Single-molecule Analyses of the Dynamics of Heat Shock Protein 104 (Hsp104) and Protein Aggregates*

Background: The process of disaggregation by heat shock protein 104 (Hsp104) and heat shock protein 70/40 (Hsp70/40) has not been elucidated. Results: We developed several methods to investigate the dynamics of Hsp104 at single-molecule levels. Conclusion: Statistical analyses revealed that Hsp70/40 affected the dynamics of Hsp104. Significance: Single-molecule approaches are a unique way to unravel the functional mechanisms of disaggregases. Hsp104 solubilizes protein aggregates in cooperation with Hsp70/40. Although the framework of the disaggregase function has been elucidated, the actual process of aggregate solubilization by Hsp104-Hsp70/40 remains poorly understood. Here we developed several methods to investigate the functions of Hsp104 and Hsp70/40 from Saccharomyces cerevisiae, at single-molecule levels. The single-molecule methods, which provide the size distribution of the aggregates, revealed that Hsp70/40 prevented the formation of large aggregates from small aggregates and that the solubilization of the small aggregates required both Hsp104 and Hsp70/40. We directly visualized the individual association-dissociation dynamics of Hsp104 on immobilized aggregates and found that the lifetimes of the Hsp104-aggregate complex are divided into two groups: short (∼4 s) and long (∼30 s). Hsp70/40 stimulated the association of Hsp104 with aggregates and increased the duration of this association. The single-molecule data provide novel insights into the functional mechanism of the Hsp104 disaggregation machine.


Hsp104 solubilizes protein aggregates in cooperation with
Hsp70/40. Although the framework of the disaggregase function has been elucidated, the actual process of aggregate solubilization by Hsp104-Hsp70/40 remains poorly understood. Here we developed several methods to investigate the functions of Hsp104 and Hsp70/40 from Saccharomyces cerevisiae, at singlemolecule levels. The single-molecule methods, which provide the size distribution of the aggregates, revealed that Hsp70/40 prevented the formation of large aggregates from small aggregates and that the solubilization of the small aggregates required both Hsp104 and Hsp70/40. We directly visualized the individual association-dissociation dynamics of Hsp104 on immobilized aggregates and found that the lifetimes of the Hsp104aggregate complex are divided into two groups: short (ϳ4 s) and long (ϳ30 s). Hsp70/40 stimulated the association of Hsp104 with aggregates and increased the duration of this association. The single-molecule data provide novel insights into the functional mechanism of the Hsp104 disaggregation machine.
Misfolded or aggregated proteins damaged by cellular stresses such as heat shock, trigger impaired cellular protein homeostasis and proteostasis, eventually leading to serious disorders in cells (1,2). To protect against these threats, molecular chaperones promote protein folding and prevent the accumulation of aggregated proteins in the cell (1,2). In the yeast Saccharomyces cerevisiae, Hsp104 is essential for thermotolerance (3,4). Hsp104, a member of the ATPases associated with diverse cellular activities (AAAϩ) superfamily, has homologs in fungi and prokaryotes, including Escherichia coli ClpB (1,2,5).
As described above, extensive efforts over the past decade have elucidated the basic mechanism of Hsp104/ClpB as a molecular machine (1,2). However, the process by which Hsp104-Hsp70 solubilizes the substrate aggregates remains poorly understood. For example, because the aggregates are heterogeneous intermolecular assemblies in nature, the ability of conventional biochemical analyses to characterize the heterogeneous substrates are limited. Indeed, conventional methods to detect the Hsp104/ClpB activity have mainly relied on socalled "end point" assays, such as centrifugation analyses, the reactivation of enzymatic activity because of the disaggregation, or electron microscopic observations (e.g. Refs. 7, 21). Alternatively, previous in vivo imaging using GFP fusions showed "snapshots" of the individual interactions of Hsp104/ ClpB with the substrates (15,19). However, so far, no attempts have been made to visualize and analyze the dynamics of the interaction of Hsp104 with the aggregates. To understand the dynamic aspects of the disaggregase reaction in more detail, methods to visualize the heterogeneous aggregates at singlemolecule/particle levels must be developed and then extended to analyze the interactions between Hsp104/ClpB and the aggregates.
