Heparin Binds 8 kDa Gelsolin Cross-β-Sheet Oligomers and Accelerates Amyloidogenesis by Hastening Fibril Extension

Abstract

Glycosaminoglycans (GAGs) are highly sulfated linear polysaccharides prevalent in the extracellular matrix, and they associate with virtually all amyloid deposits in vivo. GAGs accelerate the aggregation of many amyloidogenic peptides in vitro, but little mechanistic evidence is available to explain why. Herein, spectroscopic methods demonstrate that GAGs do not affect the secondary structure of the monomeric 8 kDa amyloidogenic fragment of human plasma gelsolin. Moreover, monomerized 8 kDa gelsolin does not bind to heparin under physiological conditions. In contrast, 8 kDa gelsolin cross-β-sheet oligomers and amyloid fibrils bind strongly to heparin, apparently due to electrostatic interactions between the negatively charged polysaccharide and a positively charged region of the 8 kDa gelsolin assemblies. Our observations are consistent with a scaffolding mechanism, whereby cross-β-sheet oligomers, upon formation, bind to GAGs, accelerating the fibril extension phase of amyloidogenesis, possibly by concentrating and orienting the oligomers to more efficiently form amyloid fibrils. Notably, heparin decreases the 8 kDa gelsolin concentration necessary for amyloid fibril formation, likely a consequence of fibril stabilization through heparin binding. Since GAG overexpression, which is common in amyloidosis, may represent a strategy to minimize cross-β-sheet oligomer toxicity by transforming them into amyloid fibrils, the mechanism described herein for GAG-mediated acceleration of 8 kDa gelsolin amyloidogenesis provides a starting point for therapeutic strategy development. The addition of GAG mimetics, small molecule sulfonates shown to reduce the amyloid load in animal models of amyloidosis, to a heparin-accelerated 8 kDa gelsolin aggregation reaction neither significantly alters the rate of amyloidogenesis nor prevents oligomers from binding to GAGs, calling into question their commonly accepted mechanism.

Glycosaminoglycans (GAGs) are linear sulfated polysaccharides that are prevalent in the extracellular matrix and are made up of repeating disaccharide units (1). GAGs are associated with virtually all extracellular amyloid deposits, formed from one of 30 different human amyloidogenic proteins linked to distinct amyloid diseases (2), including immunoglobulin light chains, transthyretin, amyloid β (Aβ), and the serum amyloid A protein (3). Thus, GAGs appear to play a vital role in amyloid deposition in mammals, by a mechanism that remains unclear (4, 5). The hypothesis that the tissue-selective distribution of different GAGs creates distinct extracellular environments with differential abilities to accelerate amyloidogenesis potentially explains why different amyloidogenic proteins lead to the degeneration of specific tissues and why different mutants of the same protein can cause proteotoxicity in distinct tissues.

It has been proposed that cells upregulate the formation of sulfated GAGs in order to reduce organismal proteotoxicity (6). In support of such a mechanism, it has been previously demonstrated that exogenous addition of various GAGs protects against Aβ₄₂ proteotoxicity (7), Dutch-mutant Aβ₄₀ proteotoxicity (8), and possibly degeneration by the human prion protein (9), although this latter effect is not always seen (10). One explanation for the protective effect of GAGs is that they stabilize amyloid fibrils and shield the surface of amyloid from the remainder of the proteome, preventing aberrant protein recruitment, thus rendering the amyloid less proteotoxic. Herein we show that GAGs can transform proteotoxic oligomers into less toxic amyloid fibrils, representing another possible mechanism of protection.

Biophysical experiments conclusively demonstrate that the addition of GAGs to human amyloidogenic proteins accelerates their amyloidogenesis in vitro. These proteins include Aβ (11) β₂-microglobulin (12, 13), transthyretin(14), and the amyloidogenic fragments of gelsolin (15). Although detailed and convincing mechanistic studies to explain the acceleration of amyloidogenesis are scarce, numerous mechanisms have been proposed to explain how GAGs accelerate amyloidogenesis. Herein, we endeavor to understand the mechanism of GAG-accelerated amyloidogenesis of the 8 kDa fragment of human plasma gelsolin, the clinically most important fragment in gelsolin amyloid disease (16).

Gelsolin has six domains, each of which can bind a Ca²⁺ ion (17). There are two splice variants of gelsolin – an 81 kDa intracellular variant that is responsible for remodeling the actin cytoskeleton by facilitating actin polymerization or fragmentation, while an 83 kDa version secreted into the plasma is responsible for scavenging actin fibrils released from injured tissue into the blood, thereby preventing increases in blood viscosity (18). Plasma gelsolin also likely has other functions (19, 20). Several types of cells secrete gelsolin into the blood, although muscles seem to be the major source of secreted gelsolin (21).

A G654 to A or T DNA point mutation changes an aspartate at position 187 to either an asparagine (D187N) or tyrosine (D187Y) upon translation, respectively. These mutations eliminate the 187 carboxylate side chain calcium binding ligand, dramatically lowering the Ca²⁺ binding affinity of the second domain of plasma gelsolin, compromising its stability (22–24). Without the stabilization afforded by Ca²⁺-binding, domain 2 can sample the unfolded state as it passes through the Golgi on its way to the cell surface – the unfolded population being susceptible to cleavage by furin in the trans Golgi, producing a secreted C-terminal 68 kDa gelsolin fragment (C68) (21–26). The secreted C68 fragment is then cleaved again, apparently in the extracellular matrix, to generate 8 and 5 kDa amyloidogenic fragments, with the former being the main component of gelsolin amyloid fibrils in humans (16) and the amyloidogenic fragment studied herein. While it is clear that membrane type 1 matrix metalloprotease (MT1-MMP) can perform this cleavage step in cultured cells (27), other proteases in the extracellular matrix may also contribute. The formation of the 8 and 5 kDa amyloidogenic peptides appear to cause Familial amyloidosis of Finnish type (FAF) through systemic amyloid deposition in the basement membrane of the skin (28), blood vessel walls (29), eyes (30), and peripheral nervous system (31).

We have recently created a murine model of FAF featuring a muscle-specific promoter to drive D187N gelsolin synthesis and secretion (32). This model recapitulates the aberrant endoproteolytic cascade and the aging-associated extracellular amyloid deposition of human FAF. Amyloidogenesis is observed only in tissues synthesizing human D187N gelsolin, despite the presence of the 68 kDa cleavage product in blood. As is the situation in humans, gelsolin fragment amyloidogenesis and accumulation is closely associated with GAGs in the extracellular matrix (32). In fact, GAGs are substantially upregulated in this mouse model as the amyloid disease progresses, suggesting that GAG overexpression may be linked to pathology or protection from the process of amyloidogenesis.

The gelsolin amyloidogenic peptides have been recently established to aggregate by a nucleated polymerization mechanism (33), and their aggregation rate is known to be profoundly accelerated by addition of heparin (15). Herein, we demonstrate that heparin does not alter the intrinsically disordered structure of the monomeric 8 kDa amyloidogenic fragment of human plasma gelsolin. We also show that the binding of monomerized 8 kDa gelsolin to heparin cannot be detected at neutral pH, even at the 50 μM concentration utilized in the NMR experiments. Complementary experiments also fail to detect a binding interaction. We provide strong evidence to support the hypothesis that cross-β-sheet oligomers and amyloid fibrils bind to heparin, apparently due to an electrostatic interaction between the positively charged regions of the 8 kDa gelsolin aggregates and the negatively charged, sulfated polysaccharide. Our observations suggest that heparin binds to 8 kDa gelsolin oligomers, hastening the fibril extension phase of the nucleated polymerization reaction, possibly by binding to post-nucleation cross-β-sheet oligomers, facilitating quaternary structural conversion into larger aggregates. Notably, the presence of heparin is able to decrease the concentration of 8 kDa gelsolin necessary for the accumulation of thioflavin T (TfT)-positive aggregates, likely by fibril binding and stabilization. We demonstrate that GAG mimetics, including those used in amyloid disease clinical trials, do not slow the rate of GAG-mediated 8 kDa gelsolin amyloidogenesis in vitro, nor do they seem to affect the solubility of the gelsolin amyloid fibrils.

