Substrate Binding to the Molecular Chaperone Hsp104 and Its Regulation by Nucleotides*

The Hsp104 protein from Saccharomyces cerevisiae is a member of the Hsp100/Clp family of molecular chaperones. It mediates the solubilization of aggregated proteins in an ATP-dependent process assisted by the Hsp70/40 system. Although the principal function of Hsp104 is well established, the mechanistic details of this catalyzed disaggregation are poorly understood. In this work, we have investigated the interaction of Hsp104 with reduced, carboxymethylated α-lactalbumin (RCMLa), a permanently unfolded model substrate. Our results demonstrate that the affinity of Hsp104 toward polypeptides is regulated by nucleotides. In the presence of ATP or adenosine-5′ -O-(3-thiotriphosphate), the chaperone formed complexes with RCMLa, whereas no binding was observed in the presence of ADP. In particular, the occupation of the N-terminally located nucleotide-binding domain with ATP seems to be crucial for substrate interaction. When ATP binding to this domain was impaired by mutation, Hsp104 lost its ability to interact with RCMLa. Our results also indicate that upon association with a polypeptide, a conformational change occurs within Hsp104 that strongly reduces the dynamics of nucleotide exchange and commits the bound polypeptide to ATP hydrolysis.

Molecular chaperones are important constituents of the cellular protein quality control system (1,2). In response to severe growth conditions, the synthesis of many chaperones is up-regulated to cope with potentially harmful consequences of cellular stress such as protein unfolding and aggregation. The Hsp100/Clp chaperone family, a subclass of the AAA ϩ 3 proteins, is involved in the unfolding and subsequent degradation of misfolded or damaged polypeptides (3,4) as well as in the resolubilization of protein aggregates (5)(6)(7). The high degree of sequence homology suggests that, despite their diverse functions, Hsp100/Clp proteins employ a common, ATP-dependent mechanism to exert their biological roles, for which the disruption of non-covalent interactions appears to be central (8).
Among the Hsp100/Clp chaperones involved in cellular protein disaggregation, Hsp104 of Saccharomyces cerevisiae and its bacterial homologue ClpB are best understood on the molecular level. Like other members of the Hsp100/Clp family, both proteins assemble into ringshaped hexamers. Several studies have shown that disaggregation is a complex process requiring the assistance of DnaK/DnaJ/GrpE in bacteria and Ssa1/Ydj1,Sis1 in the yeast system (5,6). Two models have been proposed to describe the action of Hsp104/ClpB in substrate resolubilization. According to the "molecular crowbar" hypothesis, the flexible linker arms located at the perimeter of the Hsp104/ClpB oligomer serve as binding sites for aggregated polypeptides. In response to nucleotide binding and hydrolysis, a conformational change in these arms could provide the mechanical force necessary to disrupt protein aggregates (9 -11). More recent findings favor an unfolding/threading mechanism similar to that observed for the protease-associated Hsp100 proteins ClpA and ClpX (12,13). In this model, a single polypeptide chain is extracted from an aggregate by translocation through the central channel of Hsp104/ClpB (14,15). These two mechanisms, however, are not mutually exclusive. In this case, multiple binding sites for polypeptide substrates may exist, e.g. in the linker and at the entrance and exit of the central channel.
Hsp104 and ClpB contain two nucleotide-binding domains (NBDs), the sequences of which are highly conserved among Hsp100/Clp proteins (16). Although mutational analyses have shown that both domains are essential for chaperone activity (17)(18)(19), their respective roles in disaggregation are rather ill defined. In the case of the yeast chaperone, NBD1 was shown to be highly active under steady-state conditions, whereas NBD2 is a weak ATPase and seems to have more of a regulatory function by promoting the nucleotide-dependent hexamerization of the chaperone (20,21). In the case of ClpB from Escherichia coli both NBDs seem to contribute to the steady-state ATPase to a similar extent. Also, nucleotide binding to NBD1 rather than to NBD2 was found to be important for oligomerization (19,22). To what degree these differences are relevant for the mechanism of disaggregation is arguable, as recent studies indicate an intimate cross-talk between both NBDs (23).
