Aggregates Dramatically Alter Fibrin Ultrastructure

Abstract

Among the many factors influencing fibrin formation and structure (concentration, temperature, composition, pH, etc.), it has been suggested that the polydispersity of fibrinogen may play an important role. We propose here a detailed investigation of the influence of this parameter on fibrin multiscale structure. Two commercial fibrinogen preparations were used, a monodisperse and a polydisperse one. First, the respective compositions of both fibrinogen preparations were thoroughly determined by measuring the fibrin-stabilizing factor; fibronectin; α, β, and γ intact chain contents; the γ/γ′ chains ratio; the N-glycosylation; and the post-translational modifications. Slight variations between the composition of the two fibrinogen preparations were found that are much smaller than the compositional variations necessary to alter significantly fibrin multiscale structure as observed in the literature. Conversely, multiangle laser light scattering-coupled size exclusion chromatography and dynamic light scattering measurements showed that the polydisperse preparation contains significant amounts of aggregates, whereas the other preparation is essentially monodisperse. The multiscale structure of the fibrins produced from those two fibrinogen preparations was determined by using x-ray scattering, spectrophotometry, and confocal microscopy. Results show that fibers made from the aggregate-free fibrinogen present a crystalline longitudinal and lateral structure and form a mikado-like network. The network produced from the aggregates containing fibrinogen looks to be partly built around bright spots that are attributed to the aggregate. The multiscale structure of mixtures between the two preparations shows a smooth evolution, demonstrating that the quantity of aggregates is a major determining factor for fibrin multiscale structure. Indeed, the effect of a few percent in the mass of aggregates is larger than any other effect because of compositional differences under the same reaction conditions. Finally, we propose a mechanistic interpretation of our results, which points at a direct role of the aggregates during polymerization, which disrupts the ideal ordering of monomers inside fibrin protofibrils and fibers.

Significance

Fibrin formation, structure, and properties are major biophysical topics because of their potential relevance to cardiovascular and thromboembolic diseases. Although there has been a recent surge of work in this area, most studies used the same fibrinogen provider and therefore possessed very similar compositional and polydispersity profiles. We show that different fibrinogen preparations with identical compositions but containing or not containing fibrinogen aggregates produce fibrins with different structures, uncovering a new and essential control parameter. Results strongly suggest that characterization by size exclusion chromatography should be systematically performed to ensure that results are comparable from study to study. Furthermore, the importance of the modifications due to relatively minute amounts of aggregates suggests that they may play a major role in fibrin polymerization.

Introduction

The formation of the fibrin clot is essential in the process of blood coagulation. Fibrin forms a protein scaffold that enables the organism to close off damaged blood vessels. In the first step of fibrin formation, the fibrinopeptides A and B from the protein fibrinogen are cleaved by thrombin, producing the so-called fibrin monomers, which then polymerize into a fibrin network. Fibrinogen itself is a plasma 340-kDa centrosymmetric protein. Its structure is constituted by three aligned domains. The central part of the molecule (E region) is slightly smaller (∼5 nm of diameter), whereas the distal regions (D regions) present a 6-nm diameter to form a structure ∼45 nm in length.

Although some aspects of the formation of the fibrin network from fibrin monomers are still under debate, it is known that various genetic and environmental factors influence not only the fibrin structure and function but that they can also be related with thrombotic disease (1). Indeed, many clinical studies have associated the fibrin properties with thrombosis (2,3). Among the many factors influencing fibrin polymerization, structure, and function (ionic strength, pH, concentrations, γ′ content, etc. (4, 5, 6, 7, 8, 9, 10)), Huang et al. (11,12) suggested that the size dispersity (i.e., the amount of aggregates or oligomers present in the fibrinogen preparation) may play an important role. Indeed, fibrins made from chromatography-fractionated fibrinogen exhibited final turbidities between two and three times higher than the unfractionated samples (11), and significant differences both in kinetics as well as in the apparent size of the polymerizing objects were observed in a later study (12). However, a small angle x-ray scattering study of this fractionation process showed significant in-column degradation (13), and gel filtration could change the concentration in important proteins present in the fibrinogen preparations (called co-purified proteins), such as fibronectin, fibrin-stabilizing factor (FXIII), etc. Likewise, fibrinogen isoforms (γ′ content, oxidized forms, N-/C-terminus cleaved forms, etc.) may be prone to variability during the chromatography fractionation process and therefore behave specifically vis-à-vis polymerization.