Materials and Methods

Preparation of the 8 kDa Amyloidogenic Fragment of Gelsolin

The 8 kDa amyloidogenic fragment of gelsolin (corresponding to amino acids 173-242) was prepared as previously described (33, 34).

Preparation of Monomerized 8 kDa Amyloidogenic Fragment of Gelsolin

The 8 kDa amyloidogenic fragment was monomerized by dissolving 1–4 mg of lyophilized 8 kDa gelsolin (amount dependent on desired final concentration) in 500 μL of 8 M guanidinium hydrochloride (GdnHCl), 50 mM sodium phosphate (NaPi) (pH 7.5). The solution was sonicated in a water bath sonicator for at least 2 h. It was then buffer exchanged using a 20 mL Superdex 30 gel filtration column that had been equilibrated with 50 mM NaPi, 100 mM NaCl, 0.02% NaN₃, pH 7.2 (unless otherwise specified). The monomerized peptide peak, free of chaotrope, eluted after about 9 mL and was collected and used immediately in subsequent experiments.

Circular Dichroism Experiments

Monomeric 8 kDa amyloidogenic fragment prepared in 10 mM KPi, 50 mM K₂SO₄, (pH 7.0), as this buffer allowed for better measurements at wavelengths < 205 nm, was diluted to a concentration of 20 μM, either in the absence or presence of 30 μg/mL heparin (purchased from Calbiochem, heparin sodium salt from porcine intestinal mucosa). For the quiescent experiments, the gelsolin was allowed to sit at room temperature, while for the agitated experiments, it was mixed using overhead rotation (24 rpm) at 37 °C. Periodically, 200 μl was removed, and a far-UV CD spectrum was taken using an AVIV model 202SF CD spectrometer. Quartz cells of 0.1 or 0.2 cm path lengths were used. The wavelength step size used was 1 nm with an averaging time of 2 s for each wavelength. The buffer and buffer plus heparin controls were subtracted from the 8 kDa gelsolin and 8 kDa gelsolin plus heparin data, respectively; their contribution to MRE was < 2% that of gelsolin (Supplementary Figure 1).

Fluorescence Anisotropy Studies

Heparin was fluorescently labeled according to a procedure of Glabe et al. (35). Briefly, the reducing end of heparin was activated by the addition of 15 mg of cyanogen bromide (CNBr) to a solution of 30 mg of heparin in 1 mL of water (pH 11). Once completely dissolved, it was allowed to incubate at room temperature for 10 min. The reaction mixture was then desalted using a G-25 column that had been preequilibrated in 0.2 M sodium borate (pH 8.0) in order to buffer exchange and remove excess CNBr. Fluoresceinamine (3 mg) was immediately added to the eluate and the reaction mixture was incubated overnight at room temperature. To remove unreacted fluoresceinamine, the reaction mixture was again run over a G-25 desalting column that had been preequilibrated in 50 mM NaPi, 100 mM NaCl, 0.02% NaN₃ (pH 6.8). The final concentration of the fluoresceinated heparin was 3 mg/mL.

For anisotropy studies, the monomerized 8 kDa amyloidogenic fragment of gelsolin was diluted to a concentration of 20 μM, and agitated by overhead rotation (24 rpm) at 37 °C. Periodically, 90 μL of the sample was removed and added to 10 μL of a 1:10 dilution of the fluoresceinated heparin solution and the fluorescence anisotropy examined in an AVIV fluorimeter (excitation at 485 nm, emission at 515 nm). Each anisotropy reading was taken at least five times. A Thioflavin T (TfT) fluorescence reading was taken at the same time as the anisotropy assessment to determine aggregation state. For the fluorescence readings, 20 μL of the 8 kDa gelsolin solution was added to 80 μL of 25 μM TfT (final concentration equals 20 μM) in a quartz cuvette and evaluated in a Cary Eclipse fluorescence spectrophotometer (excitation at 440 nm, emission at 485 nm).

Heparin Affinity Chromatography

Monomerized 8 kDa amyloidogenic fragment of gelsolin was diluted to a concentration of 20 μM, and agitated by overhead rotation at 37 °C. Periodically, 300 μL of the sample was removed and injected at a rate of 1 mL/min onto a 1.5 mL column of Heparin HyperD Affinity resin (Pall Corporation) that had been preequilibrated with 50 mM NaPi, 100 mM NaCl, 0.02% NaN₃, (pH 7.2). After rinsing the column with at least 5 column volumes of buffer, a gradient up to 4 M NaCl over 5 column volumes was used for elution. After the gradient, the column was rinsed with two column volumes of 7.2 M GdnHCl in 50 mM NaPi, pH 7.5. Once again, a TfT fluorescence reading was taken at the same time as the affinity chromatography assessment to determine aggregation state.

1D Proton NMR Experiments

For all experiments, heparin was diluted such that the disaccharide subunit concentration was 50 μM in a solution of 90% of 50 mM NaPi, 100 mM NaCl, 0.02% NaN₃, (pH 7.4)/10% of D₂O. For the heparin plus monomerized 8 kDa gelsolin sample, 8 kDa gelsolin was monomerized as described above and added to a final concentration of 50 μM. For the heparin plus 8 kDa gelsolin fibril sample, 50 μM freshly monomerized 8 kDa gelsolin was added to the heparin solution and rotated at 24 rpm at 37 °C for 3 h to make amyloid fibrils, the presence of which was confirmed by TfT fluorescence. A 500 MHz NMR spectrum (8 k scans with water suppression) was taken using an Avance 501 NMR.

Gel Mobility Shift Assay

Monomerized 8 kDa gelsolin was diluted to a concentration of 24 μM in 50 mM NaPi, 100 mM NaCl, 0.02% NaN₃, (pH 7.2) and rotated at 37 °C for 2 h (24 rpm) to form amyloid fibrils, the presence of which were verified by TfT fluorescence. A freshly monomerized 8 kDa gelsolin solution was allowed to incubate on ice for 2 h and did not exhibit TfT fluorescence. Fluoresceinated heparin (60 μg/mL) was then added to each of the solutions, (affording a 1:6 heparin to 8 kDa gelsolin fragment molar ratio), and the mixtures were incubated for 30 min at 25 °C. As a positive control, fluoresceinated heparin was mixed with 300 μg/mL protamine sulfate, a known binder of heparin. 2 μL of glycerol was then added to 10 μL of each of the solutions, and these were loaded into the wells of a 1.2 % (w/v) agarose gel in 90 mM Tris, 90 mM borate, 2 mM EDTA (pH 8.3). Electrophoresis was carried out at 100 V for 1 h in the same buffer. The fluoresceinated heparin was visualized by UV fluorescence.