In this study, we have investigated the interaction of an unfolded polypeptide with Hsp104. Based on these experiments, we propose a model that links ATP binding and hydrolysis by Hsp104 with substrate binding and release. Our results show that, with respect to polypeptide binding, Hsp104 cycles between a low affinity ADP-bound state and a high affinity, ATP-bound state. Although the latter form of Hsp104 is barely populated under conditions of steady-state hydrolysis, it accumulates in the presence of a slowly hydrolyzable ATP analogue such as ATP␥S or when a mutant defective in ATP hydrolysis is used.
We also determined the contribution of the NBDs to substrate binding. When ATP binding to NBD1 was impaired by mutation, Hsp104 lost its ability to interact with substrates, whereas a mutant defective in nucleotide binding to NBD2 showed no such defect. Once the substrate is bound to the chaperone, it becomes kinetically trapped. Its release depends on the hydrolysis of the bound ATP, thus ensuring that the energy generated by ATP hydrolysis can be used for the remodeling of the substrate. This tight coupling between ATP binding and hydrolysis on the one side and substrate binding and release on the other side may reflect a general feature of AAA ϩ ATPase family members.

MATERIALS AND METHODS
Cloning of Hsp104 Variants and Purification-Site-directed mutagenesis of Hsp104 was performed using polymerase chain reaction with the QuikChange TM site-directed mutagenesis kit (Stratagene). A plasmid carrying the Hsp104 gene was used as template for mutagenesis (24). The following mutations were introduced using appropriate mutagenesis primers: 1) K218T (AAG to ACC), 2) E285Q (GAA to CAG), and 3) K620T (AAA to ACC). The Hsp104 TRAP variant was obtained by introducing the mutation E687Q (GAA to CAG) into the gene of the E285Q single mutant. All mutations were verified by DNA sequencing. After overproduction in E. coli strain BL21-Codon-Plus(DE3)-RIL (Stratagene) wild type and mutant forms of Hsp104 were purified as described previously (24). All concentrations of Hsp104 are given with respect to the monomeric form.
Fluorescein Labeling of ␣-Lactalbumin-10 mg/ml ␣-lactalbumin (Sigma) were labeled by adding 2.3 mM fluorescein isothiocyanate as described by the manufacturer (Molecular Probes). The protein was purified via Sephadex G25 column (Amersham Biosciences) equilibrated in 0.2 M Tris/HCl, 2 mM EDTA, pH 7.5, concentrated to a final concentration of 1.3 mM and carboxymethylated (see below).
Preparation of Reduced, Carboxymethylated ␣-Lactalbumin (RCMLa)-1.3 mM ␣-lactalbumin was unfolded and reduced in 0.2 M Tris/HCl, 7 M guanidinium chloride, 20 mM dithiothreitol, 2 mM EDTA, pH 8.7, for 90 min at room temperature. Subsequently, iodoacetic acid was added to a final concentration of 0.1 M. After 20 min of incubation under dark room conditions, the reaction was quenched with an excess of reduced glutathione in 0.2 M Tris/HCl, pH 7.5. The derivatized protein was purified using a Sephadex G25 column equilibrated with 20 mM sodium phosphate, pH 7.5.
Fluorescence Anisotropy-Complex formation between Hsp104 and RCMLa was analyzed using fluorescence anisotropy. Hsp104 and f-RCMLa at various concentrations were incubated in standard buffer at 30°C for 5 min. Subsequently, ATP, ATP␥S, or ADP was added at different concentrations. The increase in fluorescence anisotropy upon f-RCMLa binding to Hsp104 was measured in a thermostated Fluoromax-3 fluorescence spectrometer (Jobin-Yvon-Horiba, Munich, Germany) equipped with autopolarizers. Monochromators were set to 494 nm (excitation) and 515 nm (emission), respectively. Once the anisotropy signal had reached the steady-state level, release of bound f-RCMLa was monitored by following the decrease in anisotropy after addition of a 150-fold excess of unlabeled RCMLa over f-RCMLa or a 40-fold excess of ADP over ATP.