As the relative compositions of the mono- and polydisperse fibrinogen preparations were not characterized nor the molar masses of the different fractions measured, the question of whether the observed effect is a consequence of variations in aggregate content and/or composition remains open. Furthermore, single-wavelength optical-density measurements only indicate an overall change in fibrin structure but provide no indication about the morphological changes or the scale(s) affected nor about the involved mechanisms.

In the last decade or so, ∼two-thirds of the fundamental work on fibrin formation has been performed on fibrinogen from Enzyme Research Laboratories (South Bend, IN) (see Supporting Materials and Methods, Section 1), hence with rather constant compositional and polydispersity profiles. Although this is pertinent for comparison purposes, the effect of fibrinogen’s polydispersity was mostly overlooked in the current fibrin-related literature, whereas it is a well-known factor in standard polymer science (14).

The above observations raise four important questions. First, does fibrinogen dispersity really influence the multiscale structure of fibrin, or were Huang et al.’s observations (11) a consequence of the purification procedure they used? Second, what are the scales affected by dispersity, and how are those scales affected? Third, are those differences also observed at physiological fibrinogen concentrations (2–4 mg/mL) or limited to the very low concentrations used in previous studies? Finally, are the dispersity-induced structural effects important? In other words, are they smaller or larger than those induced by varying compositional aspects of the fibrinogen preparations?

In the following, we start by describing the numerous experimental methods used in this work. To characterize the two fibrinogens, we performed refractive index and multiangle laser light scattering (MALLS)-coupled size exclusion chromatography (SEC), dynamic light scattering (DLS), reverse phase chromatography coupled to mass spectrometry (MS), and all necessary co-purified proteins quantification assays. The methods used to determine the structure of the fibrins at each scale are then described, starting with small angle x-ray scattering (SAXS), spectrophotometry, and finally, confocal microscopy.

Then, we present the detailed characterization of the physical and physicochemical properties of the different fibrinogens, showing that the two fibrinogens used in this study present identical compositional profiles while differing only in their aggregate content. We then investigate the effect of this size dispersity on the nano- and microscale structure of the fibrin, showing that the two fibrins present vastly different structures, at all scales. We finally discuss these findings in the light of recent results concerning the effect of compositional and isoform changes on fibrin structure and propose a mechanistic explanation of our results.

Materials and Methods

Materials

Human thrombin was purchased from Cryopep (Montpellier, France) as a 12 μL solution containing 298.9 IU. The solution was diluted to 200 IU/mL in a 2-(N-morpholino)ethanesulfonic buffer (20 mM 2-(N-morpholino)ethanesulfonic and 50 mM NaCl (pH 6.5)), aliquoted, and kept frozen at −80°C.

Two fibrinogens were used: Clottafact (Laboratoire Français du Fractionnement et des Biotechnologies, Les Ulis, France) and Fib1 (Enzyme Research Laboratories). In the following, Clottafact will be termed as FibWoA for fibrinogen without aggregates, whereas Fib1 will be termed as FibWA for fibrinogen with aggregates. The reason for this terminology will become apparent in the SEC and DLS results.

Fibrinogens were reconstituted using manufacturers’ guidelines: Fib1 and Clottafact fibrinogens were reconstituted by adding, respectively, 25 and 100 mL of sterile water for injection into the product vials. Then, the product vials were incubated at 37°C and gently swirled until the product was fully dissolved. Both fibrinogens were dialyzed together twice overnight against >100 volumes of HEPES buffer (140 mM NaCl, 20 mM HEPES, and 5 mM CaCl₂ (pH 7.4)), aliquoted, and kept frozen at −80°C. Fibrinogen concentrations were determined by absorbance at 280 nm using a specific absorption coefficient of E²⁸⁰ = 1.51 mL/mg⁻¹/cm⁻¹.

Dithiothreitol, iodoacetamide, urea, phosphate-buffered saline, ammonium carbonate, and HEPES were purchased from Sigma-Aldrich Chemical (St. Louis, MO). Acetonitrile (MeCN) was HPLC reagent grade and purchased from JT Baker (Philipsburg, NJ). Trifluoroacetic acid (TFA) was from Merck Biosciences (Darmstadt, Germany). All the aqueous solutions were prepared using ultrapure water (18.2 MΩ-cm resistivity at 25°C, total organic carbon <5 ppb).

Sample preparation

Fibrinogens were thawed at 37°C for 5 min, equilibrated at room temperature for another 5 min before use, and filtered with a 0.2-μm ClearLine syringe filter. Concentrations were then determined by absorbance at 280 nm and adjusted as desired. Thrombin was thawed at 37°C for 1 min, diluted in HEPES buffer, and immediately used.

