The amino-terminal domain of ClpB supports binding to strongly aggregated proteins.

Bacterial heat-shock proteins, ClpB and DnaK form a bichaperone system that efficiently reactivates aggregated proteins. ClpB undergoes nucleotide-dependent self-association and forms ring-shaped oligomers. The ClpB-assisted dissociation of protein aggregates is linked to translocation of substrates through the central channel in the oligomeric ClpB. Events preceding the translocation step, such as recognition of aggregates by ClpB, have not yet been explored, and the location of the aggregate-binding site in ClpB has been under discussion. We investigated the reactivation of aggregated glucose-6-phosphate dehydrogenase (G6PDH) by ClpB and its N-terminally truncated variant ClpBDeltaN in the presence of DnaK, DnaJ, and GrpE. We found that the chaperone activity of ClpBDeltaN becomes significantly lower than that of the full-length ClpB as the size of G6PDH aggregates increases. Using a "substrate trap" variant of ClpB with mutations of Walker B motifs in both ATP-binding modules (E279Q/E678Q), we demonstrated that ClpBDeltaN binds to G6PDH aggregates with a significantly lower affinity than the full-length ClpB. Moreover, we identified two conserved acidic residues at the surface of the N-terminal domain of ClpB that support binding to G6PDH aggregates. Those N-terminal residues (Asp-103, Glu-109) contribute as much substrate-binding capability to ClpB as the conserved Tyr located at the entrance to the ClpB channel. In summary, we provided evidence for an essential role of the N-terminal domain of ClpB in recognition and binding strongly aggregated proteins.

Thus, the mechanism of ClpB is analogous to that of ClpA, ClpX, HslU (ClpY), and a number of eukaryotic AAAϩ ATPases that translocate and unfold substrates and deliver them to the associated peptidases for degradation (12). The crucial remaining questions concern the role of DnaK/DnaJ/GrpE in the ClpB-assisted substrate unfolding/reactivation and the mechanism of substrate recognition by ClpB.
Because aggregated proteins, unlike degradation substrates, are not tagged, ClpB should possess unique aggregate-recognition capabilities that are distinct from those of ClpA, ClpX, and HslU. A protein machine that is capable of unfolding any polypeptide whose termini are exposed would be harmful rather than beneficial in vivo. Studies on peptide binding to ClpB revealed preferential interactions with positively charged and, to a lower extent, with aromatic residues (13). A tyrosine and a pair of acidic residues located at the N-terminal entrance to the ClpB channel are involved in substrate binding (13). However, that group of amino acids is located within the AAAϩ module (between Walker A and Walker B motifs) and is found in many AAAϩ ATPases, regardless of their function (13). This suggests that those conserved residues may play a role in initiating the ratcheting mechanism of substrate threading rather than in substrate recognition.
We hypothesized that the N-terminal domain of ClpB may contain sequence or structural motifs that are essential for the capability of ClpB to recognize and bind aggregated proteins. Sequence similarity between the N-terminal regions of ClpB and other AAAϩ ATPases is low and limited to a few short amino acid motifs (14). The role of the N-terminal domain of ClpB is unknown and has been under an intense discussion. Importantly, the ClpB transcript contains an alternative translation initiation site at the N terminus of the first AAAϩ module. Expression of the ClpB gene results in production of the full-length 95-kDa ClpB as well as the truncated 80-kDa ClpB⌬N (15,16). Thus, investigating the function of the N-terminal domain may also help explain the physiological role of the ClpB isoform lacking that domain.
Published data suggest that the N-terminal domain plays a role in the reactivation of aggregated substrates (17). Mutations of selected residues within the N-terminal domain of ClpB inhibit its chaperone activity and casein-induced activation of ClpB ATPase (14,18). The isolated N-terminal domain of ClpB binds to misfolded proteins and displays "chaperone-like" aggregation-suppressing properties (19). In contrast, other studies implied that the N-terminal domain is not essential for the function of ClpB. Both the full-length ClpB and ClpB⌬N contributed to the survival of bacteria during heat shock (20). Some aggregated substrates could be reactivated in vitro by the bichaperone system containing either ClpB or ClpB⌬N (21). Thus, the N-terminal domain does not appear to play a direct role in the mechanism of substrate threading, which is mediated by the channel-forming domains of the AAAϩ modules. Finally, mutant substrate-trap variants of ClpB and ClpB⌬N interacted with selected protein substrates to a similar extent (22). In summary, the role of the N-terminal domain of ClpB remains controversial.