ATP␥S Hydrolysis-Rates of ATP␥S hydrolysis were determined by incubating 5 M Hsp104 and 30 M RCMLa in standard buffer at 30°C. The reaction was started by adding 2.5 mM ATP␥S containing 32 nM [ 35 S]ATP␥S. Aliquots (5 l) were taken at 0, 20, 40, 60, and 90 min, and hydrolysis was stopped with 10 l of 50 mM Tris/HCl, pH 7.5, 100 mM EDTA. To determine ratios of ATP␥S and released thiophosphate, reaction mixtures were separated on polyethylene terephthalate-cellu-lose TLC plates using 0.4 M LiCl, 1.5 M formic acid as the mobile phase. Spot intensities were quantified on a phosphor screen with a Typhoon PhosphorImager (Amersham Biosciences).
Analytical Ultracentrifugation-Sedimentation velocity experiments were performed in a Beckman Optima XLI analytical ultracentrifuge equipped with UV-visible and interference optics. Rotation speed for sedimentation was 50,000, 55,000, or 60,000 rpm (Ti-60 rotor). Data scans were recorded continuously throughout the experiment using the absorption optical system at 280 nm or the Rayleigh interference optics. Alumina double sector centerpieces were used that contained buffer and protein solution (5 M) in separate chambers. Measurements were carried out in standard buffer at 30°C. In addition, samples contained 1 mM ATP, 1 mM ADP, or 2.5 mM ATP␥S. Data analysis was performed using the ultrascan software by B. Demeler (University of Texas Health Science Center, San Antonio, TX).
Dynamic Light Scattering-The diffusion coefficient of Hsp104 was determined by dynamic light scattering. Experiments were performed at a scattering angle of 90°using an AXIOS-150 apparatus (Triton-Hellas) with a 35-milliwatt diode laser operating at 658 nm. Conditions were 5-20 M Hsp104 WT , with and without 1 mM ADP in standard buffer at 20°C. The particle distributions were obtained using Provencher's regularized Laplace inversion CONTIN algorithm for the autocorrelation function (26).
The diffusion coefficient D and the sedimentation coefficient s 20,w were used to calculate the molecular mass (M) of Hsp104 according to the Svedberg equation, in which v is the partial specific volume of Hsp104 (0.737 ml/g) and is the density of water.

Substrate Binding to Hsp104 Is Dependent on the Presence of ATP-
Hsp104 was shown to support the refolding of aggregated enzymes such as luciferase in vitro (5), but attempts to obtain stable chaperone⅐ polypeptide complexes have been futile so far. To investigate the Nucleotide binding and hydrolysis properties of Hsp104 WT and Hsp104 TRAP Apparent K D values were determined by isothermal titration calorimetry as described previously (24). k cat of ATP hydrolysis was measured using a regenerative assay. All experiments were carried out at 30°C in standard buffer. Data for the wild type protein were taken from Ref. 24. ND, not determined.

E285Q/E687Q
Wild type requirements for polypeptide binding to Hsp104, we used reduced carboxymethylated ␣-lactalbumin (RCMLa) as a substrate. This protein is highly soluble and permanently unfolded (27,28), and no side reactions such as folding or aggregation can compete with chaperone binding. Furthermore, RCMLa has been shown to moderately stimulate ATP hydrolysis by Hsp104 (29). As expected, we were unable to detect binding of f-RCMLa to Hsp104 using size exclusion chromatography, irrespective of whether ATP, ADP, or no nucleotide was present. Only when the slowly hydrolyzable ATP analogue ATP␥S was used, a slight decrease in the elution time of f-RCMLa was observed (data not shown) indicating a transient interaction with the chaperone. This suggested that only the ATP state of Hsp104 is capable of polypeptide binding, similar to what has been found for other AAA ϩ ATPases including the bacterial Hsp104 homologue ClpB (30,31). Because Hsp104 is a fast ATPase with a k cat of ϳ70 min Ϫ1 (20,24), the lifetime of the ATP state presumably is too short to detect substrate complexes in SEC analysis. We therefore constructed an Hsp104 variant that can still bind ATP (but is unable to hydrolyze it) by replacing two conserved Glu residues in the Walker B motifs of both NBDs with Gln (E285Q/E687Q) (31,32). The data summarized in TABLE ONE demonstrate that this double mutant, Hsp104 TRAP , binds ADP with a slightly higher affinity than the wild type protein but displays no ATPase activity. Moreover, the structure of Hsp104 was found to be unaffected by the mutations we introduced, as assessed by far UV CD spectroscopy (see supplemental Fig. 1).
