Hsp104 Catalyzes Formation and Elimination of Self-Replicating Sup35 Prion Conformers

See allHide authors and affiliations

Science  18 Jun 2004:
Vol. 304, Issue 5678, pp. 1793-1797
DOI: 10.1126/science.1098007


The protein-remodeling factor Hsp104 governs inheritance of [PSI+], a yeast prion formed by self-perpetuating amyloid conformers of the translation termination factor Sup35. Perplexingly, either excess or insufficient Hsp104 eliminates [PSI+]. In vitro, at low concentrations, Hsp104 catalyzed the formation of oligomeric intermediates that proved critical for the nucleation of Sup 35 fibrillization de novo and displayed a conformation common among amyloidogenic polypeptides. At higher Hsp104 concentrations, amyloidogenic oligomerization and contingent fibrillization were abolished. Hsp104 also disassembled mature fibers in a manner that initially exposed new surfaces for conformational replication but eventually exterminated prion conformers. These Hsp104 activities differed in their reaction mechanism and can explain [PSI+] inheritance patterns.

The yeast prion state [PSI+] results from self-perpetuating, dysfunctional, prion aggregates of Sup35 (a translation termination factor) that cause stable, heritable reductions in the fidelity of translation termination (1, 2). The N-terminal, glutamine- and asparagine-rich (N) and highly charged middle (M) domains of Sup35 are necessary and sufficient for Sup35 to transition between the soluble, functional state of Sup35 in [psi] cells and the prion state of [PSI+] cells (15). In [PSI+] cells, NM (the N and M domains of Sup35) adopts an amyloid-like conformation (NM) (18). In vitro, NM spontaneously forms self-propagating, β-sheet–rich amyloid fibers after a characteristic lag phase (812). Fibers grow rapidly at their ends (9, 10) and truly embody prion conformers, because transforming [psi] cells with in vitro–generated NM fibers, but not with soluble NM, efficiently induces [PSI+] (13).

[PSI+] inheritance depends absolutely on the cellular concentration of Hsp104. Either deletion or overexpression of Hsp104 eliminates [PSI+] (16, 14, 15). Hsp104 is a protein-remodeling factor belonging to the AAA+ (Adenosine triphosphatases Associated with diverse Activities) family. Two models might explain the baffling dosage relationship (14, 6, 14, 15): (i) Hsp104 acts on soluble Sup35 to generate prion conformers through a critical oligomeric intermediate. Excess or insufficient Hsp104 would create imbalances between Hsp104 and Sup35 and perturb amyloidogenic oligomerization. (ii) The disaggregation activity of Hsp104 (16) operates on Sup35 fibers to cleave them. Excess Hsp104 would completely annihilate fibers, whereas too little would yield overly large aggregates that are poorly disseminated to progeny and have too few polymerization surfaces to sustain prion propagation. Here we define the direct effects of Hsp104 concentration on the different conformational states of NM, the prion domain of Sup35.

When Hsp104 [together with adenosine triphosphate (ATP) and an ATP regeneration system] was added to unpolymerized NM at substoichiometric concentrations, it greatly accelerated NM polymerization into amyloid fibers (Fig. 1A). Throughout this study, we confirmed the nature of the amyloid fibers by several different techniques (fig. S1) (17). Hsp104 completely eliminated the lag phase (reducing T0, the time before detection of amyloid, from 45 min to undetectable) and accelerated the assembly phase (reducing TC, the time between the first appearance of amyloid and completion of conversion, from 195 min to 45 min). Titration of several other proteins into NM fibrillization reactions did not stimulate assembly (supporting online text). Electron microscopy (EM) revealed that Hsp104-generated fibers were indistinguishable from spontaneously formed fibers, except that they were slightly shorter (1.1 ± 0.8 μm without Hsp104 versus 0.8 ± 0.4 μm with Hsp104).

Fig. 1.

