Hsp90, DnaK, and ClpB collaborate in protein reactivation

Significance

Hsp90, Hsp70, and Clp/Hsp100 are highly conserved molecular chaperones that participate in protein remodeling, activation, disaggregation, and degradation. We found that Escherichia coli homologs of these three chaperones, Hsp90_Ec, ClpB, and DnaK, together with DnaK cochaperones, function synergistically in reactivation of aggregated substrates when the concentration of DnaK is high. We observed that adenosine triphosphate (ATP) hydrolysis and substrate binding by all three chaperones are essential for the collaborative function. The work also shows that ClpB acts early in protein reactivation with DnaK and its cochaperones, and E. coli Hsp90 (Hsp90_Ec) acts at a later stage after ClpB. The results highlight the interplay among chaperones to regulate and maintain proteostasis.

Abstract

Hsp70, Hsp90, and ClpB/Hsp100 are molecular chaperones that help regulate proteostasis. Bacterial and yeast Hsp70s and their cochaperones function synergistically with Hsp90s to reactivate inactive and aggregated proteins by a mechanism that requires a direct interaction between Hsp90 and Hsp70 both in vitro and in vivo. Escherichia coli and yeast Hsp70s also collaborate in bichaperone systems with ClpB and Hsp104, respectively, to disaggregate and reactivate aggregated proteins and amyloids such as prions. These collaborations are dependent on direct interactions between ClpB/Hsp104 and Hsp70. We explored the possibility that E. coli homologs of Hsp70, Hsp90, and ClpB, referred to as DnaK, Hsp90_Ec, and ClpB, respectively, in combination with two DnaK cochaperones, DnaJ and GrpE, could promote protein disaggregation and reactivation under conditions where bichaperone systems are ineffective. Our results show that Hsp90_Ec is able to overcome the inhibition of protein disaggregation and reactivation observed when the concentration of DnaK is approaching physiological concentrations. We found that ATP hydrolysis and substrate binding by all three chaperones are essential for the collaborative function. The work further shows that ClpB acts early in protein reactivation with DnaK and its cochaperones; E. coli Hsp90 acts at a later stage after ClpB. The results highlight the collaboration among chaperones to regulate and maintain proteostasis.

All organisms possess multiple ATP-dependent molecular chaperones along with an array of cochaperones that are involved in proteostasis (1–5). Molecular chaperones help remodel other proteins by promoting protein folding, unfolding, refolding, disaggregation, and degradation. While chaperones can perform some activities alone, some also act collaboratively in pairs. The focus of the current study is on the functional interaction between three major chaperones, Hsp70, Hsp90, and Clp/Hsp100.

Hsp70 chaperones are highly conserved and essential in all organisms. They function as protomers, composed of a nucleotide-binding domain (NBD) connected via a short linker to a substrate-binding domain (SBD) (6, 7) (Fig. 1A). The SBD is made up of two subdomains: SDBβ contains the substrate-binding pocket, and SBDα provides a flexible lid for the substrate-binding pocket. Substrate binding is regulated by conformational changes in Hsp70 induced by ATP binding and hydrolysis (8–12). In the ATP-bound “open” conformation, the SBD α and β subdomains contact the NBD, and Hsp70 has low affinity for substrate. In the “closed” ADP-bound conformation, the SBDα lid covers the substrate-binding pocket, and Hsp70 has high affinity for substrate. Two cochaperones, J-domain proteins (JDP) and nucleotide exchange factors (NEF), regulate the Hsp70 cycle of substrate binding and release. JDPs bind substrate proteins and deliver them to Hsp70 (13, 14). They also stimulate Hsp70 ATP hydrolysis, which results in high affinity binding of the substrate to Hsp70 (6, 7, 13, 14). NEFs interact with Hsp70 and promote the exchange of ADP with ATP, which results in substrate release (7, 15). In Escherichia coli, the major Hsp70 is DnaK, the major JDP is DnaJ, and the sole NEF is GrpE.

Fig. 1.

Models and structures of E. coli Hsp90_Ec, DnaK, and ClpB. (A) The ADP-bound DnaK structure [PDB ID 2KHO (8)], with the DnaK NBD shown in green and the DnaK SBD subdomains, α and β, shown in gold and purple, respectively. The linker between the NBD and the SBD is in magenta. K70, T199, and V436 are depicted as red, blue, and green spheres, respectively. (B) AlphaFold model of Hsp90_Ec dimer in the apo form showing residues discussed in this paper. One protomer is gray and the other is colored as follows: NTD is green, the MD is purple, and the CTD is gold. E34A, R355L, and W467R are depicted as red, magenta, and green spheres, respectively. (C) AlphaFold model of E. coli ClpB protomer showing residues discussed in this paper. Y251, E279, K438* (K438E/K439E/R440E), Y503, Y653, and E678 are depicted as red, cyan, green, orange, red, and cyan spheres, respectively. ATP is depicted as black spheres. (D and E) Cryo-EM structure of ClpB hexamer [PDB ID 6QS6 (16)], with one protomer colored and the others in gray, depicted in a (D) side view and (E) Top view. The missing N-domain density is highlighted by a green oval, NBD I is in purple, the middle domain is in magenta, NBD II is in gold, and ATPγS is depicted as black spheres. The model substrate, casein, is depicted in red sticks. Models in (A–E) were imaged in PyMOL (17).

