Reversible stacking of lipid nanodiscs for structural studies of clotting factors

2.1 Reagents

HEPES sodium salt, Na cholate, MSP1D1, and MSP1E3D1 scaffolding proteins were purchased from Sigma (Sigma-Aldrich, St. Louis, MO, USA). PS (1,2-dioleoyl-sn-glycero- 3-phospho-L-serine, sodium salt) and GC (D-galactosyl-β-1,1′ N-nervonoyl-D-erythro-sphingosine, d18:1/24:1) were purchased from Avanti Polar Lipids Inc. (Alabaster, AL, USA). All buffer solutions – HBS (150 mm NaCl, 20 mm HEPES, pH 7.4) and HBS-Ca (HBS with 5 mm CaCl₂) – were filtered through a 0.22-μm Millex^® GP filter (Millipore Co., Cork, Ireland).

2.2 ND assembly and SEC

ND were prepared following the procedure described in refs. [7], [17]. Briefly, stock solutions of PS and GC lipids in chloroform were mixed in the desired weight ratio. The chloroform was evaporated under Argon gas, and the lipids were resolubilized in HBS buffer. The MSP1D1 and MSP1E3D1 were reconstituted in HBS buffer with 15 mm Na cholate and mixed with the lipids in the desired molar ratios. After 1-h incubation at 37°C, the Na cholate was removed by adding 1 g of Bio-Beads SM-2 per milliliter of mixture. After 4-h incubation at room temperature, the Bio-Beads were removed by centrifugation and ND were separated by SEC through a Superdex 200 HR 10/30-gel filtration column equilibrated in HBS buffer. Elution fractions of 0.5 ml were collected at 1-min intervals. The fractions corresponding to the ND peak were combined and concentrated. The final ND concentration was determined by the MSP absorption at 280 nm (extinction/280 nm 21,000 m⁻¹ cm⁻¹ for MSP1D1 and 29,400 m⁻¹ cm⁻¹ for MSP1E3D1, Sigma-Aldrich protocols: sigmaaldrich.com/etc/medialib/docs/) with a NanoDrop UV-VIS spectrophotometer (Thermo Scientific, Wilmington, DE, USA), as previously described [7]. The size of the ND was estimated from the SEC elution profiles by comparing to the elution profiles of standard proteins with known Stokes diameters: albumin (BSA) – 7.1 nm, aldolase – 9.6 nm, ferritin – 12.2 nm, and thyroglobulin – 17 nm.

2.3 Dynamic light scattering (DLS)

All DLS measurements were carried out with a Zetasizer μV particle analyzer (Malvern Instruments Ltd., Malvern, Worcestershire, UK) at a light-scattering detection angle of 90° with a 2-μl volume quartz cuvette at 21°C, as previously described [7]. All calculations were done with the Zetasizer Software [18].

2.4 Small-angle X-ray scattering (SAXS)

SAXS data were collected using a Rigaku BioSAXS-1000 camera with a FR-E++ X-ray source (Rigaku Americas, The Woodlands, TX, USA). For each measurement, 30 μl of ND sample at 1 mg/ml concentration was manually pipetted into a quartz capillary cell, sealed, placed under vacuum, and aligned in the camera. A series of 30-min exposures were collected and averaged in SAXLab to produce separate sample and buffer curves of 6–9-h exposure. Buffer subtraction, absorption correction, and molecular weight (MW) calibration were performed using the SAXNS-ES server (http://xray.utmb.edu/SAXNS). Data analysis was performed with the Primus software. The distance distribution function – P(r) – was calculated with the GNOM software, both from the ATSAS suite [18].