Hsp104 catalyzes de novo NM fibrillization. (A) Kinetics of unseeded, rotated (80 rpm) NM (2.5 μM) polymerization, without or with Hsp104 (0.01 to 0.03 μM; all Hsp104 concentrations refer to Hsp104 hexamers), plus ATP (5 mM). Fibrillization was monitored by Congo Red (CR) binding. Values represent means ± SD (n = 4 experiments). (B) NM (2.5 μM) was polymerized as in (A) without or with Hexahistidine-tagged (His6)–Hsp104 (0.03 μM) for 45 min or 4 hours. Reactions were then depleted of His6-Hsp104, sonicated, and used to seed [2% weight per weight (wt/wt)] fresh, unrotated NM (2.5 μM) polymerization reactions. Values represent means ± SD (n = 3 experiments). (C) Unseeded, rotated (80 rpm) NM (2.5 μM) fibrillization without or with Hsp104 (0.03 μM) plus ATP, AMP-PNP, ADP (5 mM), or no nucleotide. At various times, reactions were applied to nitrocellulose and probed with oligomer-specific antibody or NM-specific antibody. The far left lane contains NM in 8 M urea. (D) Unseeded, rotated (80 rpm) NM (2.5 μM) fibrillization after 6 hours with increasing concentrations of either anti-oligomer immunoglobulin G (IgG), nonspecific IgG, anti-amyloid IgM, or nonspecific IgM. Values represent means ± SD (n = 4 experiments). (E) Reactions were performed as in (A) without or with Hsp104 (0.03 μM) plus either ATP (5 mM), AMP-PNP (5 mM), ADP (5 mM), or no nucleotide. Values represent means ± SD (n = 3 experiments). (F) Seeded (2% wt/wt), unrotated NM (2.5 μM) fibrillization without or with Hsp104 (0.03 μM) plus ATP or AMP-PNP (5 mM). Values represent means ± SD (n = 3 experiments).

If Hsp104-generated NM fibers are relevant to prion propagation, they should seed the fibrillization of unpolymerized NM. We first depleted these fibers of Hsp104, because the remodeling factor would interfere with analysis of seeding efficacy. Consistent with the transient nature of Hsp104-Sup35 interactions (1, 18), Hsp104 was readily removed without codepleting NM (fig. S2). Hsp104-generated NM fibers seeded polymerization just as well as fibers that had been assembled in spontaneous reactions (Fig. 1B) (17), which can convert cells from [psi] to [PSI+] (13). Thus, Hsp104 catalyzes the acquisition of a self-replicating prion conformation.

Amyloid fibers are connected with several devastating neurodegenerative disorders, including Alzheimer's, Parkinson's, and Huntington's diseases (1921). A common feature of amyloidogenesis is the appearance of oligomeric species before fibrillization that may or may not be “on pathway” for fiber assembly (1012, 19, 20). An antibody raised against an amyloidogenic peptide associated with Alzheimer's disease, Aβ40, which had been tethered at one end to prevent fibrillization, recognizes β-sheet–rich, oligomeric intermediates of Aβ40 (20). It also recognizes oligomeric species of several other amyloidogenic polypeptides including Aβ42, lysozyme, islet amyloid polypeptide, α-synuclein, polyglutamine, insulin, and prion protein (20). It does not, however, recognize monomers or mature fibers of these proteins (20). We used this antibody to determine the role of oligomers and of Hsp104 in prion assembly.

Unlike an NM-specific antibody, the oligomer-specific antibody recognized neither NM solubilized in urea nor NM fibers (Fig. 1C). However, in spontaneous assembly reactions, it recognized a species that peaked late in the lag phase and was rapidly consumed during the assembly phase (Fig. 1C). This corresponded to an oligomeric form of NM, because the immunoreactive species did not pass through a 100-kD filter [NM is a 28.5-kD protein (fig. S3A)]. Consistent with previous studies (1012), the proportion of NM that was present in an oligomeric state and was retained by the 100-kD filter remained constant (∼10% of total NM) throughout the lag phase (fig. S3B). Hence, NM forms molten oligomeric complexes rapidly, and these gradually metamorphose into oligomeric species recognized by the conformation-specific antibody (1, 1012).

The oligomer-specific antibody severely inhibited unseeded NM polymerization, even at concentrations 100-fold lower than the NM concentration (Fig. 1D). Thus, NM oligomers recognized by the oligomer-specific antibody are crucial for nucleating polymerization at the end of the lag phase. Conversely, the antibody had no effect on NM polymerization seeded by sonicated NM fibers, even at a 100-fold molar excess over added seed (fig. S4). Therefore, the amyloidogenic oligomer recognized by this antibody is not required for polymerization once fibers have formed. That is, fibers can recruit NM that is not in this amyloidogenic oligomeric form (either monomers or immature oligomers).