Hsp70 with its cochaperones can refold and reactivate many substrate proteins and small aggregates in an ATP-dependent fashion (6, 7). However, for other substrates and aggregates, Hsp70 collaborates with Hsp90 (18, 19). Hsp90 chaperones are highly conserved across many species and essential in higher eukaryotes, although they are not known to be essential in bacteria (3, 20). Hsp90 forms dimers; each protomer is composed of an N-terminal domain (NTD), a middle domain (MD), and a C-terminal domain (CTD), where dimerization occurs (20–22) (Fig. 1B). The NTD contains a nucleotide-binding pocket related to the GHKL ATPase superfamily. This site, in conjunction with a catalytic loop in the MD, functions in ATP hydrolysis. Hsp90 dimers undergo large conformational changes from an extended open form in the absence of nucleotide to a closed form in the presence of ATP characterized by transient dimerization of the NTDs. Conformational changes promote substrate binding and release by Hsp90, and these changes are further modulated in eukaryotes by numerous Hsp90 cochaperones (20–22). Hsp90 alone has not been shown to reactivate inactive proteins, but it is able to bind substrates and prevent aggregation. However, bacterial, yeast, and human endoplasmic reticulum (ER) Hsp90s function synergistically with Hsp70s to reactivate inactive and aggregated proteins (19, 20, 23). This bichaperone reaction also requires Hsp70 cochaperones, and in yeast, an Hsp90 cochaperone, Sti1/Hop, is also required (20). The synergistic activity of the two chaperones involves a direct complex between Hsp90 and Hsp70 in vivo and in vitro in E. coli and yeast (19, 24–26). Ternary complexes between Hsp90, Hsp70, and JDP and binary complexes between Hsp90 and JDP have also been reported, indicating the complex nature of the interactions (27, 28).

To disaggregate and reactivate large aggregates, E. coli and yeast Hsp70s (DnaK and Ssa1, respectively) collaborate with ClpB and Hsp104 chaperones, respectively (29–31). ClpB and Hsp104 are members of a subfamily of Clp/Hsp100 chaperones that are abundant in bacteria, yeast, and plants, but are also present in the mitochondria of some vertebrates, including humans (32). They are composed of six protomers that form a hexameric spiral structure with a central channel. Each protomer contains an N-terminal domain (NTD) and two nucleotide-binding domains (NBD I and NBD II) that have AAA+ ATP-binding sites (33–35) (Fig. 1 C–E). ClpB and Hsp104 can prevent aggregation of non-native proteins and assembly of amyloids but are unable to act alone in protein remodeling except under unnatural in vitro conditions requiring a slowly hydrolyzable ATP analog, ATPγS (36). However, in combination with Hsp70 and cochaperones, they bind aggregated and amyloid substrates and promote ATP-dependent disaggregation by forcing substrates to unfold and pass through the central channel of the hexameric chaperone (29). The unfolded substrates are released and spontaneously refold or are bound by another chaperone. It has been shown that ClpB and Hsp104 function synergistically with their respective Hsp70 and Hsp70 cochaperones through interactions between the MD of ClpB/Hsp104 and the NBD of Hsp70 (29, 37–43).

Here, we investigated whether Hsp90, Hsp70, and ClpB/Hsp100 chaperones can function collaboratively to reactivate inactive and aggregated proteins and found that E. coli Hsp90 (encoded by the htpG gene and referred to as Hsp90_Ec), ClpB, and DnaK chaperones, plus the DnaK cochaperones, DnaJ and GrpE, function synergistically in protein reactivation.

Results

Hsp90_Ec Collaborates with ClpB, DnaK, and the DnaK Cochaperones in Protein Disaggregation and Reactivation.

In this study, we explored the possibility that Hsp90 functions in protein disaggregation and reactivation with a larger chaperone system than the previously described bichaperone system of Hsp90 and Hsp70. Specifically, we tested whether Hsp90 can collaborate with ClpB/Hsp100 and Hsp70 to disaggregate and reactivate aggregated proteins. For these studies Hsp90_Ec, DnaK, and ClpB, the E. coli homologs of Hsp90, Hsp70, and ClpB/Hsp100, respectively, and two DnaK cochaperones, DnaJ, the major E. coli JDP, and GrpE, the E. coli NEF, were used, and reactivation of chemically inactivated luciferase, a well-characterized model aggregated protein, was monitored. We observed that ClpB and the DnaK system reactivated denatured luciferase as previously shown (30, 31) (Fig. 2A). With these conditions, which included 1 µM DnaK (low DnaK), Hsp90_Ec slightly stimulated reactivation (Fig. 2A). However, at a higher concentration of DnaK (5 µM; high DnaK), there was no detectable reactivation by ClpB and the DnaK system together, and Hsp90_Ec was required to promote reactivation (Fig. 2 B and C). All three chaperones plus DnaJ and GrpE were important for collaborative reactivation. Hsp90_Ec, DnaK, and DnaJ were essential; omission of any one resulted in background levels of reactivation. ClpB and GrpE were not essential, but they stimulated reactivation 10- and 5-fold, respectively (Fig. 2 B and C and SI Appendix, Fig. S1).

Fig. 2.