In this report, we describe several methods we developed to visualize and quantitate aggregated proteins at single-molecule levels. These analyses discriminated the size distribution of the heterogeneous aggregates. Moreover, we successfully visualized real-time binding and release events between Hsp104 and the aggregated proteins at a single-molecule level. Statistical analysis revealed the detailed kinetics of the Hsp104-aggregate dynamics, which was influenced by the presence of Ssa1/Ydj1, providing novel insights into the mechanism by which Hsp104 and Hsp70/40 cooperatively function in aggregate solubilization.

EXPERIMENTAL PROCEDURES
Protein Purification-The purification of Hsp104 and the Hsp104-trap mutant was conducted as described previously, with several modifications (22)(23)(24). Hsp104 was produced in E. coli Rosetta (DE3) pLysS cells and purified by chromatography on Ni-NTA 2 Superflow and HiTrap Q FF TM columns (GE Healthcare). The poly-His tag was removed with thrombin (GE Healthcare). The purification of Ssa1 and Ydj1 was conducted as described previously, with several modifications (25,26). The cells were homogenized by a metal cone (Yasui Kikai, Osaka, Japan) using a Multi-Beads Shocker (Yasui Kikai, Osaka, Japan). Ssa1 was purified by chromatography on Ni-NTA Superflow (Qiagen) and HiTrap Q FF (GE Healthcare) columns. Ydj1 was purified by chromatography on HiTrap DEAE-FF (GE Healthcare) and Macro-Prep ceramic hydroxyapatite TYPE1 40-m (Bio-Rad) columns. Firefly luciferase was produced in E. coli BL21 (DE3), with expression induced overnight with 1 mM isopropyl 1-thio-␤-D-galactopyranoside at 23°C. Luciferase was purified by chromatography on Ni-NTA Superflow (Qiagen) and HiTrap Q FF (GE Healthcare) columns. The poly-His tag was removed with thrombin (GE Healthcare). Protein concentrations refer to the concentration of the monomer, except where indicated otherwise.
Labeling of Proteins-Luciferase was incubated with a 5-fold molar excess of Cy TM 5 mono-reactive N-hydroxysuccinimide (NHS) ester dye (GE Healthcare), 5-fold Alexa Fluor 488 carboxylic acid succinimidyl ester (Molecular Probes), or 20-fold EZ-LINK TM Sulfo-NHS-LC Biotin (Pierce) at 4°C for 90 min. We confirmed that the Cy5-labeled luciferase had a comparable enzymatic activity to that of non-labeled luciferase. Hsp104 was incubated with an equimolar amount of Cy TM 3 mono-reactive NHS ester dye (GE Healthcare) at 4°C for 90 min. Nonreacted dyes were removed by passage through a Micro Bio-Spin TM 6 column (Bio-Rad) and equilibrated with 40 mM HEPES-KOH (pH 7.4), 150 mM KCl, 20 mM MgCl 2 , 1 mM DTT, and 10% glycerol. The labeling yield was ϳ0.8 Cy3 per Hsp104 monomer. For the single-molecule analysis, the Hsp104 hexamer was prepared by incubating Cy3-Hsp104 and non-labeled Hsp104 at a molar ratio of 1:5.
Luciferase Aggregate Formation and Reactivation Assay-The recovery of aggregated luciferases was performed as described previously, with some modifications (6). Luciferase was denatured by 6 M urea for 30 min at room temperature. To prepare luciferase aggregates, the urea-denatured luciferase was diluted 100-fold into buffer (25 mM HEPES-KOH (pH 7.4), 150 mM potassium acetate, 10 mM Mg(OAc) 2 , 10 mM DTT, and 0.1 mg/ml BSA) containing an ATP-regenerating system consisting of 5 mM ATP, 10 mM phosphoenolpyruvate, and 10 g/ml pyruvate kinase for 30 min on ice. For the disaggregation assay, the solution of aggregated luciferase (100 nM in monomer concentration), with or without 1 M Hsp104, 2 M Ssa1, and 1 M Ydj1, was incubated for 30 min at 30°C. Aliquots of the solutions were mixed with the luciferase substrate solution (Promega). Luminescence was measured by luminescence PSN TM (ATTO).