Fibrin clots were formed by incubating fibrinogen (typically 0.5, 1, and 3 mg/mL) with 0.1 IU/mL thrombin (final concentration) at 37°C for 90 min.

Size exclusion chromatography

The molecular distribution of the different fibrinogens was analyzed using a Superose 6 column (GE Healthcare, Chicago, IL) on an Elite LaChrom system (Hitachi, Tokyo, Japan), coupled with a Dawn Heleos II Multi Angle Static Light Scattering system (Wyatt Technology, Santa Barbara, CA) and a Optilab T-rEX Refractometer (Wyatt Technology). The same HEPES buffer as before was used. 50 μL of fibrinogen (5 mg/mL) were applied, and the column was developed at a 0.5 mL/min flow rate. Elution (λ = 280 nm), MALLS, and refractometer profiles were analyzed using the ASTRA software from Wyatt Technology.

Dynamic light scattering

Dynamic light scattering (DLS) measurements were performed using a CGS-8FS/N069 apparatus from ALV Technologies (Manila, Philippines) with a 35 mW, 632.8-nm laser from JDS Uniphase (Milpitas, CA). Fibrinogens (1 mg/mL) were loaded in 10-mm diameter cylindrical cells and immerged in a toluene bath at 25.0 ± 0.1°C. Data were collected at 90° for 120 s. Hydrodynamic radii distributions were determined using Contin analysis.

Protein assays

Fibronectin level was assayed on a Siemens BN II nephelometer (Munich, Germany). Briefly, the sample is mixed with a polyclonal rabbit anti-human fibronectin to generate immune complexes measured by nephelometry. The fibronectin concentration is deduced by interpolation with a standard curve using the Dade Behring N protein standard PY.

FXIII levels were determined by enzyme-linked immunosorbent assay using sheep polyclonal anti-human FXIII (CL20057A; Cedarlane Laboratories, Burlington, ON, Canada) for the coating and the same polyclonal anti-human FXIII conjugated with horseradish peroxidase (CL20057HP; Cederlane Laboratories) as secondary antibodies. Standard human plasma (ORKL 17; Siemens) was used to establish the calibration curve. Results are expressed as an international unit (IU), knowing that 1 IU is equivalent to 30 μg/mL.

Reverse phase chromatography coupled to MS

Fibrinogen (100 μg) was vacuum dried and dissolved in 35 μL of an 8 M urea and 0.4 mM ammonium carbonate solution at pH 8.0. Reduction was done by adding 10 μL of a 40 mM dithiothreitol solution in water and incubating the resulting mixture for 20 min at 50°C. After cooling at room temperature, 10 μL of an 80 mM iodoacetamide solution in water were added, and the solution was incubated at room temperature for 20 min in the dark. Reverse phase high-pressure liquid chromatography (RP-HPLC) was performed using an ACQUITY UPLC system (Waters, Milford, MA). An amount of 20 μg of sample was injected on a Pursuit 3 diphenyl reverse phase column (150 × 2.0 mm, 3 μm; Agilent Technologies, Santa Clara, CA) equilibrated at 70°C and operated at a flow rate of 200 μL/min. An aqueous solution containing 0.1% TFA and MeCN containing 0.1% TFA were respectively used as buffer A and buffer B; proteins were eluted by using an increasing gradient of buffer B. After separation, reduced and alkylated fibrinogen chains were detected by UV at 280 nm, and MS analysis was achieved by interfacing the ultraviolet detector output to a SYNAPT G2-S High definition Mass Spectrometer (Waters) scanning from m/z 500 to 2000.

Small angle x-ray scattering

Small angle x-ray scattering (SAXS) experiments were performed at the ID02 line at the European Synchrotron Radiation Facility (Grenoble, France). The samples (two volumes of fibrinogen and one of thrombin) were mixed directly in a 12 × 12 × 3 mm homemade cell with 20 μm mica windows and thermostated at 37 ± 0.3°C (15). The sample-to-detector distance was set to 7 m, and the acquisition time was set to 0.1 s. To avoid radiation-induced degradation of the protein, the cell was displaced by a motorized x-y stage between each measurement in a snake-like fashion, by 2 mm horizontally and 2 mm vertically at the end of a line (i.e., each time by a distance much larger than the beam size (0.5 × 0.5 mm)). The results presented here are the average of the 15 last acquisition times corresponding to steady-state scattering curves. The constancy of the signal over several centimeters of the cell shows that the polymerization was finished and that the initial mixing was good. A comparison between two replicate experiments is presented in Fig. S6.

