The Hsp40 J‐domain modulates Hsp70 conformation and ATPase activity with a semi‐elliptical spring

Abstract

Regulatory protein interactions are commonly attributed to lock‐and‐key associations that bring interacting domains together. However, studies in some systems suggest that regulation is not achieved by binding interactions alone. We report our investigations on specific physical characteristics required of the Hsp40 J‐domain to stimulate ATP hydrolysis in the Hsp40‐Hsp70 molecular chaperone machine. Biophysical analysis using isothermal titration calorimetry, and nuclear magnetic resonance spectroscopy reveals the importance of helix rigidity for the maintenance of Hsp40 function. Our results suggest that the functional J‐domain acts like a semi‐elliptical spring, wherein the resistance to bending upon binding to the Hsp70 ATPase modulates the ATPase domain conformational change and promotes ATP hydrolysis.

Abbreviations

Δδ_c

changes in chemical shift (composite)

Δδ_H

changes in chemical shift (Hydrogen/Proton)

Δδ_N

changes in chemical shift (Nitrogen)

HSQC

heteronuclear single quantum correlation (NMR pulse program)

ITC

isothermal titration calorimetry

NMR

nuclear magnetic resonance spectroscopy

Introduction

Regulatory protein‐protein interactions frequently involve conformational changes that communicate regulatory signals. With the existence of alternative protein conformations, it has been difficult to discover how tertiary structure contributes to binding and signaling beyond the presentation of relevant contacts at the protein‐protein interface. In most cases, the role of a particular conformation appears to be to provide the necessary binding interactions for a specific function. However, in the example of Hsp70 regulation by its partner Hsp40, binding alone is not sufficient for the stimulation of the ATPase activity.1

The Hsp70/DnaK and Hsp40/DnaJ molecular chaperone machine of Escherichia coli participates in numerous cellular functions, including protein folding, protein translocation across membranes, protein complex disassembly,2 and the heat‐shock response.3, 4 These functions are achieved through the ATP‐dependent conversion of Hsp70 between conformational states with high and low affinities for unfolded polypeptides.5, 6, 7

Hsp70/DnaK8 is comprised of two subdomains: an ATPase domain9 and a client‐binding domain.10 Biochemical and biophysical experiments have revealed probable mechanisms for the interaction of these domains in DnaK.11, 12 The ATPase domain was said to be engaged in an energetic tug‐of‐war between the ATP‐bound and client‐bound conformations.11 When ATP is bound in the ATPase domain, DnaK assumes a conformation with a fast off‐rate for polypeptides in its client‐binding domain. Binding of the client polypeptide stimulates a rapid conformational change in DnaK that accelerates ATP hydrolysis.13 ATP hydrolysis stabilizes DnaK in a conformation that has a very slow off‐rate for polypeptides. Polypeptides tend to remain bound to the client‐binding domain until the nucleotide exchange factor, GrpE, replaces the ADP with ATP to restart the cycle.14, 15 DnaK's cyclic transitions between these two stable conformational states support its chaperone actions on unfolded proteins.16

Hsp40/DnaJ contains multiple client‐binding domains and a conserved J‐domain that stimulates ATP hydrolysis in Hsp70/DnaK.17 DnaJ has been found to facilitate the capture of client proteins in DnaK's client‐binding domain with the formation of DnaJ‐client‐DnaK ternary complexes. Through experiments using Jdp5, a recombinant J‐domain (Jd) fused with a client polypeptide (p5), we have shown that the J‐domain of DnaJ is sufficient for coordinating both ATP hydrolysis in the ATPase domain and client capture in the client‐binding domain.1 NMR studies have demonstrated that the ATPase domain is the site for J‐domain interaction with DnaK.18 However, the actual mechanism through which the J‐domain coordinates allostery between the two functional domains of DnaK has remained unresolved.

Experiments with recombinant forms of Jd and Kase reveal that their interaction involves more than simple binding. The substitution of asparagine for aspartate in the conserved HPD tripeptide results in a mutant (JdD35N) that was unable to stimulate DnaK action.1 This is despite JdD35N's higher‐than‐wildtype affinity for Kase.1 Greater induced‐fit conformational strain was observed in JdD35N compared to Jd when bound to Kase.19 The induced‐fit conformational strain involved the bending of J‐domain helix II (residues 18–31). Previous NMR experiments have mapped the Jd‐Kase interaction to this region of the J‐domain.18 Helix II bending was indicated by the presence of a periodic pattern of negative and positive proton chemical shift perturbations upon Jd‐Kase binding.19 This distinctive pattern was previously observed for helix‐bending in coiled‐coil structures.20, 21 Jd helices II and III form a coiled coil.22 The extent of Kase‐induced helix II bending was observed to be greater for the nonfunctional mutant JdD35N.19

