Discovery and characterization of a mammalian amyloid disaggregation activity

Abstract

The formation of amyloid, a cross-β-sheet fibrillar aggregate, is associated with a variety of aging-associated degenerative diseases. Herein, we report the existence of a mammalian amyloid disaggregase activity that is present in all tissues and cell types tested. Homogenates from mammalian tissues and cell lines are able to disaggregate amyloid fibrils composed of amyloid β (Aβ)_1–40 or the 8 kDa plasma gelsolin fragment. The mammalian disaggregase activity is sensitive to proteinase K digestion and can be uncoupled from proteolysis activity using a protease inhibitor cocktail. Amyloid disaggregation and proteolysis activities are remarkably resistant to changes in temperature and pH. Identification and manipulation of the proteins responsible for the amyloid disaggregation/degradation activities offers the possibility of ameliorating aggregation-associated diseases.

Introduction

Maintenance of the proteome in a functional, nonpathogenic state is a critical component of organismal health.^1,2 One facet of this constitutive process is the elimination of misfolded and misassembled/aggregated proteins. Many age-related diseases (including Alzheimer's disease, the transthyretin amyloidoses, Parkinson's disease, Huntington's disease, and familial amyloidosis of Finnish type) are associated with amyloidogenesis, a process by which proteins misassemble or misfold and misassemble into both soluble and insoluble cross-β-sheet fibrillar aggregates called amyloid.^3–18 Amyloid is toxic when applied exogenously to or when formed within cultured cells.^19–32 Mouse and C. elegans models of amyloid diseases show that the cytotoxicity associated with amyloidogenesis increases with organismal age.^22,33 Postnuclear supernatant (PNS) from C. elegans has recently been shown to disaggregate Aβ_1–40 amyloid fibrils, which are linked to Alzheimer's disease.^5,10,34,35 C. elegans disaggregation and proteolysis activities can be uncoupled by protease inhibition or by heating the worm homogenate to 80°C, both of which eliminate proteolysis but not disaggregation, indicating that nematode disaggregation activity does not rely on proteolysis.^34,35

In this article, we test the hypothesis that mammals also have a disaggregase activity(ies) that can be uncoupled from proteolysis. Disaggregation was studied using Aβ_1–40 fibrils (associated with Alzheimer's disease)^5,10 and 8 kDa plasma gelsolin fragment fibrils (linked to familial amyloidosis of Finnish type²²) prepared in vitro. Notably, homogenates from all mouse tissues and human cell lines evaluated exhibit amyloid disaggregase and proteolysis activities, highlighting the apparent ubiquity and necessity of these activities for the maintenance of the proteome, or proteostasis.¹ As is true for the C. elegans activities, mammalian disaggregase and proteolysis activities are notably resistant to thermal denaturation and changes in pH, although proteinase K digestion readily inactivates these activities, demonstrating that they are protein based. Mammalian disaggregase and proteolysis activities can be uncoupled in the presence of Roche Complete EDTA-Free Protease Inhibitor Cocktail, suggesting that mammalian amyloid disaggregation activity is not driven by proteolysis. Identifying the molecular machinery responsible for the mammalian disaggregase activity and, ultimately, adapting this machinery offer the prospect of ameliorating age-related degenerative diseases in which cytotoxicity is genetically linked to the process of amyloidogenesis.^{5,10–19,22,30,33–37}

Results

Mouse tissue PNS dose-dependently disaggregates Aβ_1–40 and 8 kDa gelsolin fibrils

We have previously shown that supernatant from homogenized C. elegans subjected to 3000 rpm centrifugation (PNS) is able to disaggregate Aβ_1–40 fibrils in vitro.^34,35 To ascertain whether mammalian tissue also possesses a disaggregase activity, and to determine whether this activity can extend to other amyloid aggregates, a kinetic disaggregation assay was used to monitor Aβ_1–40 and 8 kDa gelsolin disaggregation kinetics. Amyloid was quantified by the fluorescence intensity of thioflavin T (ThT), a dye that undergoes an increase in fluorescence quantum yield when bound to amyloid.³⁸ Mouse tissues were Dounce homogenized, and lysates were clarified by low-speed centrifugation (1,000 g, 3 min). The resulting PNS was added to Aβ_1–40 or 8 kDa gelsolin fibrils prepared as described in the “Materials and Methods” section. PNS from murine brain, heart, kidney, and liver disaggregates both Aβ_1–40 fibrils (65 μg/mL) [Fig. 1(A)] and 8 kDa gelsolin amyloid fibrils (31 μg/mL) [Fig. 1(B)].

Figure 1.