Plate Reader Aggregation Assay

The 8 kDa amyloidogenic fragments were monomerized as described above and diluted to the desired concentrations in 50 mM NaPi, 100 mM NaCl, 0.02% NaN₃ (pH 7.2). For experiments involving GAGs or GAG mimetics, the oligosaccharides were dissolved in the same buffer and added to the solution to achieve the reported concentration. For examining TfT fluorescence, the solution also contained a final concentration of 20 μM TfT.

All components were mixed together in an Eppendorf tube and 97 μL of each solution was transferred to a well of a 96-well microplate in triplicate (black plate with clear bottom, Corning Inc, Corning, NY). The plate was sealed with an airtight lid to minimize evaporation and then placed into a Gemini SpectraMax EM fluorescence plate reader (Molecular Devices, Sunnyvale, CA). For the duration of the experiment, the plate was incubated at 37 °C. Every 10 min, the plate was shaken for 5 s and the fluorescence was measured from the bottom of the plate (ex. 440 nm, em. 485 nm). Each reading was the mean of thirty individual scans.

Using Wolfram Mathematica 8, the dataset from each individual well was fit to the Finke-Watzky model of nucleation according to the equation:

$$F_t=F_f+(F_i-F_f)\ast \frac{\frac{k_1}{k_2}+{[A]}_0}{1+\frac{k_1}{k_2{[A]}_0}exp(k_1+k_2{[A]}_0)t}$$ (1)

where F_t is fluorescence at time t, F_f is the fluorescence of the completely aggregated sample (the end of the assay), F_i is the fluorescence of the freshly monomerized sample (the beginning of the assay), and [A]₀ represents the concentration of 8 kDa gelsolin used in the experiment. By fitting the data using this equation, the t₅₀, or time required to reach 50% of fibril formation, could be determined. Notably, the constant k₁ is proportional to the inverse of the lag phase time, and the constant k₂ is proportional to the slope of the curve during the fibril extension phase, affording quantitative information about these phases of the 8 kDa gelsolin aggregation reaction (36).

Atomic Force Microscopy (AFM)

The 8 kDa amyloidogenic fragment of gelsolin was pretreated to afford monomer as described above, diluted to a concentration of 24 μM, either in the absence or in the presence of 10 μg/mL heparin, and agitated by overhead rotation (24 rpm) at 37 °C to form amyloid fibrils. Periodically, aliquots of the reaction were removed and examined by TfT fluorescence as described above and by AFM to discern aggregate morphology. For the AFM sample, 2 μL of the 8 kDa amyloidogenesis reaction was diluted into 18 μL of buffer, and the resulting solution was adsorbed to a surface of freshly cleaved mica for 1 min. The liquid was then wicked off the surface of the mica with filter paper. Salt and unbound material was removed by washing three times with 20 μL of distilled water that was immediately wicked off the surface of the mica with filter paper. AFM Images were recorded using a Digital Instruments multimode scanning probe microscope with a Nanoscope IIIa controller and force modulation etched silicon probes.

Analytical Size Exclusion Chromatography Coupled to Light Scattering

The 8 kDa amyloidogenic fragment of gelsolin was monomerized as described above, diluted to a concentration of 24 μM, and agitated by overhead rotation (24 rpm) at 37 °C. Experiments were either performed using a 20 mL Superdex 75 analytical gel filtration column with in-line 0.1 μm filters or a 2 mL Superdex 75 analytical gel filtration column without any filters. For the larger, 20 mL column, periodically, 400 μL of the 8 kDa amyloidogenesis reaction was removed and loaded into a sample loop for loading onto the column. For the smaller, 2 mL column, periodically, 75 μL of the sample was removed, 75 μL buffer was added for dilution, and the solution was spun at 15,000 g for two minutes to pellet any large aggregates. The supernatant was then loaded into a 25 μL sample loop and loaded onto the column. In each case, the mobile phase of the Superdex 75 column was 50 mM NaPi, 100 mM NaCl, 0.02% NaN₃ (pH 7.2). The column eluate was then directly injected into a Dawn EOS light scattering photometer (Wyatt Technology, Santa Barbara, CA) that was connected to the FPLC for measurements of 90° static light scattering as well as dynamic light scattering. In both cases, a TfT fluorescence reading was taken at the same time as the analytical size exclusion chromatography to determine aggregation state, as described above.

Fitting to Nucleated Polymerization with Competing Off-Pathway Oligomerization Model

Data from a plate reader aggregation assay in triplicate wells of a 96-well plate was arithmetically averaged and was plotted as TfT fluorescence intensity vs. time. The baseline was subtracted, and the data up to either 25% or 50% of the TfT fluorescence at completion of the amyloidogenesis reaction was fit to a second order polynomial using Microsoft Excel. The residuals were calculated by subtracting the best fit second order polynomial from the experimental data.

Results

Addition of heparin does not change the intrinsically disordered structure of 8 kDa gelsolin

Far-UV circular dichroism (CD) spectroscopy was employed to determine whether the presence of heparin affects the intrinsically disordered structural ensemble of the monomerized 8 kDa amyloidogenic fragment of gelsolin. When freshly monomerized 8 kDa gelsolin (20 μM) was incubated quiescently without heparin for 3 h, no discernible change in the far-UV CD spectrum was observed (Figure 1A). A strictly analogous quiescent sample with heparin added at a concentration of 30 μg/mL (molecular weight of 13.5–15 kDa) exhibited an unaltered CD spectrum as well (Figure 1B), indicating no discernible change in the intrinsically disordered structural ensemble of 8 kDa gelsolin. When 8 kDa gelsolin (20 μM) in the absence of heparin was agitated by overhead rotation (24 rpm), a procedure we previously employed to produce amyloid fibrils from this intrinsically disordered protein (33), aggregation was observed, as reflected by the conversion of a random coil to a cross-β-sheet quaternary structure over the course of 5 h (Figure 1C). In a strictly analogous sample with heparin added (30 μg/mL, Figure 1D), the conversion to a cross-β-sheet quaternary structure occurred on a slightly accelerated time scale (cf. 0.66 h CD spectrum of Figure 1C and D).

Figure 1.

Effect of heparin on the secondary/quaternary structure of the 8 kDa gelsolin fragment. The 8 kDa gelsolin fragment (20 μM) was incubated quiescently at 37 °C either in the absence (A) or presence of 30 μg/mL heparin (B) and was periodically examined by far-UV CD (cell path length 0.1 cm). No changes were observed over the course of 3 h. C. When 8 kDa gelsolin (20 μM) was agitated by overhead rotation (24 rpm) at 37 °C, the secondary structure converted from random coil to cross-β-sheet quaternary structures (cell path length 0.2 cm). D. The addition of heparin (30 μg/mL) decreased the time required for the formation of cross-β-sheet quaternary structures (cell path length 0.2 cm). E. The increase in cross-β-sheet quaternary structures, as measured by mean residue ellipticity (MRE) at 216 nm (open circles, with error bars representing standard deviation in the measurement), occurs at the same time as the increase in TfT fluorescence (filled squares, with error bars representing standard deviation in the measurement smaller than the size of the points), both in the absence (black datapoints) and presence (red datapoints) of heparin, suggesting that the 8 kDa fragment of gelsolin does not alter its random coil secondary structure until it begins to aggregate and bind TfT.