Addition of Hsp104 plus ATP to soluble NM caused the immediate appearance of mature oligomers that reacted with the oligomer-specific antibody. This species was rapidly consumed upon fibrillization (Fig. 1C). Thus, Hsp104 eliminates the lag phase in NM polymerization by catalyzing the nascence of the critical amyloidogenic NM oligomer that elicits fibrillization.

We also employed an amyloid-specific antibody, raised against Aβ40 fibers, which also recognizes fibers formed by several other amyloid proteins (21). We found that it recognized NM fibers but not unassembled protein (Fig. 1D and fig. S4). In contrast with the oligomer-specific antibody, it inhibited both unseeded and seeded NM fibrillization (Fig. 1D and fig. S4), reinforcing the importance of amyloid conformers in the conversion of NM to the prion state (13).

Nonhydrolyzable ATP analogs adenosine 5′-(β-γ-imido) triphosphate (AMP-PNP) and methylene-adenosine 5′-triphosphate (AMP-PCP) supported Hsp104-catalyzed oligomer maturation and fibrillization (Fig. 1E) (17), even with hexokinase and glucose present to eliminate trace contaminating ATP (22). Adenosine diphosphate (ADP) did not support these activities. These findings were extended with several Hsp104 AAA+ point mutants defective in ATP binding and/or hydrolysis at nucleotide binding domain (NBD) 1 or 2, which have previously been shown to affect [PSI+] inheritance in vivo (4, 15, 23, 24) (supporting online text and fig. S5). Hsp104 mutants that interfered with hexamerization, or with the ability to bind ATP at either NBD, failed to induce NM fibrillization (fig. S5). Mutants that could bind but not hydrolyze ATP accelerated polymerization, but not as well as wild-type protein (fig. S5). Thus, ATP hydrolysis is not required per se, but hydrolysis at both NBDs maximizes the rate of Hsp104-catalyzed NM fibrillization.

To investigate the unusual dosage relationship between Hsp104 and prion replication, we tested higher Hsp104 concentrations. When the stoichiometry of NM monomers to Hsp104 hexamers was altered from 250:1 to 15:1, NM polymerization was abolished (Fig. 2A and fig. S6) (17). Hsp104 blocked fibrillization by coupling ATP hydrolysis to the elimination of amyloidogenic NM oligomers (Fig. 2B). At high concentrations, Hsp104 also eliminated fibrillization with AMP-PNP, ADP, and even without nucleotide. However, successively higher Hsp104 concentrations were required in each case (fig. S6A) (17). Without ATP, Hsp104 did not eliminate oligomers but simply prevented their maturation (Fig. 2B) (22). Corroboratively, Hsp104 point mutants with reduced hexamerization or ATPase activity inhibited NM fibrillization, but with decreased efficiency (supporting online text and fig. S6B). Thus, Hsp104 can passively inhibit NM fibrillization, perhaps through transiently binding NM. Because AMP-PNP allows a more severe inhibition at lower Hsp104 levels [median inhibitory concentration (IC50) ∼ 0.2 μM] than ADP (IC50 ∼ 3.1 μM) or in the absence of nucleotide (IC50 ∼ 7.6 μM), Hsp104 may preferentially engage NM in an ATP-bound conformation. However, inhibition is potentiated when coupled to ATP hydrolysis (IC50 ∼ 0.1 μM) and contingent oligomer remodeling.

Fig. 2.

Hsp104 abolishes NM fibrillization. (A) Kinetics of unseeded, rotated (80 rpm) NM (2.5 μM) polymerization without or with Hsp104 (0.17 to 0.83 μM) plus ATP (5 mM). Fibrillization was quantified by CR binding. Values represent means ± SD (n = 4 experiments). (B) Unseeded, rotated (80 rpm) NM (2.5 μM) fibrillization without or with Hsp104 (0.3 μM) plus ATP or AMP-PNP (5 mM). At various times, reactions were applied to nitrocellulose and probed with oligomer-specific antibody or NM-specific antibody.