PNS from mouse tissues disaggregates amyloid fibrils. (Error bars indicate standard deviations in three independent disaggregation experiments; * denotes P < 0.05 and ** denotes P < 0.01). A: Mouse brain, heart, kidney, and liver PNS (31 μg/mL total protein) disaggregate Aβ_1–40 fibrils (86 μg/mL), as compared with BSA (31 μg/mL) or buffer alone controls. B: Mouse brain, heart, kidney, and liver PNS (45 μg/mL total protein) disaggregate 8 kDa gelsolin fibrils (31 μg/mL), as compared with BSA (45 μg/mL) or buffer alone controls. C: The extent of Aβ_1–40 fibril (65 μg/mL) disaggregation (measured at 40 h) depends on the total protein concentration of mouse tissue PNS (3–30 μg/mL total protein), with higher concentrations leading to more disaggregation. All samples have a statistically significantly lower fluorescence than the buffer alone control except for BSA and 3 μg/mL liver PNS, with P < 0.01. D: The extent of 8 kDa gelsolin fibril (31 μg/mL) disaggregation is similarly dependent on total PNS protein concentration (3–30 μg/mL total protein). All samples have a statistically significantly lower fluorescence than the buffer alone control except for BSA, with P < 0.01. E: AFM confirms a decrease in the amount of Aβ_1–40 fibrils (65 μg/mL) following disaggregation by mouse heart PNS (30 μg/mL total protein) (bottom panels), when compared with a buffer alone control (top panels). F: AFM demonstrates a decrease in 8 kDa gelsolin fibrils (31 μg/mL) following disaggregation by mouse kidney PNS (30 μg/mL total protein) (bottom vs. top panels). In each case, a representative image of the fewest fibrils observed in the 10 or more fields examined is shown in the left panels, and a representative image of the most fibrils observed is shown in the right panels.

The dose dependency of the disaggregase activity was assessed using three concentrations of total PNS protein for each tissue (3, 10, and 30 μg/mL). The disaggregase activity is qualitatively dose dependent, in that higher concentrations of total PNS protein result in more Aβ_1–40 and 8 kDa gelsolin disaggregation [Fig. 1(C,D)], as measured by final ThT fluorescence after 40 h. The linear range of the disaggregation assay was thoroughly characterized across the range of PNS total protein concentrations tested and with several different Aβ_1–40 and 8 kDa gelsolin fibril preparations [Supporting Information Fig. S1]. Different fibril preparations can exhibit different linear ranges of disaggregation; for this reason, in all experiments where the extent of disaggregation is compared, the linear range of disaggregation for the fibril preparation used is carefully characterized, and ratios of [PNS]:[gelsolin] used are within this range.

Atomic force microscopy (AFM) confirms a substantial decrease in the number of Aβ_1–40 and 8 kDa gelsolin fibrils following treatment with mouse heart and kidney PNS [Fig. 1(E,F), respectively]. A minimum of 10 fields were examined for each sample. Shown are representative images of fields with the most fibrils (right panels) and fields with the fewest fibrils (left panels) for each PNS-treated sample (bottom panels) and buffer alone control (top panels). Although there is a dramatic decrease in the number of fibrils following disaggregation with mouse tissue PNS, some fields still contain small fibrils or other aggregates, which explains the residual ThT fluorescence at 40 h.

To support the claim that a decrease in ThT fluorescence signifies disaggregation, as opposed to an effect of PNS on the fluorophore, Aβ_1–40 aggregates were pretreated with glutaraldehyde to prevent disaggregation. Addition of mouse brain PNS (5 μg/mL total PNS protein) to chemically crosslinked Aβ_1–40 amyloid fibrils (32 μg/mL) attenuates the decrease in ThT fluorescence associated with fibril disaggregation [Supporting Information Fig. S2(A)], supporting the conclusion that the kinetic assay monitors disaggregation. PNS alone (from all mouse tissues and cultured cells tested) contributes minimally to the absolute ThT signal [Supporting Information Fig. S2(B)], further supporting the claim that ThT fluorescence monitors the amount of Aβ_1–40 fibrils and is negligibly affected by the components of PNS.

To assess whether the disaggregase activity relies on proteins, mouse kidney PNS (300 μg/mL total protein) was pretreated with proteinase K (PK, 2 μg/mL) for 2 h. Phenylmethylsulfonyl fluoride (PMSF, 1 mM) was then added to inhibit remaining PK activity. Samples were diluted (10 μg/mL total PNS protein concentration) and added to Aβ_1–40 fibrils (65 μg/mL). PK pretreatment eliminates the disaggregase activity [Fig. 2], suggesting that proteins or a protein complex(es) within the PNS are responsible for amyloid fibril disaggregation.