Hsp90_Ec collaborates with DnaK and ClpB in protein disaggregation and reactivation. (A) Reactivation of urea-denatured luciferase in the presence of low DnaK (1 µM), Hsp90_Ec (1.2 µM), ClpB (0.1 µM), DnaJ (0.1 µM), and GrpE (0.05 µM) as indicated, was monitored by the increase in luminescence over time. (B) Reactivation of urea-denatured luciferase in the presence of high DnaK (5 µM), Hsp90_Ec (1.2 µM), ClpB (0.1 µM), DnaJ (1 µM), and GrpE (0.2 µM) as indicated, was monitored by the increase in luminescence over time. (C) Bar graph showing the amount of luciferase reactivated after 20 min from the data in panel (B). (D) Reactivation of urea-denatured luciferase in the presence or absence of Hsp90_Ec (1.2 µM) with varying concentrations of DnaK. All reactions contained ClpB (0.1 µM), DnaJ (0.5 µM), and GrpE (0.2 µM). The amount of luciferase reactivated after 50 min is shown. (E) Reactivation of heat-denatured GFP-38 in the presence of low DnaK (1 µM), Hsp90_Ec (2 µM), ClpB (0.1 µM), DnaJ (0.1 µM), and GrpE (0.05 µM) as indicated, was monitored by the increase in fluorescence with time. (F) Reactivation of heat-denatured GFP-38 in the presence of high DnaK (10 µM), Hsp90_Ec (2 µM), ClpB (0.1 µM), DnaJ (1 µM), and GrpE (0.2 µM) as indicated, was monitored by the increase in fluorescence with time. Assays were carried out as described in Materials and Methods. In (A–F), the data are means ± SD (n ≥ 3).

To further investigate inhibition of reactivation by ClpB and the DnaK system by high concentrations of DnaK, we titrated DnaK in the presence and absence of Hsp90_Ec. We observed that in the absence of Hsp90_Ec, as the DnaK concentration was increased, luciferase reactivation decreased. However, in the presence of Hsp90_Ec, as the DnaK concentration was increased the inhibition of reactivation was prevented (Fig. 2D and SI Appendix, Fig. S1C). A previous study of the bichaperone system of DnaK and Hsp90_Ec by Morán-Luengo et al. showed that reactivation by the DnaK system was inhibited by high concentrations of DnaK and Hsp90_Ec overcame the inhibition (44). Their work further suggested that the inhibition by high DnaK concentrations was due to unfolded substrates becoming trapped in a cycle of rapid binding and release from DnaK and implied that Hsp90_Ec was able to break the cycle and allow the non-native polypeptides to refold in native conformations (44). Consistent with the previous work, one attractive interpretation of our observations is that, under high concentrations of DnaK, Hsp90_Ec functions synergistically with ClpB and the DnaK system to rescue partially folded substrates from being trapped in cycles of binding and release from DnaK.

To determine whether Hsp90_Ec, ClpB, DnaK, and DnaK cochaperones were able to collaboratively reactivate other inactive non-native proteins besides luciferase, we tested another previously used model substrate protein, heat-denatured GFP-38, a GFP derivative with a 38 amino acid extension at the C terminus (45). As reported before, ClpB and the DnaK system collaboratively reactivated GFP-38 when a low concentration of DnaK was used (1 µM) (Fig. 2E) (46). However, when the concentration of DnaK was high (10 µM), Hsp90_Ec in addition to ClpB, DnaK, DnaJ, and GrpE was required to reactivate denatured GFP-38 (Fig. 2F). All three chaperones were required for maximal reactivation as was the DnaJ cochaperone (Fig. 2F). In the absence of ClpB or GrpE, there was detectable reactivation of GFP-38, as seen with denatured luciferase (Fig. 2 B and F). Altogether, the results demonstrate that Hsp90_Ec, ClpB, and the DnaK system can function synergistically to reactivate non-native proteins when the concentration of DnaK is elevated.

The Collaborative Function of Hsp90_Ec, ClpB, and DnaK Requires the ATP-Dependent Chaperone Activity of All Three Chaperones.

We next tested whether known chaperone activities of the three chaperones were essential for their collaboration. To assess the importance of ATP hydrolysis by Hsp90_Ec, a known inhibitor of Hsp90 ATP binding, geldanamycin (GA) (47), was added to luciferase reactivation assays. Under conditions with high concentrations of DnaK, GA blocked luciferase disaggregation and reactivation by Hsp90_Ec, ClpB, DnaK, DnaJ, and GrpE (Fig. 3A and SI Appendix, Fig. S2A). GA also blocked the residual reactivation by Hsp90_Ec, DnaK, DnaJ, and GrpE in the absence of ClpB (SI Appendix, Fig. S2A). Control experiments demonstrated that GA was specific for Hsp90_Ec. It had no effect on ATP hydrolysis by ClpB or DnaK (SI Appendix, Fig. S3A) and no effect on protein reactivation by DnaK, DnaJ, and GrpE or by ClpB, DnaK, DnaJ, and GrpE with a low concentration of DnaK (SI Appendix, Fig. S3 B and C). Additionally, when wild-type Hsp90_Ec was substituted with a mutant known to be defective in ATP hydrolysis, Hsp90_Ec E34A (Fig. 1B) (19, 48–50), reactivation was prevented (Fig. 3A and SI Appendix, Fig. S2B). Together these results show that ATP hydrolysis by Hsp90_Ec is essential for synergistic reactivation by Hsp90_Ec, ClpB, DnaK, DnaJ, and GrpE, ruling out a solely protein-holding function of Hsp90_Ec.

