|Home | About | Journals | Submit | Contact Us | Français|
Familial amyloidosis of Finnish type (FAF), or gelsolin amyloidosis, is a systemic amyloid disease caused by a mutation (D187N/Y) in domain 2 of human plasma gelsolin, resulting in domain 2 misfolding within the secretory pathway. Upon passage through the Golgi, furin endoproteolysis within domain 2 occurs as a consequence of the abnormal conformations that enable furin to bind and cleave, resulting in the secretion of a 68-kDa C-terminal plasma gelsolin fragment (amino acids173–755, C68). The C68 fragment is cleaved upon secretion from the cell by membrane type 1 matrix metalloprotease (MT1-MMP), affording the 8 and 5 kDa fragments (amino acids 173–242 and 173–225, respectively) comprising the amyloid fibrils in FAF patients. Herein, we show that the 8 and 5 kDa gelsolin fragments form amyloid fibrils by a nucleated polymerization mechanism. In addition to demonstrating the expected concentration dependence of a nucleated polymerization reaction, the addition of preformed amyloid fibrils, or “seeds”, was shown to bypass the requirement for the formation of a high energy nucleus, accelerating 8 and 5 kDa D187N gelsolin amyloidogenesis. The C68 fragment can form small oligomers, but not amyloid fibrils, even when seeded with preformed 8 kDa fragment plasma gelsolin fibrils. Because the 68 kDa fragment of gelsolin does not form amyloid fibrils in vitro or in a recently published transgenic mouse model of FAF, we propose that administration of an MT1-MMP inhibitor could be an effective strategy for the treatment of FAF.
Over thirty human proteins can misassemble into cross-β-sheet aggregates, termed amyloid fibrils, by the multi-step process of amyloidogenesis that appears to cause the degenerative diseases known as amyloid diseases (1–4). Some human amyloidogenic proteins are intrinsically disordered, whereas others adopt well-defined folded structures (1–7). Amyloidogenesis and aging together appear to lead to proteotoxicity, which is thought to cause the degeneration characteristic of the human amyloid diseases (8–12), clinically important examples of which include Alzheimer’s disease, light chain amyloidosis and the transthyretin amyloidoses (8, 13–17). Amyloidogenesis by wild-type proteins or their fragments appear to cause sporadic amyloid diseases (3, 12, 18). In contrast, amyloidogenesis by mutant proteins or their fragments (which may or may not contain the mutation) appear to cause familial amyloidoses (12, 19).
Gelsolin amyloid disease, also called Familial Amyloidosis of Finnish type (FAF), appears to require a mutation in the plasma gelsolin protein (20–23) since there are no reports of wild type plasma gelsolin aggregation leading to a sporadic amyloid disease. FAF is autosomal dominantly inherited, consistent with amyloidogenesis by a mutant plasma gelsolin fragment leading to a gain-of-toxic-function (19, 20). The mechanism of proteotoxicity, including whether intra- and/or extracellular amyloidogenesis lead to the degeneration characteristic of FAF, remains unclear (19). An aging-associated decline in cellular protein homeostasis capacity may contribute to the onset of the amyloidoses, including FAF, and may explain why aging is a dominant risk factor for these diseases (8, 10–12, 19).
FAF is characterized by systemic amyloid deposition, apparently facilitated by the extracellular matrix (19, 24). Gelsolin fragment amyloid accumulation is particularly noticeable in the basement membrane of the skin (25), blood vessel walls (26), eyes (27), and the peripheral nervous system (28). Patients often present with corneal lattice dystrophy and a slowly progressive cranial and peripheral polyneuropathy. As the disease progresses, the skin becomes thickened and loose and the neuropathy is exacerbated, often with mild autonomic nervous system involvement (26). Ultimately, the amyloid deposition and the resulting inflammation in vital organs becomes life threatening, and thus, the primary causes of death are typically pneumonia, cerebral hemorrhage or nephrotic syndrome (28, 29).
The amyloidogenic fragments that deposit in FAF patients derive from the protein gelsolin, a six-domain, Ca2+-binding protein that regulates actin polymerization and fibril fragmentation (30, 31). Gelsolin has two splice variants – an 81 kDa intracellular variant responsible for remodeling the actin cytoskeleton, and a secreted 83 kDa version responsible for scavenging actin fibrils released from injured tissue into the blood (thereby preventing increases in blood viscosity.) Plasma gelsolin may also have other functions (32, 33). Plasma gelsolin is secreted by several cell types, with muscle being a major source (34).