In this study, we tested the hypothesis that the above discrepancy among results from different assays comparing the full-length ClpB with ClpB⌬N arises from differences in properties of aggregates presented to ClpB as potential substrates. We examined the chaperone activity and aggregate-binding capability of ClpB and ClpB⌬N using a substrate with a controllable extent of aggregation. We found that the N-terminal domain of ClpB significantly contributes to the aggregatebinding affinity and becomes particularly essential for binding large protein aggregates.

EXPERIMENTAL PROCEDURES
Proteins-Previously published procedures were used to produce and purify the Escherichia coli chaperones, ClpB and ClpB⌬N (17), DnaK (23), and DnaJ (24). GrpE was obtained from StressGen Biotechnologies (Victoria, Canada). Site-directed mutagenesis of ClpB was performed using the QuikChange method (Stratagen), and the mutated ClpB variants were purified as described for wild type ClpB (17). Glucose-6phosphate dehydrogenase (54 kDa, monomer) (G6PDH) 3 from Leuconostoc mesenteroides was obtained from Sigma. Protein concentrations were determined spectrophotometrically and are given in monomer units.
G6PDH Aggregate Production-To prepare aggregated G6PDH, 5 l of the 600 M stock solution was mixed with an equal volume of the heated denaturation buffer (10 M urea, 16% glycerol, and 40 mM dithiothreitol). The mixture was incubated at 47°C for 5 min, at which time 90 l of the refolding buffer (50 mM triethanolamine/Cl, pH 7.5, 20 mM Mg(OAc) 2 , 30 mM KCl, 1 mM ␤-mercaptoethanol, and 1 mM EDTA) without or with 2 mM ATP or ADP was added, and the G6PDH sample was mixed vigorously and incubated at 47°C for a variable period of time. The solution was then mixed briefly and incubated on ice for 2 min followed by centrifugation at 4°C for 10 min at 13,000 ϫ g.
Gel-filtration Chromatography-Soluble G6PDH aggregates were chromatographed with or without ClpB or its variants. Protein samples were injected onto a 25-cm Pharmacia HR10/30 column packed with Superose 6 gel-filtration medium (Amersham Biosciences) equilibrated with the refolding buffer with 2 mM ATP or ADP or without nucleotides. Elution at 0.5 ml/min was performed at room temperature using a Shimadzu HPLC LC10ATvp equipped with a SPD-M10Avp diodearray detector. Gel-filtration standards were from Bio-Rad.
One-minute fractions (0.5 ml) were collected from the column and analyzed by SDS-PAGE followed by Coomassie staining or Western blotting. Some eluted fractions were concentrated by precipitation with 5% trichloroacetic acid overnight at 4°C. After 13,000 ϫ g centrifugation for 30 min at 4°C, the supernatant was removed, and 1 ml of ice-cold acetone was added to the pellet. The samples were again centrifuged for 30 min under the same conditions. After removing the acetone, 20 l of 5ϫ SDS loading buffer was added to the pellet, the sample was boiled for 2 min, and analyzed by SDS-PAGE. For Western blotting, the samples were transferred to nitrocellulose membrane and incubated with rabbit polyclonal anti-ClpB antibody (7) and goat antirabbit secondary antibody conjugated to horseradish peroxidase (Southern Biotechnology Associates, Birmingham, AL). Blots were visualized by SuperSignal West Pico Chemiluminescent Substrate (Pierce).
G6PDH Reactivation Assay-Aggregates of G6PDH (22 M) were diluted 8-fold into the refolding buffer with 5 mM ATP containing no chaperones, 1 M DnaK with 0.5 M DnaJ and 0.5 M GrpE (KJE), KJE with 1.5 M ClpB, or KJE with 1.5 M ClpB⌬N. After incubation at 30°C, aliquots were withdrawn, and the G6PDH activity was measured.

RESULTS
We produced aggregates of G6PDH using a procedure analogous to that developed by Goloubinoff and co-workers (27). Native G6PDH was first unfolded in urea at 47°C and then rapidly diluted into a refolding buffer and further incubated at 47°C. As has been shown before (27), G6PDH does not spontaneously refold under such conditions, instead it misfolds and starts to aggregate. The aggregation process can be effectively arrested at different stages by rapid cooling of the protein samples. The resulting G6PDH solutions exhibited variable turbidity, but their soluble fractions showed consistent elution patterns in gel-filtration chromatography. As shown in Fig. 1, the apparent size of the aggregates depends on the time of refolding at 47°C. Immediately after the initiation of refolding, G6PDH eluted with an apparent size of ϳ200 kDa, which is consistent both with the expanded size of unfolded monomeric G6PDH and with formation of small aggregates. After 1 min of refolding, the elution pattern showed a broad distribution of particles with a maximum at ϳ1,000 kDa. After 15 min of refolding, the majority of G6PDH eluted in the void volume fraction corresponding to Ͼϳ2,000 kDa with a minor fraction of smaller aggregates.