Figure 2.

PNS-mediated amyloid disaggregase activity is sensitive to proteolytic digestion. Proteinase K (PK) (2 μg/mL) pretreatment of mouse kidney PNS (300 μg/mL total protein) eliminates its ability (at 10 μg/mL total protein) to disaggregate Aβ_1–40 fibrils (65 μg/mL, blue), when compared with untreated control (green).

To discern the subcellular localization of the disaggregase activity, 100 μL mouse heart PNS (1 mg/mL) was subjected to centrifugation at 5,000 g, 50,000 g, or 150,000 g for 1 h at 4 °C. The resulting supernatants were separated from pellets, and the pellets were resuspended in 100 μL PBS. Both supernatants and pellets were subjected to sonication for 1 h, after which protein concentrations were determined by BCA assay. Fractionated PNS (5 μg/mL total protein) was applied to Aβ_1–40 fibrils (65 μg/mL). All supernatants and pellets disaggregate Aβ_1–40 fibrils [Supporting Information Fig. S3], indicating that disaggregases are present in both purely cytosolic (e.g., S150K; no ER marker, see Western blot) and membrane-containing (e.g., P150K; no cytosolic marker, see Western blot) subcellular fractions.

To test the hypothesis that the mammalian amyloid disaggregase activity is an ATP-driven process, liver PNS (200 μg/mL) was pretreated with 1.5 U apyrase (high) for 30 min at 30 °C, conditions sufficient to hydrolyze nearly all accessible ATP [Supporting Information Fig. S4(A)]. ATP depletion has no effect on liver PNS-mediated disaggregation [Supporting Information Fig. S4(B)], indicating that mammalian disaggregase activity is likely not ATP dependent or that ATP inaccessible to apyrase (e.g., vesicular ATP) provides energy for the disaggregase activity.

Human cell PNS dose dependently disaggregates Aβ_1–40 and 8 kDa gelsolin fibrils

Experiments were next performed to analyze the amyloid disaggregation capability of human cell line PNS. Cell lines derived from human kidney, liver, and neurons (HEK 293, Huh7, and IMR-32, respectively) were cultured to 90% confluence, manually scraped, and Dounce homogenized. The lysate was clarified by low-speed centrifugation (1,000 g, 3 min), and the resulting PNS (3, 10, and 30 μg/mL total protein) was applied to Aβ_1–40 fibrils (65 μg/mL) or 8 kDa gelsolin fibrils (31 μg/mL). PNS from HEK 293, Huh7, and IMR-32 cells disaggregates both Aβ_1–40 [Fig. 3(A)] and 8 kDa gelsolin fibrils [Fig. 3(B)]. Human cell PNS disaggregation kinetics are similar to those of mouse tissue PNS, and the extent of disaggregation of Aβ_1–40 [Fig. 3(C)] and 8 kDa gelsolin fibrils [Fig. 3(D)] is qualitatively dependent on the concentration of total PNS protein added. A decrease in the number of fibrils after 40 h of disaggregation is confirmed by AFM. Representative fields with the most (right panels) and fewest (left panels) fibrils are shown for buffer alone controls (top panels) and cultured cell PNS-treated fibrils (bottom panels), with a minimum of 10 fields examined for each sample [Fig. 3(E,F)].

Figure 3.

PNS from human cell lines disaggregates amyloid fibrils. (Error bars indicate standard deviations in three independent disaggregation experiments; * denotes p < 0.05 and ** denotes p < 0.01). A: HEK 293, Huh7, and IMR-32 PNS (30 μg/mL total protein) disaggregates Aβ_1–40 fibrils (65 μg/mL), when compared with BSA (30 μg/mL) or buffer alone controls. B: HEK 293, Huh7, and IMR-32 PNS (30 μg/mL total protein) disaggregates 8 kDa gelsolin fibrils (31 μg/mL), as compared with BSA (30 μg/mL) or buffer alone controls. C: The extent of Aβ_1–40 fibril (65 μg/mL) disaggregation (measured at 40 h) depends on the total protein concentration of human cell PNS (3–30 μg/mL total protein), with higher concentrations leading to more disaggregation. All samples have a statistically significantly lower fluorescence than the buffer alone control except for BSA, with P < 0.01. D: The extent of 8 kDa gelsolin disaggregation (31 μg/mL) also depends on the total protein concentration of human cell PNS (3–30 μg/mL total protein). All samples have a statistically significantly lower fluorescence than the buffer alone control except for BSA and 3 μg/mL Huh7 PNS, with P < 0.05. E: AFM confirms a decrease in the amount of Aβ_1–40 fibrils (65 μg/mL) following disaggregation by Huh7 PNS (30 μg/mL total protein) (bottom panels), when compared with a buffer alone control (top panels). F: AFM demonstrates a decrease in 8 kDa gelsolin fibrils (31 μg/mL) following disaggregation by HEK 293 PNS (30 μg/mL total protein) (bottom vs. top panels). In each case, a representative image of the fewest fibrils observed in the 10 or more fields examined is shown in the left panels, and a representative image of the most fibrils observed is shown in the right panels.