FAF is most often caused by a G654 to A point mutation in one gelsolin allele. This mutation changes an aspartate at position 187 to an asparagine (D187N), eliminating a calcium binding ligand and dramatically lowering the Ca2+ binding affinity of the second domain of plasma gelsolin, compromising its stability (35–37). A less common G654 to T point mutation, resulting in a tyrosine substitution at position 187 (D187Y), also causes FAF. Without Ca2+-binding-mediated stabilization, domain 2 is conformationally heterogeneous, and the more extended conformers are susceptible to cleavage by furin in the trans Golgi, affording a C-terminal 68 kDa gelsolin fragment (C68), corresponding to amino acids 173–755 (34, 35, 37–39). Only a fraction of D187N/Y gelsolin is cleaved by aberrant furin endoproteolysis. The remainder is secreted as full length plasma gelsolin that can adopt a folded tertiary structure similar to wild type, and appears to be adequately functional. Because D187N/Y homozygotes do not seem to exhibit a loss-of-function phenotype, and because gelsolin knockout mice are viable and exhibit only a mild, primarily hematological phenotype (40), the symptoms of FAF described above appear to derive from the gain-of-toxic-function of mutant gelsolin. The C68 fragment is then cleaved by membrane type 1 matrix metalloprotease (MT1-MMP) or possibly other related MMPs in the extracellular matrix (41). The MMP(s) responsible likely derive from either the same cell as or cells adjacent to those secreting C68 (19). The MT1-MMP endoproteolysis of C68 results in the formation of 8 and 5 kDa amyloidogenic fragments, the latter is not formed without the presence of the former. The 8 kDa fragment is the main component of FAF-associated fibrils (19, 42).
The amyloidogenic 8 kDa plasma gelsolin fragment (amino acids 173–243 in the native full length protein) incorporates the D187N mutation. This segment in the natively folded protein is made up of four β-strands and an α-helix, but it is unlikely that the native structure is maintained in the fragment since endoproteolytic cleavage eliminates an internal β-strand (43). The 5 kDa amyloidogenic fragment (amino acids 173–225 in the native full length protein), also incorporating the D187N mutation, is likely also intrinsically disordered as a monomer (42, 44).
While the amyloidogenicity of the 8 kDa plasma gelsolin fragment has been previously studied (although not with rigorously monomerized peptides) (24, 45), its amyloidogenicity relative to the other plasma gelsolin fragments has not been examined, nor has its mechanism of amyloid fibril formation been established. In addition, we wanted to know whether the secreted C68 fragment could form amyloid, as it is conceivable that the 8 and 5 kDa plasma gelsolin fragments composing human amyloid fibrils could result from C68 amyloidogenesis followed by further proteolytic digestion. Herein, we report that the 8 kDa fragment is more amyloidogenic than the 5 kDa fragment and that both the 8 and 5 kDa gelsolin fragments aggregate by a nucleated polymerization mechanism. We also demonstrate that C68 is not amyloidogenic, even upon attempted seeding by 8 kDa gelsolin fragment amyloid fibrils.
The 8 kDa fragment (173–242) and the 5 kDa fragment (173–225) of plasma gelsolin were biosynthesized as recombinant fusion proteins to BCL-XL-1/2, an aggregation-prone polypeptide that sequesters the attached gelsolin sequences into inclusion bodies within BL21 (DE3) E. coli, as previously described (46). Briefly, E. coli were grown to an OD600nm of 1.2, induced with 5 mL of 20% arabinose, and grown for an additional 3 h. The bacteria were lysed by sonication in lysis buffer (50 mM Tris, 100 mM NaCl, 5 mM EDTA, 0.02 % NaN3, pH 8.0 with 0.1% Triton X added). After centrifugation, the supernatant was removed, fresh lysis buffer was added to the pellet, and sonication was repeated to ensure complete lysis. After centrifugation, the remaining insoluble material, containing mostly inclusion bodies, was rinsed by sonication-facilitated resuspension followed by centrifugation once in 50 mM Tris, 100 mM NaCl, 5 mM EDTA, 0.02% NaN3, pH 8.0 and then twice in 50 mM Tris, 50 mM MES, 100 mM NaCl, 0.02% NaN3, pH 5.0. After the rinse steps, the inclusion bodies were dissolved in 7.2 M guanidine HCl for 5 to 7 days at 4 °C using rotary agitation. The protein was then purified using a Co column preequilibrated in 7.2 M guanidine HCl, 50 mM Tris, pH 8.0 utilizing the His7 tag on the N-terminus of the BCL-XL-1/2 fusion protein. The protein was eluted in 7.2 M Guanidine HCl, 50 mM Tris, 50 mM MES, pH 5.0 through a reduction in pH. The fusion protein was then purified by reverse phase HPLC using a C-4 preparative column. When available, acetonitrile was used as the organic phase solvent, but a 50:50 mixture of isopropanol and methanol was also used as an alternative organic phase solvent when acetonitrile was commercially unavailable. The fusion protein was loaded onto the column in water containing 0.2% trifluoroacetic acid and eluted with a linear gradient (20% to 80%) of increasing organic phase (containing 0.2% trifluoroacetic acid). The fusion protein was then lyophilized. At this stage, the BCL-XL-1/2 sequence was removed by cyanogen bromide cleavage. For the cleavage reaction, the lyophilized fusion protein was dissolved in 7.2 M guanidine HCl, 0.1 M HCl, pH < 1.0 at a concentration of 10 mg/mL, an equivalent volume of 5 mg/mL (roughly 100 molar equivalents) CNBr in the same buffer was added, and the reaction proceeded overnight at room temperature. The crude cleavage product was purified by reverse phase HPLC using the solvent system described above, revealing > 95% purity. The 8 and 5 kDa plasma gelsolin fragments were characterized by LC-MS, and their masses matched the expected masses of 7860 Da and 6067 Da, respectively. A representative HPLC chromatogram and LCMS characterization data from the purified 8 kDa fragment is shown in Supplementary Figures 1 and 2. The addition of the phosphine reducing agent, TCEP (4 mM), resulted in a 2 Da increase in mass, consistent with reduction of an intramolecular disulfide bond between cysteines at 188 and 201 that exists in the wild type full length gelsolin.