Spectrophotometry

Fibrin gels were formed in 96-well Immulon 2 HB Plates (Thermo Fisher Scientific, Illkirch, France) by mixing two volumes of fibrinogen (120 μL) and one of thrombin (60 μL). Optical-density spectra (500–800 nm) were measured after 90 min at 37 ± 0.3°C using a SPECTROstar Omega (BMG LABTECH, Ortenberg, Germany). For the total volume used in the experiments, the path length was 4.38 mm as determined from calibration using a range of known absorbance.

Optical-density data were analyzed using a corrected version of Yeromonahos’s (7,15,16) model. A detailed discussion concerning the model choice, pertinence, and potential inaccuracies is presented in Supporting Materials and Methods, Section 4, with raw data (Fig. S3) and fit residuals (Fig. S4, left). It should be kept in mind that the protofibrils numbers, fiber radii, and protein densities are not directly measured but are the result of the fitting of this simple model to the spectral data. Therefore, those data, although quantitative, must be viewed with some caution as the use of a different model may give significantly different results.

Each data point represents the mean and SD calculated from at least three individual experiments.

Confocal microscopy

Microscopic images of fibrin networks were obtained using Alexa Fluor 488 fluorescent fibrinogen (Invitrogen, Breda, the Netherlands) mixed with the unlabeled fibrinogens at a 1:10 ratio. Each mixture was filtered with a 0.2 μm ClearLine syringe filter, and the concentrations were then determined by absorbance at 280 nm. Samples were polymerized using 0.1 IU/mL of thrombin and various fibrinogen concentrations (0.5, 1, and 3 mg/mL). Fibrinogen and thrombin were mixed in Eppendorf Tubes 3810X, rapidly injected in SecureSeal Hybridization Chambers (Grace Bio-Labs, Bend, OR), and polymerized during 60 min at 37°C. Confocal image stacks of the fibrin networks were acquired using a Zeiss LSM710 confocal microscope with a 63×/1.2 water immersion objective (ZEISS, Oberkochen, Germany). The three-dimensional (3D) stacks (100 images) were 67.5 μm in the x-y direction and 25 μm in the z direction with resolutions of 100 and 250 nm, respectively. The confocal image stacks of the fibrin networks are homogeneous both in the imaging plane and also in depth (data not shown).

For the visualization and presentation of those stacks, we use a recently developed ImageJ plugin: Smooth Manifold Extraction (SME) (17). This method, unlike z-maximal intensity projections, provides a robust two-dimensional representation of 3D objects, which preserves local spatial relationship, particularly in branching (see Supporting Materials and Methods, Section 6.1 for details and examples).

Finally, the pore size distribution was determined for each frame using the bubble method (18,19) on each frame and calculated for the complete stack. For details about the segmentation and pore size analysis, see Supporting Materials and Methods, Section 6.2 and Supporting Materials and Methods, Section 6.4.

Results

Physicochemical characterization

Molecular size distribution

The SEC elution profiles of the two fibrinogens are presented in Fig. 1 A. They both show a main peak at an elution volume of 11 mL as well as a small broad peak around 9.3 mL, albeit with different intensities (Fib1 > Clottafact). In addition, Fib1 presents a well-defined high molecular weight peak occurring at an elution volume of ∼7.5 mL. The analysis of the MALLS and index refraction profiles obtained simultaneously to the elution profile (see Materials and Methods) yielded the molar masses of the species corresponding to each of those elution peaks as well as their relative concentrations.

Figure 1.

(A) Molecular size distribution of Clottafact and Fib1 determined by SEC. (B) Multiscale structure of the fibrin clot. To see this figure in color, go online.

A (weight-averaged) molar mass of 325 kDa is obtained for the 11 mL elution peak observed in both fibrinogens, which corresponds closely to the molar mass of fibrinogen monomers (13). The ratio of the weight-averaged mass to the number-averaged mass for this peak is below 1.01, indicating good monodispersity. For the broad small peak at 9.5 mL also observed in both fibrinogens, a molar mass of ∼1 MDa is obtained, with a significant polydispersity, corresponding probably to a distribution of oligomers. Finally, a mass of ∼5 MDa can be assigned to the large peak observed at 7.5 mL in Fib1, corresponding most likely to fibrinogen aggregates. The elution profile and molar masses obtained for Fib1 are in excellent agreement with those of Brookes et al. (Fig. 3; (13)), which were also obtained on Fib1 fibrinogen, but using a SEC column coupled to a SAXS cell.

Figure 3.