The present studies employ variants of the DnaK ATPase domain that inhibit aggregation and modify its ATPase activity. The DnaK ATPase domain was reported to form dimers in solution, and cysteine 15 was found to be essential for dimerization.23 Cysteine 15 is not universally conserved. Alanine replaces cysteine at the position corresponding to C15 in several bacterial homologs.24 A study by Winter et al. showed that a C15A mutant of DnaK has less activity than its wildtype form, but is able to complement wild‐type DnaK when expressed in dnaK7 and ΔrpoH strains, which lack functional DnaK.23 To focus the investigation on the details of DnaJ (i.e., Hsp40) and DnaK (i.e., Hsp70) monomer interactions, a cysteine‐to‐alanine mutation was introduced at residue 15 of the DnaK ATPase domain to generate KaseC15A. Subsequent experiments were conducted on the monomer‐biased KaseC15A and its derivatives, KaseC15AK70A and KaseC15AT199A. The K70A mutation was previously shown to inhibit conformational changes in DnaK.25 The T199A mutation was previously shown to inhibit ATP hydrolysis in DnaK, without disrupting conformational change.25

Our investigations centered on the role of J‐domain conformational strain in mediating regulatory protein interactions with the DnaK ATPase domain. This study involved the development of methodologies for modulating binding affinity of the recombinant DnaJ J‐domain (Jd) and recombinant DnaK ATPase domain (Kase) and testing the effects on the protein interactions. We find that Jd function depends on the rigidity of its helix II, which pushes back against a bend imposed by binding to Kase to promote ATP hydrolysis.

Results

Modulation of Jd‐Kase binding interactions

The present study supports the initial observation that composite backbone amide ¹⁵N‐¹H chemical shift perturbations in Jd correlated with the free energies of binding for different Jd and Kase pairings.19 Results with four Jd‐Kase pairs (Jd‐Kase; JdD35N‐Kase, Jd‐KaseR167A and JdD35N‐KaseR167A) were presented in the original study. Three additional Jd‐Kase pairs (Jd with KaseYND, KaseEV, and KaseDRQ) are included here. The amino‐acid substitutions in KaseYND (Y145A, N147A, D148A), KaseEV (E271A,V218A), and KaseDRQ (D148A, R151A, Q152A) were found to disrupt the ability of DnaJ to stimulate the DnaK ATPase (Ref. 31 and data not shown). The combined results demonstrate a linear relationship between chemical shift perturbation (Δδ_c) in Jd and free energy of binding to Kase (ΔG°′_binding), and that relationship persists even when the most extreme values for JdD35N‐Kase pairings are omitted (Fig. 1). The perturbation‐affinity correlation was interpreted as evidence of conformational rigidity in the wild‐type Jd.19 This feature may be likened to the elastic modulus of a macroscopic material.26 The Jd‐KaseYND pairing exhibited lower than expected perturbation, and this may be due to its particularly severe effect on the putative Jd‐Kase interface.27 The potential for mutations in the protein‐protein interface to confound the perturbation‐affinity correlation motivated a search for a different method of varying binding affinity.

Figure 1.

Graph of integrated composite amide ¹H‐¹⁵N NMR chemical shift perturbations (ΣΔδ_c) versus free energies of binding (ΔG°′_binding) for pairs of the DnaJ J‐domain ([¹⁵N]Jd or [¹⁵N]Jd D35N) and DnaK ATPase domain (Kase, KaseDRQ, KaseEV, KaseR167A, or KaseYND). A linear relationship with positive slope was observed. Much larger ΔG°′_binding and ΣΔδ_c were observed for domain pairs involving the defective JdD35N. Nevertheless, the linear relationship holds when the data are restricted to domain pairs involving only the functional J‐domain (dashed line). In our previous study, the positive slope observed with only Jd, JdD35N, Kase, and KaseR167A was attributed to an incrementally greater conformational strain in the Kase‐bound J‐domain conformation at increasing binding affinity.19

Potassium chloride concentration modulated J‐domain binding affinity for Kase and KaseC15A

Isothermal titration calorimetry (ITC) was used to determine the binding affinities of Jd for Kase in different buffer conditions. Kase has been observed to dimerize in solution (data not shown), and this process interferes with the interpretation of ITC data. To prevent dimer formation, a nondisruptive Cys‐to‐Ala substitution23 was incorporated to yield KaseC15A. ITC experiments revealed that buffer potassium chloride concentration modulates J‐domain binding affinity for both Kase and KaseC15A. The observed slope for the Jd‐KaseC15A interaction was significantly steeper than the slope for the Jd‐Kase interaction [Fig. 2(A)]. These results suggested that KaseC15A has greater susceptibility than Kase for binding affinity modulation by potassium chloride. Jd‐KaseC15A binding affinity modulation was observed to occur with the variation of either sodium ion (Na⁺) or potassium ion (K⁺) concentration, but the effect of K⁺ was greater [Fig. 2(B)].