1–2 mg of lyophilized 8 or 5 kDa gelsolin fragment, purified as described above, (the exact amount being dependent on the desired final concentration) was dissolved in 8M guanidine HCl buffered with 50 mM Sodium Phosphate (pH 7.5) and sonicated in a water bath sonicator for 2 h to break up any aggregates or oligomers. The peptide solution was then loaded onto a Superdex 30 gel filtration column (20 mL volume) preequilibrated with 50 mM sodium phosphate, 100 mM NaCl (pH of buffers used was between 6.0 and 7.5, depending on the experiment) to buffer exchange and remove any aggregated species. Oligomers eluted from the column in the void volume at about 8 mL, while the monomerized protein eluted at 12 mL. The monomerized peptide in the desired buffer, free of chaotrope, was collected and used immediately in the amyloidogenicity experiments.
Recombinant C68 was made as an N-terminal glutathione S-transferase (GST) fusion protein featuring a His6 tag N-terminal to the GST sequence (His6 – GST – TEV cleavage site – C68), as previously described (45), with minor modifications. The TEV protease cleavage site was added for the specific and efficient removal of the solubilization protein and purification tag from the 68 kDa fragment of interest, and the fusion protein was purified using a Ni column, as it was determined that the fusion protein did not bind well to the glutathione resin. Bacterial lysate was loaded onto the Ni column preequilibrated in 20 mM Tris, 10 mM imidazole, and was eluted using a linear gradient to 250 mM imidazole over three column volumes. The fractions containing the fusion protein were collected and pooled. TEV protease was added in a 1:100 molar ratio to cleave His6 – GST from C68, and glycerol was added to a final concentration of 20% to prevent protein aggregation. The cleavage reaction was mixed and then incubated overnight, without stirring, at 4 °C. A reverse Ni column strategy was then utilized to purify C68 from the other cleavage products and the TEV protease (which also had a His6 tag). Ni resin was added directly to the cleavage reaction and incubated for 2 h at 4 °C. With the addition of the glycerol and the Ni resin, the imidazole concentration was substantially diluted, allowing the His6 – GST purification tag and the TEV protease to bind efficiently to the Ni resin. The flow through, containing the purified C68, was then collected and concentrated using a centrifugal protein concentrator. Because C68 is a large, multidomain protein with folded domains, it was not pretreated with guanidine, as the 8 kDa and 5 kDa fragments were, for the purpose of monomerization. However, it was still subjected to a final purification and monomerization step by passage through a Superdex 200 gel filtration column (20 mL volume column) that had been preequilibrated with 50 mM sodium phosphate, 100 mM NaCl (pH of buffers used was between 6.5 and 7.5, depending on the experiment) to buffer exchange and remove any aggregated species immediately before any experiments were performed. SDS-PAGE revealed >95% purity (Supplementary Figure 3).
The pET3a construct for PG, (obtained as a kind gift from the Yin laboratory, UT Southwestern), was expressed in BL21 (DE3) E. coli cells and purified by a method based on the one reported by the Yin laboratory (47). Briefly, E. coli transfected with the construct were grown at 37 °C to an OD600nm of 0.6, induced by addition of 1 mM isopropyl-1-thio-β-D-galactopyranoside, and allowed to grow for an additional 3 h. Cells were lysed by sonication in multiple 5 s bursts on ice, aided by the addition of 1 mg of lysozyme. The lysate was centrifuged for 10 min at 10,000 × g at 4°C. Because SDS-PAGE revealed most of the PG to be in the supernatant, the supernatant was filtered with a 0.45 µm filter and loaded onto a Source Q anion exchange column preequilibrated with 20 mM Bis-Tris, 0.02% NaN3 (pH 7.0). The bound PG was eluted with a gradient of NaCl (0 to 0.5 M). Because PG is a large, multidomain protein with folded domains, it was not pretreated with guanidine. Instead it was passed through a Superdex 200 gel filtration column (20 mL volume column) that had been preequilibrated with 50 mM sodium phosphate, 100 mM NaCl (pH of buffers used was between 6.5 and 7.5, depending on the experiment) to buffer exchange and remove any aggregated species immediately before any experiments were performed. SDS-PAGE revealed >90% purity (Supplementary Figure 4).