Centrifugation assay-Cy5-luciferase aggregates were incubated with or without Hsp104/Ssa1/Ydj1, for 30 min at 30°C. An aliquot of the mixture was centrifuged at 45,000 rpm in a TLA 100.3 rotor (Beckman Coulter) for 1 h, to separate the supernatant and the precipitate. The total fraction and the supernatant fractions were separated by SDS-PAGE, and the band intensities were detected by Cy5 fluorescence, using an FLA7000 bioimager (Fujifilm). The band intensities were measured by the MultiGauge software (Fujifilm), and the ratio of the supernatant to the total fraction was defined as the solubility.
Dynamic Light Scattering (DLS)-Luciferase aggregates were prepared by diluting the urea-denatured luciferase as described above. All DLS measurements were performed at 30°C with a Zetasizer Nano ZSP system (Malvern).
Highly Inclined and Laminated Optical Sheet (HILO) Microscopy-Aliquots of the solution of Cy5-luciferase aggregates, with or without Hsp104/Ssa1/Ydj1 and 5 mM ATP, were placed on a non-coated glass slide (Matsunami) at 30°C. The fluorescence images of Cy5-luciferase aggregates were observed on an inverted fluorescence microscope (IX-70, Olympus) with a He-Ne laser (633 nm) and recorded by an electron multiplying charge-coupled device camera iXon897 (Andor Technology) at 100 ms/frame. The fluorescence intensities of the Cy5-luciferase aggregates were measured by ImageJ software.
Fluorescence Correlation Spectroscopy-All FCS measurements were performed at 25°C with an LSM780 microscope (Carl Zeiss) as described previously with some modifications (27). Alexa Fluor 488 fluorescence was excited at 488 nm by adjusting the acousto-optical tunable filter to 0.5%. All fluorescence autocorrelation functions (FAFs, G()) in aqueous solutions were measured three times for 100 s. The chamber was coated with 3 mg/ml BSA and washed with PBS to prevent the nonspecific absorption of proteins. The solutions of 100 nM Alexa Fluor 488-luciferase aggregates were incubated at 30°C with or without Hsp104/Ssa1/Ydj1 and 5 mM ATP and measured.
Total Internal Reflection Fluorescence (TIRF) Microscopy-The experimental procedure described previously was performed (28,29), with some modifications. A flow cell, shown schematically in Fig. 5A, was made from a glass slide and a coverslip, with double-sided tape acting as a spacer. The experimental procedure for coating the surface of the glass slide with 0.5% biotinylated PEG has been described previously (30,31). To immobilize Cy5-luciferase aggregates on the glass surface, 0.1 mg/ml neutravidin was flowed into the cell. After washing, 100 nM biotinylated Cy5-luciferase aggregates, used immediately after preparation by diluting the urea-denatured luciferase as described above, were flowed into the cell and incubated for 5 min. After the excess amounts of the luciferase aggregates were washed out, 100 nM Cy3-Hsp104, 200 nM Ssa1, 100 nM Ydj1, and 5 mM ATP, with the ATP-regenerating system con-taining the oxygen scavenger system (4.5 mg/ml glucose, 50 unit/ml glucose oxidase, 50 unit/ml catalase, and 10 mM DTT), were preincubated for 10 min at 37°C and flowed into the cell. The Cy5-luciferase aggregates were illuminated with a He-Ne laser (633 nm), and Cy3-Hsp104 was illuminated with a green solid laser (532 nm, Sapphire 532 LP, Coherent). Images were separated by using Dual-View (Optical Insights) and recorded by an iXon897 camera at 100 ms/frame. At least two fields of images were recorded for each assay. The positions of the Cy5luciferase aggregates and the association times of Cy3-Hsp104 with the Cy5-luciferase aggregates were analyzed by ImageJ software and a homemade program. The plots of the complexes between Hsp104 and the aggregates were fit by using the following equation: C[exp(Ϫkt)]ϩCЈ[exp(ϪkЈt)], in which k and kЈ are the rate constants of the fast and slow components, respectively.