To probe whether this slightly hastened conversion to a cross-β-sheet quaternary structure was due to an effect of heparin on the intrinsically disordered ensemble of conformers of monomeric 8 kDa gelsolin or to an accelerated post-nucleation step, the kinetics of the increase in β-sheet character, as measured by mean residue ellipticity at 216 nm (θ₂₁₆), was compared to the kinetics of amyloid fibril formation, as measured by Thioflavin T (TfT) fluorescence (Figure 1E; excitation at 440 nm, emission at 485 nm). Both in the absence and in the presence of heparin, the increase in β-sheet quaternary structure detected by far-UV CD strictly coincides with the formation of amyloid monitored by TfT fluorescence. That the heparin-accelerated time course of amyloid fibril formation and the heparin-accelerated time course of β-sheet quaternary structure formation are very similar, if not identical (red symbols), provides further evidence that it is unlikely that heparin affects the conformational ensemble of monomerized 8 kDa gelsolin or any pre-nucleation step.

8 kDa Gelsolin Amyloid Fibrils bind to heparin, whereas monomeric 8 kDa Gelsolin does not

To further scrutinize the hypothesis that monomeric 8 kDa gelsolin does not bind to heparin, 1D 500 MHz ¹H NMR experiments were performed. The ¹H NMR spectrum of polydisperse heparin (13,500 Da – 15,000 Da, at a disaccharide repeat subunit concentration of 50 μM) was recorded, revealing two diagnostic singlet peaks at 5.31 and 5.48 ppm that are in a region of the spectrum where there are no resonances from 8 kDa gelsolin (Figure 2A, Supplementary Figure 2A). Upon addition of monomerized 8 kDa gelsolin (50 μM), there was no change in the chemical shifts or intensity of the heparin peaks at 5.31 and 5.48 ppm, which would be expected to change if monomeric 8 kDa gelsolin was binding to heparin (Figure 2B, Supplementary Figure 2B). In striking contrast, when heparin was mixed with 8 kDa gelsolin and agitated by overhead rotation (24 rpm, 37 °C) to form amyloid fibrils, the peaks at 5.31 and 5.48 ppm were not observed (Figure 2C, Supplementary Figure 2C). The formation of 8 kDa gelsolin cross-β-sheet amyloid fibrils in this rotated sample was confirmed by TfT fluorescence. Thus, it appears that an interaction between soluble heparin and 8 kDa gelsolin cross-β-sheet structures dramatically increased the T₂ spin-spin relaxation time of heparin, leading to substantial heparin peak broadening.

Figure 2.

1D ¹H 500 MHz NMR spectra with water suppression. A. The proton NMR spectrum of heparin (50 μM disaccharide subunit concentration) has two characteristic resonances at 5.3 and 5.5 ppm which do not overlap with any ¹H NMR resonances from the 8 kDa fragment of gelsolin (50 μM) in solution. B. These resonances do not change intensity or chemical shift when an equimolar concentration of monomerized 8 kDa gelsolin is added. C. When a solution of 50 μM heparin and 50 μM monomerized 8 kDa gelsolin is agitated by overhead rotation at 37 °C for 3 h, so that amyloid fibrils are formed, the characteristic heparin resonances are no longer observed.

To further probe the interaction between heparin and the 8 kDa fragment of gelsolin, we performed fluorescence anisotropy experiments with fluoresceinated heparin. Fluoresceinated heparin was able to accelerate the aggregation of the 8 kDa amyloidogenic fragment of gelsolin to the same extent as unlabeled heparin in a TfT fluorescence plate reader-based aggregation assay (Supplementary Figure 3). Fluoresceinated heparin exhibited a significantly higher anisotropy than fluorescein by itself (Figure 3A).

Figure 3.

Heparin was labeled with fluorescein to probe its binding to various forms of 8 kDa gelsolin utilizing anisotropy. A. The fluorescence anisotropy after 5 h of 8 kDa gelsolin aggregation, once amyloid fibrils had been formed (bar on far right), was significantly greater than the anisotropy at the beginning of the experiment when all of the 8 kDa gelsolin was monomeric (second bar from right). Fluoresceinated heparin (second bar from left) and fluorescein alone (far left bar) are included as controls. B. The increase in anisotropy paralleled the formation of 8 kDa gelsolin (20 μM) amyloid fibrils discerned by TfT fluorescence binding-associated fluorescence, demonstrating binding of the amyloid fibrils to fluoresceinated heparin.

The anisotropy of fluoresceinated heparin in the presence of monomerized 8 kDa gelsolin was not significantly different from the anisotropy of fluoresceinated heparin in buffer (Figure 3A, cf. height of second and third bars), although only a small increase in anisotropy was anticipated if the fluoresceinated heparin (13.5 – 15 kDa) were to bind the 8 kDa gelsolin monomers, as the molecular weight of the complex would be 23 kDa. When monomerized 8 kDa gelsolin (20 μM) was agitated by overhead rotation (24 rpm) at 37 °C to enable amyloidogenesis, the anisotropy was significantly higher (Figure 3A, far right bar). These results were confirmed using a 96-well fluorescence polarization plate reader assay (Supplementary Figure 4). The increase in fluoresceinated heparin anisotropy paralleled the formation of amyloid fibrils, as reflected by TfT fluorescence (Figure 3B). Because the heparin anisotropy increased as the cross-β-sheet oligomers and fibrils were being formed, it can be concluded that the post-nucleation cross-β-sheet oligomers and fibrils bind to fluoresceinated heparin.

To confirm the binding of 8 kDa gelsolin amyloid fibrils to heparin, aliquots (400 μL) of an 8 kDa gelsolin (24 μM) amyloidogenesis reaction were removed at various time points and loaded onto a 2 mL heparin affinity column. Heparin binding 8 kDa gelsolin structures were eluted with a gradient up to 4 M NaCl before the column was cleaned with a denaturing solution (7.2 M GdnHCl, 50 mM NaPi (pH 7.5)) (Supplementary Figure 5). Freshly monomerized 8 kDa peptide did not bind to the resin, as all of it flowed through before the gradient was started (Supplementary Figure 5B). After 2.5 h of rotation, when TfT-positive fibrils and/or cross-β-sheet oligomers were present in the sample, the flow-through peak was dramatically decreased in intensity (Supplementary Figure 5C). Integration reveals that the flow-through peak of the 8 kDa gelsolin fibrillar sample was 22% of the area of the flow-through peak of the monomerized 8 kDa gelsolin sample, indicating substantial binding. No additional peaks were observed when a gradient up to 4 M NaCl was applied, suggesting very strong binding between the 8 kDa gelsolin oligomers and/or amyloid fibrils and the heparin (Supplementary Figure 5C). However, when the column was cleaned with 7.2 M guanidinium chloride, an additional peak was observed corresponding to the 8 kDa gelsolin fragment comprising cross-β-sheet quaternary structures. This observation is consistent with the expectation that 8 kDa gelsolin cross-β-sheet structures and fibrils bind very tightly to heparin and the only means to elute 8 kDa gelsolin aggregates is to utilize chaotrope solutions that partially or fully denature the amyloid fibrils.

Because it is conceivable that the cross-β-sheet oligomers and amyloid bind nonspecifically to the sepharose core of the heparin affinity resin, the same experiment was performed using a Superdex 75 sepharose-based size exclusion column as a control. With this resin, both freshly monomerized 8 kDa and the 8 kDa gelsolin oligomers and/or fibrils after 2.5 h of mixing were able to flow through the column (Supplementary Figures 5E and 5F respectively). Integration showed that the Superdex 75 flow-through peaks of the monomerized and fibrillar 8 kDa gelsolin samples were 95% and 93%, respectively, of the area of the monomerized 8 kDa gelsolin heparin-affinity column flow-through peak, demonstrating a lack of binding. No additional peaks were observed when 4 M NaCl or 7.2 M guanidine were applied. The volume required for elution of the 8 kDa gelsolin monomer was not significantly different from that of the 8 kDa gelsolin fibrils because the volume of the size exclusion column (2 mL) was small compared to the volume of the samples examined (400 μL).