Monomerized 8 or 5 kDa gelsolin fragments (20 µM) were aggregated in 50 mM sodium phosphate, 100 mM NaCl, 0.02% NaN3 (pH 7.0) at 37 °C in an microfuge tube (wherein the final volume was at least 600 µL) under constant agitation using an overhead shaker (24 rpm) for at least 16 h. It was critical that the final volume was at least 600 µL otherwise the surface tension does not allow the solution to mix with each rotation. The presence of fibrils was verified by thioflavin T (TfT) fluorescence to ensure that the fluorescence signal was at least 10 times that of a solution of 20 µM TfT with no protein. Preformed fibrils were used in experiments within 3 days, after which the TfT signal decreased dramatically. Immediately before use in the seeding experiments, the gelsolin fibrils were sonicated for 20 minutes in a water bath sonicator, as this has been previously shown with amyloid β to disrupt long fibrils into more uniform 50–100 nm fibrils (previously characterized by AFM) more suitable for a seeding experiment (8, 48).
The monomerized amyloidogenic gelsolin fragments were diluted to the desired concentrations in 50 mM Sodium Phosphate buffer with 100 mM NaCl, 0.02% NaN3 and 20 µM TfT (pHs of buffers used were between 6.0 and 7.5, as reported in the results section). For assessing the amenability to seeding, a fraction of the amyloidogenic gelsolin fragment concentration (ranging from 0.1% to 10%, as reported in the results section) was added as preformed fibril seeds, prepared as described above, instead of monomerized peptide. For experiments involving extracellular matrix components, the oligosaccharides were dissolved in buffer and added to the solution to achieve the reported concentration. 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) and incubated at 37 °C for the duration of the experiment. Every 10 min, the fluorescence (excitation at 440 nm and emission at 485nm) was measured from the bottom of the plate after 5 s of shaking. Each reading was the mean of thirty individual scans. The data from the triplicate wells were averaged and plotted as TfT fluorescence vs. time.
To confirm the validity of the data obtained from the plate reader, additional experiments were performed using a fluorescence spectrometer. The monomerized solution of gelsolin fragment was diluted to the desired concentration in 50 mM Sodium Phosphate buffer with 100 mM NaCl, 0.02% NaN3 (pHs of buffers used were between 6.5 and 7.5, as reported in the results section). For experiments involving extracellular matrix components, the oligosaccharides were dissolved in the same buffer and added to the solution. All components were mixed together in an Eppendorf tube wherein the final volume was at least 600 µL. The tubes were placed in a 37 °C incubator and mixed using overhead rotation at 24 rpm. Periodically, 15 µL of the solution was removed, added to 85 µL of a 23.5 µM TfT solution (final TfT concentration was 20 µM), and transferred to a fluorescence cuvette. The fluorescence (excitation at 440 nm and emission at 485 nm) was then measured in a Cary Eclipse fluorescence spectrophotometer, and the data were plotted as fluorescence vs. time. Aliquots could also be taken from these samples for examination by other techniques, as described below.
Aliquots from samples prepared for the fluorimeter-based Thioflavin T aggregation assay described directly above were examined by AFM to discern aggregate morphology. Twenty µL of the sample 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 25 µL of distilled water that was immediately wicked off of the surface of the mica with filter paper. AFM Images were recorded in tapping mode using a Digital Instruments multimode scanning probe microscope with a Nanoscope IIIa controller and force modulation etched silicon probes.
Samples prepared for the fluorimeter-based Thioflavin T aggregation assay described above were examined by AUC. Sedimentation velocity experiments were performed on a Beckman XL-I analytical ultracentrifuge using both absorbance and interference optics. Protein samples were loaded into 1.2 cm cells and were allowed to equilibrate for 1 h at 25 °C prior to centrifugation at 50,000 rpm in an An60-TI four-hole rotor. Data were analyzed using the c(s) method in Sedfit (49).
Monomerized 8 or 5 kDa gelsolin (20 µM) was subjected to overhead rotation (24 rpm) at 37 °C for 16 h at pH 7.0 to form fibrils, as in the fluorimeter-based Thioflavin T aggregation assay described above. The presence of fibrils was verified by TfT fluorescence. Proteinase K (3 µL of a 0.1 mg/mL solution) was added to 100 µL of the fibril solution and incubated at 37 °C with continuous orbital shaking (250 rpm) for various lengths of time before inactivating the proteinase K by boiling for 10 min.
For SDS-PAGE analysis, gel loading buffer was added to the proteinase K treated fibrils, and the samples were boiled for 10 min. Samples were run on a 16.5% tris/tricine peptide gel to separate the small peptide fragments (50), and the gel was stained with Coomassie blue for analysis. For mass spectrometry analysis, 90 µL of 8 M guanidine HCl (pH 4.0) was added to 10 µL of the digested fibrils and incubated for 6 h before being examined by LC-MS.