RESULTS
The Hsp70/40 System Solubilizes, but Does Not Reactivate, Aggregated Luciferase-We used firefly luciferase aggregates, prepared by diluting urea-denatured luciferase, as a model substrate for the disaggregation reaction (6). Before conducting the single-molecule experiments, we examined which chaperones were responsible for the reactivation of the luciferase aggregates. As reported previously (e.g. Ref. 6), the luciferase aggregates were reactivated only in the presence of both Hsp104 and Ssa1/Ydj1 in an ATP-dependent manner (Fig. 1A). We centrifuged the luciferase aggregates, which were labeled with Cy5, to monitor their aggregation status. The centrifugation assay revealed that substantial fractions of luciferase were soluble in the presence of either Hsp104-Ssa1/Ydj1 or Ssa1/Ydj1 (Fig. 1B), showing that the Hsp70/40 system is sufficient to maintain the solubility of luciferase. The result that Ssa1/Ydj1 maintained the solubility but did not reactivate the protein is consistent with previous data from the bacterial DnaK/DnaJ system (7,32).
Luciferase Aggregates Analyzed by Dynamic Light Scattering-To determine the molecular mechanism for the chaperone-dependent disaggregation, the aggregates were subjected to a detailed characterization. At first, we conducted DLS experiments to characterize the luciferase aggregates. DLS is a particle size analysis method applicable to protein aggregates (33,34). The DLS measurements revealed that the luciferase aggregates at the initial stage had a heterogeneous size distribution (Fig. 2). The hydrostatic radius of the main species was ϳ200 nm, and this species accumulated after a 30-min incubation (Fig. 2). Note that the ϳ200 nm hydrostatic radius corresponds to ϳ9.5 ϫ 10 5 kDa if the particles are assumed to be spherical. The polydispersities of the aggregates were ϳ0.96 at the initial stage, which shifted to ϳ0.27 after the 30 min incubation, indicating an apparent reduction of the heterogeneity. Further characterization of chaperone-dependent disaggregation requires the inclusion of chaperones during the DLS measurement. However, the experiment was not straightforward because the inclusion of a stoichiometric amount of the relatively large-sized chaperones, such as the Hsp104 hexamers, perturbed the DLS measurement of the luciferase aggregates.

Distribution of Luciferase Aggregates Revealed by Fluores-
cence Microscopy-To directly characterize the aggregates, we adopted the HILO illumination method, which illuminates fluorescent molecules by a sharply inclined laser at a low background ( Fig. 3A) (35). Using HILO illumination, we clearly observed visible dot-like fluorescent particles when we immobilized the Cy5-luciferase aggregates. Because there were no such dot-like particles in the buffer containing native luciferase   MARCH 20, 2015 • VOLUME 290 • NUMBER 12

Single-molecule Analyses of Hsp104 in Disaggregation
( Fig. 3B), we concluded that the dot-like particles are aggregated forms of Cy5-luciferase. Quantification of the fluorescence intensities of these dot-like particles permitted us to estimate the sizes of the aggregates and to generate histograms of the size distributions (Fig. 3C). The aggregated luciferase was then incubated in the presence of chaperones and ATP. The presence of Hsp104 and ATP did not significantly affect the size distribution of the aggregates (Fig. 3C). In contrast, the addition of either Ssa1/Ydj1 or Hsp104/Ssa1/Ydj1 drastically changed the size distribution of the aggregates, as detected by the reduction in the number of high fluorescence intensity particles (Fig.  3C), suggesting that Ssa1/Ydj1, regardless of the presence of Hsp104, prevents the accumulation of large-size aggregates. However, the histograms with the addition of Hsp104/Ssa1/ Ydj1 or Ssa1/Ydj1 were distinct from those of native luciferase because of the residual presence of smaller aggregates (Fig. 3C). Solubilization of Luciferase Aggregates Revealed by FCS-Because the reactivation of luciferase activity absolutely required Hsp104 in addition to Ssa1/Ydj1 (Fig. 1A), we expected, at first, that the size distributions of the aggregates would be different between Hsp104/Ssa1/Ydj1 and Ssa1/Ydj1. However, the size distribution of the aggregates under the Hsp104/Ssa1/Ydj1 conditions was indistinguishable from that under the Ssa1/Ydj1 conditions (Fig. 3C), probably because of a limitation in the dynamic range of the microscopy. Therefore, we took another approach, FCS, to characterize the smaller aggregates that could not be distinguished by HILO