Finally, to further demonstrate that 8 kDa gelsolin amyloid fibrils can interact with heparin, while 8 kDa gelsolin monomer cannot, a gel mobility shift assay was performed. In this assay, free, unbound fluoresceinated heparin migrates rapidly towards the anode unless complexed with gelsolin structures, which alter the migration by both increasing the size and altering the overall charge of the complex, enabling separation and subsequent visualization by fluorescence (37). The addition of monomerized 8 kDa gelsolin (24 μM) to fluoresceinated heparin (50 μg/mL) did not appreciably change the mobility of fluoresceinated heparin (Supplementary Figure 6, cf. lanes 1 and 2). The addition of 8 kDa gelsolin amyloid fibrils, however, exhibited a substantial effect on the mobility of a fraction of fluoresceinated heparin. The fluoresceinated heparin remaining at the top of the gel was retained by 8 kDa gelsolin amyloid fibril binding (Supplementary Figure 6, lane 3). As a positive control, protamine sulfate, a positively-charged and known binder of heparin was also examined, which prevented the migration of fluoresceinated heparin (Supplementary Figure 6, lane 4).

The Importance of Electrostatic Interactions for Quaternary Structural Conversion

To test the hypothesis that the 8 kDa gelsolin oligomer-amyloid fibril heparin interaction is predominantly an electrostatic interaction, the salt concentration of the buffer was varied in a plate reader aggregation assay. When the osmolarity of the aggregation reaction buffer is increased from the standard physiological conditions of 300 mOsm to 670 mOsm by addition of NaCl, the rate at which 8 kDa gelsolin aggregates in the absence of heparin is slightly increased (Figure 4A; cf. blue vs. black traces), while the rate at which 8 kDa gelsolin aggregates in the presence of heparin is clearly reduced (Figure 4A; cf. green (high salt) vs. red (low salt) traces). Further increasing the osmolarity of the aggregation reaction buffer to 1410 mOsm almost eliminates heparin acceleration of the 8 kDa gelsolin amyloidogenesis, in that the kinetics of gelsolin aggregation is nearly the same in the absence and presence of heparin (Figure 4B; cf. green vs. blue traces).

Figure 4.

Effect of salt concentration on heparin-dependent acceleration of 8 kDa gelsolin amyloidogenesis. Various volumes of a 2M NaCl solution were added to the amyloidogenesis reactions to obtain an osmolarity of 670 mOsm (A) or 1.41 Osm (B). The data from each individual well was fit to the Finke-Watzky model (equation 1) and the constants k₁ and k₂, as well as the t₅₀ shown in Table 1 depict the arithmetic average of data from triplicate wells. Collectively, these data demonstrate that increasing the ionic strength dramatically decreases heparin acceleration of 8 kDa gelsolin amyloidogenesis.

Each trace in Figure 4 is the arithmetic average of data from triplicate wells of a representative experiment. We recognize that these types of TfT fluorescence aggregation assays performed in a plate reader can have large variation and thus exhibit reproducibility challenges. However, our rigorous pretreatment method, combined with the inherent biophysical properties of the 8 kDa fragment of gelsolin, render these experiments reproducible. To demonstrate the low variability in the time courses, data from each individual well from the experiment presented in Figure 4 is shown in Supplementary Figure 7. In addition, to confirm that the observations made in this experiment are reproducible, the assay was performed a total of three times on three separate occasions using two different preparations of the 8 kDa fragment of gelsolin. Separate plate reader aggregation assays are difficult to compare because the 8 kDa gelsolin fragment was freshly monomerized in each assay, leading to slight alterations in osmolarity and pH that the amyloidogenicity time courses of 8 kDa gelsolin are very sensitive to (33). However the same trends in the data – increasing osmolarity increases the aggregation rate of 8 kDa gelsolin in the absence of heparin while decreasing the aggregation rate in the presence of heparin – were robust and always observed.

The data from each individual well from the representative experiment was also fit to the Finke-Watzky model of nucleation according to equation 1, and the values of k₁ and k₂, along with the calculated t₅₀ are shown in Table 1 (The individual datasets and their best fit curves are shown in Supplementary Figure 8). Although the Finke-Watzky model probably oversimplifies the mechanism of amyloidogenesis and is primarily a phenomenological model, some quantitative information can be gleaned from fitting the data to this model via equation 1 (36). The k₁ values, theoretically proportional to the inverse of the lag phase of the reaction, are exquisitely sensitive to the data fitting early in the reaction, and these early time points did not always fit well to equation 1 (See Supplementary Figure 8). Therefore, we cannot make any significant observations about the correlation of k₁ values to lag time. However, we could make some interesting observations about constant k₂, which is proportional to the slope of the curve during the fibril extension phase (36). Under physiological salt conditions (300 mOsm), the presence of heparin results in k₂ values that are significantly higher. As the concentration of salt is increased to 670 mOsm and then 1410 mOsm, the difference between the k₂ values of the samples containing heparin and the samples without heparin decreases. The expected trend was also observed for the calculated t₅₀ values. A large and significant difference was observed between the t₅₀ of the sample containing heparin and the sample without heparin at the physiological salt concentration. This difference was decreased with the addition of NaCl to 670 mOsm, and there was no significant t₅₀ difference between the heparin-containing sample and the sample without heparin at 1410 mOsm.

Table 1.

Constants k₁ and k₂ and t₅₀ values from fitting data from the plate reader aggregation assays to the Finke-Watsky model, equation 1. The plots of the fits are shown in Supplemental Figure 8.^a

Experiments for Figure 4	k1 (h−1)	k2 (μM−1 h−1)	t50b
16 μM 8 kDac	1.63 (± 0.09) × 10−2	1.57 (± 0.02) × 10−2	10.68 (± 0.11)
16 μM 8 kDa + 0.2 μg/mL heparin	2.27 (± 0.13) × 10−3	6.17 (± 0.16) × 10−2	6.15 (± 0.09)
16 μM 8 kDa in 670 mOsm	5.40 (± 0.52) × 10−3	3.14 (± 0.06) × 10−2	8.97 (± 0.09)
16 μM 8 kDa in 670 mOsm + 0.2 μg/mL heparin	2.00 (± 0.09) × 10−3	4.92 (± 0.13) × 10−2	7.58 (± 0.17)
16 μM 8 kDa in 1410 mOsmd	1.15 (± 0.13) × 10−2	3.30 (± 0.01) × 10−2	7.17 (± 0.20)
16 μM 8 kDa in 1410 mOsm + 0.2 μg/mL heparin	5.60 (± 0.35) × 10−3	4.06 (± 0.03) × 10−2	7.29 (± 0.14)
Experiments for Figure 5
16 μM 8 kDac	9.29 (± 0.61) × 10−3	8.73 (± 0.14) × 10−3	19.04 (± 0.29)
16 μM 8 kDa + 1 μg/mL heparin	8.38 (± 0.23) × 10−3	2.04 (± 0.06) × 10−2	11.08 (± 0.17)
16 μM 8 kDa + 1 mM eprodisate	7.17 (± 1.40) × 10−3	1.08 (± 0.07) × 10−2	18.20 (± 0.49)
16 μM 8 kDa + 1 mM eprodisate + 1 μg/mL heparin	8.20 (± 0.18) × 10−3	2.10 (± 0.03) × 10−2	10.91 (± 0.14)
16 μM 8 kDa + 1 mM homotaurined	4.30 (± 2.28) × 10−3	1.22 (± 0.29) × 10−2	19.86 (± 1.13)
16 μM 8 kDa + 1 mM homotaurine + 1 μg/mL heparin	5.58 (± 0.53) × 10−3	2.44 (± 0.11) × 10−2	10.81 (± 0.35)

^aThe values shown are average k₁, k₂, and t₅₀ values (with standard devations in parentheses) of two or three individual fits of data to equation 1 from triplicate wells from a representative plate reader aggregation assay.