The aggregation propensity of the monomerized gelsolin fragments was examined as a function of pH using the plate reader TfT fluorescence amyloidogenicity assay, as described in the Materials and Methods Section. It was previously shown that aggregation of the 8 kDa plasma gelsolin fragment is very sensitive to pH – amyloidogenicity is enhanced as the pH decreases, consistent with the calculated pI (5.36) of this fragment (45). However, in the prior study, a rigorous monomerization pretreatment was not performed; hence it is important to collect revised data to ensure that fibrillar seeds were not contributing to the previous observations. Decreasing the pH should lower the net charge on this intrinsically disordered fragment, rendering it more assembly competent at a given concentration (51). Figure 1A illustrates the pH dependence of the amyloidogenicity of the 8 kDa fragment. Reducing the pH dramatically accelerates 8 kDa plasma gelsolin amyloid fibril formation.
Because the aggregation kinetics exhibited by the 8 kDa fragment of plasma gelsolin are highly pH dependent, a comparison of the relative amyloidogenicities of the C68, 8, and 5 kDa gelsolin fragments was performed at several physiologically relevant pHs (pH 6.5–7.5). The physiological concentration of gelsolin in human plasma averages 200 µg/mL in healthy volunteers (2.5 µM), with higher levels often seen (34). We therefore decided to examine amyloidogenicity of the fragments at a concentration of 4 µM as this concentration is in the physiologically relevant range, is high enough to facilitate aggregation on a convenient laboratory time scale, and allows for an excellent signal to noise ratio in the TfT-monitored amyloidogenicity time courses to be observed.
At the lowest pH examined (pH 6.5), the 8 and 5 kDa gelsolin fragments exhibited nearly identical amyloidogenicity rates, while the C68 fragment exhibited only a slight increase in TfT binding (Figure 1B). Both the 8 and 5 kDa fragments exhibit time courses expected of a nucleated polymerization, featuring a lag or nucleation phase (~2 h), a growth phase with a t50 of 6.5 h, and a stationary phase, reached when the monomer is depleted below the critical concentration. The C68 curve does not show these characteristic phases.
At pH 6.8, the 8 kDa fragment again exhibited an amyloidogenicity time course consistent with a nucleated polymerization, although aggregation was slower (Figure 1C). In contrast, the 5 kDa fragment did not aggregate at pH 6.8, implying that at this pH the concentration of the 5 kDa fragment is now below its critical concentration, that the time frame of the experiment was too short to observe any aggregation and increase in TfT signal, or that the 5 kDa fragment simply does not aggregate at pH 6.8. The C68 fragment exhibited an increase in TfT fluorescence with time, and although it reached the amplitude of that of the 8 kDa fragment, it did not exhibit a lag phase followed by a growth phase. Two explanations for the increase in TfT fluorescence are that C68 is forming reversible oligomers or that it is aggregating into fibrils by a downhill polymerization like transthyretin does (52). It is well established that TfT is an environment-sensitive fluorophore that reliably reports on amyloid fibril formation (53), but there are also examples of TfT exhibiting fluorescence when bound to a native or near-native protein (54). These possibilities will be examined below. Full length recombinant plasma gelsolin (PG) was also examined at pH 6.8, revealing that TfT fluorescence did not increase with time. It is notable that the amplitude of the TfT fluorescence at t = 0 from both C68 and PG are nearly identical and substantially greater than the buffer control or the monomeric 8 and 5 kDa fragments at the beginning of their aggregation reactions, indicating that there are one or more TfT binding sites in PG and C68, likely in a folded domain(s) common to both.
The 68, 8 and 5 kDa gelsolin amyloidogenicity experiments were also carried out at pH 7.0 and 7.4. The 5 kDa fragment was not amyloidogenic on this timescale, in contrast to the 8 kDa fragment, which exhibited a slight increase in TfT fluorescence, consistent with inefficient fibril formation at these pHs (Supplementary Figure 5). The C68 fragment exhibited a similar TfT fluorescence increase with time as observed at the lower pHs, as discussed above (Supplementary Figure 5).
The amyloidogenicity of the 8 and 5 kDa fragments was further examined at pH 7.4 as a function of gelsolin fragment concentration (Figure 1D). The 8 kDa fragment formed amyloid much more quickly than the 5 kDa fragment at these higher concentrations. With the exception of the data collected at pH 6.5, the 8 kDa fragment is more amyloidogenic than the 5 kDa fragment under all other conditions explored (physiological concentrations and pH).
It is important to note that the first few points of the TfT plate reader amyloidogenicity assays have a higher fluorescence reading than the baseline. It was determined that this is an artifact of the instrument and does not represent TfT binding to beta sheet structures, as buffer with TfT added also exhibits the higher initial reading (Figure 1). Because the initial reading is always about 20 raw fluorescence units higher than the baseline, the artifact is more pronounced when the curve has a smaller maximum fluorescence, as in the experiments with gelsolin fragments at a concentration of 4 µM or less (e.g., Figures 1B, E, and F).