When the trap mutant was incubated with f-RCMLa and ATP, we detected a substrate-specific peak ( ϭ 494 nm) at the position of the Hsp104 hexamer in SEC analysis, i.e. the formation of stable Hsp104 TRAP ⅐ATP⅐RCMLa complexes (Fig. 1A). Apparently, the inability of the mutant to hydrolyze bound ATP traps it in a state with a high affinity for polypeptides. When ADP or no nucleotide was present dur-ing incubation, no complexes were observed, indicating that substrate binding to Hsp104 TRAP is strictly ATP-dependent.
SEC-HPLC has the disadvantage of being a slow non-equilibrium method, and labile complexes may irreversibly dissociate during the time course of analysis. Thus, we employed fluorescence anisotropy as an alternative probe to monitor binding of f-RCMLa to Hsp104. The anisotropy signal of a fluorophore depends on its size or more precisely on its rotational correlation time R (33). Hence, free f-RCMLa (15 kDa) should display a smaller fluorescence anisotropy compared with the large Hsp104⅐f-RCMLa complex (Ͼ600 kDa). The anisotropy signal of f-RCMLa did not change when Hsp104 TRAP was added without nucleotide (data not shown), supporting our finding that substrate binding is ATP-dependent. Upon addition of ATP fluorescence anisotropy strongly increased, indicating an increase in R of f-RCMLa because of its binding to the chaperone (Fig. 1B). Again, ADP was not able to trigger complex formation. When Hsp104 WT was used instead of the trap mutant, binding of RCMLa was only observed in the presence of ATP␥S but not ATP (Fig. 1C), consistent with our SEC-HPLC data.
We also examined whether the different substrate affinities observed in the presence of ADP and ATP are related to changes in the oligomeric state of the chaperone. In both ADP and ATP, analytical ultracentrifugation yielded sedimentation coefficients, s, of ϳ16 S (Fig. 2 and TABLE TWO), in agreement with data obtained for hexameric ClpB (34). The integral distribution plot (Fig. 2B) shows that Hsp104 sediments as a single species. Nevertheless, an s value of ϳ16 S appears to be remarkably small for a 612-kDa protein such as hexameric Hsp104. An accurate determination of the molecular weight from sedimentation velocity data also requires knowledge of the translational diffusion coefficient D. Using dynamic light scattering, we obtained a value of D ϭ 2.5⅐10 Ϫ7 cm 2 /s. In combination with s ϭ 16 S, this yields a molecular mass of 596 kDa, demonstrating that Hsp104 is completely hexameric under the conditions of our experiments. This conclusion is supported by sedi-   mentation equilibrium analysis of Hsp104, in which Hsp104 was found to consist of a single species of 611 kDa (see supplemental Fig. 2). Thus, hexamer formation alone (as in the presence of ADP) is not sufficient to trigger substrate binding. Rather an ATP-induced conformational change within the hexamer must be responsible.
NBD1 Regulates the Affinity of Hsp104 for Polypeptides-As Hsp104 contains two nucleotide-binding domains, we tried to assess their roles in polypeptide binding. The mutation of two conserved lysine residues in the Walker A motifs of Hsp104, Lys 218 and Lys 620 , strongly reduces the affinity of the affected NBD for nucleotides, 4 similar to what has been found for Thermus thermophilus ClpB (35). Consequently, the respective mutants only bind nucleotides in the wild type-like NBD but not in the mutated NBD. Fig. 3A shows that the K620T variant of Hsp104, in which only NBD1 can be occupied by nucleotide, still forms complexes with f-RCMLa in the presence of ATP␥S. The K218T mutation, in contrast, results in a complete loss of polypeptide binding. These findings strongly suggest that polypeptide substrates can only bind to Hsp104 when NBD1 is occupied with ATP or ATP␥S.
To confirm the importance of NBD1, we employed a more stringent assay for Hsp104 function, the disaggregation of luciferase. The K218T mutant, which was unable to bind RCMLa in our anisotropy experiments, also failed to recover any luciferase activity (Fig. 3B). The K620T variant exhibited a significant although reduced disaggregation activity when compared with Hsp104 WT . This demonstrates that whereas a functional NBD2 may be dispensable for substrate interaction, it is necessary for efficient disaggregation. The inability of the NBD1 mutant to resolubilize protein aggregates is readily explained by our finding that substrate interaction is crucially dependent on ATP binding to NBD1.