Next, we asked if the ability of Hsp104 to accelerate NM fibrillization during the assembly phase is due to the same activity that eliminates the lag phase (that is, the production of amyloidogenic oligomers) or represents a distinct activity. We found that Hsp104 promoted polymerization in reactions that did not require the production of new oligomers, because they were seeded with preformed fibers. However, this activity required ATP hydrolysis. Hsp104 plus AMP-PNP, which was able to catalyze de novo assembly of oligomeric intermediates, did not accelerate seeded assembly (Fig. 1F) (17). Moreover, promoting assembly was independent of amyloidogenic oligomers, because blocking their maturation with oligomer-specific antibody had no effect on seeded assembly (fig. S4). Thus, Hsp104 also promotes fiber assembly by a reaction mechanism distinct from nucleation.

We postulated that accelerating the assembly phase might involve an effect of Hsp104 on NM fibers. Indeed, when Hsp104 (with ATP and an ATP regeneration system) was added to NM fibers, it disassembled them (Fig. 3, A to C, and fig. S7). The reaction exhibited a steep Hsp104 concentration dependence, implying a cooperative reaction mechanism (Fig. 3, A and B, and fig. S7). Hsp104 briskly diminished the mean fiber length from ∼1.1 ± 0.8 μmto ∼0.2 ± 0.1 μm after 5 min, and to ∼0.1 ± 0.05 μm after 10 min (Fig. 3C). This was superficially reminiscent of sonication, which generates short fibers and creates additional polymerization surfaces (10). Indeed, the short fibers produced by brief Hsp104 treatments markedly increased seeding activity (Fig. 3D) (17).

Fig. 3.

Hsp104 couples ATP hydrolysis to NM fiber and oligomer disassembly. (A) NM (2.5 μM) was incubated for 6 hours with rotation (80 rpm) to generate fibers and then incubated with Hsp104 (0 to 2 μM) for 20 min at 25°C. Disassembly was assessed by sedimentation. Values represent means ± SD (n = 3 experiments). (B) Kinetics of Hsp104 (0 to 2 μM) induced NM fiber disassembly monitored by CR binding. Values represent means ± SD (n = 3 experiments). (C) NM fibers (2.5 μM monomer) were incubated with Hsp104 (2 μM) for 0 to 30 min, and reactions were processed for EM at various times. Scale bars, 0.25 μm. (D) NM fibers (2.5 μM monomer) were either left untreated, sonicated, or treated with His6-Hsp104 (0.3 μM) for 2 min or 30 min. Reactions were then depleted of Hsp104 and used to seed (2% wt/wt) fresh, unrotated NM (2.5 μM) polymerization. Values represent means ± SD (n = 3 experiments). (E) NM fibers (2.5 μM monomer) were incubated without or with Hsp104 (2 μM) plus ATP (5 mM) or AMP-PNP (5 mM) for 0 to 60 min. At various times, reactions were applied to nitrocellulose and probed with oligomer-specific antibody or NM-specific antibody. (F) NM fibers (2.5 μM monomer) were incubated (for 30 min at 25°C) with Hsp104 (2 μM) and no nucleotide, ADP (5 mM), ATP (5 mM), AMP-PNP (5 mM), or AMP-PCP (5 mM). Values represent means ± SD (n = 3 experiments).

With longer incubations, fibers were completely obliterated (Fig. 3, B and C, and fig. S7). The final disassembly products were devoid of seeding activity (Fig. 3D) (17), distinguishing them from short fibers. In the early phases of disassembly, Hsp104 released amyloidogenic NM oligomers from fibers (Fig. 3E). Later, these oligomers were no longer apparent (Fig. 3E), which correlates with the annulment of seeding activity. Thus, when Hsp104 disassembles NM fibers, it initially creates additional polymerization surfaces as well as new amyloidogenic oligomers. However, Hsp104 eventually destroys seed, emancipating NM from the self-replicating prion conformation.

Unlike the formation of amyloidogenic oligomers, fiber shortening and the eradication of seeding activity by Hsp104 required ATPase activity. It was not supported by AMP-PNP, AMP-PCP, ADP, or the absence of nucleotide (Fig. 3F). Furthermore, hydrolysis was required at both NBD1 and NBD2. Hsp104 mutants defective in ATP hydrolysis at either NBD could not depolymerize fibers (fig. S8).

