Lipid nanotechnologies for structural studies of membrane-associated clotting proteins by cryo-electron microscopy

1 Introduction

The biological processes localized to cell membranes are of great interest for scientific investigation as they are the primary targets for pharmaceutical interventions [1], [2], [3], [4]. Understanding the detailed mechanism and regulation of these processes requires the knowledge of the functional structure of the involved membrane and membrane-associated proteins at close to native lipid conditions. Such studies are often contrived by the in vitro systems meant to mimic the membrane environment of these proteins in vivo. To address that issue, suitable mimetic membrane nanostructures have been developed that allow functional and structural determination at near-physiological conditions [5], [6], [7]. Membrane proteins can be divided into peripheral and integral. Integral membrane proteins span the lipid bilayer, whereas peripheral membrane proteins associate with the membrane surface without permanently binding to it [8], [9]. The reversible attachment of these peripheral proteins to the biological membranes has been shown to regulate various cell functions through a diversity of mechanisms: spatially, by bringing closer membrane-associated enzymes to their substrate, and functionally, by promoting conformational changes, rearrangement, and dissociation in protein structural domains [8], [10].

Cryogenic transmission electron microscopy (CryoEM) is a high-resolution imaging technique that can generate subnanometer-scale 3D reconstructions of individual macromolecules and macromolecular complexes in solution. This is achieved by averaging 2D projection of the macromolecular complexes (single particles) in different orientations from CryoEM images, known as single-particle CryoEM (SPCryoEM) or by averaging 3D volumes of the macromolecular complexes from cryo-electron tomograms (CryoET), known as single-particle tomography (SPT) [11], [12], [13]. The flash-freeze approach employed for CryoEM sample preparation retains the aqueous environment preserving the biological macromolecules fully hydrated in amorphous ice below −180°C. Therefore, CryoEM is particularly suitable for structural analyses of aqueous formulations of protein complexes in a lipid environment, as well as of the lipid nanostructures such as liposomes, nanodiscs, and nanotubes. Recent advances in the development of direct electron (DE) imaging have contributed significantly to bring the resolution of SPCryoEM closer to the realm of macomolecular structure determination by X-ray crystallography. The quality of the CryoEM densities resolved today is remarkable and had led to an exponential increase in the near-atomic resolution structures resolved over the last 5 years [14], [15], [16]. The combination of high-resolution SPCryoEM with novel lipid nanotechnologies, such as nanodiscs (ND), has shown the potential of the CryoEM approach for structure determination of functional protein conformations in a membrane environment and at near-atomic resolution [13], [17], [18], [19].

The traditional approach for structural and biophysical studies of integral membrane proteins in vitro is the use of detergents [20], [21], which allows their structural characterization in micelles, bicelles [7], [20], [22], [23], [24], [25], and various cubic phases by X-ray crystallography and NMR [6], [22], [26]. Recently, near-atomic structures of solubilized membrane proteins stabilized by detergent or amphipoles have been resolved by SPCryoEM [27], [28], [29], [30]. In the case of peripheral membrane proteins, detergents are not recommended for structural studies as they do no reproduce accurately the lipid membrane integrity and fluidity. Detergents can also modify the membrane surface if mixed with lipids and affect the lipid-protein interactions, such as to impact the macromolecular organization and function of the associated protein complexes [21], [23], [31]. Therefore, there has been a need to optimize model lipid nanosystems specifically suited for structural studies of peripheral membrane protein complexes essential to understand the associated biology.

In this review, we present our achievements in developing lipid nanotechnologies for the structure determination of membrane-bound clotting proteins and specifically blood coagulating factor VIII (FVIII) by CryoEM. The designed experiments and consecutively resolved membrane-bound FVIII structures reviewed here [11], [32], [33], [34], [35], [36], [37], [38], [39] are based on our previous experience in organizing membrane-associated proteins, such as cholera toxin, annexin V, and coagulation factors V, VII, IX, and X [38], [40], [41], [42], [43], [44] on phosphatidylserine (PS)-rich monolayers and liposomes. The developed approach for structure determination of functional membrane-bound blood coagulation proteins can be applied for the structure-function investigations of various membrane-bound protein systems in a close to physiological environment. The described methodology and its application for the blood clotting complexes and specifically FVIII can open a new avenue for drug design, exploiting the possibility to modulate the function of the membrane-associated proteins at the lipid-protein interface.