To further probe the apparent aggregation of the C68 fragment, we examined its aggregation under conditions known to accelerate amyloidogenesis (Figures 1E–H). All experiments were performed using 1.5 µM C68 at a pH of 6.5. First, heparin (10 µg/mL) was added to the reaction mixture, as heparin has been shown to accelerate 8 kDa gelsolin fragment amyloid fibril formation (24). When compared to the control (Figure 1E), no acceleration or amplitude increase in the TfT time course was observed (Figure 1F). Nearly all amyloidogenicity reactions are accelerated by sulfonated or sulfated glycosaminoglycan polymers (55, 56), making it unlikely that C68 is forming amyloid. Secondly, sonicated 8 kDa fibrils were added to monomerized C68 in an attempt to nucleate amyloid fibril formation. TfT did bind to the 8 kDa fibril seeds, as can be seen by the increase in the TfT fluorescence baseline relative to the other experiments, but these seeds clearly did not template the fibrillization of C68 (Figure 1G), providing further evidence that it is unlikely that the C68-mediated increase in TfT signal with time is a consequence of amyloid fibril formation. Finally, 0.5 M guanidine was added to C68, because chaotropes are known to destabilize initially folded amyloidogenic proteins, or collapsed intrinsically disordered proteins, accelerating the process of fibril formation (57). However, no increase in the TfT signal, linked to amyloidogenicity, was observed (Figure 1H).
Next, we attempted to maximize the probability for C68 amyloid fibril formation by using a different agitation method, as the method of agitation can be an important factor in inducing amyloidogenesis (58). The C68 fragment (4 µM) was subjected to constant rotation at 37 °C (24 rpm) both in the presence and absence of heparin (10 µg/mL), at pH 6.5, utilizing the fluorimeter-based TfT assay described in the Materials and Methods section. Constant rotation maximizes the chance for aggregation because of the constant exposure of C68 to the denaturing air/water interface (59–61). A slight increase in TfT fluorescence was observed, comparable to that derived from the plate reader assays described above, but no evidence for amyloidogenesis of C68 was noted (Supplementary Figure 6A).
To determine whether this small increase in the C68 TfT signal with time was due to the oligomerization of C68 and TfT binding, the solution was examined by atomic force microscopy (AFM) and analytical ultracentrifugation (AUC) at the conclusion of the fluorimeter-based TfT assay. AFM did not detect the presence of any large oligomers or amyloid fibrils, as the surface of the mica looks very similar to the buffer control (Supplementary Figure 6, cf. B to C). In contrast, when a solution of the 8kDa gelsolin fragment (4 µM) is subjected to constant rotation at 37 °C (24 rpm) overnight, fibrils are formed at pH 6.5 (Supplementary Figure 6D).
To analyze potential oligomerization by AUC, an initially monomeric solution of C68 (4 µM) that was rotated for 16 h (24 rpm) at 37 °C was compared to a freshly monomerized C68 solution (4 µM). Although there was no evidence for a rapidly sedimenting species that would be observed for amyloid fibrils or large oligomers, the sedimentation velocity experiment is fully supportive of reversible soluble oligomer formation. For freshly monomerized C68 (Figure 2A), a peak with a sedimentation coefficient of 4.0 S, representing 92% of the sample, corresponds to the molecular weight of monomer. A small peak at 6.8 S, representing 6% of the sample, is consistent with the molecular weight of a C68 dimer. If the sample is rotated overnight, a much higher percentage is oligomeric (Figure 2B). The peak with a sedimentation coefficient of 3.8 S, consistent with the molecular weight of a monomer, represents only 45% of the sample. A substantial dimer peak at 6.8 S is present, along with shoulders, likely representing higher order oligomers. All together, the oligomers represent 42% of the sample. The small peak with a sedimentation coefficient of 2.0 S is likely due to proteolysis.
Because no fibrils or high molecular weight structures were observed by AFM, because the TfT fluorescence does not increase significantly, and because no rapidly sedimenting species were observed in the AUC experiments, it can be concluded that the 68 kDa fragment does not form cross-β-sheet amyloid fibrils. Instead, it remains mostly monomeric, transiently populating dimers, trimers and possibly tetramers.
A nucleated polymerization amyloidogenesis reaction is characterized by a lag phase, during which the amyloidogenic peptide forms a high-energy oligomer or nucleus, which is the rate-limiting step of fibril formation. Once the nucleus is formed, additional monomers can add to it in a fibril extension reaction, a thermodynamically favorable process characterized by a relatively rapid growth phase. The rapidity of the growth phase is also influenced by fibril fragmentation efficiency (62–64). Nucleation and fibril extension are concentration dependent (62, 65) up to a certain protein concentration beyond which off-pathway aggregation may begin to occur, slowing the rate of fibril formation (66).
As the concentration of the monomerized 8 kDa plasma gelsolin fragment increases from 2 µM up to 32 µM, the lag phase shortens and the rate of fibril extension increases (Figure 3). However, as the concentration of the monomerized 8 kDa fragment increases from 32 to 48 µM, the lag phase is slightly longer and the rate of amyloidogenesis begins to plateau, if not decrease, suggesting the onset of off-pathway aggregation. The lag phase associated with the aggregation of the 5 kDa peptide also shortens as the concentration increases (Supplementary Figure 7). However, a minimum in the nucleation period or a maximum in the extension rate was not observed at the highest concentration employed (40 µM).