Fig. 3.

The ATP-dependent chaperone activities of Hsp90_Ec, ClpB, and DnaK are required for reactivation. Reactivation of urea-denatured luciferase in the presence of Hsp90_Ec (1.2 µM), ClpB (0.1 µM), DnaK (5 µM), DnaJ (1 µM), and GrpE (0.2 µM) using (A) Hsp90_Ec WT and mutants; (B) ClpB WT and mutants; and (C) DnaK WT and mutants. K438* refers to ClpB K438E/K439E/R440E. Geldanamycin (30 µM) and GuHCl (20 mM) were added where indicated. In each panel, the amount of luciferase reactivated after 20 min is shown as means ± SD, (n = 3). Assays were carried out as described in Materials and Methods.

We also tested whether the interaction between Hsp90_Ec and DnaK was necessary for Hsp90_Ec to act collaboratively with ClpB, DnaK, and the DnaK cochaperones in luciferase reactivation by substituting Hsp90_Ec R355L (26) (Fig. 1B), a previously described mutant that is defective in its interaction with DnaK, for wild-type Hsp90_Ec. Hsp90_Ec R355L was defective, demonstrating the importance of direct interaction between Hsp90_Ec and DnaK (Fig. 3A and SI Appendix, Fig. S2B). Additionally, a substrate-binding defective Hsp90_Ec mutant, W467R (51) (Fig. 1B), was defective in luciferase reactivation with ClpB and the DnaK system, indicating substrate binding by Hsp90_Ec was also important (Fig. 3A and SI Appendix, Fig. S2B).

Chaperone activities of ClpB were also assessed for their importance for protein reactivation by Hsp90_Ec, ClpB, DnaK, DnaJ, and GrpE. We observed that luciferase reactivation by a ClpB mutant defective in ATP hydrolysis in both NBD I and NBD II due to Walker B substitutions in both domains, ClpB E279A/E678A (Fig. 1C) (52), was very low, similar to that observed by the combination of Hsp90_Ec, DnaK, DnaJ, and GrpE (Fig. 3B and SI Appendix, Fig. S2C). The effect of guanidinium chloride (GuHCl), a specific inhibitor of ClpB chaperone activity, was also tested. Previous work showed that low concentrations of GuHCl inhibit the chaperone functions of ClpB, Hsp104, and other ClpB homologs in vivo (53–55) and in vitro (56, 57) by binding to ClpB NBD I adjacent to the nucleotide-binding site (58). When tested for its effect on luciferase reactivation by ClpB, Hsp90_Ec, DnaK, DnaJ, and GrpE, we observed that reactivation was reduced to the level observed by Hsp90_Ec, DnaK, DnaJ, and GrpE (Fig. 3B and SI Appendix, Fig. S2C). In control experiments, GuHCl had no effect on protein reactivation by the DnaK system alone or by Hsp90_Ec in collaboration with the DnaK system (SI Appendix, Fig. S4 A and B). As expected, GuHCl did inhibit luciferase reactivation by ClpB, DnaK, DnaJ, and GrpE using a low concentration of DnaK (SI Appendix, Fig. S4C).

In addition, the importance of substrate threading activity of ClpB was determined by using a mutant with substitutions in two essential tyrosines in pore loops 1 and 2, ClpB Y251A/Y653A (46, 59) (Fig. 1C). Reactivation by the double pore loop mutant was reduced to the amount observed in the absence of ClpB, demonstrating that polypeptide translocation through the central channel of ClpB is essential (Fig. 3B and SI Appendix, Fig. S2C). We also tested whether mutants in the MD of ClpB that are defective or partially defective in DnaK interaction, including Y503A (60) and K438E/K439E/R440E (40) (Fig. 1C), were able to function with Hsp90_Ec, DnaK, and the DnaK cochaperones. For both MD mutants, luciferase reactivation was similar to the amount observed in the absence of ClpB (Fig. 3B and SI Appendix, Fig. S2C), suggesting the importance of the interaction between ClpB and DnaK. Additionally, a ClpB mutant lacking the NTD, ΔN ClpB (61), and known to be defective in binding some substrates with the DnaK system (46), was also tested. ΔN ClpB was similar to wild-type ClpB in its ability to reactivate luciferase in combination with Hsp90_Ec and DnaK (Fig. 3B and SI Appendix, Fig. S2C). ClpB Y251A, which is slightly defective in substrate engagement (46) was not defective in luciferase reactivation (Fig. 3B and SI Appendix, Fig. S2C). However, a ΔN ClpB/Y251A mutant previously shown to be partially defective in substrate interaction (46) was defective in luciferase reactivation, suggesting that substrate recognition by ClpB is critical for luciferase reactivation (Fig. 3B).

Finally, we tested whether the known DnaK chaperone activities were required for reactivation by Hsp90_Ec, ClpB, and the DnaK system. To probe the importance of ATP hydrolysis by DnaK, two DnaK mutants, K70A (62) and T199A (62) (Fig. 1A), which are impaired in ATP hydrolysis, were tested for their ability to reactivate luciferase synergistically with Hsp90_Ec, ClpB, DnaJ, and GrpE. Both mutants were inactive (Fig. 3C and SI Appendix, Fig. S2D). In addition, a DnaK mutant defective in substrate binding, V436F (63) (Fig. 1A), was inactive, showing that substrate binding by DnaK is important (Fig. 3C and SI Appendix, Fig. S2D).