2.5 Electron microscopy (EM)

The negatively stained and cryogenic EM experiments were conducted, as previously described in Refs. [7], [19]. Shortly, the ND samples were diluted to 0.005 mg/ml in HBS or HBS-Ca buffer and 5 μl deposited on a freshly carbon coated plasma-treated hexagonal copper EM grid (300 mesh, Ted Pella Inc., Redding, CA, USA). The sample was stained with 1% Uranyl acetate for 2 min. Digital electron micrographs were collected with a 200-kV JEM2100-LaB6 electron microscope (JEOL USA Inc., Peabody, MD, USA) equipped with a 4k Ultrascan CCD camera (Gatan Inc., USA) at low electron doses (<16 e⁻ Å⁻² s⁻¹) and a final magnification of 56,400× (2.9 Å/pixel). The size of the ND was evaluated with the Digital Micrograph (DM) software (Gatan Inc., Pleasanton, CA, USA).

2.6 ND stacking experiments

ND stacks were obtained from ND assembled with MSP1E3D1 at MSP-to-lipid ratios of 1:80 and 1:150 in the presence of 5 mm CaCl₂. The number and length of ND stacks were followed at seven different Ca²⁺ concentrations ranging from 10 μm to 15 mm Ca²⁺. The ND were first incubated with the CaCl₂ solutions for 10, 30, and 60 min; diluted to 0.005 mg/ml and negatively stained as previously described in Refs. [7], [19]. EDTA was added in two times the excess of the CaCl₂ concentration to reverse the stacking and incubated for 10 min before negative staining. Digital electron micrographs were collected as described for the single ND, and the length and frequency of the ND stacks were evaluated with the DM software (Gatan, Inc., USA).

2.7 Cryo-electron microscopy (EM)

For the cryo-EM experiments, the ND stacks were formed at 10 mm CaCl₂ from ND assembled at MSP1E3D1-to-lipids ratio of 1:150 after 60-min incubations. The cryo-EM grids were prepared similarly to the negative stain samples, except for replacing the carbon-coated grids with holey grids (Ted Pella Inc., USA) and the negative staining with flash-freezing with a Vitrobot Mark IV (FEI Inc., USA), as described previously [13], [19]. Digital cryo-EM images were recorded at low electron doses (<16 e⁻ Å⁻² s⁻¹) and close to liquid nitrogen temperature (−174°C) with the same electron microscope and at the same EM conditions as the negatively stained samples.

3.1 Size distribution of ND assembled with different MSP and MSP-to-lipid ratio

All ND were assembled from a two-component lipid mixture of 80% PS and 20% GC that has been previously optimized for structural studies of clotting proteins by cryo-EM [7], [13], [15], [16]. The type of PS and PC was defined in our early studies of binding FVIII to LNT with different lipid composition [20], [21]. We further repeated this experiments with vesicles and ND assembled with MSP1D1 at MSP-to-lipids ratio of 1:47 [7], [14], [19] and confirmed that the 80% PS and 20% GC lipid composition is optimal for structural studies of membrane-bound FVIII in vitro [15], [16]. The ND assembled at a MSP1D1-to-lipids ratio of 1:40 showed a similar SEC elution profile to the one previously assembled at a MSP1D1-to-lipids ratio of 1:47 [7] (Figure 1). The theoretical area calculated for the circular arrangement of two MSP1E3D1 surrounding the ND bilayer is twice the size of the area calculated for circular arrangement of two MSP1D1 [2], [3], [10]. Therefore, for the assembly of ND with MSP1E3D1, we started with an initial MSP-to-lipids ratio of 1:80. The SEC elution profiles of the ND assembled at a MSP1D1-to-lipids ratio of 1:40 and at a MSP1E3D1-to-lipids ratio of 1:80 were quite similar, presenting a predominant elution peak of ND with a Stokes’ diameter centered around 10 nm (Figure 1, Table 1). To obtain ND with a larger diameter, we gradually decreased the MSP1E3D1-to-lipid ratio to 1:150. A shift to larger diameters was observed for the main eluted ND fraction at a MSP1E3D1-to-lipids ratio of 1:125 showing a broad elution peak spanning from 10- to 12-nm Stokes’ diameter. Further decreasing the MSP1E3D1-to-lipids ratio to 1:150 resulted in a well-defined elution peak of ND with a Stokes’ diameter centered around 11 nm (Figure 1, Table 1). As SEC is not always accurate for size determination of discoid (non-globular) particles, the size and morphology of the ND particles collected from the fractions indicated on Figure 1 (with dashed lines) were further evaluated by conventional EM, dynamic light scattering (DLS), and SAXS [7] (Table 1). The SEC elution profiles also showed that the ND population is quite polydisperse for the GC-PS lipid mixtures, type of MSP, and MSP-to-lipids ratio employed in this study (Figure 1). The correlation of SEC, DLS, SAXS, and EM methods presented in this work shows the potential of these techniques for the characterization of heterogeneous ND population in the context of ND optimization for structural studies of membrane-bound clotting factors by cryo-EM (Table 1).