Another feature of a nucleated polymerization aggregation reaction is that the lag phase can be dramatically shortened or even eliminated by adding preformed fibril seeds that bypass the requirement for nucleation (67). To determine whether gelsolin fragment amyloidogenesis exhibits this feature, increasing quantities of either 5 or 8 kDa gelsolin fragment preformed fibril seeds were added to the amyloid fibril formation assay. The addition of increasing quantities of 8 kDa seeds was shown to significantly decrease the nucleation period associated with 8 kDa gelsolin fragment amyloidogenesis (Figure 4A). 5 kDa seeds were shown to similarly decrease the 8 kDa gelsolin fragment amyloidogenesis nucleation period (Figure 4B). In addition, 5 kDa gelsolin fragment fibril formation was accelerated and the nucleation period was proportionately decreased by addition of either 5 kDa and 8 kDa preformed fibrillar seeds (Figures 4C and D, respectively).
The time-dependence of the morphology of the aggregates formed by the 8 kDa gelsolin fragment (16 µM) was examined by AFM at three different time points in the nucleated polymerization aggregation reaction (Figure 5A). Figure 5B depicts the morphology of the aggregates formed at the beginning of the growth phase (4 h, 1st red data point). Figure 5C depicts the 8 kDa fibrillar aggregate morphology at the completion of the growth phase (12 h, 2nd red data point). Figure 5Dshows the laterally associated fibrils that are typically formed after a long stationary phase (64 h, 3rd red data point). These aggregate morphologies as a function of time are precisely those expected from a nucleated polymerization fibrillization reaction. Unfortunately, the 5 kDa fibrils do not adhere to the mica, and therefore could not be studied analogously by AFM.
The amino acid substructure of the 8 and 5 kDa gelsolin fragments comprising gelsolin’s cross-β-sheet amyloid core is unknown. Since the portion of a peptide comprising the cross-β-sheet core of an amyloid fibril is resistant to proteolysis (68, 69), the sequence comprising the gelsolin amyloid fibril core was probed by the established method of limited proteolytic digestion using proteinase K, a heat and chaotrope stable protease that is largely non-specific, but does exhibit some preference for cleaving after aliphatic and aromatic amino acids (70, 71).
The proteinase K concentration and incubation period was first optimized via SDS-PAGE, and it was determined that an overnight incubation (16 h) of the fibrils using a proteinase K concentration of 3 µg/mL was optimal (Supplementary Figure 8). After digestion of the 8 kDa fibrils for 1 h, mass spectrometry revealed the presence of two peptides, one corresponding to a mass of 7860, which is the mass of the 8 kDa (amino acids 173–242) peptide itself, and one corresponding to a mass of 6831, which is the 173–232 60-mer sequence. After overnight digestion, mass spectrometry revealed the presence of another peptide exhibiting a mass of 6595, corresponding to a 58-mer (amino acids 173–230), as the major product. The sequences of these peptides are shown in Table 1. As a control, the freshly monomerized 8 kDa gelsolin fragment (40 µM) was subjected to the same concentration of proteinase K, and it was efficiently degraded by proteinase K, as no peaks were observed by ESI-MS (data not shown).
When the 5 kDa gelsolin fragment amyloid fibrils were digested overnight, no other band besides the 5 kDa fragment appeared in the SDS-PAGE gel, and mass spectrometry revealed only one species of mass 6067, corresponding to the 5 kDa (173–225) amyloidogenic fragment itself. Once again, the experimental masses of these peptides were two less than the calculated theoretical masses of the reduced peptides due to disulfide bond formation between the cysteines at positions 188 and 201. These data together suggest that the amyloid core is the 58-mer corresponding to amino acids 173–230.
A number of complementary approaches have been utilized to characterize the relative amyloidogenicities of the plasma gelsolin 68, 8, and 5 kDa fragments observed in FAF patients, as well as the mechanism by which the 8 and 5 kDa amyloidogenic fragments aggregate. While the C68 fragment exhibits a slight increase in TfT fluorescence with time, and the sedimentation velocity data convincingly demonstrated reversible oligomerization, no evidence for C68 amyloid fibril formation could be obtained, despite attempting a wide variety of experiments that often afford amyloid fibrils. In contrast, the 8 and 5 kDa FAF-associated gelsolin fragments were observed to aggregate into amyloid fibrils under a wide variety of conditions, by unseeded and seeded aggregation reactions, and by variable agitation methods. Concentration dependent aggregation reactions and seeding assays strongly suggest that the 8 and the 5 kDa gelsolin FAF-associated fragments form amyloid fibrils by a nucleated polymerization mechanism. In this mechanism, the highest energy species on the aggregation pathway is the nucleus, and after nucleation is achieved, fibril elongation occurs in a thermodynamically favorable fashion. Our kinetic observations regarding the 8 kDa gelsolin fragment amyloidogenesis reaction, when considered together with the visualization of oligomers by AFM at the beginning of the rapid growth phase of the polymerization, strongly support a nucleated polymerization mechanism.