2 Lipid nanotechnologies for structural studies of membrane-associated proteins by CryoEM

To study the structure of membrane-associated proteins by CryoEM, several lipid nanotechnologies have been developed (Figure 1):

Figure 1:

Lipid nanotechnologies for structural studies of membrane-associated proteins. The lipid nanotechnologies optimized in our laboratory for structural studies of factor VIII (FVIII) by CryoEM are shown. FVIII binds specifically to phosphatidylserine (PS)-rich lipid membranes, which is overexpressed on the activated platelet membrane in vivo. Nanotubes are single bilayer lipid nanotubes (LNTs) self-assembled from galactosylceramide (GC) that can incorporate a specific ligand similar to the lipid layer technique [39], [45]. Nanodiscs (NDs) are bilayer patches held together by membrane scaffolding proteins (MSPs) that can be assembled from various phospholipid mixtures, including GC at different contents of PS and lipids-to-MSP ratio [32], [35]. Liposomes are single bilayer lipid vesicles that can be assembled from various lipid mixtures and used to complement the lipid monolayer structural studies at the lipid compositions required for coagulation factors [38], [40]. Monolayers are lipid monolayers stabilized at the air/buffer interface that have been assembled at the same PS content as LNT and ND for 2D crystallization of membrane-bound coagulation proteins [37], [38], [40], [44]. Bilayers are lipid bilayers supported by a solid surface and have proven to be more suitable for AFM [46].

1. Supported lipid bilayers that represent planar lipid bilayers on a solid surface (Mica, Carbon films, SiO₂) [47], [48]. Advantages: obtaining flat and stable lipid bilayers. Disadvantages: lack of compartmentalization and curvature. This approach has not been very successful for CryoEM structure determination and has shown to be more suitable for atomic force microscopy (AFM) studies [46], [49], [50], [51].

2. Lipid monolayers that are planar lipid monolayers formed on an air/water interface. These monolayers are suitable for 2D crystallization of membrane-associated proteins by attaching the proteins from the liquid phase to a specific ligand incorporated into the lipid monolayer [38], [40], [41], [52], [53], [54], [55], [56]. Advantages: the membrane-associated proteins are kept fully hydrated and in functionally significant conformation, as they are bound to a specific ligand; require significantly smaller amounts of protein than 3D crystallography; suitable for high-resolution structure analysis by CryoEM [57], [58]. Disadvantages: 2D crystallization trials are time consuming and experimentally challenging; require tilting of the specimen in the electron microscope, which is limited to ±70° [47].

3. Phospholipid vesicles (liposomes) that are lipid bilayer enclosed compartments of aqueous solution with diameters from 30 nm to few microns. Advantages: similar to the bilayer curvature and compartmentalization of cells in vivo [59]; can mimic closely the lipid bilayer composition in vivo; suitable for CryoEM visualization of the lipid bilayer and morphology at subnanometer resolution [60], [61]. Disadvantages: high heterogeneity in the vesicles’ size and lamellarity; challenging for structural studies by Cryo-EM due to crowding of the membrane-associated proteins at the vesicles’ surface [13], [33], [62]. These lipid nanosystems have shown to be more suitable for biophysical characterizations – dynamic light scattering, circular dichroism [63], [64].

4. Lipid nanotubes (LNTs) are single bilayer galactosylceramide (GC) lipid nanotubes with ~20–24 nm in diameter that self-assemble in solution and are suitable for helical crystallization of soluble proteins and complexes [45]. Advantages: close to biological bilayer systems (curvature and compartmentalization); can mimic the lipid bilayer composition in vivo; more uniform in diameter than liposomes; require significantly less (~1000 times less) protein than 3D crystallization, suitable for structural studies by CryoEM [65]; do not require tilting of the specimen, as all views in a helical crystal are presented in the planar images. Disadvantages: require high-quality helical crystals, which is hard to achieve for membrane-associated protein complexes.