Figure 1:

SEC of ND with 80% PS and 20% GC lipid composition assembled from either MSP1D1 or MSP1E3D1 at different MSP-to-lipid ratios. The grey areas indicate the fractions, which were collected for further characterization. The elution time of standard proteins with known Stokes’ diameters is shown: albumin (BSA) – 7.1 nm, aldolase – 9.6 nm, ferritin – 12.2 nm, and thyroglobulin – 17 nm.

For the EM experiments, the ND particles were negatively stained as absorbed on amorphous carbon, and the particles’ size and shape were evaluated from ~15 EM micrographs. Only micrographs showing homogenously distributed and evenly stained ND particles were selected for the statistics shown on Figure 2B. Less than 5% of the ND particles showed an oval shape due to incomplete or unsuccessful assembly. Less than 10% of the ND particles showed “side-on” views when absorbed on the amorphous carbon film (Figure 2A). Therefore, the diameter of each particle was determined by measuring the longest dimension of the ND directly from the digital micrographs, when the particles are adsorbed “face on” the carbon (Figure 2A). More than 500 representative ND particles were selected from 15 micrographs and for each MSP-to-lipid ratio. The ND particles were separated into four ND size groups with a bandwidth of 4 nm: 8±2 nm, 12±2 nm, 16±2 nm, and >18 nm (Table 1). The number of particles in each ND group was plotted as percentage of the total amount of measured particles (Figure 2B, Table 1). Direct visualization by negatively staining EM showed that the main population (~80%) of the ND assembled at a MSP1D1-to-lipids ratio of 1:40 and at a MSP1E3D1-to-lipids ratio of 1:80 has a diameter of ~12 nm (Figure 2B, Table 1). The ND assembled at a MSP1E3D1-to-lipids ratio of 1:150 showed nearly equal distribution of ND with diameter of ~12-nm diameter (36%) and ~16 nm (45%). A significantly larger population of ND (~17%) showed a diameter >18 nm (Figure 2B, Table 1). The ND assembled with MSP1E3D1-to-lipids ratio of 1:125 eluted with a broad peak after SEC at the same conditions as the ND assembled at the MSP1E3D1-to-lipids ratio of 1:80 and 1:150 (Figure 1). Our attempt to obtain more homogenous populations with ND assembled at a MSP1E3D1-to-lipids ratio lower than 150 was unsuccessful and not suitable for further characterization by the methods employed in this study.

Figure 2:

Negative staining electron microscopy (EM) of ND with 80% PS and 20% GC lipid composition assembled from either MSP1D1 or MSP1E3D1 at different MSP-to-lipid ratio. (A) EM micrographs of negatively stained ND absorbed on amorphous carbon film. Scale bar: 50 nm. (B) Size distribution graphs as measured directly from the EM micrographs.