When amyloid from FAF patients is isolated and characterized, over 85% of the peptides composing the fibrils are found to be the 8 kDa fragment, with the remainder being the 5 kDa fragment (42). Notably, no C68 is observed in amyloid taken from FAF patients. This data is strictly consistent with the lack of C68 in amyloid extracted from a transgenic gelsolin amyloidosis mouse model (19). In the muscle of the D187N FAF mouse model, the 8 kDa fragment deposits as early as 6 weeks of age, whereas the 5kDa fragment is not observed until about twelve months of age. All of these observations are consistent with our in vitro data, demonstrating that the 8 kDa fragment is more amyloidogenic than the 5 kDa fragment. Once the highly amyloidogenic 8 kDa fibril extension reaction begins in vivo, it could seed the fibrillization of the lower concentration of the 5 kDa monomer, as well as the 8 kDa monomer. This hypothesis, supported by the results outlined herein, offers an explanation for how amyloid load increases and likely explains why the 5 kDa peptide is not observed until later in the course of the disease – its in vivo concentration is lower and it forms amyloid very slowly, if at all, until seeded by 8 kDa fibrils.
Integrating the data presented here with the biological data obtained from the FAF mouse model, it becomes possible to draw some conclusions about the pathogenesis of gelsolin amyloidosis and to make some predictions. Full-length D187N/Y plasma gelsolin is cleaved within domain 2 in the trans Golgi network, where furin is active, to form the C68 fragment. While it is possible that C68 forms oligomers, as evidenced by the AUC and the TfT data, these oligomers are not fibrillar in nature and it is unlikely they form in vivo, owing to the lack of an air/water interface.
It seems clear that the 8 and 5 kDa fragments are formed by the cleavage of C68 by active MT1-MMP (and possibly other proteases) on cells making C68, or on nearby cells, because in the mouse model, only the tissues that synthesize C68 cleave it into the amyloidogenic 8 and 5 kDa fragments, and only these tissues deposit 8 kDa gelsolin amyloid (19). It appears that high concentrations of sulfated oligosaccharides accelerate 8 kDa gelsolin aggregation, as the increasing population of glycosaminoglycans and 8 kDa amyloid correlate as the mice age (19, 24). It is quite interesting that the core of the 8 kDa gelsolin fragment amyloid fibril is very similar to that of the 5 kDa fragment, yet the 8 kDa fragment aggregates faster, suggesting that the five extra amino acids in the β-sheet core (amino acids226–230) and/or a subsequence not integral to the 8 kDa cross-β-sheet structure make it aggregate faster.
The generation of the amyloidogenic 8 kDa plasma gelsolin fragment, and to a lesser extent the 5 kDa fragment, is clearly toxic to cells, apparently owing to the process of amyloidogenesis, although the precise mechanism of proteotoxicity is anything but clear (19). The generation of these fragments and likely the subsequent process of amyloidogenesis ultimately lead to intracellular inclusion bodies made up of multiple proteins that contribute to muscle cell death, as observed in the FAF mouse model (19).
These biophysical studies, when considered in concert with the data from the FAF mouse model, support the hypothesis that the administration of an MT1-MMP inhibitor that is able to reach the muscle tissue should be effective at inhibiting the cleavage of C68 that forms the 8 and 5 kDa amyloidogenic fragments. Because it has been demonstrated that the C68 fragment is not amyloidogenic and does not appear to play a major part in FAF pathogenesis, it is likely that this treatment strategy would be effective in ameliorating FAF.
The authors thank Colleen Fearns and Lesley Page for helpful discussions and help with manuscript preparation, Huiqiao Susan Sun and Helen Lu Yin for full length plasma gelsolin construct and purification advice, and M. R. Ghadiri for use of his AFM.
†This research was supported by NIH Grant AG018917, The Skaggs Institute for Chemical Biology, and the Lita Annenberg Hazen Foundation
Supporting Information Available
Supplementary Figures are available, as referred to in the text. (1) Representative trace from the final reverse phase HPLC purification step of the 8 kDa gelsolin fragment resulting from CNBr cleavage. (2) Example LC-MS data from the purified 8 kDa gelsolin fragment. (3) SDS-PAGE depiction of the purification methodology of C68. (4) SDS-PAGE depiction of final purification step of full length plasma gelsolin. (5) Relative amyloidogenicity of the 68, 8 and 5 kDa gelsolin fragments as measured by the increase in TfT fluorescence in the plate reader TfT aggregation assay at pH 7.0 and 7.4. (6) AFM analysis of a solution of C68 (4 µM) that was mixed by overhead rotation (24 rpm) at 37 °C overnight. The AFM is compared to a negative control of buffer alone and a positive control of a solution of 8 kDa (4 µM) that was mixed by overhead rotation (24 rpm) at 37 °C overnight. (7) Concentration dependence of the 5 kDa plasma gelsolin fragment amyloidogenesis reaction analyzed at pH 6.5 using the TfT plate reader assay. Concentrations of 5 kDa fragment ranged from 2 µM to 40 µM, as indicated. (8) SDS-PAGE of proteinase K digested 8 and 5 kDa fibrils. This material is available free of charge via the Internet at http://pubs.acs.org.