(A) Qualitative geometry of the fibrin networks. Smooth manifolds were extracted from 3D confocal microscopy stacks at [Fg] = 3 mg/mL. White arrows show curved fibers, whereas red arrows show bright dots. Frames are 67.5 μm wide. The insets are ×2 enlargements of a portion of the original SME image (12 × 12 μm). (B) Number of protofibrils, protein density, and radius (inset) versus content in FibWA. Closed circles represent protofibrils number; open circles represent protein density. Inset: fibers average radius is shown. Error bars are standard deviations calculated from at least three individual experiments. To see this figure in color, go online.

Those results demonstrate the presence of both oligomers and large aggregates in Fib1, whereas the Clottafact preparation is close-to-monodisperse with a small number of oligomers and no detectable aggregates. The relative mass of the aggregates present in Fib1 is found to be of ∼4% of the total mass of fibrinogen.

Dispersity analysis by DLS

DLS was used to confirm the above results by determining the average hydrodynamic radius of the different fibrinogens. Because the SEC-MALLS results showed a tridisperse size distribution, we tried to use a triple exponential fit of the raw correlation spectra to obtain the hydrodynamic size distribution of the molecules. However, those fit were not robust as the result was strongly dependent on the initial values used for the fit. Therefore, we only present the average hydrodynamic radii and the polydispersity indexes obtained by a Contin analysis.

FibWoA fibrinogen has a hydrodynamic radius of 11 nm, a hydrodynamic size corresponding well to that of fibrinogen monomers (11,12,20,21). The measured polydispersity index is of 25%, close to the value of 20%, below which samples are considered truly monodisperse. Conversely, FibWA fibrinogen has a hydrodynamic radius of 22 nm, with a polydispersity index of 52%. This result confirms that FibWA fibrinogen presents mixtures of higher molecular weight species.

RP-HPLC-MS analysis

The two fibrinogens were further analyzed by RP-HPLC-MS after reduction and separation of the α, β, and γ chains. MS was used to determine the sequence integrity of the different chains as well as their post-translational modifications. The resulting chromatograms (UV detection) for the two fibrinogen preparations are essentially identical in all respects (see Fig. S1; Table S2 for a description and summary of the MS-based identifications). Finally, sodium dodecyl sulfate polyacrylamide gel electrophoresis of reduced fibrinogens showed similar γ′/γ chains ratio with 14% for FibWoA and 10% for FibWA (data not shown).

Co-purified proteins content

Co-purified proteins (FXIII and fibronectin) concentrations measured as described in Materials and Methods are shown in Table 1 (see Fig. S2). The main finding is that FibWoA and FibWA have an almost identical composition in co-purified proteins. The only slight difference is the fibronectin content (0.05 vs 0.02 mg/mg), which is low compared to the normal plasmatic value (0.12 mg/mg) and much smaller than the range for which significant structural variations are observed in the literature (0.8 mg/mL) (21,22) (see Table S3).

Table 1.

Physicochemical Characteristics of the Two Fibrinogens

	FibWoA	FibWA
Presence of aggregates	Not detectable	4% in total Fg mass
Average hydrodynamic radius (nm)	11	22
FXIII (U/mg Fg)	0.26	0.25
Fibronectin (mg/mg Fg)	0.05	0.02
Post-translational modifications	Identical	Identical
Intact α-chains (% versus FibWoA)	100%	95%
Respective intact β and γ chains	100%	100%
γ′/γ chains ratio (%)	14	10

The results of the physicochemical analyses are summarized in Table 1 and show that the composition of the two fibrinogens is almost identical except for the presence of large molecular weight aggregates.

Fibrin’s ultrastructure

Because polydispersity could influence fibrin structure at the molecular, fiber, and network scales, each scale was investigated with the appropriate tool (see Fig. 1 B).

Fibrin fibers internal structure

We used SAXS to determine the molecular organization inside fibrin fibers. Fig. 2 A shows x-ray intensity spectra obtained on mature fibrin clots. The general shape of the scattering curves is identical to those obtained in the literature (7,23). A replicate experiment was performed, yielding an identical spectrum (see Fig. S6), showing the excellent reproducibility of the protocols and methods. Spectra from each fibrin gel showed a main peak at a q-vector ∼0.3 nm⁻¹, corresponding to the usual 22.5-nm periodicity of half-staggered fibrin monomers. However, significant differences are observed around this peak (i.e., between 0.2 and 0.5 nm⁻¹).

Figure 2.

(A) Scattered intensity of SAXS for fibrin clots (1 mg/mL). (B and C) Relative intensity of around the 22.5-nm periodicity peak for respective FibWoA and WibWA. Continuous lines are the total fit, whereas dashed lines are the individual peaks. The R2 values are the goodness-of-fit coefficients. (D) Final number of protofibrils and (E) final protein density inside fibrin fibers. Inset: fibers radii versus fibrinogen concentration. Error bars are standard deviations calculated from at least three individual experiments. To see this figure in color, go online.