Goloubinoff and co-workers (27) concluded that small aggregates of G6PDH could be reactivated by the DnaK/DnaJ/GrpE system alone, whereas the reactivation of large aggregates required ClpB. We compared the rates of G6PDH reactivation by different chaperones for the aggregate population prepared with 1-and 15-min refolding time (see Fig. 1). As shown in Fig. 2A (open circles), the 1-min aggregate population showed measurable G6PDH activity that did not increase over time, which indicated that the aggregate population contains a fraction of refolded G6PDH. The amount of refolded G6PDH did not increase spontaneously, which is consistent with the irreversible nature of protein aggregation. Either DnaK/DnaJ/GrpE alone or in cooperation with ClpB reactivated aggregated G6PDH, but the reactivation rate was 3 The abbreviation used is: G6PDH, glucose-6-phosphate dehydrogenase. ϳ4-fold higher when ClpB was present. The rate of G6PDH reactivation was ϳ2-fold lower when the full-length ClpB was replaced with ClpB⌬N.
In contrast to the 1-min aggregate sample, the 15-min aggregate population did not show G6PDH activity (Fig. 2B, open circles), which indicates that essentially all G6PDH aggregated during 15 min of refolding. The DnaK/DnaJ/GrpE system alone did not reactivate G6PDH, which is consistent with previous observations (27). Either ClpB or ClpB⌬N reactivated G6PDH, but the reactivation rate in this case was ϳ5-fold lower with ClpB⌬N than with the full-length ClpB. Collectively, the results in Fig. 2 indicate that the chaperone activity of ClpB⌬N is weaker than that of ClpB. The ClpB⌬N deficiency in processing aggregated G6PDH is exacerbated when the effective size of the aggregates increases (compare Figs. 1, 2A, and 2B).
To further investigate the reasons for the ClpB⌬N deficiency in reactivating large aggregates, we compared the amounts of ClpB and ClpB⌬N bound to the aggregates. Large G6PDH aggregates were prepared after the 15-min refolding, incubated with ClpB or ClpB⌬N in the presence of ATP with the ratio [G6PDH]/[ClpB] as in Fig. 2, and analyzed with gel-filtration chromatography. Without G6PDH, ClpB and ClpB⌬N eluted as hexamers with an apparent molecular weight ϳ600 kDa and did not contain large aggregated particles (Fig. 3, A, B, and D).
In the presence of aggregated G6PDH, a small amount of ClpB, but not ClpB⌬N was detected in the aggregate fractions after trichloroacetic acid precipitation (Fig. 3, C and E). This result indicates that the lower G6PDH reactivation rate of ClpB⌬N versus ClpB might be because of a lower substrate-binding affinity.
Because AAAϩ ATPases interact with their substrates in the ATPbound state (28), the amount of ClpB or ClpB⌬N found in gel-filtration fractions containing aggregated G6PDH may be affected by either substrate-binding affinity or the lifetime of the ATP-bound conformation of the ATPase. To observe the interactions between ClpB or ClpB⌬N and the aggregates in the absence of ATP hydrolysis, we employed ATPase-deficient mutants of ClpB. Weibezahn et al. (22) described a substrate-trap variant of ClpB with mutations of Walker B motifs in both AAAϩ modules (E279A/E678A). We produced analogous variants of ClpB and ClpB⌬N with the mutations of essential Walker B glutamates (E279Q/E678Q). We also produced a ClpB variant with mutations of sensor-1 motifs in both AAAϩ modules (T315A/N719A). It has been postulated that hydrogen-bonding residues of sensor-1 play a role in synchronization of ATP hydrolysis in all catalytic sites within the AAAϩ ring (29), and their mutations might, therefore, inhibit the substrate threading reaction. Indeed, single sensor-1 mutations have been found to inhibit the ATPase of yeast Hsp104 (30).  Both the double Walker B and double sensor-1 mutants of ClpB formed ATP-dependent oligomers, similar to wild type ClpB and ClpB⌬N (data not shown). However, neither mutant rescued the growth of the clpB-null strain of E. coli at 50°C (data not shown), which indicates that mutations of conserved motifs in both AAAϩ modules strongly inhibit the chaperone activity of ClpB. As shown in Fig. 4A, the sensor-1 mutations inhibited partially, and the Walker B mutations completely, the ATPase of ClpB in the presence of casein, an activator of ATPase and a substrate analog. The amount of the double sensor-1 mutant ClpB bound to G6PDH aggregates was similar to that of wild type ClpB (Fig. 4C). However, the amounts of the double-Walker B mutant ClpB eluting with G6PDH aggregates were significantly increased (Fig. 4E). We concluded that the lifetime of stable ClpB com-plexes with the aggregates is limited as long as the ClpB ATPase is active, as shown by the sensor-1 mutant, and that the inactive double Walker B mutant is indeed an efficient trap of aggregated proteins. In agreement with the properties of the ClpB trap mutant described by Weibezahn et al. (22), the interaction of double Walker B mutant ClpB with large G6PDH aggregates required ATP (data not shown). The formation of complexes of the ClpB trap with the aggregates represents the initial event of ClpB-assisted disaggregation, and its efficiency is determined by the ClpB-aggregate-binding affinity.