5. Lipid nanodiscs (NDs) that are self-assembled discoid lipid bilayers (8–16 nm in diameter) stabilized by a belt of two amphipathic helical scaffold proteins in solution [66], [67]. NDs are suitable for structure-function studies of membrane and membrane-associated protein and complexes [68], [69] by SPCryoEM [17], [18], [19], [70], [71]. Advantages: bilayer systems that can mimic lipid composition; monodisperse in solution, do not require ordered assemblies for structural studies by CryoEM. Disadvantages: lack compartmentalization and curvature; small size that may limit native protein-protein interactions and complex assembly.

In the light of the recent advances in near-atomic structure determination of macromolecular complexes by SPCryoEM, NDs are the most promising lipid nanotechnology. As nanometer-scale lipoprotein mimetic of membranes, NDs have proven to be of great utility for the purification, stabilization, and functional studies of integral membrane proteins [67]. NDs are chemically engineered self-assembled systems based on a transient form of the high-density lipoprotein (HDL) particles, named “nascent discoidal HDL particles” that are stabilized by the amphipathic ApoA1 apolipoprotein [72]. NDs are also a two-component lipid-protein system where the lipid component is organized as a bilayer, in which lipid composition and phase transition can be controlled. The protein component is a derivative of ApoA1 called membrane-scaffolding protein (MSP) that can be genetically engineered [73], [74]. The MSP render the NDs soluble and biocompatible due to its amphipathic nature. Numerous studies have been conducted to investigate the effect of the length of the MSP, composition of lipids and MSP-to-lipid ratio on the structure, and physicochemical properties of the NDs, showing that the lengths of the MSP and MSP-to-lipid ratio are critical for the size and monodispersity of the ND.

The size and homogeneity of the NDs are routinely defined by size exclusion chromatography (SEC) and dynamic light scattering (DLS) followed by small-angle X-ray scattering (SAXS) studies [66], [75], [76], [77]. These studies have demonstrated that for saturated dipalmitoylphosphtidylcholine DPPC lipids, increasing the length of the MSP from 200 to 280 amino acid residues produced ND with increased size (from 9.8 nm to 12.9 nm in diameter) [76]. Further optimizing the MSP to lipid ratio and synthesizing longer MSP (up to 400 residues) increases the ND diameter to 16–17 nm [78]. It has been suggested that the phase of the PC lipids in the ND bilayer (gel or liquid crystalline) does not affect significantly the ND size, the length of the MSP being the main determinant [79]. The phase diagram for saturated and unsaturated PC and different-length MSPs has been further characterized by differential scanning calorimetry (DSC) and recently with electron paramagnetic resonance (EPR) [80], [81]. The melting enthalpy measured for NDs showed significantly lower values than the melting enthalpy measured for liposomes with the same lipid composition [82] indicating that there is a loss of cooperativity due to the presence of boundary lipids, stabilized by the MSP (not participating in the phase transition) and the small size of the ND (160–300 lipids per ND). The 40% boundary lipids determined for the ND membranes is close to the 60% boundary lipids determined for native cell membranes, which are due to the high content of integral membrane proteins. Thus, the physicochemical properties of the ND bilayer are closer to the native cell membranes than those of liposomes. Several membrane proteins have been successfully reconstituted in NDs and subjected to various biophysical and structural studies [67]. Recent advances in single-particle SPCryoEM has contributed to near-atomic structures of few membrane proteins reconstituted in NDs: the TRVP1 ion channel at 3.2 Å [18], the Ryanodine receptor (RyR1) at 6.1 Å [17], and the Mg²⁺channel CorA at 7.1 Å [19].