To determine whether autophagy is responsible for PNS-mediated Aβ_1–40 fibril disaggregation, intact Huh7 cells were treated for 8 h with an autophagy inducer (50 pM rapamycin), an autophagy inhibitor (5 mM 3-methyladenine), or their respective vehicle controls (DMSO and EtOH). PNS was then obtained from drug- and vehicle-treated Huh7 cells. Autophagy induction or inhibition was verified by changes in the ratio of LC3-II to LC3-I, as revealed by Western blot. An increase in the LC3-II:LC3-I ratio indicates increased autophagy, and a decrease indicates decreased autophagy. Neither induction nor inhibition of autophagy discernibly affect the disaggregase activity of Huh7 PNS, indicating that autophagic disaggregation is not principally responsible for Aβ_1–40 fibril disaggregation [Supporting Information Fig. S5(A,B)].

Mammalian PNS dose-dependently degrades Aβ_1–40 and 8 kDa gelsolin, depleting the pelletable fibril population

Following 40 h of mouse tissue PNS-mediated Aβ_1–40 fibril (65 μg/mL) or 8 kDa gelsolin fibril (31 μg/mL) disaggregation, samples were denatured in 8 M guanidine hydrochloride with sonication in a Fisher Scientific FS60 Sonic Cleaner. Using RP-HPLC, the resulting Aβ_1–40 peak areas were compared with that of the Aβ_1–40 fibrils + buffer control (no PNS) to quantify the extent of Aβ_1–40 proteolysis [Supporting Information Fig. S6(A)]. Mouse tissue PNS (3, 10, and 30 μg/mL total protein) degrades Aβ_1–40 in a dose-dependent fashion, with higher concentrations of total PNS protein leading to increased Aβ_1–40 proteolysis [Fig. 4(A)]. Following mouse tissue PNS-mediated disaggregation (3, 10, and 30 μg/mL total protein) and denaturation in 8 M guanidine hydrochloride, the total amount of the 8 kDa plasma gelsolin fragment was also quantified by RP-HPLC [Supporting Information Fig. S6(B)], revealing that its proteolysis is also dose dependent [Fig. 4(B)].

Figure 4.

Mammalian PNS degrades Aβ_1–40 and 8 kDa gelsolin from fibrils, depleting high-molecular weight species. (Error bars indicate standard deviations in three independent proteolysis experiments; * denotes p < 0.05 and ** denotes p < 0.01.) A: Quantification of Aβ_1–40 reversed phase liquid chromatography peak areas indicates that mouse tissue PNS degrades Aβ_1–40 (65 μg/mL), and the extent of proteolysis is proportional to the concentration of total PNS protein (3–30 μg/mL total protein). All 10 and 30 μg/mL PNS samples have statistically significantly less Aβ_1–40 remaining, with P < 0.05. B: Quantification of 8 kDa gelsolin reversed phase liquid chromatography peak areas shows that mouse tissue PNS (3–30 μg/mL total protein) also degrades 8 kDa gelsolin (31 μg/mL) in a dose-dependent fashion. All 10 and 30 μg/mL PNS samples have statistically significantly less gelsolin remaining, with P < 0.05. C: Human cell PNS (5 μg/mL total protein) depletes the pelletable (200,000 g for 1 h) population (pels) of Aβ_1–40 but does not affect the supernatant Aβ_1–40 population (sups). D: Removal of soluble Aβ_1–40 following a 200,000 g spin at t = 0 (sup discarded, blue) does not repopulate the supernatant pool from a 200,000 g spin after a 40-h incubation period (cf. short blue bar, sups, to short green bar, sups).

Two potential models of PNS-mediated disaggregation and degradation exist. In model 1, the amyloid disaggregase machinery pulls monomers and possibly oligomers from fibrils, rendering them sensitive to proteolysis. In model 2, proteases degrade soluble monomers and possibly oligomers, depleting the soluble population below the critical concentration for fibril formation, taking advantage of the fibril dissociation equilibrium to drive amyloid fibril disaggregation.