Because the complexes of the trap ClpB and G6PDH aggregates formed very efficiently, they could be detected with Coomassie-stained polyacrylamide gels (Fig. 5). Whereas the full-length ClpB trap eluted in the fractions containing strongly aggregated G6PDH ( Fig. 5A and a control and Fig. 6C), no trap ClpB⌬N was detected in such fractions ( Fig. 5B and a control and Fig. 6D). We concluded that the removal of the N-terminal domain significantly decreases the binding affinity of ClpB to large aggregates.
Next, we asked whether trap ClpB⌬N would preferentially interact with a specific size of G6PDH aggregate. We prepared a broad distribution of aggregates by using a 1-min refolding time (see Fig. 1). As shown in Fig. 6, B and D, the binding of trap ClpB⌬N to G6PDH aggregates not much larger than ClpB itself is evident for the elution at 19 and 20 min. However, the amounts of trap ClpB⌬N decrease significantly for earlier elution times that correspond to larger G6PDH aggregates. The full-   length trap ClpB binds with a similar efficiency to all but the largest G6PDH aggregates (Fig. 6, A and E) The binding affinity of trap ClpB⌬N is lower than that of the full-length ClpB for all aggregate sizes, and it is at least an order of magnitude lower for very large aggregates (Figs. 6E and 7). We concluded that whereas some aggregate-binding site(s) do reside in ClpB⌬N, presumably at the channel entrance (13), the N-terminal domain of ClpB contributes significantly to the aggregate-binding capability of the chaperone and supports binding to both small and large aggregates.
We have shown earlier that the conserved residues within the N-terminal domain of ClpB, Thr-7, Asp-103, and Glu-109, support the chaperone activity (14). We asked whether the interactions of trap ClpB with aggregated G6PDH are affected by mutations in the N-terminal domain. Because positively charged peptides bind preferentially to ClpB (13), we focused on the pair of acidic residues and produced a quadruple mutant D103A/E109A/E279Q/E678Q, i.e. the ClpB trap with mutations of the Asp/Glu pair in the N-terminal domain. As has been shown before, the mutations in the N-terminal domain do not destabilize its structure nor do they affect the oligomerization and the basal ATPase activity of ClpB (14). We also produced the ClpB trap variant with the mutation of the conserved Tyr at the entrance to the channel: Y251A/ E279Q/E678Q. As shown in Fig. 7, both the N-terminal mutant and the channel mutant bind less efficiently to large aggregates of G6PDH than the full-length ClpB, but more efficiently than ClpB⌬N. We concluded that the pair of acidic residues in the N-terminal domain of ClpB contributes significantly to the aggregate-binding affinity.

DISCUSSION
It is believed that the multitude of functions performed by AAAϩ ATPases is achieved by a variability of substrate-recognition domains covalently attached to the energy-transducing AAAϩ modules (6). In the case of ClpB, its N-terminal domain forms an independently folded structural unit that is attached to the first AAAϩ module with a flexible linker (8). Thus, the N-terminal domain of ClpB is a candidate for the substrate-sensing "attachment" domain, but its role in substrate binding has not been demonstrated yet. In this work, we have shown directly that the binding of ClpB to protein aggregates is supported to a high extent by molecular contacts provided by the N-terminal domain. We have also identified two residues within the N-terminal domain, Asp-103 and Glu-109, that participate in substrate binding. Fig. 8 shows the monomer structure of ClpB from Thermus ther-mophilus (8). The N-terminal domain is located at the top of the figure.