Together, these results demonstrate that all three chaperones utilize their ATP hydrolytic abilities as well as their substrate-binding capabilities to collaborate in protein reactivation. In addition, the polypeptide translocation activity of ClpB is important for the collaboration, as well as direct interactions between Hsp90_Ec and DnaK and between ClpB and DnaK.

Knowing that all three chaperones function synergistically and that there are pairwise interactions between DnaK and Hsp90_Ec (26) and between ClpB and DnaK (29, 37–43), we tested for direct interactions between ClpB and Hsp90_Ec. However, we were unable to detect Hsp90_Ec–ClpB complexes by pulldown experiments using assay conditions that readily detected complex formation between DnaK and ClpB (SI Appendix, Fig. S5 A and B). In addition, while Hsp90_Ec associated with biotinylated DnaK, as previously shown (26, 27) (SI Appendix, Fig. S5C), it did not stabilize or compete with the ClpB–DnaK complex (SI Appendix, Fig. S5 C and D). ATP hydrolysis by the mixture of Hsp90_Ec and ClpB was additive, consistent with a lack of direct interaction of the two proteins (SI Appendix, Fig. S5E). Moreover, Hsp90_Ec had no effect on protein unfolding by ClpB in the presence of a 1:1 mixture of ATP:ATPγS, a condition known to promote unfolding by ClpB alone (36, 45) (SI Appendix, Fig. S5F). ATP, which does not support protein unfolding by ClpB alone, did not promote protein unfolding by ClpB in the presence of Hsp90_Ec (SI Appendix, Fig. S5F).

ClpB and the DnaK System Are Required Early in Protein Reactivation Before Hsp90_Ec.

To explore the synergistic action of Hsp90_Ec, ClpB, and the DnaK system in protein reactivation, we tested whether the chaperones act together or sequentially by performing delayed addition experiments (Fig. 4A). Urea-inactivated luciferase was first preincubated for 30 min with one or two of the chaperones and cochaperones in the presence of ATP. The missing chaperone or chaperones were then added, and luciferase reactivation was monitored over time. The expectation was that if one or two of the chaperones included in the preincubation step acted early in the reactivation pathway, the rate of reactivation upon the addition of the omitted chaperone(s) after the preincubation step would be greater than if all the components were added together after the 30-min preincubation.

Fig. 4.

ClpB and DnaK act early in protein reactivation before Hsp90_Ec. (A) Schematic of the experiments shown in (B) and (C). (B) Urea-denatured luciferase was preincubated with DnaK (5 µM), DnaJ (1 µM), GrpE (0.2 µM), ClpB (12.5 nM), Hsp90_Ec (1.2 µM), or buffer for 30 min in the presence of ATP as indicated. Following the 30-min preincubation, the remaining reaction component(s) were added at t = 0 and luciferase reactivation was monitored over time. (C) The rates of luciferase reactivation from (B) were determined using the initial linear phase following t = 0 and are shown as means ± SEM. Each reactivation rate was compared to KJE, B/90 using one-way ANOVA, n = 4, **0.001 < P < 0.01, ***0.0001 < P < 0.001. (D) Reactivation of urea-denatured luciferase in the presence of DnaK (5 µM), Hsp90_Ec (1.2 µM), ClpB (0.1 µM), DnaJ (1 µM), and GrpE (0.2 µM) in the presence or absence of GA (30 µM). GA was added where indicated at t = 0 or t = 10 min. Assays were carried out as described in Materials and Methods. In (B) and (D), the data are means ± SD (n ≥ 3).

In a control experiment, when all three chaperones, ClpB, Hsp90_Ec, and the DnaK system, were added together in the second step, there was a lag phase followed by a slow rate of reactivation, ~1% per minute. Strikingly, the highest reactivation rate was achieved when ClpB was incubated with the DnaK system in the first reaction and Hsp90_Ec was added in the second step; the rate was ~3.8% per minute, nearly 4-fold greater than when all three chaperones were added together in the second step (Fig. 4 B and C). Moreover, reactivation started immediately and without a lag when Hsp90_Ec was added. When the other two combinations of chaperones were tested in the two-step reaction, either the DnaK system in the first step with Hsp90_Ec and ClpB added in the second step, or the DnaK system and Hsp90_Ec in the first step with ClpB added in the second step, we observed rates of reaction similar to the slow rate observed when all three chaperones were added in the second reaction, although the lag phases were shorter (Fig. 4 B and C).

Together these observations suggest that ClpB and the DnaK system act early in the protein reactivation reaction and Hsp90_Ec acts later. Consistent with this observation, luciferase reactivation by Hsp90_Ec, ClpB, and the DnaK system was immediately inhibited by the addition of geldanamycin at 10 min, confirming that Hsp90_Ec is required at a later stage of reactivation (Fig. 4D).

ClpB Is Only Required Early in the Reactivation Pathway.