To investigate which Aβ_1–40 population (fibril vs. soluble) is degraded by human cell PNS, Aβ_1–40 fibrils (43 μg/mL) disaggregated by human cell PNS (5 μg/mL total protein) were spun at 200,000 g for 1 h. Supernatants were separated from pellets, and both were dissolved in 8 M guanidine hydrochloride with sonication for 3 h. Aβ_1–40 RP-HPLC peak areas indicate that the pelletable, high-molecular weight population is depleted following disaggregation, while the soluble population remains constant [Fig. 4(C)]. These data are consistent with both models. Since amyloid is protease resistant,^39–41 it is unlikely that the fibrils themselves are proteolyzed by human cell PNS. Model 2 could explain these results if the soluble Aβ_1–40 population is proteolyzed but repopulated by spontaneous fibril dissociation to maintain the critical concentration of soluble Aβ_1–40. To test whether depletion of soluble monomers and possibly oligomers by PNS proteases can drive spontaneous disaggregation of fibrils on the timescale observed, preaggregated Aβ_1–40 was spun at 200,000 g for 1 h. Resulting supernatants were discarded from half of the samples, and the pellets (from the supernatant-discarded samples) or the supernatants and pellets (from the supernatant-retained samples) were resuspended with sonication and allowed to equilibrate for 40 h in aggregation buffer (50 mM sodium phosphate, pH 7.4, 150 mM NaCl, 0.02% NaN₃). Samples were again spun at 200,000 g for 1 h, and the amount of Aβ_1–40 present in the supernatants and pellets was quantified by RP-HPLC [Fig. 4(D)]. Consistent with model 1, the amount of Aβ_1–40 present in the pelletable fraction (fibrillar fraction) remains constant with (sup discarded, blue, pels) or without (sup retained, green, pels) prior depletion of soluble Aβ_1–40 [Fig. 4(D)]. This experiment demonstrates that the spontaneous disassembly of fibrils is not fast enough to repopulate the monomer to the critical concentration during the 40 h disaggregation assay [cf., sup discarded, blue, sups to sup retained, green, sups in Fig. 4(D)]. Within the timescale of the disaggregation assay, the initial equilibrium between fibrils and monomer, as determined by both aggregation and disassembly rates, is not reestablished, eliminating model 2 as a possibility. This result, coupled with the protease inhibitor results described below, demonstrates that proteolysis alone, in the absence of disaggregase activity, is not responsible for mammalian PNS-mediated disaggregation.

Protease inhibitors uncouple PNS-mediated disaggregation from proteolysis

The C. elegans amyloid disaggregation and degradation activities can be uncoupled using protease inhibitors or heat pretreatment of worm homogenate. In the presence of PMSF, a serine protease inhibitor, or Roche Complete EDTA-Free Protease Inhibitor Cocktail (PIC), known to inhibit serine proteases, metalloproteases, cysteine proteases, and other proteases, C. elegans PNS-mediated proteolysis of Aβ_1–40 is inhibited, but disaggregation is not.^34,35

To determine whether the disaggregation and proteolysis activities from mammalian PNS can be uncoupled by protease or proteasome inhibitors, Huh7 PNS (30 μg/mL total protein) was added to Aβ_1–40 fibrils (65 μg/mL) in the presence of protease and proteasome inhibitors (10 μM epoxomicin, 20 μM phosphoramidon, 1 mM PMSF, and PIC (one tablet per 50 mL of solution). Proteasome (epoxomicin), metalloprotease (phosphoramidon), and serine protease (PMSF) inhibitors have little or no effect on Huh7 PNS-mediated Aβ_1–40 disaggregation [Fig. 5(A)] or proteolysis [Fig. 5(B)]. In this way, mammalian PNS-mediated proteolysis differs from that of C. elegans, which is strongly inhibited by PMSF.^34,35 Only PIC, a mixture of protease inhibitors known to inhibit serine proteases, metalloproteases, cysteine proteases, and other proteases, prevents Huh7 PNS-mediated Aβ_1–40 proteolysis [Fig. 5(B), cyan bar]. Notably, the presence of PIC only minimally affects the extent of fibril disaggregation after 40 h [Fig. 5(A), cyan bar], compared to complete inhibition of Aβ proteolysis [Fig. 5(B), cyan bar]. Thus, PIC successfully uncouples Aβ_1–40 disaggregation from proteolysis, supporting the idea that there is a distinct mammalian amyloid disaggregase activity that can function independently of proteolysis.

Figure 5.

