing the interaction between fVa and fXa. The 120- and 170-helices of the PD of fXa are positioned at the interface with the A2 domain of fVa. The mainchain segments of these 2 helices fit accurately within the map density (larger inset). The smaller inset shows the map density around residues 655 to 674 of the A2 domain of fVa, which weave over the top of the PD of fXa. Specific contacts forming the fXa-fVa interface are listed in Table 2 and shown in supplemental Figure 4.

The fXa-fVa interface

Overall, the fXa-fVa complex is similar to that bound to prothrombin on nanodiscs³³ (9CTH, root mean square deviation = 2.5 Å over 1457 Cα atoms), where the atomic model of several portions of the PD of fXa and the 655 to 674 segment of fVa was updated to supersede the earlier entry 7TPQ³⁰ and prothrombin docking was optimized with all-atom refinement (supplemental Note 1). Hence, the switch in cleavage site from R320 to R271 produces only modest changes in the enzyme-cofactor assembly. The interface between fXa and fVa is stabilized by contacts involving the PD of fXa and A2 domain of fVa and, to a minor extent, the EGF2 of fXa and A3 domain of fVa (Figures 1 and 3; Table 2; supplemental Figures 3 and 4). We stress that modeling side chains in flexible maps of moderate to low resolution is challenging and may produce uncertainty in the estimated distances. Of particular importance in stabilizing the interface is the 170-helix of fXa that aligns parallel to the long turn comprising residues 500 to 514 of fVa. The 170-helix also makes several contacts with residues located in the 575 to 579 and 626 to 631 turns of fVa (Table 2; Figure 3; supplemental Figures 3 and 4). The relevance of the interaction of the 170-helix of fXa with the 500 to 514 turn of fVa validates the results of site-directed mutagenesis studies of fXa^{63, 64, 65} and fVa.^{66, 67, 68} Interestingly, a critical role for the 170-helix has also been reported in the interactions of fIXa with fVIIIa^{69, 70, 71} and of fVIIa with tissue factor,^{72, 73, 74} suggesting that vitamin K–dependent proteases containing EGF domains⁷⁵ evolved a common epitope for cofactor-dependent stimulation of catalysis.⁷⁶

Table 2.

H-bonds and salt bridges between fVa and fXa bound to mzTΔF1 (9YQ8) or prothrombin (9CTH)

fVa	fXa∗	Distance [Å]	fVa	fXa∗	Distance [Å]
H-bonds (9YQ8)			H-bonds (9CTH)
S502 [OG]	S355 (173c) [O]	2.6	S502 [OG]	S355 (173c) [O]	2.8
D513 [OD1]	Q360 (178c) [NE2]	3.9	A511 [O]	R347 (165c) [NH1]	3.4
D513 [OD1]	R347 (165c) [NH2]	3.0	D577 [OD2]	R347 (165c) [NH1]	2.4
T579 [OG1]	N348 (166c) [N]	3.9	D661 [OD1]	R424 (240c) [NH2]	2.6
T579 [OG1]	R347 (165c) [NH1]	2.5	E662 [OE1]	R273 (93c) [N]	3.2
T579 [O]	R347 (165c) [NH1]	3.3	D663 [OD1]	R424 (240c) [NH2]	2.7
T624 [OG1]	N348 (166c) [OD1]	2.5
T626 [OG1]	N348 (166c) [ND2]	3.4	Salt bridges (9CTH)
C656 [O]	R306 (125c) [NH2]	3.7	fVa	fXa	Distance [Å]
D661 [OD1]	R273 (93c) [NE]	3.1	D577 [OD1]	K414 (230c) [NZ]	3.6
D663 [O]	N272 (92c) [ND2]	3.6	D577 [OD2]	K414 (230c) [NZ]	3.1
D663 [O]	R273 (93c) [NH1]	2.8	D577 [OD2]	R347 (165c) [NH1]	2.4
S664 [O]	R424 (240c) [NH2]	2.7	D661 [OD1]	R424 (240c) [NH1]	2.9
E666 [O]	K270 (90c) [N]	3.3	D661 [OD1]	R424 (240c) [NH2]	2.6
F668 [O]	V268 (88c) [N]	2.9	D661 [OD2]	R424 (240c) [NH2]	3.1
F668 [N]	V268 (88c) [O]	3.8	E662 [OE1]	H271 (91c) [ND1]	3.3
P671 [O]	R243 (63c) [NE]	2.8	D663 [OD1]	R424 (240c) [NH1]	2.7
S1546 [OG]	D92 [OD1]	2.9	D663 [OD2]	R424 (240c) [NH1]	3.7
S1546 [OG]	D92 [N]	3.6	K655 [NZ]	D307 (126c) [OD1]	3.8
Q1629 [OE1]	K87 [NZ]	2.6
Salt bridges (9YQ8)
fVa	fXa	Distance [Å]
D513 [OD1]	R347 (165c) [NH1]	3.9
D513 [OD1]	R347 (165c) [NH2]	3.0
D661 [OD1]	R273 (93c) [NH2]	3.7
D661 [OD1]	R273 (93c) [NE]	3.1
D661 [OD1]	H271 (91c) [NE2]	3.8
D661 [OD2]	H271 (91c) [NE2]	3.4

Calculated using PISA.⁵³

^∗Chymotrypsin numbering of fXa residues is shown in parentheses.