Knowing that Hsp90_Ec was only required late in the refolding pathway, we wanted to assess whether ClpB was only required early in the reactivation pathway. To address this question, heat-denatured GFP-38 was incubated with ClpB, the DnaK system, and Hsp90_Ec. At various times after the start of reactivation GuHCl was added to inhibit ClpB (Fig. 5A and SI Appendix, Fig. S6). When GuHCl was added at 0 min, reactivation was inhibited to the rate observed when ClpB was omitted from the reaction (Fig. 5A and SI Appendix, Fig. S6). As the time of addition of GuHCl following the start of the reactivation reaction was increased between 0 and 3 min, inhibition of reactivation rapidly decreased (Fig. 5A and SI Appendix, Fig. S6). Importantly, there was no inhibition of reactivation when GuHCl was added after 4 or more minutes, suggesting that ClpB function is only necessary during the first few minutes of the reactivation process (Fig. 5A and SI Appendix, Fig. S6).

Fig. 5.

ClpB is only required early in the reactivation pathway. (A) Heat-denatured GFP-38 was incubated in the presence of Hsp90_Ec (2 µM), ClpB (0.1 µM), DnaK (10 µM), DnaJ (1 µM), and GrpE (0.2 µM), and the rate of reactivation was monitored. GuHCl (20 mM) was added at the times indicated. (B) Heat-denatured GFP-38 was incubated in the presence of Hsp90_Ec (2 µM), ClpB (0.1 µM), DnaK (10 µM), DnaJ (1 µM), and GrpE (0.2 µM) with and without 20 mM GuHCl, and reactivation was monitored over time. A subset of components was added at t = 0 min as indicated and the remaining components were added at t = 10 min. (C) Reactivation of urea-denatured luciferase in the presence of Hsp90_Ec (1.2 µM), ClpB (0.1 µM), DnaK (5 µM), DnaJ (1 µM), and GrpE (0.2 µM), and 20 mM GuHCl was monitored over time. A subset of components was added at t = 0 min and the remaining components were added at t = 5 min. Assays were carried out as described in Materials and Methods. In (A–C), the data are means ± SD (n = 3).

Additional delayed addition experiments confirmed that ClpB was only important during the early phase of reactivation. ClpB, the DnaK system, and heat-denatured GFP-38 were incubated for 10 min, during which time there was little reactivation (Fig. 5B). Then, after 10 min of incubation, GuHCl and Hsp90_Ec were added; we observed immediate and robust reactivation (Fig. 5B). Similar robust reactivation was observed when Hsp90_Ec alone was added after a 10-min incubation of ClpB and the DnaK system with denatured GFP-38 (Fig. 5B). In contrast, when the DnaK system, Hsp90_Ec, and heat-denatured GFP-38 were incubated for 10 min and then ClpB and GuHCl were added, the level of reactivation was low, similar to that observed when GuHCl and ClpB were present during the initial 10 min (Fig. 5B). As expected, when Hsp90_Ec and the DnaK system were incubated for 10 min and then ClpB alone was added, reactivation increased but the initial rate of reactivation appeared slightly slower than when ClpB and the DnaK system were present during the first 10 min. However, the extent of reactivation was similar to that seen when all components were present from the start of the reaction (Fig. 5B).

Additional delayed addition experiments performed using urea-denatured luciferase as the substrate yielded similar results. ClpB and the DnaK system were incubated with denatured luciferase for 5 min and then Hsp90_Ec was added, with or without GuHCl. The rate and extent of reactivation following the addition of Hsp90_Ec were the same with or without the addition of GuHCl (Fig. 5C). When Hsp90_Ec and the DnaK system were preincubated and then ClpB and GuHCl added at 5 min, luciferase reactivation was similar to that seen when GuHCl was present from the onset of reactivation; when Hsp90_Ec and the DnaK system were incubated for 5 min and then ClpB was added, the initial rate of reactivation was slower than when ClpB was present during the initial 5 min, but the extent of reactivation was similar (Fig. 5C). Altogether the results show that ClpB functions early in reactivation with DnaK and its cochaperones and is dispensable during the later steps of reactivation that depend on Hsp90_Ec.

Direct Interaction between ClpB and DnaJ.

While probing for interactions among the three chaperones and the DnaK cochaperones, we found that ClpB directly interacts with DnaJ. Using biotinylated DnaJ, we observed that ClpB bound DnaJ in a concentration-dependent manner (Fig. 6A and SI Appendix, Fig. S7A). The interaction between ClpB and DnaJ was stabilized by the slowly hydrolyzable ATP analog, ATPγS. About 3-fold more DnaJ–ClpB complex was detected in the presence of ATPγS than in the absence of nucleotide or in the presence of ATP or ADP (Fig. 6B and SI Appendix, Fig. S7B). Analysis of the ClpB–DnaJ complex by bio-layer interferometry (BLI) additionally demonstrated a strong interaction between the two proteins in the presence of ATPγS with a K_d value of 55.2 ± 9.9 nM (SI Appendix, Fig. S7 C–E).

Fig. 6.

Direct interaction between ClpB and DnaJ. (A) In vitro interaction between DnaJ-biotin (2 µM) and ClpB (1 to 4 µM) in the presence of ATPγS was monitored using a pull-down assay as described in Materials and Methods. (B) In vitro interaction between DnaJ-biotin and ClpB (1 µM) in the presence of ATP, ATPγS, ADP, or no nucleotide was monitored using a pull-down assay as in (A). In (A) and (B), a representative gel of three replicates is shown. (C) Schematic depicting sites of interaction between Hsp90_Ec and DnaJ identified by DSSO cross-linking followed by tandem mass spectrometry, n = 3. Identified intermolecular cross-links are denoted using solid lines. Red lines indicate cross-linking between the DnaJ J-domain and the ClpB NBDI or MD. (D) AlphaFold ClpB protomer model with colored spheres showing locations of cross-linked residues identified in (C). (E) AlphaFold DnaJ dimer model with colored spheres showing positions of cross-linked residues identified in (C).