Mammalian amyloid disaggregation and proteolysis activities are uncoupled by a protease inhibitor cocktail. (Error bars indicate standard deviations in three independent disaggregation or proteolysis experiments; * denotes p < 0.05 and ** denotes p < 0.01. A: Phosphoramidon (PA, 20 μM) and Roche Complete EDTA-Free Protease Inhibitor Cocktail (PIC, at 1 tablet per 50 mL of solution) slightly reduce disaggregation of Aβ_1–40 fibrils (65 μg/mL) by Huh7 PNS (30 μg/mL total protein), increasing the average relative ThT fluorescence at 40 h from 0.14 (Huh7 PNS, green) to 0.19 (Huh7 PNS, PA, blue) and 0.16 (Huh7 PNS, PIC, cyan). Epoxomicin (epox, 10 μM) and PMSF (1 mM) do not significantly affect disaggregation of Aβ_1–40 fibrils by Huh7 PNS. All samples have a statistically significantly lower fluorescence than the buffer alone control, with P < 0.01. B: Only Roche Complete EDTA-Free Protease Inhibitor Cocktail (PIC, cyan) prevents Huh7 PNS-mediated Aβ_1–40 proteolysis under the same conditions.

Mammalian amyloid disaggregase and proteolysis activities are heat resistant

Heat pretreatment of worm PNS at 80° C significantly reduces Aβ_1–40 proteolysis activity but not disaggregation activity, whereas both activities are inactivated at 95° C.^34,35 To assess the thermal sensitivity of the mammalian PNS-mediated disaggregation and proteolysis activities, mouse heart PNS (300 μg/mL total protein) was pretreated for 2 h at temperatures ranging from 37 to 95° C. These thermal pretreatments had a small but statistically significant effect on disaggregase activity but did not fully inhibit fibril disaggregation [Fig. 6(A)] by 10 μg/mL PNS. Similarly, thermal pretreatments were unable to dramatically inhibit PNS-mediated Aβ_1–40 proteolysis [Fig. 6(B)], underscoring the robustness of both of these activities in mammalian tissue. The proteins or protein complexes responsible for disaggregation and proteolysis appear to be remarkably thermally stable. There is precedent for this, in that mammalian RNase is routinely boiled to remove activity from DNase impurities.⁴² Another explanation for the apparent thermal stability is that the disaggregase(s) and protease(s) may efficiently refold with the aid of chaperones and other refolding machinery present in the PNS between the time boiling ends and the disaggregation assay begins (∼30 min).

Figure 6.

Mammalian amyloid disaggregase and proteolysis activities are resistant to high-temperature pretreatment. (Error bars indicate standard deviations in three independent disaggregation or proteolysis experiments; * denotes p < 0.05 and ** denotes p < 0.01.) A: Pretreatment (2 h) of mouse heart PNS (10 μg/mL, low) at temperatures up to 95 °C decreases disaggregation of Aβ_1–40 fibrils (65 μg/mL) but does not completely inhibit the disaggregase activity. There is a statistically significantly lower ThT signal from fibrils to which a higher concentration of PNS was added (10 μg/mL, high), corroborating that the ratio of 65 μg/mL Aβ_1–40 fibrils: 10 μg/mL mouse heart PNS is within the linear range of the disaggregation assay. All samples have a statistically significantly lower fluorescence than the buffer alone control, with P < 0.01. B: Pretreatment (2 h) of mouse heart PNS (10 μg/mL) at temperatures up to 95 °C does not prevent proteolysis of Aβ_1–40 (65 μg/mL). All samples have a statistically significantly lower Aβ_1–40 peak area than the buffer alone control, with P < 0.05.

Mammalian amyloid disaggregase and proteolysis activities are sensitive to pH changes

In a further attempt to characterize the mammalian disaggregation and proteolysis activities, HEK 293 PNS (1 mg/mL total protein) was dialyzed for 16 h at 4° C into buffers at pHs ranging from 4.0 to 10.0. Dialyzed samples were then added to Aβ_1–40 fibrils (65 μg/mL) at a final total PNS protein concentration of 10 μg/mL. Although low pH (and to a lesser extent high pH) pretreatment attenuates HEK 293 PNS-mediated Aβ_1–40 fibril disaggregation, Aβ_1–40 proteolysis is inhibited similarly. The amount of Aβ_1–40 remaining after 40 h is always inversely proportional to the observed extent of disaggregation; thus, there is a linear correlation between ThT fluorescence and relative RP-HPLC Aβ_1–40 peak area (Fig. 7). This sensitivity to low pH pretreatment suggests that the proteins or protein complexes responsible for mammalian amyloid disaggregation and/or proteolysis have acidic isoelectric points, leading to partial pH denaturation as the pH is decreased. Because PNS is a crude preparation, it is likely that more than one type of disaggregase and proteolysis machinery exists within the PNS. Some of the proteins responsible for the amyloid disaggregation/proteolysis activities may have a basic isoelectric point and thus be denatured by higher pH pretreatment, explaining the slight sensitivity to high pH pretreatments. The tight linkage of disaggregation and proteolysis may be due to a physical linkage of proteins or protein complexes that carry out these activities. Additionally, because amyloid is known to be protease resistant,^39–41 it is likely that disaggregation is required for proteolysis, and this relationship would necessitate a linear correlation between disaggregation and proteolysis (Fig. 7).