Flexible diaphragms extending into the central channel in the oligomeric ClpB could not be resolved in the structure, but their positions are indicated by the flanking residues Glu-246 and Gln-651 shown in purple. The top diaphragm contains Tyr-251 that was mutated in the experiment shown in Fig. 7, and its position defines the entrance to the channel. The ClpB channel continues through the second diaphragm at Gln651 toward its exit at the C-terminal surface of the protein (bottom of Fig. 8). As shown in this monomer structure, the Asp-103 and Glu-109 pair is located at the surface of the N-terminal domain and faces the channel entry. In the oligomeric ClpB, the ring of N-terminal domains could create a funnel-like surface involved in binding aggregated polypeptides and guiding an exposed terminus of the substrate toward the channel entrance where Tyr-251 resides.
In the crystal structure of ClpB from T. thermophilus, the orientation of the N-terminal domain varied among crystallized monomers (8), which demonstrates high mobility of the link between the N-terminal domain and the first AAAϩ module. It is possible that different orientations of the N-terminal domain help the chaperone seek and bind the recognition sites on the surface of protein aggregates. Whether the rotations of the N-terminal domains in oligomeric ClpB occur in a concerted way and whether they are coupled to the ATP-hydrolysis cycle of the AAAϩ modules remains to be investigated. It is also possible that the N-terminal domain contains multiple substrate-recognition sites, as has been postulated for ClpA (31) and rotates to expose a site with the highest affinity for a particular substrate. Fig. 9 shows the structure of the N-terminal domain of ClpB from E. coli (18). The two conserved acidic residues investigated in this work (Asp-103 and Glu-109) reside at the surface of the N-terminal domain in close proximity of Thr-7, which has been also identified as an essential  residue for the ClpB chaperone activity (14) and Phe-105, which is conserved among ClpB and yeast Hsp104. A group of conserved hydrophobic amino acids (green) forms a groove adjacent to the site formed by Thr-7, Asp-103, Phe-105, and Glu-109. It is possible that the region of the N-terminal shown in Fig. 9 represents a substrate-binding site capable of using a combination of electrostatic and hydrophobic forces to attract protein aggregates. The role of Asp-103 and Glu-109 in substrate binding is consistent with the propensity of ClpB to bind positively charged peptides (13). Interestingly, whereas the folding topology of the N-terminal domain is similar in ClpB and ClpA (31), the latter protein contains an arginine in place of Asp-103 and an alanine in place of Glu-109, which may account for differences in substrate recognition between ClpA and ClpB.
This work also helps explain a discrepancy between previously published results on the chaperone activity of the full-length ClpB and ClpB⌬N. Goloubinoff and co-workers showed that small protein aggregates were reactivated by the DnaK/DnaJ/GrpE system alone, but large aggregates also required ClpB for efficient reactivation (27). In previous studies on ClpB, whenever protein aggregates could be reactivated by the DnaK/DnaJ/GrpE system alone, the difference in activity between ClpB and ClpB⌬N was small (21). Conversely, when a strong inhibition of ClpB activity upon the deletion of the N-terminal domain was observed, the aggregates were not reactivated by DnaK/DnaJ/GrpE alone (17). ClpB⌬N does bind to the aggregates, but its binding affinity drops significantly as the aggregate size increases (see Fig. 6). Unlike in this work, no analysis of aggregate sizes was performed in previously published studies. In our opinion, which is supported by Fig. 2, small aggregates could have predominated in the former (21) and large aggregates in the latter (17) experiments. Why would an aggregate size matter for their recognition by a chaperone? Larger aggregates have a lower surface-to-volume ratio than smaller ones and may, therefore, expose a lower concentration of chaperone-binding sites (per mg of protein). Further studies on the mechanism of substrate recognition by ClpB will help clarify this result.
Recently, it has been shown that small heat-shock proteins are the third essential component of the aggregate-reactivation machinery besides DnaK/DnaJ/GrpE and ClpB (32,33). Protein aggregates bound to small Hsps interacted with the full-length ClpB trap as well as with ClpB⌬N trap (22). This result suggests that small Hsps may provide the "missing" aggregate-binding affinity to ClpB⌬N in the absence of the N-terminal domain. The concerted action of DnaK/DnaJ/GrpE and small Hsps, which prevents the formation of large aggregates before their reactivation becomes necessary, may explain why either the full-length ClpB or ClpB⌬N supports the survival of bacteria under heat shock (20,34). However, a loss of ClpB⌬N activity, but not that of the full-length ClpB is observed in vivo in the background of a defective DnaK. 4