To identify the specific site or sites of interaction between DnaJ and ClpB we performed cross-linking using disuccinimidyl sulfoxide (DSSO), an amine-reactive cross-linker with a 10.3 Å spacer arm followed by tandem mass spectrometry (XL-MS) (Fig. 6 C–E and SI Appendix, Fig. S7F and Table S1). The majority of the sites identified were in the NBD I or MD of ClpB and the J-domain and CTD I of DnaJ, suggesting several sites of interaction on both ClpB and DnaJ (Fig. 6 C–E and SI Appendix, Fig. S7F and Table S1). Consistent with these results, ClpB ΔN interacted with DnaJ in pull-down assays, although the binding was ~40% that of wild-type ClpB (SI Appendix, Fig. S7 G and H).

Since the ClpB MD is also a known interaction site for DnaK, we tested whether DnaJ would compete for the interaction between DnaK and ClpB. In pulldown assays containing biotinylated DnaK, ClpB, and a 1:1 mix of ATP:ATPγS, the DnaK–ClpB interaction decreased with increasing concentrations of DnaJ (SI Appendix, Fig. S7 I and J). This suggests that DnaJ binding to either ClpB or DnaK leads to reduced interaction between ClpB and DnaK. Further research is needed to demonstrate whether the ClpB–DnaJ interaction is important for collaboration between ClpB and the DnaK system or for collaboration between ClpB, the DnaK system, and Hsp90_Ec.

Discussion

We identified a system of chaperones that involves an interplay between E. coli Hsp90_Ec, DnaK, and ClpB to reactivate aggregated and inactive proteins in the presence of high concentrations of DnaK where neither bichaperone system, DnaK–ClpB, nor Hsp90_Ec–DnaK, can reactivate proteins alone. The results suggest that collaborative systems of chaperones may be important to modulate proteostasis when individual chaperones or pairs of chaperones are inadequate.

The results show that the ATP-dependent chaperone activities of Hsp90_Ec, DnaK, and ClpB are important for the synergistic reactivation of aggregated substrates. Therefore, the chaperones do not function solely in a protein-holding capacity, but instead utilize their ATP hydrolytic and substrate-binding abilities to promote protein disaggregation. Moreover, we deciphered the order of chaperone function in the pathway of protein disaggregation. ClpB, DnaK, and DnaK cochaperones act before Hsp90_Ec, demonstrating a temporal separation in the association of chaperones with the aggregated substrate in the pathway of disaggregation. These results in combination with previous studies showing that DnaK and its cochaperones act before ClpB in the ClpB–DnaK bichaperone system (64) and also act before Hsp90_Ec in the Hsp90_Ec–DnaK bichaperone system (19), suggest that the DnaK system acts both early and late in protein reactivation with ClpB and Hsp90_Ec. Altogether these results suggest that the bichaperone systems of DnaK–ClpB and Hsp90_Ec–DnaK are more tightly coupled than initially considered.

Consistent with our observation that the dependence on Hsp90_Ec in protein reactivation by ClpB and the DnaK system is elicited under high concentrations of DnaK, Morán Luengo et al. (44) observed that high concentrations of DnaK inhibit reactivation of substrates that are reactivated by the DnaK system alone. The authors suggested that simultaneous binding events occur between multiple DnaK molecules and multiple sites on a single substrate before the native state is reached. They posited that DnaK binding to the substrate is a DnaK concentration-dependent process whereas dissociation from the client is a stochastic, DnaK concentration-independent process. Thus, at high concentrations of DnaK, the rate of substrate rebinding to DnaK would be greater than the rate of dissociation, preventing productive client folding and maturation. They showed that the inhibition by high concentrations of DnaK is overcome by Hsp90_Ec and concluded that the role of Hsp90_Ec is to rescue partially folded substrates from sequestration in cycles of binding to DnaK followed by release and rapid rebinding (44). Our observation that Hsp90_Ec overcomes the inhibition of reactivation by ClpB and high concentrations of DnaK may be explained by a similar mechanism. Following the action of the DnaK system and ClpB on aggregated proteins, unfolded substrates released from ClpB may be captured by DnaK, under conditions involving high concentrations of DnaK, and become trapped in a cycle of binding to DnaK followed by release and rapid rebinding. Hsp90_Ec may break the cycle by transiently interacting with DnaK, which could lead to substrate handoff from DnaK to Hsp90_Ec. Once transferred to Hsp90_Ec, remodeling of the partially folded substrate could be completed.

The physiological concentration of E. coli DnaK in vivo has been reported to be 27 µM when cells are growing at 30 °C and increases twofold upon heat shock (65). Therefore, the high concentrations of DnaK that lead to inhibition of reactivation in vitro (5 to 10 µM) approach the physiological concentrations of E. coli DnaK. However, it is unknown how much of the cellular DnaK is free and how much is bound to substrates during normal growth or during stress conditions. It is also not known whether there are in vivo conditions during which ClpB, Hsp90_Ec, DnaK, and the DnaK chaperones collaborate. Thus, it remains unknown whether the in vitro conditions reflect the in vivo conditions.