Figure 7.

Mammalian amyloid disaggregase and proteolysis activities are sensitive to low and high pH pretreatment. Dialysis of HEK 293 PNS (1 mg/mL total protein) for 16 h at 4 °C into buffers at low pH (and to a lesser extent high pH) decreases its ability (at a final total PNS concentration of 10 μg/mL) to disaggregate Aβ_1–40 fibrils (65 μg/mL) and to degrade Aβ_1–40.

Discussion

In amyloid diseases, proteotoxicity is thought to arise from the process of amyloidogenesis.^{1–10,23,33–37,43–45} The prevention of amyloidogenesis by a kinetic stabilizer approach has recently been demonstrated to halt the progression of nervous system degeneration in familial amyloid polyneuropathy associated with the amyloidogenesis of transthyretin. Thus, it is logical that the removal of protein aggregates would be crucial for protein homeostasis.^1,34,35 It is expected that activities leading to the disaggregation of amyloid and the degradation of amyloidogenic peptides or proteins exist within all tissues and cell types, consistent with their recent discovery in C. elegans^34,35 and in all of the mammalian cell lines and tissues we have investigated thus far. Different cell types and tissues are subject to distinct proteomic pressures and aggregation-associated stresses, and the observed differences in disaggregase activity [Figs. 1(A,B) and 3(A,B)] may be necessary for successful tissue- and cell-specific protein homeostasis.

Mammalian amyloid disaggregase and proteolysis activities are carried out by proteins, as evidenced by the sensitivity of these activities to proteinase K digestion; however, these activities appear to be remarkably resistant to high temperature. This stability is likely necessary for successful responses to a variety of denaturing insults, including heat shock associated with fever. Amyloid disaggregase and proteolysis activities may be coupled, as evidenced by matched sensitivities to changes in pH. This coupling suggests that the proteins or protein complexes responsible for disaggregation and proteolysis may work in synchrony to detoxify aggregates, possibly as part of the same macromolecular complex. Only a mixture of protease inhibitors uncouples disaggregation from proteolysis, confirming that mammalian amyloid disaggregases can function independently of proteases. Given the established resistance of amyloid to proteolysis, it seems likely that the mammalian disaggregase activity precedes the proteolysis activity,^39–41 and the proteolysis activity is crucial for preventing reaggregation.³⁵

The discovery and partial characterization of the mammalian amyloid disaggregase activity(ies) are a step towards identification of its molecular underpinnings. Manipulation of the proteins or protein complexes responsible for amyloid detoxification has the potential not only to increase our understanding of the maintenance of the proteome, but also to prevent aging-associated amyloid diseases linked to tissue degeneration.

Materials and Methods

Preparation of Aβ_1–40 and 8 kDa gelsolin fibrils

Aβ_1–40 was synthesized on Wang solid phase resin using an FMOC-based chemical synthesis strategy, purified by reversed-phase high-performance liquid chromatography (RP-HPLC), and lyophilized, as previously described.²³ The lyophilate was dissolved at 1 mg/mL in 5 mM NaOH by sonication for 4 h, to monomerize Aβ_1–40. Freshly monomerized Aβ_1–40 (173 μg/mL) in aggregation buffer (50 mM sodium phosphate, pH 7.4, 150 mM NaCl, 0.02% NaN₃) was rotary agitated (24 rpm) for 3 days at 37 °C, yielding fibrils. In one experiment, fibrillar Aβ_1–40 (32 μg/mL) was crosslinked with 0.4% glutaraldehyde for 10 min at 37 °C. Crosslinking was terminated by addition of 100 mM Tris, pH 8.