To isolate those differences, we adapted the Missori et al. (23) method to our data; the log-log scattering curves were first fitted with an order 3 polynomial. This polynomial was then subtracted to the scattering data, yielding the curves shown in Fig. 2, B and C. The data were then fitted to two Gaussian distributions (23), one for the main ∼22.5-nm peak and the other Gaussian for the broad peak (dashed lines).

It is obvious both in Fig. 2 B and in Fig. 2 C that FibWoA fibrin exhibits a much higher amplitude and narrower main peak than FibWA fibrin. FibWoA fibrin also shows a well-defined secondary broad peak that is not observed for FibWA fibrin (Fig. 2 C).

FibWoA (Fig. 2 B) sharp peak is at q₁ = 0.284 nm⁻¹ with a full-width at half maximum (FWHM) of 0.0026 nm⁻¹, and the second broader peak is at q₂ = 0.327 nm⁻¹ with an FWHM of 0.094 nm⁻¹. The sharp peak can be assigned to a Bragg diffraction from a repeat distance of d₁ = 2π/q₁ = 22.2 nm that corresponds well to the usual periodicity of the half-staggered arrangement of the protofibrils (see Fig. 1 B). Its FWHM gives a persistence length of 2.4 μm for this longitudinal feature (i.e., along the length of the fiber). This shows that the fibers are straight over this length and without defects in the 22.2-nm half-staggered arrangement. The second, broader peak can be assigned to a repeat distance of d₂ = 2π/q₂ = 19.1 nm, which corresponds accurately to the lateral dimension of the unit cell of the protofibrils crystal described theoretically by Yang et al. (24) and very close to the one observed experimentally by Freyssinet et al. (25). This result indicates that the fibers are also close to crystalline laterally (see Fig. 1 B).

FibWA fibrin (Fig. 2 C) shows a small peak at ∼0.285 nm⁻¹ (i.e., a longitudinal repeat of 22.1 nm with a FWHM of ∼0.01 nm⁻¹), giving a persistence length of ∼0.63 μm, about four times smaller than for FibWoA. This indicates that FibWA fibers are less longitudinally organized and/or shorter and/or curvier fibers than FibWoA’s. The absence of the large secondary peak observed for FibWoA shows also a much less organized internal lateral structure.

All those results indicate a major structural change between FibWA and FibWoA at the nanoscale. Indeed, the molecular organization of fibrin fibers made from FibWoA is close to crystalline both longitudinally and laterally, whereas the fibrin fibers made from FibWA are much less organized both longitudinally and laterally.

To extend the scope of the investigation, we performed spectrophotometry measurements (7) for a wider range of experimental conditions and numerous replicates, experiments that would have been unacceptably time-consuming on a synchrotron. Fig. 2 D shows that fibers from FibWoA clots have about twice as many protofibrils than those from FibWA. This result holds from 0.4 to 3 mg/mL (i.e., up to the normal physiological range). The fitting of the scattering model to the spectrophotometry data also provides the average radius of the fibers from which the protein density within the fibers can be deduced from $\varrho =\mu /\left(\pi r^2\right)$ (7). Fig. 2 E shows that the protein ensity inside FibWoA fibrin fibers is significantly larger (30–80%) than the density of fibers from FibWA. The range of density values obtained here are in agreement with previous work (16,26).

To summarize the nanostructural findings, SAXS indicates that FibWoA fibrin fibers are close to crystalline, whereas the internal structure of fibers produced from FibWA fibrinogens is significantly less organized (e.g., amorphous or fractal as was advocated previously) (7, 27, 28). Spectrophotometry results point in the same direction, showing that FibWoA fibers possess a significantly higher internal density than FibWA fibers.

Fibrin microstructure: confocal microscopy

Finally, to check whether the fibrinogen preparation differences could also impact the microstructure, we investigated the network organization of fibrin clots at the micron scale. The 3D microstructure of the different fibrins obtained from confocal 3D stacks is visualized using the SME (Supporting Materials and Methods, Section 6.1; 17). Major differences in the networks geometry are observed (Fig. 3 A).

First, bright spots are observed in FibWA, whereas there are almost none in FibWoA (red arrows in insets of Fig. 3 A, right). Second, differences in fibers geometry are also observed. FibWoA fibrin shows needle-like fibers that are always perfectly straight over several μm, whereas FibWA produces networks with shorter and some curved fibers (highlighted by the white arrows in FibWA inset). The observed straightness of FibWoA fibers is in perfect agreement with the large persistence length obtained in the SAXS measurements, whereas FibWA fibers relative shortness or curviness is also in agreement with the lower persistence length obtained from the SAXS measurements.