^bt₅₀ was calculated as the time at which the sigmoidal curve reached 50% of maximal fluorescence.

^cThe experiments were performed at different pH (7.0 for the Figure 4 experiment and 7.2 for the Figure 5 experiment), which accounts for the difference in values for similar experiments. It should be noted, however, that while amyloidogenesis is exquisitely sensitive to reaction conditions, the observations presented and discussed here and in Figures 4 and 5 – that increasing salt concentration reduces the effect of heparin on aggregation rate and that GAG mimetics have no effect on aggregation rate either in the presence or absence of heparin – were observed in at least three independent experiments.

^dOne of the data sets for this condition was poor (See Supplementary Figure 7), and so these values reflect an average of only duplicate wells.

It should also be noted that there is a substantial difference in the amplitude of the TfT fluorescence amongst some of the samples (Figure 4). The quantum yield of TfT is affected by the polarity and rigidity of its microenvironment, and it has been proposed that TfT interacts with the cross-β-sheet structures of amyloid (38). Many possibilities exist, however for the difference in microenvironment reflected by the difference in quantum yield – including fibril structure, bound glycosaminoglycans, etc., and it is difficult to be more specific about these amplitude changes without structural data to guide our hypotheses.

Effect of GAG Mimetics

GAG mimetics, sulfonated small molecules that mimic the sulfated sugars of the extracellular matrix, are hypothesized to ameliorate human amyloidoses by inhibiting the interaction between amyloidogenic peptides and GAGs in the extracellular matrix, thus reducing the extent of GAG-mediated acceleration of amyloid fibril formation (39–41). If true, the addition of the GAG mimetics eprodisate sodium and homotaurine should reduce the heparin-dependent acceleration of the rate of the 8 kDa gelsolin amyloidogenesis reaction as examined by TfT fluorescence, provided the GAG mimetics themselves do not accelerate amyloidogenesis. The GAG mimetics themselves did not alter the rate of 8 kDa gelsolin amyloidogenesis (Figure 5, cf. orange and green traces in the presence of the mimetics to the gelsolin alone trace (black)). Moreover, even at a concentration of 1 mM, much higher than that of heparin (1 μg/mL or 2 μM based on disaccharide repeats), the GAG mimetics only minimally affected the heparin-dependent acceleration of the rate of 8 kDa gelsolin amyloidogenesis (Figure 5, cf. red, blue, and purple traces). Again, each trace in Figure 5 is the arithmetic average of data from triplicate wells from a representative experiment, and data from each individual well is shown in Supplementary Figure 7C. To confirm the observations described, this experiment was performed a total of three times on three separate occasions using two different preparations of the 8 kDa fragment of gelsolin, and no significant effect of the GAG mimetics on aggregation kinetics was ever observed.

Figure 5.

The influence of GAG mimetics eprodisate sodium (1 mM) and homotaurine (1 mM) on the aggregation rate of the amyloidogenic 8 kDa fragment of gelsolin in the absence and presence of 1 μg/mL of heparin. The data from each individual well was fit to the Finke-Watzky model (equation 1) and the constants k₁ and k₂, as well as the t₅₀ shown in Table 1 are the arithmetic average of data from triplicate wells. Collectively, these data suggest that GAG mimetics have no effect on 8 kDa gelsolin amyloidogenesis.

To confirm the lack of any significant influence of the GAG mimetics on 8 kDa gelsolin amyloidogenesis, the data from individual wells from the representative experiment were again fit to the Finke-Watzky model (equation 1). The values of k₁, k₂, and t₅₀ shown in Table 1 (fitted curves shown in Supplementary Figure 8B) reveal that the addition of the GAG mimetics, either in the presence or in the absence of heparin, shows no significant effect on either k₂ or t₅₀. Thus, the GAG mimetics tested were unable to inhibit the interaction of heparin with the amyloidogenic 8 kDa fragment of gelsolin in vitro, consistent with the expectation that the interaction between heparin and gelsolin oligomers is polyvalent. In contrast, the 8 kDa gelsolin aggregation reactions that occurred in the presence of heparin exhibit k₂ values that are significantly higher and the t₅₀ values are significantly lower when compared to the reactions in the absence of heparin.

Examining the Early Stages of Gelsolin Amyloidogenesis

Analytical size exclusion chromatography coupled with light scattering was employed next to probe the early stages of the 8 kDa gelsolin amyloidogenesis reaction. 8 kDa gelsolin (24 μM) in 50 mM NaPi, 100 mM NaCl, 0.02% NaN₃ (pH 7.2) was mixed by overhead rotation (24 rpm) to form amyloid fibrils. At various time points into the amyloidogenesis reaction, aliquots were removed and loaded onto a Superdex 75 size exclusion column with a mobile phase of the same buffer. Upon eluting 8 kDa gelsolin from the column, the static light scattering at 90° and dynamic light scattering were measured. The TfT fluorescence of the aliquot was also examined to determine presence of amyloid fibrils or oligomeric cross-β-sheet aggregates.

We first reproduced the experiments previously described by Suk et al (15), using a 20 mL Superdex 75 column with an in-line 0.1 μm filter. In the absence of heparin, the monomer peak decreases with time, but no oligomers can be observed directly, as they are likely filtered out of solution before they are able to flow through the column (Supplementary Figure 9A). The decrease in the monomer peak occurs concomitantly with an increase in TfT fluorescence, suggesting the presence of fibrils. In the presence of heparin, however, a significant oligomer peak is observed as soon as TfT positive oligomers are observed (Supplementary Figure 9B, cf. absorbance and 90° light scattering traces), in addition to the previously observed decrease in the monomer peak. It should be noted that the 8 kDa gelsolin fragment used in the current work corresponds to amino acids 173-242, whereas in the previous work, an 8 kDa gelsolin fragment of amino acids 173-243 was used; however, this change is not expected to change aggregation propensity (15). In addition, the monomerization procedure is more rigorous in the current work, using GdnHCl and size exclusion chromatography to completely remove any oligomers.

An analogous experiment was performed with a slightly altered setup, in that a 2 mL Superdex 75 column without any in-line filters was used instead of the 20 mL Superdex 75 column. In addition, a much smaller sample was injected (see Methods). In this enhanced experimental paradigm, in the absence of heparin, cross-β-sheet oligomers were observed as soon as the TfT fluorescence began to increase (Figure 6A, cf. absorbance and 90° light scattering traces), although the mass balance was still poor (< 40%), likely due to fibrils being retained on top of the column. In the presence of heparin, more cross-β-sheet oligomers were observed (Figure 6B) and the mass balance was better (> 80%). Dynamic light scattering reveals that the hydrodynamic radius of the oligomers in the oligomer peak, both in the presence and absence of heparin is between 60 and 80 nm. It should be noted that the light scattering of the oligomers going through the column does not line up exactly with the absorbance peaks and is spread out because the volume of the sample was small compared to the volume of the flow cell. Collectively, these experiments demonstrate that no oligomeric species can be seen by size exclusion chromatography until the solution exhibits TfT fluorescence, an observation that, along with the far-UV CD spectroscopy data presented in Figure 1E, suggests a very fast transition from monomeric random coil 8 kDa gelsolin into cross-β-sheet oligomers.