Release of Bound Polypeptide Is Triggered by ATP Hydrolysis-After having determined the requirements for substrate binding to Hsp104, we next investigated the mode of substrate release. In principle, polypeptides could either dissociate from Hsp104 before ATP hydrolysis occurs or remain associated during the hydrolysis step. Likely, only the second pathway is productive, as current models of disaggregation suggest that the energy provided by ATP hydrolysis is used to induce a conformational change in the bound polypeptide (10,36).
To determine the fate of the bound substrate, ternary complexes between Hsp104 and f-RCMLa were formed as described above. Once fluorescence anisotropy had reached a plateau indicating steady state, we chased the complexes with either (i) a 150-fold excess of unlabeled RCMLa over f-RCMLa or with (ii) a 40-fold excess of ADP over ATP. In both cases, f-RCMLa should be released irreversibly, either because (i) Hsp104 will preferentially reassociate with unlabeled RCMLa or because (ii) excess ADP prevents the binding of ATP necessary to maintain the high affinity state of the chaperone. When experiments were carried out with hydrolytically inactive Hsp104 TRAP , the anisotropy signal did not change after the addition of unlabeled RCMLa, irrespective of whether complexes had been formed with ATP or ATP␥S. Apparently, the dissociation rate of f-RCMLa is very small and prevents the exchange of bound f-RCMLa with free, unlabeled substrate (Fig. 4A). A very similar result was obtained when the chase was carried out with excess ADP (Fig. 4B).
In contrast to the trap mutant, Hsp104 WT ⅐ATP␥S was able to release bound f-RCMLa after addition of either unlabeled RCMLa or ADP, as indicated by the decrease in the anisotropy signal (Fig. 4, filled squares). Intriguingly, the rate constant of dissociation (ϳ0.12 min Ϫ1 ) was the same for both types of chase experiments suggesting that the same molecular step triggers release. Furthermore, our observation that only hydrolytically active Hsp104 can exchange the bound substrate suggests that polypeptide release is tightly coupled with ATP hydrolysis and the concomitant transition of the chaperone into its low affinity ADP state. Under the conditions of our anisotropy experiments, Hsp104 WT hydrolyzes [ 35 S]ATP␥S ϳ30 times more slowly (k app ϭ 0.9 min Ϫ1 ) than ATP. This rate is ϳ6 times higher than the rate of substrate release, suggesting that multiple hydrolysis events may be required.
One could argue that the release of f-RCMLa from Hsp104 WT ⅐ATP␥S simply reflects that these ternary complexes are more dynamic than 4 V. Grimminger and S. Walter, unpublished results.  Hsp104 TRAP ⅐ATP complexes and is not related to hydrolysis of the bound nucleotide. Because the rate of ATP␥S hydrolysis is significantly larger than the rate of substrate dissociation, ATP␥S becomes hydrolyzed before the polypeptide is released.
Polypeptide Binding to Hsp104⅐ATP Reduces the Dynamics of Nucleotide Exchange-The observation that ADP is not able to trigger the dissociation of f-RCMLa from ternary Hsp104 TRAP ⅐ATP complexes (Fig. 4B) is remarkable and implies that the exchange of nucleotide in NBD1 must be very slow. Otherwise, ADP should displace ATP and induce the low affinity state of Hsp104. To investigate whether slow nucleotide exchange is a direct consequence of the bound polypeptide, we carried out order-of-addition experiments (see flow chart in Fig. 5).