Another epitope stabilizing the fXa-fVa interface is in the C-terminal region of the heavy chain of fVa, where the long segment of residues 655 to 674 from the A2 domain is positioned over the PD of fXa and forms extensive contacts with residues in the 88 to 93 turn, 120-helix and 240-helix of fXa (Table 2; Figure 3; supplemental Figure 4). The conformation of this segment is similar to that reported in the cryo-EM structure of the prothrombinase complex on nanodiscs³³ and the structures of prothrombinase from the venom gland of Pseudonaja textilis³² and an engineered fXa bound to fVa.⁷⁷ Interestingly, the segment of residues 654 to 670 engages the PD of APC as an exosite ligand in the cryo-EM structure of the fVa-APC complex⁶² and may function as a latch to capture target proteins. Progressive deletions of segments from the C-terminal region of the heavy chain of fVa have shown that binding of fXa is most perturbed upon deletion of residues 658 to 678,⁷⁸ in agreement with our cryo-EM structure. Significant perturbations in prothrombinase activity are also observed upon mutation of fXa residues that interact with the 654 to 674 segment of fVa, including R306 (R125c) in the 120-helix and K414 (K230c) and R424 (R240c) in the 240-helix.^63,65 Potential sulfation of Y665⁷⁹ (Figure 3) may establish interaction with R424 (R240c) nearby. No density could be detected past residue T674 of fVa and only few contacts were present between the EGF2 domain of fXa and A3 domain of fVa in the fXa-fVa interface (Table 2; Figure 3; supplemental Figure 4), in agreement with recent cryo-EM structures^30,33 where the fXa-fVa interaction is mediated mainly by contacts of the PD of fXa and A2 domain of fVa. Overall, the orientation of the fXa-fVa complex in the prothrombin and mzTΔF1 bound structures are similar except for a subtle translational movement that brings the PD of fXa closer to the A2 domain of fVa (supplemental Figure 3). Superposition of the 2 structures reveal small differences in the positions of the 120-, 170- and 240-helices of fXa, but these changes may be caused either by binding of mzTΔF1 or by weak binding of fXa to fVa detected by 3D-variability analysis (supplemental Video 1).

The mzTΔF1-fVa-fXa interface

The substrate mzTΔF1 bears most of the rearrangements observed in the ternary complex and assumes a tilted orientation, with the PD facing the active site of fXa and kringle2 positioned along the 305 to 320 loop of fVa located at the border between the A1 and A2 domains (Figures 1, 2, and 4). Binding of mzTΔF1 to this loop imposes steric constraints on R306 of fVa that preclude cleavage by APC, consistent with biochemical data.^80,81 Our structure predicts that only kringle2 of mzTΔF1 is close enough to fVa for potential interaction (Figures 1, 2, and 4), but the poor resolution (∼5-6 Å) of this domain (supplemental Figure 2) makes it difficult to identify residues that interact with fVa. Except for kringle2, no other structural segment of mzTΔF1 comes close enough to fVa to contribute to binding, which explains the modest cofactor effect of fVa in accelerating fXa cleavage of mzT (30-fold) compared to cleavage of prothrombin (>2000-fold).¹⁵

Figure 4.

The mzTΔF1-fXa interface. (A) Cartoon representation of the atomic model of mzTΔF1 (purple) bound to fXa (yellow surface) with the cryo-EM density around the 271-loop of mzTΔF1 shown as a mesh. (B) Close-up view of the 271-loop and of the interaction between R271 of mzTΔF1 and D189c of fXa. (C) Residue I268 of mzTΔF1 docking into the aromatic S4 site of fXa composed of the side chains of Y99c, F174c, and W215c. The cryo-EM density in panels B-C is shown as a gray mesh. (D) smFRET efficiency profiles of the prothrombin (proT) and mzTΔF1 mutants R271C/S525A/Q576C labeled with Alexa Fluor 555 and 647 fluorophores. The 2 profiles indicate that the loop housing R271 has a similar conformation in prothrombin and mzTΔF1.

The cryo-EM structure shows that mzTΔF1 is productively bound to fXa. Electron density for residues 267 to 286, comprising the site of cleavage at R271, could not be modeled in the X-ray structure of mzTΔF1 due to pronounced flexibility.²⁹ On the other hand, the cryo-EM structure identifies the 267 to 286 segment inside the primary specificity pocket of fXa (Figure 4A-B) at 3.5 Å resolution around the site of cleavage at R271 (supplemental Figures 2 and 5), thereby giving confidence to the assignment of the interaction of R271 with D373 (D189c) of fXa (Figure 4B). The backbone O atom of R271 also forms H-bonds with the side chains of the catalytic H236 (H57c) and Q240 (Q61c) of fXa, whereas its guanidinium group H-bonds to G402 (G218c) of fXa. Another contact is established by the side chain of I268 of mzTΔF1 (local resolution ∼3.5 Å, supplemental Figure 5) with the S4 subsite of fXa comprising the aromatic side chains of Y279 (Y99c), F356 (F174c) and W399 (W215c) (Figure 4C).⁸² The interaction confirms the importance of this subsite in binding direct oral anticoagulants⁸³ and the deleterious effect of mutating Y99c and F174c of fXa in the activation of prothrombin.^84,85