microscopy. FCS is a method to analyze the diffusion of fluorescent molecules in a microscopic detection volume (under 1 femtoliter) and, therefore, estimate the diffusion constants, which correlate with the molecular sizes (36). We first examined native and aggregated forms of Alexa Fluor 488-luciferase by FCS. FCS analysis revealed that the FAF, which correlates with the size distribution, of the Alexa Fluor 488-luciferase aggregates was shifted to the right compared with the FAF of native luciferase (Fig. 4A), confirming that the aggregates are larger than native luciferase in FCS. The aggregates were then incubated in the absence or presence of chaperones, and aliquots of the mixtures were measured by FCS over time. After the 30-min incubation, no reliable FCS data could be obtained under the chaperone-free conditions because of the frequent appearances of spike noises (data not shown), which were too large to analyze by FCS, reflecting further self-association of aggregates, as shown in the DLS analysis (Fig. 2). In contrast, the FAF in the presence of Ssa1/Ydj1 did not change, even after 30 min (Fig. 4B), suggesting that Ssa1/Ydj1 prevented further association of large aggregates, and maintained the size of the aggregates. Strikingly, the presence of Hsp104 in addition to Ssa1/ Ydj1 shifted the FAF to the left with time (Fig. 4C), reflecting the size reduction of the aggregates because of their solubilization. The Hsp104/Ssa1/Ydj1-dependent size reduction eventually reached that of native luciferase (Fig. 4C), implying that the solubilized luciferase has enzymatic activity, as shown in Fig. 1A.
The FAFs fit well into a two-component model, indicating that the major species can be categorized as slower (larger) and faster (smaller) molecules. For example, two-component fitting of the FAFs in the presence of Ssa1/Ydj1 (Fig. 4B) showed that the solution contained ϳ53% slower diffusing species (diffusion time, ϳ1.5 ms) and ϳ47% faster species (diffusion time, ϳ210 s). Assuming that the molecules are globular in shape, the slower and faster molecules correspond to ϳ4.3 ϫ 10 4 kDa and ϳ120 kDa, respectively. Notably, we found that the fractions of the slower species decreased in the presence of Hsp104/Ssa1/ Ydj1 (Fig. 4D), further confirming that the Hsp104-Hsp70 bichaperone systems solubilized the aggregates. In summary, we developed methods to investigate the size distribution of protein aggregates using HILO microscopy and FCS.
Visualization of the Interaction between Hsp104 and Aggregated Luciferase-Hsp104 is essential for solubilizing aggregates with the aid of Ssa1/Ydj1. A detailed characterization of the interaction between Hsp104 and aggregates is necessary to understand the mechanism underlying the disaggregase function. Previous efforts to investigate the interaction between Hsp104 and substrates have been restricted to ensemble-averaged techniques using model peptides or proteins, not protein aggregates (22,37), or static interactions between Hsp104 and amyloid fibrils (24). Therefore, we applied single-molecule fluorescence imaging to visualize the individual dynamics between Hsp104 and aggregates to obtain information that cannot be gained from ensemble-averaged experiments. We extended our previous single-molecule experiments on the E. coli chaperonin GroEL-GroES dynamics (28,29) to the Hsp104 system and visualized the association/dissociation events of Hsp104 on the luciferase aggregates immobilized on a glass surface by TIRF microscopy (Fig. 5A).
Luciferase aggregates labeled with both Cy5 and biotin were immobilized on a glass surface through a biotin-neutravidin linker. Wild-type Hsp104 was labeled with Cy3. The Cy3-labeled Hsp104 behaved like non-labeled Hsp104 in the steadystate ATP hydrolysis, which was accelerated in the presence of ␤-casein as a model substrate, and ATP-dependent disaggregation of luciferase with the aid of Ssa1/Ydj1 (data not shown). To minimize the nonspecific absorption of Hsp104 on the surface, the glass slide was coated with aminosilane and PEG. At first we used an ATPase-deficient Hsp104 mutant (Hsp104 trap ) that binds ATP but does not hydrolyze it and, hence, forms a stable complex with the substrates (22,38). The Cy3-labeled Hsp104 trap mutants colocalized with the Cy5-luciferase aggregates in the presence of ATP (Fig. 5B). The time course of the fluorescence intensity from Cy3-Hsp104 trap associated with the aggregates revealed that the association persisted for at least 200 s, confirming that ATP-bound Hsp104 has a high affinity for the aggregates (Fig. 5C).