Finally, FibWoA’s fibrin shows usually a rather small number of branches per branching node, typically looking like the contact point between needle-like fibers, whereas FibWA shows star-like patterns, with several asymmetrical branches per node.

Discussion

The first aim of this article is to assess whether or not fibrinogen dispersity influences the ensuing fibrin ultrastructure and how important it is with respect to compositionally induced changes. A possibility to obtain such products would have been to replicate the fractionation process proposed by Huang et al. (11), followed by a complete determination of the fractionated fibrinogen composition. However, this fibrinogen fractionation method has been shown at least once to produce significant amounts of in-column degradation products (13), which contaminated the elution peaks of the undegraded material. Besides, such a method allows obtaining only very low fibrinogen concentrations (below 1 g/L). This forbids any study in the physiological range. In consequence, we chose to test existing commercial fibrinogens preparations for dispersity. Among those we tested, only FibWoA (Clottafact) showed a close-to-monodisperse profile. For the second, polydisperse fibrinogen, we chose FibWA (Fib1) fibrinogen, which is the de facto reference because it has been used in the large majority of works dealing with purified fibrin structure and rheology (see Supporting Materials and Methods, Section 1).

In the first part of this work, we determined the FibWA and FibWoA compositions by measuring their FXIII; fibronectin; α, β, and γ intact chains content; γ/γ′ ratio as well as their N-glycosylation; and post-translational modifications (see Table 1). To evaluate if the variations in composition observed between FibWA and FibWoA may have a significant impact on fibrin structure, we compiled the compositional effects on fibrin structure observed in the literature. This compilation (Table S3.1) shows that the measured FibWoA-FibWA compositional differences are (very) small compared to the compositional range of variation necessary to obtain measurable structural changes in fibrin structure. So, the slight compositional variance between FibWoA and FibWA cannot explain the large structural differences observed between FibWoA and FibWA fibrins. In consequence, those structural differences can only be the consequence of the significant quantity (4% in mass) of large aggregates found in FibWA because FibWoA does not possess such aggregates. Surprisingly, aggregates-induced structural differences are as large as or larger than those induced by compositional changes under constant reaction conditions, as shown in Table S3.1. Therefore, the presence of aggregates is a major determinant of fibrin polymerization and structure, and its effect persists at all concentrations used in this work, from 0.5 to 3 mg/mL.

Those aggregates correspond probably to the bright spots observed in the FibWA images only. Furthermore, because aggregates are known to accelerate the initial growth of the fibers (12), their main role should be to nucleate new fibers independently of the normal fiber formation mechanism. In consequence, the larger the aggregates content is, the more fibers should be nucleated. Because the total fibrinogen mass is conserved, it further derives that the larger the aggregates content is, the smaller the average number of protofibrils per fiber should be in the ensuing fibrin.

Because the FibWoA and FibWA have practically identical compositions, this hypothesis can be checked by mixing increasing quantities of FibWA with FibWoA, an experiment which amounts to increasing the aggregates content. Fig. 3 B shows a large (∼100%) linear decrease in the number of protofibrils per fiber as a function of the aggregate (∼FibWA) content, as expected from the simple argument presented above. Confocal images of the same mixtures (Fig. S10) show also a progressive change from straight fibers without bright spots to shorter, curvier fibers linking bright aggregates, supporting the above explanation. Fig. 3 B further shows that the internal density of the fibers is significantly modified by the presence of aggregates, albeit in a less dramatic fashion than the number of protofibrils.

As deduced from the confocal images above, a plausible hypothesis is that a significant part of the FibWA fibers actually grow from the activated fibrinogen present in the aggregates, hindering the lateral growth through several phenomena. First, there will be steric hindrance because of the local presence of the aggregate and of the several neighboring fibers, as shown in Fig. 4 A. Second, to link between aggregates, the fibers can stick upon each other, or the fiber can bend. In the latter case, such a bending will discourage lateral growth because the 22.5 nm periodicity will be partly lost. Finally, recent results suggest that it is the twisted conformation of both fibrinogen monomers and protofibrils that promotes the assembly of protofibrils into fibers (12), in agreement with observations by Weisel’s group (29). Consequently, aggregates could disrupt the initial fibers natural helical geometry (12), forbidding a proper local lateral aggregation and generating defects or holes in the structure of the fibers. All of these possible mechanisms lead to the same result: a net decrease in the internal density of the fibers. This interpretation also explains the absence of a well-defined lateral structure in FibWA fibers as observed in the SAXS experiments. Conversely, in the absence of aggregates, ideal needle-like fibers can grow and then stick together when they meet. So, once again, the larger the aggregates content is, the larger the number of perturbed, low density fibers will be, giving an average decrease in fiber density, as observed in Fig. 3 B.