Figure 6.

Analytical size exclusion chromatography (2 mL superdex 75 column) of 8 kDa gelsolin amyloidogenesis reactions. For each experiment, an 8 kDa gelsolin amyloidogenesis reaction was performed in which 8 kDa gelsolin (24 μM) without GAGs or GAG mimetics (A), with heparin (10 μg/mL) (B), with eprodisate sodium (1 mM) (C), or with both heparin (10 μg/mL) and eprodisate (1 mM) (D) was mixed by overhead rotation (24 rpm) to form amyloid fibrils. At the time points indicated, an aliquot of the reaction was examined by size exclusion chromatography, and 8 kDa gelsolin elution was detected by absorbance (280 nm), 90 degree static light scattering, and dynamic light scattering. The left panel of each set shows the overlaid absorption traces at indicated time points (bottom portion) and overlaid 90° light scattering traces (top portion). Dynamic light scattering revealed the hydrodynamic radius of the oligomers eluting in the void volume to be between 60 and 80 nm. The far right panels depict the integration of the absorbance peaks at each time point (top right) and the TfT fluorescence at each time point (bottom right).

It is particularly interesting that more oligomers are able to pass through the column in the presence of heparin. A possible explanation is that the cross-β-sheet oligomers and fibrils, once formed, are bound to heparin in the solution, thus inhibiting their lateral association into insoluble material that is retained by the column (15).

Because heparin appears to “solubilize” the aggregates and prevent them from associating and forming large structures, it was therefore hypothesized that the GAG mimetics, such as eprodisate sodium may function in vivo by a similar mechanism, and they were examined in this experimental paradigm. The addition of eprodisate sodium did not increase solubility of the 8 kDa gelsolin oligomers in the absence of heparin, as roughly the same amount of oligomers are able to make it through the column as without eprodisate (cf. Figure 6C to Figure 6A). The presence of eprodisate sodium in the reaction with heparin also did not affect the solubility of the oligomers. (cf. Figure 6D to Figure 6B). It therefore does not seem that GAG mimetics affect the solubility of oligomers to a great extent.

To determine if heparin had any effect on the morphology of the amyloid fibrils being formed, either early or in the late stages of the aggregation reaction, samples of 8 kDa gelsolin fibrils (24 μM, mixed by overhead rotation (24 rpm)) made either in the absence or in the presence of heparin (10 μg/mL) were examined by AFM and by TfT fluorescence as a function of time. In the absence of heparin, nothing could be observed by AFM after 40 min (Figure 7A) and the TfT signal was roughly equivalent to that of buffer, suggesting that there are no cross-β-sheet oligomers or fibrils. In the presence of heparin, however, a few small oligomers or protofibrils could be observed by AFM after 40 min (Figure 7B), when the TfT fluorescence had just begun to increase and was 2% of maximum. As the aggregation continues, fibrils could be seen both in the absence of heparin (Figure 7C (after 100 min), TfT fluorescence 60 % of maximum) and in the presence of heparin (Figure 7D (after 80 min) TfT fluorescence 55% of maximum). After 180 min, both in the absence (Figure 7E) and presence of heparin (Figure 7F) the amyloid fibrils have continued to extend, and their further lateral association can be observed, more so in the absence of heparin. In general, however, heparin did not seem to cause an appreciable difference in amyloid fibril morphology.

Figure 7.

Representative tapping mode atomic force microscopy images of 8 kDa gelsolin amyloidogenesis reactions. 8 kDa gelsolin (24 μM) was agitated by overhead rotation (24 rpm) either in the absence of heparin (A, C, and E) or in the presence of heparin (10 μg/mL) (B, D, and F). Images were taken 40 min (A), 100 min (C), and 180 min (E) after the start of the amyloidogenesis reaction for the samples examined in the absence of heparin and 40 min (B), 80 min (D), and 180 min (E) for the samples examined in the presence of heparin (10 μg/mL). Thioflavin T fluorescence readings were taken at the same time as the samples were prepared for AFM analysis.

Heparin accelerates the Fibril Extension of the Amyloidogenesis Reaction and Reduces Critical Concentration

The k₂ values obtained by fitting the 8 kDa gelsolin aggregation TfT-fluorescence monitored time courses to the Finke-Watzky model affords quantitative information about the fibril extension phase of the amyloidogenesis reaction, but the model is a simplified model of protein aggregation (36). Because it was previously demonstrated that the 8 kDa fragment of gelsolin aggregates by a nucleated polymerization mechanism with off-pathway aggregation (33), we further examined the heparin mechanism via the Powers and Powers model, which more accurately mathematically models this mechanism (42, 43). This model is characterized by a lag phase, during which a high energy oligomer or nucleus is formed, after which amyloid fibrils are extended in a thermodynamically favorable fibril extension reaction. In addition to the simple on-pathway fibril nucleus growth, off-pathway amorphous aggregates likely form. If the 8 kDa fragment of gelsolin aggregates by this mechanism, the lag and growth phases of the experimental data should fit to the equation:

$$X=c_1{t}^2+c_{2}t$$ (2)

where X is the TfT fluorescence intensity (42,43).

The fits of the data from a plate reader 8 kDa gelsolin aggregation assay (24 μM) in the absence of heparin up to 25% and 50% of maximum fluorescence (Figure 8A, green and blue curves, respectively) fit very well to the experimental data. On the contrary, when heparin is added, the fit up to 25% completion fits poorly to the rest of the time course (Figure 8B). Notably, the fibril extension phase of the reaction seems to be faster than expected from the fit to 25% completion. The plot of the residuals of the fits to 50% maximum fluorescence further emphasizes that the fit to the above equation is worse when heparin is added (Figure 8C).

Figure 8.

A–B. An 8 kDa gelsolin amyloidogenesis reaction (24 μM) either in the absence (A) or presence (B) of heparin (10 μg/mL) was performed and the data up to either 25% or 50% of maximum TfT fluorescence were then fit to a second degree polynomial (green curve and blue curves, respectively). C. The residuals calculated as experimental data minus best fit curves are plotted. D. The amyloidogenicity of 1 μM 8 kDa gelsolin examined in the absence (black curve) or in the presence (red curve) of heparin (10 μg/mL). The data for the 8 kDa gelsolin amyloidogenesis reaction up to 50% of maximum fluorescence were then fit to a second degree polynomial (blue curve).

The presence of heparin reduces the concentration of 8 kDa gelsolin necessary to form amyloid. When 1 μM 8 kDa gelsolin was examined in a plate reader amyloidogenesis reaction in the absence of heparin, the TfT fluorescent signal did not change over the 80 h timescale of the aggregation reaction (Figure 8D, black curve). However, an analogous reaction solution that contained heparin (10 μg/mL) was able to form TfT positive species (Figure 8D, red curve), suggesting that cross-β-sheet oligomers can more efficiently build up to detectible levels in the presence of heparin. Even at this low concentration, the experimental data in the presence of heparin deviated from the fit of the quadratic equation given above beyond 50% completion (Figure 8D).