Next we visualized wild-type Hsp104 labeled with Cy3 in the chamber containing Ssa1/Ydj1 and ATP and monitored the Cy3-Hsp104 spots on the positions of the Cy5-luciferase aggregates (Fig. 5D). Real-time monitoring of single Hsp104 molecules revealed that Hsp104 was bound repeatedly to and released from the immobilized aggregates (Fig. 5, D and E, and supplemental Movie S1).
We quantitated the frequency of the binding/release events with or without Ssa1/Ydj1. The frequency in this experiment means the percentage of total Hsp104 molecules attached to and detached from immobilized aggregates during an observation time (200 s). On average, Hsp104 associated with ϳ5% of the aggregates in the absence of Ssa1/Ydj1 (Fig. 6A). The association of Hsp104 with the aggregates increased ϳ2.5-fold in the presence of Ssa1/Ydj1 (ϳ13% of aggregates) (Fig. 6A), indicating that Ssa1/Ydj1 stimulates this association. The singlemolecule experiment also enabled us to analyze the association time (dwell time) of Hsp104 with the aggregates. A statistical analysis of hundreds of events revealed that the complex between Hsp104 and the aggregates decayed exponentially with time, regardless of the presence of Ssa1/Ydj1, although the curve in the presence of Ssa1/Ydj1 showed a slower rate (Fig.  6B). The plots were not fit by a single exponential but were fit satisfactorily by a two-component reaction: fast and slow components. The half-times (t1 ⁄ 2 ) of the fast and slow components were calculated as ϳ4 and ϳ30 s, respectively, and were almost the same in the presence or absence of Ssa1/Ydj1 (Fig. 6B). However, the proportions of the two components changed depending on the presence of Ssa1/Ydj1. The proportion of the slow component in the presence of Ssa1/Ydj1 (ϳ72%) was larger than that in the absence of Ssa1/Ydj1 (ϳ42%), explaining why Ssa1/Ydj1 increased the overall duration of the Hsp104 association with the aggregates (Fig. 6B).

DISCUSSION
Establishment of Experimental Systems to Analyze Protein Aggregates at Single-particle Levels-In this report, we described the development of several experimental systems to analyze protein aggregates and Hsp104 at single-molecule levels. Using the single-molecule methods, which can uniquely unmask protein behaviors that are usually inaccessible with conventional ensemble methods, we directly analyzed the aggregates and visualized the processive binding of Hsp104 on the aggregates.
Protein aggregates are notorious for their heterogeneity and precipitation-prone properties, which perturb quantitative biochemical/biophysical methods in solution. Although centrifugation assays are frequently used to detect disaggregation, they only show whether the proteins are soluble or insoluble (e.g. Ref. 7). Light scattering or density gradient analyses are more quantitative but are still at ensemble levels (32,39). By visualizing individual aggregates, we investigated which sizes of aggregates were responsible for the disaggregation process by the Hsp104/Hsp70 disaggregation machines.
Regarding a technical issue, we used HILO illumination to visualize the fluorescently labeled aggregates. Conventionally, TIRF illumination is used to visualize single-molecule fluorescence (e.g., Refs. 28,29). However, because TIRF microscopy only illuminates immobilized fluorescent molecules at ϳ150 nm from the surface, reliable quantitative observations of molecules larger than ϳ150 nm, which are common in protein aggregates, are not possible. HILO can obtain high signal-tonoise images without restrictions of the illumination field and the sample size (35). Indeed, TIRF microscopy only detected a portion of the luciferase aggregates, whereas HILO microscopy illuminated all of the aggregates, leading to the quantification of their sizes by fluorescence intensities.