Finally, NDs are self-assembled nanobioparticles that can be tuned at the molecular level for targeted drug delivery [83]. The human Apo A-I origin of MSPs suggests that NDs could be useful for therapeutic applications and have been successfully optimized for the delivery of a hydrophobic bioactive compound [84] and as carriers for small interfering RNA (siRNA) [85], active peptides, and therapeutic phospholipids [86], [87]. In our hands, binding of FVIII to PS-rich NDs prior to injecting in hemophilia A mice increases the therapeutic effect of the protein and its half-life time [88].

3 Structural studies of membrane-associated blood-clotting factors by Cryo-EM

The formation of membrane-bound complexes between specific coagulation factors at different cell surfaces is required for effective blood clotting [89], [90]. Two highly homologous coagulation complexes are assembled on the phosphatidylserine (PS)-rich activated platelet’s surface during the propagation phase of coagulation: the intrinsic tenase (FVIIIa-FIXa) and prothrombinase (FVa-FXa) complexes (Figure 2). The mechanism of assembly and function of these complexes is highly specific and depends on the lipid composition and charge of the membrane surface [91], [92], [93]. In the case of the intrinsic tenase complex, alteration in the sequence of either FVIIIa or FIXa causes mild to severe bleeding disorder known as hemophilia A and B, respectively. Therefore, knowing the macromolecular organization of the FVIIIa-FIXa complex in the presence of phospholipids and its synergy with the homologous FVa-FXa complex is critical to understand blood coagulation. Such knowledge will enable the design of novel pro- and anti-coagulant therapeutics and personalized treatments for blood disorders.

Figure 2:

The propagation phase of coagulation. The intrinsic tenase complex between the co-factor, factor VIIIa (VIIIa), and the serine protease, factor IXa (IXa), is highly homologous to the prothrombinase complex between the co-factor, Factor Va (Va), and the serine protease factor Xa (Xa). Both complexes assemble on the activated platelet surface rich in phosphatidylcholine (PS) securing the amplification of thrombin production (thrombin burst) required for successful blood clotting. The pro-enzyme (zymogene) factor X (X) is proteolytically cleaved by IXa to Xa that is the product of the tenase and the enzyme of the prothrombinase complex. Xa further cleaves the zymogene prothrombin (factor II, II) to thrombin (Factor IIa, IIa), which holds a central role in the blood-coagulation process. The size of the proteins and lipids is drawn to scale; the lipid bilayer thickness is estimated at 5 nm.

We have developed lipid nanotechnologies that mimic the negatively charged activated platelet membrane in order to investigate the membrane-bound organization of the blood-clotting proteins and complexes in the propagation phase of coagulation by CryoEM (Figures 1 and 2). These nanotechnologies have been optimized to study how the lipid composition and geometry of the cell membrane can modulate the macromolecular organization and function of membrane-associated complexes between serine proteases and their co-factors during the blood-clotting process at close to physiological conditions [11], [32], [33], [34], [35], [36], [37], [38], [40], [42], [44], [94], [95].

4 The membrane-associated FVIII structures

Factor VIII is a multidomain plasma glycoprotein, which when activated serves as a co-factor to the serine protease FIXa within the membrane-bound FVIIIa-FIXa complex [96]. Human FVIII is a 2332-amino acid protein organized in six domains arranged as A1-A2-B-A3-C1-C2 [97]. Recombinant FVIII is expressed without the B domain (FVIII-BDD), as the B domain is not essential for the FVIII procoagulant function (Figure 3). After purification, FVIII exists as a mixture of heterodimers of a heavy chain (HC) of the A1-A2 domains with parts of the linker expressed instead of the B domain, and a constant length light chain (LC) of the A1-C1-C2 domains non-covalently linked via divalent Ca²⁺ cation(s). After activation of FVIII by thrombin, FVIIIa becomes a heterotrimer composed of non-covalently linked A1, A2 domains and the LC (Figure 3). Recombinant porcine FVIII lacking the B domain (FVIII-BDD) has 84% sequence identity to the human analog and is expressed at much higher yield (10- to 14-fold) [99]. The active form, porcine FVIIIa, is significantly more stable in vitro especially in its membrane-bound form and assembles equally with human FIXa on PS-rich membranes, therefore, an excellent construct for structural and functional studies [100]. Both human and porcine FVIII-BDD are approved as therapeutics for the treatment of Hemophilia A.