With the methods used, we did not observe a direct interaction between Hsp90_Ec and ClpB, although Hsp90_Ec–DnaK and ClpB–DnaK interactions were observed and are well documented (19, 24–26, 37, 38, 40–43). The lack of an Hsp90_Ec–ClpB interaction is consistent with our finding that the roles of Hsp90_Ec and ClpB are temporally separated in protein disaggregation. However, we did observe direct binding between ClpB and DnaJ.

The results presented here in combination with studies from many laboratories (3, 6, 7, 34, 42, 66) suggest a working model of protein disaggregation and reactivation by the collaborative action of ClpB, the DnaK system, and Hsp90_Ec. DnaK and the DnaK cochaperones initially engage with the aggregated substrate (Fig. 7, Step 1). ClpB is then recruited to the aggregate (Fig. 7, Step 2) where the substrate is transferred to ClpB for unfolding and translocation through the central channel (Fig. 7, Step 3). If the concentration of DnaK is low, the unfolded polypeptide is released as it emerges from the channel (Fig. 7, Step 4), and either refolds spontaneously into its active conformation or enters additional cycles of chaperone-mediated refolding by the DnaK system or another chaperone before attaining the native conformation (Fig. 7, Step 5). If the concentration of DnaK is high, the released polypeptide is trapped in a cycle of rapid binding to and release from DnaK (Fig. 7, Step 6) (44). The working model suggests the cycle can be broken when the partially folded substrate is handed off from DnaK to Hsp90_Ec through transient complexes between Hsp90_Ec, DnaK, and DnaJ (Fig. 7, Steps 7 and 8) (24–27, 67). An alternate model is that DnaK and Hsp90_Ec directly compete for the substrate as it exits the ClpB channel, although this model does not explain why the interaction between Hsp90_Ec and DnaK is important and why ATP hydrolysis by Hsp90_Ec is required for reactivation. In the final steps of the pathway, the substrate is released from Hsp90_Ec and refolds in its active conformation (Fig. 7, steps 9 and 10). If it does not reach its native form spontaneously after release from Hsp90_Ec, it may undergo additional chaperone-mediated refolding. The working model is speculative, and further research will be necessary to refine the mechanism of protein refolding by this complex, multichaperone system.

Fig. 7.

Working model for protein reactivation. DnaK binds protein aggregates in a reaction stimulated by DnaJ and GrpE and may promote some remodeling (1). ClpB is recruited to the misfolded and/or aggregated substrate (2) and promotes substrate unfolding and translocation through the central ClpB channel (3). Under conditions of low DnaK concentration, unfolded polypeptides are released from ClpB (4) and refold spontaneously (5). Some substrates that do not spontaneously refold may be bound by other chaperones for further chaperone-mediated refolding. Under conditions of high DnaK concentration, unfolded polypeptides released from ClpB are bound by DnaK and enter a cycle of binding and release from DnaK (6). Hsp90_Ec, through interactions with DnaK and DnaJ, can interrupt the cycle (7) and facilitate hand-off of the substrate from DnaK to Hsp90_Ec (8). Finally, the substrate is released from Hsp90_Ec, either in its native conformation or in a form that refolds spontaneously (9, 10). Many aspects of the working model are speculative and further investigation is required.

Materials and Methods

Proteins.

His-tagged Hsp90_Ec wild-type, Hsp90_Ec mutants (19), DnaK wild-type and mutants (68), ClpB wild-type and mutants (69), DnaJ (68), GrpE (68), E. coli ribosomal protein L2 (70), GroEL trap (71), GFP-15 (72), and GFP-38 (45) were isolated as described. All proteins were >95% pure as determined by SDS-PAGE. Luciferase (E1701) and luciferin (E1603) were from Promega. Concentrations given are for Hsp90_Ec, DnaJ, GrpE dimers, ClpB hexamers, GroEL trap tetradecamers, and DnaK, luciferase, GFP-38, and GFP-15 monomers.

Protein Reactivation.

Luciferase and GFP-38 reactivation were performed as previously described (40) and as described in SI Appendix.

ATPase Assay.

Steady-state ATP hydrolysis was measured at 37 °C in 25 mM Hepes, 50 mM KCl, 0.5 mM EDTA, 10 mM MgCl₂, 2 mM DTT, 0.005% Triton X-100 (vol/vol), and 2 mM ATP using a pyruvate kinase/lactate dehydrogenase enzyme-coupled assay as described (48) and 1 μM Hsp90_Ec, 0.2 μM ClpB, 1 μM DnaK, 0.2 μM DnaJ, 0.1 μM GrpE, 1 μM E. coli ribosomal protein L2, and 30 μM geldanamycin as indicated.

Pull-Down Assay.

DnaJ and Hsp90_Ec were biotin labeled as previously described (67) and as described in SI Appendix.

DSSO Cross-Linking Mass Spectrometry.

DSSO (Thermo Fisher A33545) cross-linking reactions contained 20 mM Hepes, 75 mM KCl, 10 mM MgCl₂, 3 μM ClpB wild type, 12 μM DnaJ wild type, and 1 mM ATPγS. Proteins were cross-linked in the presence of 0.19 mM DSSO at 23 °C for 30 min. Samples were analyzed as described in SI Appendix.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

All study data are included in the article and/or SI Appendix.