Special attention was devoted to modeling the loop housing R271 using the structure of prothrombin,³⁹ given that residues 252 to 286 are missing in the X-ray structure of mzTΔF1.⁶⁶ The model was validated with independent smFRET measurements after labeling the R271C/Q576C mutants of prothrombin and mzTΔF1 with Alexa Fluor 555/647 fluorophores. The interprobe distances measured for prothrombin (45 Å) and mzTΔF1 (43 Å) are in good agreement with the Cβ-Cβ atoms distance of 39 Å between residues R271 and Q576 in the cryo-EM structure and show that the loop housing R271 assumes a similar conformation in the 2 proteins (Figure 4D; supplemental Figure 6).

Discussion

In the mzT pathway of prothrombin activation, cleavage at R320 is followed by cleavage at R271.^1,17 Because the 2 residues are far away from each other,^30,33,39 a significant rearrangement in the orientation of mzT and/or prothrombinase is expected to enable R271 binding to the active site of fXa (supplemental Figure 1). Comparison of the cryo-EM structures of the fXa-fVa complex bound to prothrombin^30,33 and mzTΔF1 presented here defines the structural changes that govern the switch from R320 to R271. The transition is illustrated by a video and shows a pronounced change in orientation of mzTΔF1 that moves the guanidinium group of R271 more than 20 Å toward the active site of fXa (supplemental Video 3; Figures 1 and 5). The downward shift of the PD of mzTΔF1 brings the C-terminal barrel closer to fXa and positions the 271 loop in the active site of fXa (Figures 4 and 5). Small changes were noted in the orientation of fXa in the complex when compared to that bound to prothrombin,^30,33 but they may be caused by the propensity of fXa to dissociate from fVa (supplemental Video 1). Indeed, previous smFRET measurements failed to detect significant differences in the conformation of fXa when bound to prothrombin or mzT.⁴⁰

Figure 5.

Conformational changes linked to positioning R271 of mzTΔF1 in the active site of fXa after cleavage of prothrombin at R320. Shown is the PD of fXa (yellow cartoon) and fVa (light teal surface) bound to (A) prothrombin (purple) and (B) mzTΔF1 (purple). Residue R271 (magenta spheres) in prothrombin is positioned more than 20 Å away from D189c (yellow spheres) in the primary specificity pocket of fXa, whereas R271 in mzTΔF1 comes in direct interaction with D189c. Docking of R271 in the active site of fXa requires a downward motion of the PD of mzTΔF1 toward fXa that causes little changes in the conformation of fVa and fXa.

The cryo-EM structure of the prothrombin-fVa-fXa complex^30,33 suggests that the first step of prothrombin activation along the mzT pathway involves a rigid body association of the 3 components. Bound prothrombin in the cryo-EM structure is practically identical to the X-ray structure of the free form³⁹ and the fXa-fVa complex bound to prothrombin has essentially the same conformation as its free form on nanodiscs.^30,33 Assembly of prothrombinase is itself a rigid body association because both fXa and fVa assume similar conformations in the complex and the free form.^30,33,40,86 Binding through rigid body association minimizes entropy losses due to linked conformational transitions³⁴ and ensures high affinity and rapid association, as observed experimentally in the assembly of prothrombinase and in the rate of cleavage of R320 of prothrombin by prothrombinase.^1,17 Cleavage of prothrombin at R320 in the first step of the activation mechanism generates mzT as a new substrate for prothrombinase and prepares the second cleavage at R271. The switch in cleavage site from R320 to R271 involves a significant reorientation of the PD of mzTΔF1 required to move R271 into the primary specificity pocket of fXa. Overall, the 2 steps of prothrombin activation along the mzT pathway are brokered by a change in the relative orientation of the substrate rather than the enzyme and do not involve large conformational transitions of the components involved in the interaction, contrary to the conclusions from previous structural and kinetic models.^{7,12,14, 15, 16,18,31,32}

The results presented in this study add to recent cryo-EM successes in the analysis of coagulation factors and their complexes^{30,50,62,74,86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97} that are producing a structure-based view of the response to vascular injury.⁹⁸ A new paradigm emerging from these studies suggests that the PD plays a dominant role in protein-protein interactions. The molecular contacts in the interfaces involving such domains revealed by cryo-EM validate the results from mutagenesis studies, provide valuable insight into clinical phenotypes and offer suitable targets for therapeutic intervention.^{30,50,62,87,96} Auxiliary Gla, kringle and EGF domains seem less involved in protein-protein interactions and likely function as scaffolds that position substrate and enzyme for optimal docking.

Conflict-of-interest disclosure: The authors declare no competing financial interests.