Figure 3:

Domain organization of factor VIII. Factor VIII (FVIII) is a six-domain 2332-amino acid residue single-chain protein, consisting of three homologous A domains, two homologous C domains, and a heavily glycosylated B domain. FVIII-BDD is the B domain deleted recombinant FVIII molecule expressed without the B domain, which is replaced with a short linker of ~20-amino acid residues (light green). FVIII heterodimer is the form FVIII exists in solution where the heavy (HC) and light (LC) chains are non-covalently linked via divalent metal ions (Ca²⁺). FVIIIa heterotrimer is the thrombin-activated form of FVIII where the B domain/linker is fully excised by proteolytic cleavage (red arrows) at the acidic sequences (in orange) located between the A and B domains. The A1 and A2 domains of the heavy chain are held solely by hydrophobic interactions. FVIII-3D is the FVIII structure in solution, as organized in 3D crystals 3CDZ.PDB [98]. FVIII-2D is the membrane-bound FVIII structure from electron crystallography of FVIII organized in 2D crystals bound to a lipid monolayer surface 3J2Q.PDB [37]. FVIII-LNT is the membrane-bound FVIII structure from CryoEM of helically assembled FVIII molecules bound to LNT 3J2S.PDB [11], [36]. The five FVIII domains are color coded as in the primary sequence. The residues identified at the C2 domain-membrane interface are shown with a green surface. The sequences identified at the A2 and A3 domain-FIXa interface are indicated with purple ribbons. The three FVIII structures are shown with the FVIII heterotrimer in the same orientation.

The macromolecular organization and synchronization of the FVIII domains in the membrane-bound state are the driving force behind FVIII activity and the assembly of the FVIIIa-FIXa complex on the activated platelet surface in the propagation phase of blood coagulation (Figure 3). While the FVIII crystal structure in solution shows that the C1 and C2 domains are juxtaposed [98], [101], [102], a different domain organization was resolved by CryoEM for the membrane-bound FVIII organized in 2D and helical crystals, strongly affecting the C1 and C2 domain orientation. The rearrangement upon membrane binding of the A3-C1-C2 domains from the FVIII-light chain (LC) when organized in membrane-bound 2D crystals (FVIII-2D) or helical crystals (FVIII-LNT) is due to the flexibility of the molecule (Figure 4). The differences in the FVIII domain organization in the membrane-bound state are energetically allowed, as the macromolecular interfaces between the C1 and C2 domains, and the A1 and C2 domains, are strongly hydrophilic. The extensive hydrophobic interface formed between the A3 and C1 domains is preserved in all three FVIII membrane-bound organizations and stabilizes the orientation of the A domain heterotrimer, such as to secure the FVIIIa-FIXa interaction in the tenase complex [11], [36], [101] (Figure 4).

Figure 4:

Assembly of the tenase complex. The membrane-bound FVIII heterodimer is shown as a surface based on the fitting of the FVIII-LNT (Figure 3) within the CryoEM density resolved for the helically organized FVIII-BDD on LNT [11]. The density of the LNT bilayer resolved by CryoEM is shown as a gray mesh. From left to right are shown orthogonal views of the membrane-bound FVIII dimer – perpendicular and along the LNT surface. The FVIII light chain is shown as a dark gray surface and the heavy chain as a light gray surface. The residues at the FVIII monomer-monomer interface at a distance less than 8.7 Å are colored green and orange, respectively. The C2 domain-membrane-binding residues are shown as a light green surface. The A2 and A3 domain residues identified to bind to the FIXa are shown as magenta spheres. The FIXa molecules-PDB:1PFX [103] are shown as ribbons. The FIXa heavy chain (protease domain) is colored in dark blue, and the light chain holding the two EGF-like domains and the membrane-binding Gla domain is shown in cyan. The Asp39, His57, and Cys195 residues forming the FIXa proteolytic (active) site are shown as red spheres. The identified primary and secondary FIXa binding sites for FVIIIa are shown with green and yellow spheres, respectively. The residues from the membrane-binding FIXa-Gla domain are shown as cyan spheres.