First, Hsp104 TRAP was incubated with ATP for 10 min to generate the high affinity state. When f-RCMLa was added at this stage (Fig. 5C, left branch of the chart), we observed a large increase in fluorescence anisotropy because of the binding of the labeled polypeptide. As described above, even a 40-fold excess of ADP was not able to cause a significant release of f-RCMLa. This suggests that the ATP in the ternary complex can exchange only very slowly with free ADP. However, when Hsp104 TRAP ⅐ATP was allowed to undergo 30 s of nucleotide exchange with a 40-fold excess of ADP before f-RCMLa was added (Fig. 5C, right branch of the chart), fluorescence anisotropy remained at low levels. Apparently, this short time of exchange was sufficient to convert virtually all Hsp104 TRAP molecules into their low affinity ADP form. Accordingly, replacement of ATP by ADP in NBD1 must occur very rapidly in the absence of a bound polypeptide. To rule out that the slow nucleotide exchange is because of an unusually high polypeptide affinity of the trap mutant, we performed similar order-of-addition experiments with Hsp104 WT in the presence of ATP␥S, i.e. under steady-state conditions (Fig. 5B). Again, an ADP chase of 30 s prior to the addition of f-RCMLa was sufficient to block substrate binding. When ADP was added after ternary complexes had been formed, the quantitative release of the bound substrate required more than 1000 s. As described above, this low dissociation rate presumably reflects ATP␥S hydrolysis rather than exchange of ATP␥S against ADP.

DISCUSSION
Substrate Binding to Hsp104 Is ATP-dependent-A central feature of molecular chaperones is the controlled binding and release of their client proteins (1). In the case of ATP-dependent chaperones, such as GroEL (Hsp60) or DnaK (Hsp70), this is often achieved by conformational changes induced by either nucleotide binding or hydrolysis (37,38). Clearly, substrate binding to the chaperone Hsp104 is also controlled by nucleotides. In the presence of ATP, Hsp104 adopts a conformation with high affinity for unfolded polypeptides, whereas binding is weak in the presence of ADP or in the absence of nucleotide (cf. model in Fig. 6). This nucleotide-triggered affinity switch seems to be conserved among Hsp100/Clp chaperones, as very similar findings have been reported for ClpA, ClpB, ClpX, and a number of other AAA proteins such as N-ethylmaleimide-sensitive factor (30 -32, 39, 40).
Strikingly, binding of RCMLa to Hsp104 WT was only detected in the presence of ATP␥S but not ATP. Because we used fluorescence anisotropy to monitor complex formation, the transient nature of substrate binding due to ongoing ATP hydrolysis cannot account for this result.
Rather, it appears that under conditions of ATP hydrolysis, the steadystate concentration of the ternary complex is below the detection limit. There are two possible explanations: (i) If ATP hydrolysis is much faster than nucleotide exchange, the low affinity ADP state of Hsp104 will predominate under steady-state conditions. This view is corroborated FIGURE 6. Model for polypeptide binding to Hsp104. Hsp104 rapidly cycles between an ADP state with low affinity toward polypeptides (top right) and an ATP state with high affinity toward polypeptides (bottom right). The cycling time, as judged by the k cat for ATP hydrolysis under steady-state conditions, is in the range of 1 s, and nucleotide exchange should be at least as fast. When a polypeptide substrate is present, it specifically binds to the high affinity form of the chaperone generating the ternary Hsp104⅐ATP⅐substrate complex (left). In the case of Hsp104 WT , this reaction appears to be quite inefficient, presumably because ATP hydrolysis competes with binding. Our data suggest that substrate binding induces a conformational change in Hsp104 that results in markedly reduced exchange dynamics for both substrate and nucleotide. As a consequence, the ternary complex is committed to hydrolysis. This lock-in mechanism ensures that the energy provided by ATP hydrolysis can be efficiently used for substrate processing. by the unusually high K m value of Hsp104 for ATP, which was determined to be in the range of 5-10 mM (24,41). (ii) ATP hydrolysis may be significantly faster than polypeptide binding, and thus the chaperone will switch back to the low affinity ADP state before RCMLa can bind (Fig. 6). These explanations are not mutually exclusive, and in both cases, a decrease in the hydrolysis rate, either by replacing ATP with ATP␥S or by replacing Hsp104 WT with the trap mutant, will cause an increase in substrate binding as observed in our experiments.
ATP Hydrolysis Is Required for Substrate Release-Once the polypeptide substrate is associated with Hsp104, it remains bound until ATP hydrolysis occurs. When hydrolysis was blocked as in the case of the Hsp104 TRAP mutant, no spontaneous release of labeled RCMLa could be observed. From a biological point of view, this commitment of the ternary complex ensures that the substrate remains associated with Hsp104 throughout hydrolysis. Therefore, the energy provided by ATP hydrolysis can be efficiently transferred to the substrate and used for its disaggregation, unfolding, or threading through the central pore. Commitment requires that not only the substrate is bound tightly but also ATP. Otherwise, ATP could be replaced by ADP before hydrolysis occurs, resulting in the non-productive release of the polypeptide because of the transition of chaperone to the low affinity state. Our experiments with Hsp104 TRAP show that once a ternary complex (Hsp104 TRAP ⅐ATP⅐RCMLa) has been formed, it cannot be dissociated by the addition of excess ADP, indicating that nucleotide exchange is very slow in this complex. Importantly, these statements only apply to NBDs that must maintain their ATP state to keep the substrate bound. Nucleotide exchange in "non-relevant" NBDs is not detected with our experimental setup.