The DLS and SAXS experiments showed similar values for the size of the ND assembled at a MSP1D1-to-lipid ratio of 1:40 and at a MSP1E3D1-to-lipid ratio of 1:80. These values correspond well to the ones measured by SEC and EM for the predominant ND population (Table 1). As the ND population is quite heterogeneous in size, the ND diameters calculated from the DLS values for the hydrodynamic radius of the ND particles in the Z direction are quite similar to the ND diameters measured from the EM micrographs for the main ND fractions (Figure 2B, Table 1). The SAXS scattering curves presented a characteristic shape for discoidal particles consisting of a sharp minimum and a broad maximum with identical radii of gyration (R_g)~5.2 nm and D_max of ~9.6 nm (Figure 3, Table 1) [2]. While the SAXS weighted averages for the radius of gyration (R_g) values are determined at high accuracy, they are well within the error of the DLS data and show an average for the ND diameter of 10.4 nm, which is slightly smaller than the one calculated from the DLS data of 11.2 nm (Table 1). Although the SAXS scattering curves for both ND populations (assembled with MSP1D1 or MSP1E3D1 at MSP-to-lipids ratio of 1:40 and 1:80, respectively) are very similar, they are not identical and display systematic variations with a Chi-squared metric of 2.9 up to q of 3 nm⁻¹ (Figure S1). The nearly identical distance distribution function – P(r) curves – suggests that these variations reflect changes in the organization of the MSP that circumscribe the lipids (Figure 3).

Figure 3:

Small-angle X-ray scattering (SAXS) of ND with 80% PS and 20% GC lipid composition assembled at MSP1D1-to-lipids ratio of 1:40 and at MSP1E3D1-to-lipids ratio of 1:80. (A) SAXS curves in reciprocal (Fourier) space and GNOM fits for the MSP1D1 (, ) and MSP1E3D1 (, ) ND. I(q) is radially symmetric, where I is the intensity of the scattered beam and q=2π/d is the scattering vector. 1/d is the reciprocal distance in Å⁻¹. The curves are offset for clarity. (B) Corresponding pair-wise distance distribution function – P(r) curves for MSP1D1 () and MSP1E3D1 () in real space. (r) is the inter-atomic distance in Å, and D_max is the maximum dimension of the particles.

3.2 Stacking of ND in the presence of CaCl₂

As the blood clotting proteins require 5 mm Ca²⁺ for function in vitro [7], [14], we undertook to investigate the effect of CaCl₂ on the ND stacking and aggregation. It is well known that introducing milimolar concentrations of Ca²⁺ to a solution of PS-rich liposomes induces strong aggregation of the vesicles [22], [23]. Therefore, we did expect changes in the monodispersity of the ND at millimolar Ca²⁺ concentration. What we observed, however, was not large aggregates but the formation of well-ordered ND stacks. We followed the ND stacking induced by the presence of 5 mm CaCl₂ for ND assembled at a MSP1E3D1-to-lipids ratios of 1:80 and 1:150, as representatives for the predominant ND size groups of ~12-nm and ~16-nm diameter, respectively (Figure 2). The ND were absorbed on amorphous carbon film and negatively stained after 10-min incubation with 5 mm CaCl₂ at room temperature (Figure 4). The length and average diameter of the ND stacks for each MSP1E3D1-to-lipids ratio were measured directly from ~15 EM micrographs. Increasing the incubation time to 30 and 60 min did not significantly affect the appearance and length of the ND stacks at the MSP1E3D1-to-lipid ratio of 1:80. A substantial increase in the length (~180 nm) and number of stacks per micrograph was observed for the ND assembled at a MSP1E3D1-to-lipids ratio of 1:150 after 60-min incubation with 5 mm CaCl₂ (Figure 4). The ND stacks were formed predominantly from the ND with a larger diameter (~16 nm) for both MSP1E3D1-to-lipids ratios (1:80 and 1:150). No stacks were observed for the ND with ~12-nm diameter assembled at a MSP1E3D1-to-lipids ratio of 1:150, and only short stacks (~25 nm in length) were observed for the MSP1E3D1-to-lipids ratio of 1:80 (Figure 5). For larger ND with diameter >18 nm, only a few short ND stacks were observed at a MSP1E3D1-to-lipids ratio of 1:80, whereas at a MSP1E3D1-to-lipids ratio of 1:150, a sizable population of ND stacks was formed. These stacks did not exceed 100 nm in length even after 60-min incubation (Figure 5).

Figure 4:

Negative staining EM of ND stacks formed from ND assembled at a MSP1E3D1-to-lipids ratio of 1:80 (row A) and a MSP1E3D1-to-lipids ratio of 1:150 (row B) after 10, 30, and 60 min of incubation with 5 mm CaCl₂. Scale bar: 80 nm. The insets show magnified views of the ND stacks from the respective micrographs, denoted with white dashed squares.