We have studied both human and porcine FVIII-BDD forms bound to PS-rich liposomes, NDs, and LNTs by CryoEM to characterize how differences in sequence might affect the macromolecular interactions governing the membrane-bound assembly of these proteins and their complexes with human FIXa [11], [32], [33], [34]. Based on our helical and SPT CryoEM reconstructions for membrane-bound porcine FVIII-BDD [11] and the known crystal structures for the site-inhibited FIXa [103] and its membrane-binding Gla domain [104], [105], we have built a dimer model for the FVIIIa-FIXa complex membrane-bound organization in silico (Figure 4). Mapping the residues identified at the FVIIIa-FIXa and FVIIIa-FIXa-membrane interfaces [98], [106] further validated the proposed membrane-bound FVIII heterodimer organization and consecutive tenase complex assembly on the activated platelet surface (Figure 4). The amino acid residues from the FVIII monomer-monomer interface within the membrane-bound heterodimer were at a distance less than 8.7 Å, mapped and compared to both human and porcine amino acid sequences (Table 1). As the helical organization of human and porcine FVIII-BDD on PS-rich LNTs differs consistently [34], knowing the critical residues at the respective FVIII monomer interfaces will help generate functional FVIII constructs that differ in their membrane-bound oligomer organization while retaining their clotting function. This knowledge will give us the ability to safely modulate FVIII activity for therapeutic applications through the combination of genetic engineering and lipid nanotechnology optimization.

The proposed “super” Tenase complex dimer organization also supports the six orders of magnitude amplification of FXa generation, when both components – FIXa and FVIIIa – are membrane bound [107]. The geometry of the membrane-bound FVIII dimer further secures the space and positioning required to accommodate the membrane-bound substrate – Factor X (FX) in proximity to the identified FX interaction site on the A1 and A2 domains such as to allow proteolysis by the FIXa active site (Figures 2 and 4). The assembly of “super” clotting complexes on lipid nanotechnologies that mimic the activated platelet surface can not only elucidate the macromolecular organization of the proteins involved in the propagation phase of coagulation but also bring the treatment of genetically diseases such a hemophilia A and thrombosis to a new level.

5 Conclusion

Combining lipid nanotechnologies with structure determination by CryoEM is a powerful approach to resolve the macromolecular organization of membrane-associated protein complexes at near-atomic resolution and at close to physiological conditions. This approach will advance and facilitate the characterization of protein-lipid and protein-protein interfaces that govern macromolecular assembly in vivo. These “druggable” interfaces can be further targeted to modulate the function of critical protein complexes assembled in a lipid environment, which will open the possibility to regulate biological processes at different cellular interfaces.

Understanding the mechanism of membrane-bound assembly between the clotting proteins in the tenase and prothrombinase complexes is of great importance in understanding the blood coagulation process as a whole. The membrane-bound domain organization of FVIII on different lipid nanotechnologies such as NDs and LNTs, as resolved by CryoEM, is essential to understand and modulate the function of these complexes on the activated platelet surface. By incorporating different lipid components in the ND and LNT bilayer, we can further characterize the macromolecular interactions leading to the functional membrane-bound organization of the clotting proteins in the tenase and prothrombinase complexes. The combination of complimentary lipid nanotechnologies such as LNTs, NDs, and liposomes that have different shapes and curvatures will help to further elucidate the macromolecular organization of the clotting proteins and gain control over the propagation phase of blood coagulation. The presented approach and developed methodology can be further optimized and applied for structure-function study of many membrane-associated protein complexes involved in fundamental biological processes.