Because HILO observation has a limitation on the dynamic range of fluorescent molecules, the effect of Hsp104 was not obvious from the histograms (Fig. 2C). To overcome the limitation of HILO in the observation of the smaller aggregates, we utilized FCS. FCS is a powerful method to measure the size of molecules at single-molecule sensitivity in solution or in living cells (40,41). The FCS measurement enabled us to investigate the smaller aggregates, which were further disaggregated by Hsp104/Hsp70. Taken together, the combination of the two fluorescence methods, HILO observation and FCS, allowed us to discriminate the roles of the chaperones in the disaggregation process, where Hsp70/40, even without Hsp104, were sufficient to prevent the accumulation of the large-size aggregates, but further solubilization of the aggregates required both Hsp104 and Hsp70/40.
The Role of Hsp70/40 in Controlling Protein Aggregate Size-Our experiments showed that larger aggregates did not accumulate in the presence of Hsp70/40 (Figs. 1, 3, and 4). The Hsp70/40 effect can be detected without Hsp104, as reported previously for bacterial Hsp70/40 (7,32). There are two possible explanations for the results. One is that Hsp70/40 maintain the size of aggregates by preventing the accumulation of larger aggregates, in agreement with the interpretation of a previous centrifugation assay (7). The other is that Hsp70/40 actively dissolve larger aggregates into smaller ones in an ATP-dependent manner (32). We cannot discriminate between the two mechanisms, which might not be mutually exclusive, by our experiments.
The Dynamics between Hsp104 and the Aggregates-We successfully visualized the individual association/dissociation events of Hsp104 on the immobilized aggregates at the singlemolecule level (Fig. 5). The single-molecule experiments on Hsp104 dynamics revealed several novel findings, providing insights into the functional mechanism of Hsp104. First, statistical analysis showed that Ssa1/Ydj1 drastically stimulated the association of Hsp104 with the aggregates (Fig. 6A), which is consistent with a model in which the Ssa1/Ydj1 bichaperone system recruits Hsp104 to the aggregates (14,15).
Second, the fitting of the dwell times clearly showed that the lifetimes of the Hsp104-aggregate complex are divided into two groups: short (ϳ4 s) and long (ϳ30 s), regardless of the presence of Ssa1/Ydj1 (Fig. 6B). A statistical analysis of the individual events observed in the single-molecule experiment extracted these two kinetic parameters, which would be hidden in ensemble methods. Our biochemical assays showed that the turnover rate of ATP hydrolysis by Cy3-Hsp104 is around ϳ16/min per Hsp104 monomer in the presence of ␤-casein, in which an Hsp104 monomer hydrolyzes one ATP every ϳ4 s. Considering that the ATP-bound Hsp104 bound to the substrate with high affinity, as revealed by the experiment using the Hsp104 trap mutant (Fig. 5, B and C), the short lifetime (ϳ4 s) would correspond to a transient binding event in which Hsp104 dissociates from the aggregates just after the completion of the single turnover of the Hsp104 monomer. In contrast, the long lifetime (ϳ30 s) would reflect multiple turnovers of Hsp104 monomers where Hsp104 associates with the aggregates without being released after every single turnover. We suggest that the short lifetime is not responsible for the disaggregation because the association of a single Hsp104 monomer with the substrate would be too short for the ATP-dependent threading of substrates. In contrast, the long lifetime (ϳ30 s) would reflect the functional association of Hsp104 with the aggregates for the disaggregation, where the monomers in the Hsp104 hexamer would "processively" bind to the aggregates.
Finally, Ssa1/Ydj1 increased the long lifetime fraction of the Hsp104-aggregate complex (Fig. 6B). As mentioned above, the lifetimes themselves were not affected by the presence of Ssa1/ Ydj1. How can we reconcile these single-molecule findings with previous biochemical data? A recent detailed analysis of E. coli ClpB revealed that the activation of ClpB is regulated by the binding of DnaK (19). The conformations of ClpB exist in an equilibrium between the "repressed" and "derepressed" states, which are associated with the low-and high-activity states, respectively (42). DnaK binding stabilizes the derepressed state of ClpB, leading to efficient disaggregation. On the basis of the ClpB scenario, we propose that the long lifetime would be linked to the derepressed state of Hsp104. Therefore, the presence of Ssa1 increased the long lifetime fraction of Hsp104 because Ssa1 binding to Hsp104 shifts the equilibrium to the derepressed state of Hsp104.