Collectively, these observations suggest that the presence of heparin is able to decrease the concentration of an amyloidogenic peptide necessary for the accumulation of TfT positive aggregates. In addition, heparin seems to increase the rate of the fibril extension phase of the aggregation reaction to be faster than a nucleated polymerization with off pathway aggregation in the absence of GAGs, an observation supported by the k₂ constants in the Finke-Watzky model. We hypothesize this is the case because heparin binds to 8 kDa gelsolin cross-β-sheet oligomers, increasing the efficiency by which they can convert to high molecular weight amyloid fibrils, presumably as a consequence of oligomer concentration, alignment, and fusion.

Discussion

The deposition of amyloid in mammals is strongly associated with several extracellular matrix components, including glycoproteins and GAGs. The distribution of these components in mammals creates different biological environments, which may affect the extent and tissue distribution of amyloid deposition. Understanding the interaction between GAGs and amyloidogenic peptides and the mechanism by which GAGs accelerate amyloidogenesis provides important mechanistic insights that help us think about new therapeutic strategies. Several mechanisms have been posited to explain how GAGs accelerate amyloidogenesis and these mechanisms are not necessarily mutually exclusive. In fact, GAGs could exert different roles for distinct amyloidogenic peptides.

Some have suggested that GAGs are a “scaffold” to which monomeric amyloidogenic peptide or proteins can bind, enhancing amyloid fibril formation by increasing local concentration and orienting the precursors to facilitate amyloid fibril formation (44). Our finding that heparin does not alter the intrinsically disordered structure of gelsolin and that monomeric gelsolin does not bind to heparin seems to eliminate this as a possible mechanism whereby heparin accelerates the rate of gelsolin aggregation.

Alternatively, others have hypothesized that GAGs bind to small amorphous aggregates or cross-β-sheet oligomers leading to an accelerated conformational conversion to cross-β-sheet fibrils (14). We have no evidence for the formation of unstructured pre-nucleus oligomers in the absence or presence of heparin. The isosbestic point observed in the far-UV CD experiments is fully consistent with a nucleated polymerization, whereby a very low concentration of pre-nucleus oligomers with alternative structures exist. The isosbestic point also strongly suggests a common post-nucleus cross-β-sheet quaternary structure in the oligomers and in the fibrils. Such a mechanism is also consistent with the far-UV CD transition occurring concomitantly with the increase in TfT fluorescence intensity associated with cross-β-sheet quaternary structure or amyloid formation. Moreover, the appearance of aggregate species by AFM and the increase in TfT fluorescence intensity occur simultaneously. Finally, analytical size exclusion chromatography did not reveal significant oligomer formation until the TfT fluorescence also began to increase, both in the presence and absence of heparin. Overall, the lack of pre-nucleus non-cross-β-sheet quaternary structural intermediates suggests that any off pathway aggregate formation is so minor as to be undetectable.

Our data favors the hypothesis that GAGs bind to already formed cross-β-sheet amyloid-like oligomers, or protofibrils, as they are sometimes called, accelerating their growth into amyloid fibrils, while preventing fibril disassembly (Figure 9) (45). We show that heparin was able to bind to cross-β-sheet oligomers and/or amyloid fibrils of 8 kDa gelsolin, and that the fibril extension phase of the amyloidogenesis reaction was accelerated by mathematically fitting our TfT fluorescence data to two different mathematical models of amyloidogenesis.

Figure 9.

Mechanism for heparin acceleration of 8 kDa gelsolin amyloidogenesis that is consistent with all of the data described and presented within. Blue squares represent monomeric 8 kDa gelsolin adopting a random coil or non-β-sheet ensemble of structures. Green circles represent post-nucleation 8 kDa gelsolin oligomers exhibiting a cross-β-sheet quaternary structure. The polymerized red hexagons represent heparin. Cross-β-sheet oligomers formed by a nucleated polymerization mechanism (not influenced by heparin) accumulate post-nucleation and then bind to heparin, which subsequently accelerates the fibril extension phase of the amyloidogenesis reaction by binding, concentrating, aligning, and enabling the fusion of oligomers into amyloid fibrils.

While the acceleration of fibril extension may occur by a few mechanisms, the most likely scenario involves concentrating and aligning oligomers and thus enhancing the frequency of oligomer fusion (Figure 9). Because monomeric 8 kDa gelsolin does not bind to heparin, it is unlikely that fibril extension is accelerated by enhanced recruitment of protein monomers. Increasing salt concentration dramatically reduces the effectiveness of heparin at accelerating the rate of amyloidogenesis, suggesting the importance of electrostatic interactions between the negatively charged heparin and the positively charged regions of gelsolin oligomers.

Some investigators have proposed that GAGs are key to causing amyloidogenesis and proteotoxicity in humans. Thus, substantial efforts have been expended toward the development of GAG mimetics, sulfonated small molecules that mimic the sulfated sugars of the extracellular matrix, to antagonize the GAG-amyloid interaction. These molecules were envisioned to antagonize the GAG-mediated acceleration of amyloidogenesis hypothesized to cause human amyloid diseases. These compounds have exhibited promise in clinical trials for the treatment of inflammation-associated amyloid disease, although the clinical trial results have not attained the expected clinical significance (39–41). However, the GAG mimetics, eprosidate sodium and homotaurine, were not effective at inhibiting the heparin-mediated acceleration of 8 kDa gelsolin amyloidogenesis, nor did they affect the solubility of the oligomers. It therefore seems unlikely that sulfonated small molecules are able to generally antagonize the polyvalent effect of GAGs (Figure 9), which hastens the fibril extension phase of the amyloidogenesis reaction.

All of our data is consistent with the hypothesis that 8 kDa gelsolin aggregates by a nucleated polymerization mechanism in the absence or presence of heparin (Figure 9). In this mechanism, the random coil monomers very inefficiently form a high energy β-sheet rich oligomeric nucleus structure during the lag phase of the amyloidogenesis reaction – a step not influenced by the presence of heparin. The nucleus adds monomers and matures into a cross-β-sheet oligomer with kinetics that must be faster than nucleation, as the fibril extension phase of the aggregation dominates the TfT time course. Heparin appears to accelerate the growth phase of the aggregation reaction not by facilitating monomer addition to the cross-β-sheet oligomers (heparin does not bind 8 kDa gelsolin monomer), but by binding and concentrating cross-β-sheet oligomers, apparently facilitating their fusion into higher molecular weight cross-β-sheet structures, including amyloid fibrils, more efficiently than in the absence of heparin (Figure 9). That the kinetics for the nucleated polymerization growth phase are faster than expected is consistent with this hypothesis. All evidence points only to the growth phase being accelerated by heparin. Lowering of the critical concentration of the aggregation is most likely a consequence of fibril stabilization through heparin binding. It is possible that the reason that GAGs are upregulated in human amyloidoses and in amyloidosis animal models is that GAGs play a vital biological protective role by reducing proteotoxicity from small soluble cross-β-sheet oligomers through facilitating their fusion into fibrils that are rendered less proteotoxic through GAG mediated fibril stabilization in the affected tissue, until fibrils start displacing tissue.

Abbreviations

Aβ

amyloid beta

AFM

atomic force microscopy

C68

C-terminal 68 kDa fragment of human plasma gelsolin

CD

circular dichroism

CNBr

cyanogen bromide

FAF

Familial Amyloidosis of Finnish Type

GAG

glycosaminoglycan

GdnHCl

guanidinium hydrochloride

MRE

mean residue ellipticity

MT1-MMP

membrane type-1 matrix metalloprotease

NaPi

sodium phosphate

TfT

thioflavin T