Figure 4.

Proposed fibrin formation mechanism, (A) shown in the presence of aggregates (FibWA) and (B) shown in the absence of aggregates (FibWoA). To see this figure in color, go online.

The main limitation of this work lies in the use of the simple model proposed by Yeromonahos et al. to analyze the turbidity data (6). Recently, a complex turbidity model was proposed by Ferri et al. to analyze similar data (26), a model based on the hypothesis that fibrin gels are properly described by fractal networks. This model shows that variations in the fractal dimension and, to a lesser extent, the pore size of filamentous gels can strongly affect the determination of the fibers parameters derived from turbidity data (26). If large differences in fractal dimension were determined between the fibrin gels produced in our work, it would undermine significantly the proposed mechanism of influence of the aggregates on the fibrin gels ultrastructure.

However, we doubt that those fibrin gels can be properly represented by fractal networks because no scale-recursivity is observed on the confocal images. Furthermore, the use of Ferri’s model requires measuring reliably the 3D fractal dimension from confocal microscopy data. Unfortunately, very large differences in methods and results between authors can be found in the literature (see Supporting Materials and Methods, Section 4.2). For those reasons, we chose to use the much simpler Yeromonahos model, the consistency of which is favorably checked against pore size measurements while supposing that the fractal dimension of the network is 1 (see Supporting Materials and Methods, Section 7.3 and Fig. S12 for details). This important issue of the fractality of filamentous gels and the potential need for a fractal model to analyze turbidity data will be the subject of a forthcoming article.

Conclusion

We have first shown that Clottafact’s fibrinogen is composed almost exclusively of fibrinogen monomers, whereas Fib1 possesses a significant amount of large aggregates. Conversely, the detailed physicochemical analyses of Clottafact and Fib1 show that their compositions are sufficiently similar not to influence the structure of the ensuing fibrins. SAXS demonstrates that the internal structure of Clottafact’s (FibWoA) fibrin fibers is close to crystalline, both laterally and longitudinally with a longitudinal persistence length of several microns. Conversely, SAXS also shows that Fib1 (FibWA) fibers are much more disorganized with no perceptible lateral organization and a longitudinal persistence length of a few hundreds of nanometers. Those nanostructural observations are confirmed at the microscale by direct confocal imaging. Clottafact fibrin displays long, needle-like fibers, whereas Fib1 fibrin shows short and curved fibers in agreement with the SAXS persistence length measurements. Those results are further strengthened by spectrophotometry measurements, which confirm the significantly larger density of the Clottafact fibers with respect to Fib1 ones. Those structural differences hold at all studied concentrations including physiological ones and demonstrate that the size dispersity of fibrinogen in purified systems is one of the main parameters determining fibrin fibers ultrastructure, a parameter that has been mostly overlooked. Indeed, the effect of modifying the dispersity is surprisingly large, larger than changing any other aspect of the fibrinogen preparation.

Those results open outstanding questions regarding the precise mechanisms by which the presence of aggregates can influence to such an extent the polymerization of fibrin and, therefore, concerning fibrin polymerization itself. Furthermore, given the extensive structural modifications accompanying the presence of aggregates, the mechanical properties of fibrin clots formed from monodisperse fibrinogens may be significantly different from the polydisperse ones and could be of paramount interest in tissue engineering on fibrin scaffolds as well as for the design of fibrin glues.

Finally, the actual amount of aggregates in circulating fibrinogen is not known because only a few studies investigated the multiscale structure of fibrin formed in plasma. Such investigations could be of outstanding interest in hemostasis, particularly in relation to thrombosis-related diseases.

Author Contributions

X.G., L.S., and F.C. conducted and analyzed the experiments in small angle x-ray scattering, multiwavelength spectrophotometry, dynamic light scattering, confocal microscopy, and gel-filtration chromatography. G.C. and N.B. conducted and analyzed the experiments in reverse phase chromatography, MS, and enzyme-linked immunosorbent assay dosages. F.C., B.P., and Z.T. conceived and directed the research. X.G., B.P., and F.C. wrote the manuscript. All authors contributed to the final version of the manuscript.

Editor: Jill Trewhella.