The ability of Hsp104 to promote the assembly and disassembly of NM fibers was highly specific. Hsp70 and Hsp40 chaperones can also affect [PSI+] in vivo (1, 2). They also exerted effects on NM fibrillization (25), but they were not required for either assembly or disassembly of NM fibers by Hsp104. Moreover, they could not, on their own or in combination, promote NM fiber assembly or disassembly (25). A prokaryotic homolog of Hsp104, ClpB, and another eukaryotic AAA+ protein, Cdc48p, were also ineffective in promoting NM fiber assembly or disassembly (22).

Our results establish the mechanisms by which Hsp104 may control the formation, replication, and curing of [PSI+]. By uncovering multiple Hsp104 remodeling activities on NM, the prion domain of Sup35, we have recreated the unusual dosage relationship between [PSI+] and Hsp104 in vitro and reconciled previous models concerning the role of Hsp104 in prion inheritance (14, 6, 14, 15). Two activities promote prion formation and replication: (i) At low concentrations, Hsp104 acts on soluble NM and catalyzes assembly of critical oligomeric intermediates that nucleate fibrillization; and (ii) Hsp104 fragments amyloid fibers to create new ends for polymerization and facilitate partitioning of seeds to progeny. Three activities promote prion curing: (i) At high concentrations, Hsp104 passively inhibits oligomer maturation; (ii) Hsp104 couples ATP hydrolysis to the elimination of amyloidogenic oligomers; and (iii) Hsp104 couples ATPase activity to the disassembly of fibers into non-amyloidogenic species. These activities employ different modes of Hsp104 action. We previously showed that Hsp104 binds polylysine in a highly cooperative manner, triggering a cascade of events that couple ATP hydrolysis at NBD2 to conformational change in the coiled-coil middle domain and to hydrolysis at NBD1 (26). The M region of Sup35, which resides on the exterior of NM fibers (10), is lysine-rich (8). Cooperative interactions between M and Hsp104, coupled to additional interactions with the glutaminerich N domain (27), may serve as a fulcrum for force application by Hsp104 to separate the intermolecular β-sheet interfaces of N that maintain fiber integrity (10, 26, 28). Consistent with this, changes in the M region alter the relationship between Hsp104 and [PSI+] inheritance (5). Hsp104 promotes assembly of amyloidogenic oligomers that nucleate fibrillization, requiring ATP binding at both NBDs but not hydrolysis. Hsp104 may provide a catalytic surface on which NM molecules transiently converge to attain the amyloidogenic oligomeric conformation. Alterations in the Sup35:Hsp104 concentration ratio or Hsp104 activity (e.g., by heat shock) can radically shift the balance between these activities and either promote or eliminate self-replicating prion conformers and the [PSI+] phenotype (1, 4, 14, 15, 23, 24). The effects of Hsp104 on NM fibrillization and disassembly could be reconstituted with full length Sup35, reinforcing their physiological significance (25).

We also establish that an amyloidogenic oligomer is an obligate intermediate for nucleating prion formation de novo. Previously, we postulated (1012) that intrinsically unfolded NM monomers may have too many accessible conformations to find a stable fold and that molten oligomers might allow NM to sample intermolecular interactions that stabilize β-strands to form amyloidogenic nuclei (1012). It may be that many amyloids assemble by a related mechanism (10, 19, 20). An antibody that recognizes a common conformational feature of oligomers observed for many disease-associated amyloids also recognizes the amyloidogenic intermediate of NM. Oligomers may also be the most toxic species in the protein misfolding diseases (19, 20). Rapid conversion of oligomers to amyloid fibers driven by Hsp104 may explain why the conformational conversion of Sup35 to a prion state is not toxic in yeast, because the oligomeric species would be short-lived. It seems likely that Hsp104 coevolved with Sup35 to regulate [PSI+], functioning as the minimal machinery controlling prion conformation and inheritance. [PSI+] profoundly changes the spectrum of conditions in which yeast cells thrive, and Hsp104 may therefore provide one link between heritable changes in phenotype and environmental contingency (1, 2). Our studies reinforce that amyloids are not necessarily intractable, pathogenic entities but can be harnessed and tightly regulated for advantageous purposes (1, 2, 19, 29). An accurate understanding of this regulation will enlighten analogous complexities in amyloidogenic events that underpin protein misfolding diseases (19), protein-based genetics (1, 2), and molecular memories (29).

Supporting Online Material

Materials and Methods

SOM Text

Figs. S1 to S8

References and Notes

References and Notes

View Abstract

Stay Connected to Science

Navigate This Article