The 8 kDa plasma gelsolin fragment (amino acids 173–242) was expressed as a recombinant protein fused to BCL-XL1/2, a peptide that strongly prefers inclusion body formation in E. coli. Purification was carried out according to previously described methods.⁴⁶ For monomerization, lyophilized 8 kDa gelsolin (1 mg/mL) was dissolved in 8M guanidine hydrochloride, 50 mM sodium phosphate, pH 7.4, and sonicated for 2 h. Gel filtration utilizing Superdex 30 resin buffer exchanged the gelsolin fragment into 50 mM sodium phosphate, pH 7.4, 100 mM NaCl. The monomer peak was collected, diluted to a concentration of 20 μM, and rotary agitated (24 rpm) for 24 h at 37°C, yielding fibrils.

Preparation of mammalian homogenates

C57BL/6 mouse tissues were Dounce homogenized in PBS and centrifuged at 1,000 g for 3 min. The pellets were discarded, and total protein concentration of the resulting PNSs was determined using the Pierce BCA Assay. HEK 293, Huh7, and IMR-32 cells were grown to 90% confluency in 10 cm dishes in Dulbecco's Modified Eagle Medium supplemented with 10% fetal bovine serum at 37 °C. Cells were washed with PBS and manually scraped from the dish. They were then Dounce homogenized in PBS and centrifuged at 1,000 g for 3 min. The pellets were discarded, and the total protein concentration of the resulting PNS was determined using the Pierce BCA Assay. Proteinase K (PK) pretreatment of kidney PNS (300 μg/mL total protein) was accomplished by digestion with PK (2 μg/mL) for 2 h at 37 °C. PMSF (1 mM) was then added to inhibit remaining protease activity.

Kinetic disaggregation assay

Aβ_1–40 and 8 kDa gelsolin fibrils, prepared as described above, were sonicated for 30 min to afford a uniform fibril length distribution of 50–100 nm.³⁹ The kinetic disaggregation assay was conducted in aggregation buffer (50 mM sodium phosphate, pH 7.4, 150 mM NaCl, 0.02% NaN₃) containing ThT (40 μM). Mammalian tissue PNS (3–45 μg/mL total protein, as specified) or human cell PNS (3–45 μg/mL total protein, as specified) was added to the specified concentration of sonicated Aβ_1–40 fibrils or sonicated 8 kDa gelsolin fibrils in Corning 96-well clear bottom plates (final volume 100 μL/well). Every 10 min, plates were shaken for 5 s, and fluorescence (excitation at 440 nm, emission at 485 nm) was monitored using a Spectramax Gemini EM fluorescence plate reader. For quantitative analysis of disaggregation, relative thioflavin T fluorescence at 40 h is defined as absolute fluorescence at 40 h divided by absolute fluorescence at 0 h for each sample, which corrects for slight differences in sample volumes (pipetting error) and slight differences in fluorescence from higher concentrations of PNS.

Western blotting

SDS-PAGE was performed under reducing conditions using 4–20% tris-glycine gels (Invitrogen EC60285BOX). For subcellular fractionation experiments (Supporting Information Fig. S3), 10 μL of each fraction was loaded per lane. For autophagy experiments (Supporting Information Fig. S4), 10 μg total protein for each Huh7 PNS sample was loaded per lane. Samples were transferred onto nitrocellulose membranes, which were probed using anti-calnexin (Assay Designs SPA-860, 1:5000), anti-MEK1/2 (Cell Signaling 9126, 1:1000), anti-LC3 (Novus Biologicals NB100–2220, 1:2000), or anti-α actin (Millipore MAB1501R, 1:5000) antibodies. Blots were developed using an ECL system (Thermo Scientific 34078) with appropriate secondary antibodies conjugated to horseradish peroxidase.

ATP hydrolysis

To hydrolyze ATP, liver PNS (200 μg/mL) was incubated with 0.5 U or 1.5 U apyrase (Sigma Aldrich A6410) for 30 min at 30 °C. ATP concentrations were quantified by luciferase luminescence (Perkin Elmer 6016943), as measured by a TECAN Safire II plate reader.

Atomic force microscopy

Following 40 h of kinetic disaggregation assay, samples (25 μL) were applied to freshly cleaved mica mounted on a metal sample holder. AFM images were taken as previously described using a Veeco Nanoscope III AFM setup.⁴⁷

Proteolysis assay

Following 40 h of disaggregation, samples (without or with centrifugation at 200,000 g for 1 h and subsequent resuspension in 8 M guanidine hydrochloride) were analyzed by RP-HPLC. Samples were injected in 100% H₂O, 0.03% NH₄OH on a Phenomenex Gemini-NX C18 column and eluted with a linear gradient of 50% methanol, 50% isopropanol, and 0.03% NH₄OH. Resulting Aβ_1–40 or 8 kDa gelsolin peak areas (220 nm absorbance) were quantified and normalized to a standard curve (buffer alone disaggregation negative control).