Figure 5:

Distribution of the ND stacks by length and diameter formed from ND assembled at MSP1E3D1-to-lipids ratios of 1:80 (A) and 1:150 (B) after incubation with 5 mm CaCl₂ for 10, 30, and 60 min.

3.3 Effect of the Ca²⁺ concentration on the ND stacks formation

To evaluate the effect of Ca²⁺ concentration on the ND stacking, we followed the length of the ND stacks from ND assembled at MSP1E3D1-to-lipids ratio of 1:150 after 60-min incubation with CaCl₂ at concentrations ranging from 0.1 to 15 mm. The ND stacks were assessed by several parameters: number of stacks per EM micrograph, average length of the stack, and diameter of the ND in the stack. The average ND stacks’ length at low CaCl₂ concentrations (<2.5 mm) corresponds to two-disc stacks (~14 nm). The average length of the ND stacks increases to ~19 nm at 2.5 mm CaCl₂ concentration, corresponding to three-disc stacks (Table 2, Figure 6A). A five-fold increase of the number of ND stacks per EM micrograph was observed at 5 mm CaCl₂ concentration (Table 2, Figure S2). Increasing the CaCl₂ concentration above 5 mm resulted in a pronounced increase in the length of the ND stacks with simultaneous decrease of the number of stacks per micrograph. Above 10 mm Ca²⁺, Y-branching and side-to-side bundling of the ND stacks were observed (Figure 6A) leading to the formation of large aggregates of ND stacks above 15 mm CaCl₂ concentrations that are not suitable for further EM characterization. Adding millimolar concentrations of EDTA to the ND stacks resulted in the reversing of the ND stacking after 10-min incubation (Figure S3). An increased number (~30%) of ND absorbed “side-on” on the amorphous carbon field was observed after adding EDTA, which can be useful for single particle reconstruction from 2D projections [7] (Figure S3). Cryo-EM of the ND stacks formed after a 60-min incubation with 10 mm CaCl₂ showed that the stacks could reach microns in length if suspended in vitreous water (Figure 6B). The ND stacks showed the same morphology and diameter when either negatively stained or frozen-hydrated, confirming that the negative staining procedure employed in our work is free of the artifact observed in other studies using slightly different staining procedures (Figure 6).

Figure 6:

Negative staining and cryo-EM micrographs of ND stacks formed from ND assembled at a MSP1E3D1-to-lipids ratio of 1:150 in the presence of 10 mm CaCl₂ after a 60-min incubation. Scale bar: 50 nm. (A) EM micrographs of negatively stained Y-branched and side-bundled ND stacks. Scale bar: 10 nm. (B) Cryo-EM micrographs of frozen-hydrated ND stacks suspended in amorphous ice (vitreous water) over the holes of a lacey carbon film coating an EM grid. Scale bar: 50 nm.

Free ND (not involved in stacks) for the ND assembled at a MSP1E3D1-to-lipids ratio of 1:150 at different CaCl₂ concentrations were observed even for high CaCl₂ concentrations (>10 mm) and long incubation time (>60 min) (Figure S4). The number and diameter of the free ND (not included in the stacks) per EM micrograph after 60-min incubation at different Ca²⁺ concentrations were also assessed (Figure S4). At low Ca²⁺ concentrations, the size distribution of the free ND corresponded roughly to the control population without CaCl₂. The number of free ND with diameter of 16±2-nm discs steadily declined as the CaCl₂ concentration exceeded 7.5 mm until fully disappearing at 15 mm CaCl₂ (Figure S4). These results support the observed increase in the length of the ND stacks with increasing CaCl₂ concentration (Table 2). The significant decrease in the number of free ND with a diameter of ~16±2 nm at greater than 5 mm CaCl₂ concentration confirms our observations that at 5 mm CaCl₂, the ND stacks are predominantly formed by ND with diameters >16 nm (Figure 2).