In the absence of a bound polypeptide, nucleotide exchange in Hsp104 must be at least as fast as the rate of steady-state hydrolysis, i.e. in the range of 0.1-1 s Ϫ1 . Indeed, when we chased a high affinity Hsp104 TRAP ⅐ATP complex with an excess of ADP, nucleotide exchange and the concomitant conversion to the low affinity ADP form were complete within seconds. These differences in nucleotide exchange kinetics suggest that polypeptide binding induces a conformational change in Hsp104, which prevents the dissociation of ATP from "relevant" NBDs. This altered conformation may also be the reason why RCMLa and other substrates stimulate the ATP hydrolysis of Hsp104 and ClpB (29).
NBD1 Is the Primary Regulator for RCMLa Binding to Hsp104-Like all class I Hsp100/Clp chaperones, Hsp104 possesses two nucleotidebinding domains, both of which could be involved in the control of polypeptide binding. Our results provide evidence that NBD1 serves as the primary regulator. First, a mutant that does not bind nucleotide in NBD1 was also unable to associate with RCMLa, whereas the corresponding mutation in NBD2 did not interfere with substrate binding. The importance of NBD1 is further supported by the inefficient binding of RCMLa to Hsp104 WT ⅐ATP. In our model, this is because ATP hydrolysis in NBD1 is significantly faster than either nucleotide exchange and/or polypeptide binding (see above). Because previous studies (20,21) have shown that NBD2 is a very slow ATPase under steady-state conditions, this points to NBD1 as the domain responsible for substrate affinity.
Hexamer formation was shown to be essential for Hsp104 chaperone function. A potential concern therefore is that the inability of the K218T mutant to bind polypeptide may be caused by an oligomerization defect. Two points strongly argue against this. (i) Hexamerization alone is not sufficient for polypeptide binding. Hsp104 is hexameric in the presence of ADP and even in the absence of nucleotides at the protein concentrations we used (see TABLE TWO). Under none of these conditions was substrate binding observed. (ii) Lindquist and co-workers (16) demonstrated previously that the K218T mutant displays normal oligomerization behavior, whereas the K620T mutant had an oligomerization defect, which is apparent only at low protein concentration. With respect to polypeptide binding, the effects of the mutations are reversed. The K218T mutant shows a severe defect, whereas the K620T mutant behaves similarly to the wild type protein.
The exact location of the polypeptide binding site on Hsp104/ClpB is not known, although several attempts have been made to address this issue. (i) Cashikar et al. (29) showed that poly-L-Lys binds to the C-terminal domain of Hsp104. This binding site, however, appears to be different from the one investigated in our study, as no dependence on ATP has been reported and a truncated version of Hsp104 lacking NBD1 (and the linker domain, see below) was still able to bind poly-L-Lys. (ii) Based on their crystal structure of T. thermophilus ClpB, Tsai and co-workers (10) suggested that the middle/linker domain of the chaperone recognizes aggregated proteins and pries them in a crowbarlike mechanism. Because the linker is inserted in the C-terminal portion of NBD1, a change in the nucleotide status of NBD1 may alter the polypeptide affinity of the linker, consistent with our data. (iii) Recently, Mogk and colleagues (42) showed for ClpB from E. coli that peptide substrates can be cross-linked in an ATP-dependent fashion to aromatic residues of NBD1 facing the central pore. Assuming similar roles of NBD1 in both proteins, this would provide a direct structural link between nucleotide binding and polypeptide affinity. The importance of the N-terminal region for substrate binding is also evident from a recent study on ClpB. Mutants, in which the N-terminal domain preceding NBD1 was deleted, showed a marked decrease in affinity toward aggregated polypeptides (43).