Figure 2.

Modulation of Jd‐Kase binding affinities by salt concentration. (A) Change in potassium chloride concentration (Δ[KCl]) modulates the binding of Jd to both Kase and KaseC15A. (B) Change in sodium chloride concentration elicits a smaller effect (P < 0.05), suggesting a potassium‐ion specific reaction. (C) Changes in [KCl] modulates the binding of Jd and JdD35N to the different Kase variants used in the study. Data for Jd‐KaseC15A with KCl were the same in panels A‐C.

Variation of K⁺ concentration modulated binding affinities of different Jd‐Kase pairings

ITC experiments were conducted to determine the effect of change in potassium‐chloride concentration (Δ[KCl]) on the binding affinities of different Jd‐KaseC15A pairings (Jd‐KaseC15A, Jd‐KaseC15AK70A, JdD35N‐KaseC15A, and JdD35N‐KaseC15AK70A). Raising the KCl concentration reduced binding affinity between the different Jd and KaseC15A pairs to similar degrees within the 2–10 mM KCl range [Fig. 2(C)]. The following NMR experiments were limited to this range of KCl concentration.

Effect of Kase binding on Jd conformation

The mutant JdD35N experienced greater conformational strain than wildtype Jd when bound to KaseC15A

Chemical shift perturbations revealed differences in the modes of interaction for Jd and JdD35N binding to KaseC15A at the same binding stress (ΔG°′_binding = −7 kcal). KaseC15A binding to Jd produced chemical shift perturbations that were concentrated in helix II; whereas, KaseC15A binding to JdD35N produced chemical shift perturbations in helix II, loop II‐III, and helix III [Fig. 3(A)]. Furthermore, where perturbations were shared by the two J‐domains, they were of greater magnitude for JdD35N than for Jd.

Figure 3.

Structural distribution of Δδ_c for the [¹⁵N]J‐domain—KaseC15A interaction. (A) Bars indicate Δδ_c for the individual backbone amides. For Jd, significant Δδ_c (greater than 0.05 ppm) were mainly focused on helix II. For JdD35N, significant Δδ_c were observed in both helices II and III. Inset graphs indicate Δδ_c versus ΔG°′_binding at representative residues for binding in various [KCl]. (B) Bars indicate slopes of Δδ_c versus ΔG°′_binding for each amide. Slopes for amides in both Jd and JD35N were largely uniform and negative, suggesting that [KCl] modulates a conformational change in KaseC15A that is responsible for the change in Δδ_c.

Potassium chloride modulated KaseC15A conformation and Jd induced fit

The interactions of Jd and JdD35N with KaseC15A were found to have a similar dependence on [KCl], as revealed by chemical shift perturbations. Individual amide resonance perturbations were affected to similar extents for both Jd and JdD35N. Slopes of Δδ_c versus ‐ ΔG°′_binding for 55 of the 58 monitored crosspeaks were indistinguishable for Jd and JdD35N [Fig. 3(B)]. Since increased [KCl] was found to reduce the binding affinity between the two proteins, we expected it to result in smaller chemical shift perturbations. Interestingly, the opposite effect was observed. Raising the potassium chloride concentration resulted in larger chemical shift perturbations for both Jd and JdD35N, as revealed by the negative slopes of Δδ_c versus –ΔG°′_binding [Fig. 3(A), insets, 3(B)]. We combined the chemical‐shift perturbations into a single composite value (∑Δδ_c), which served as a measure of the overall effect of Δ[KCl] on KaseC15A interaction. Perturbations for JdD35N were larger than for Jd; nevertheless, similar slopes of ∑Δδ_c versus ‐ ΔG°′_binding were observed for the two J‐domains [Fig. 5(A)]. We hypothesized that increased potassium chloride concentration promoted a conformational change in KaseC15A. This translated to increased conformational strain for bound J‐domains. The similar slopes observed for Jd and JdD35N suggested that the KCl‐induced variation in ∑Δδ_c was governed by the conformation of KaseC15A, rather than by the affinity for KaseC15A.

Figure 5.

Modulation of J‐domain chemical shift perturbations. (A) Higher potassium chloride concentrations induced greater chemical shift perturbations for both Jd and JdD35N concomitant with reduced affinity for KaseC15A (negative slopes of ∑Δδ_c versus –ΔG°′_binding). The gray arrow highlights a hypothetical potassium‐dependent conformational change in KaseC15A. (B) Use of a conformationally restricted mutant KaseC15AK70A results in greater J‐domain chemical shift perturbations with increased Jd‐Kase binding affinity (positive slopes of ∑Δδ_c versus –ΔG°′_binding). The black arrow highlights the increased bend in J‐domain helix II at low potassium concentration.

The K70A mutation reversed the slope of ∑Δδ_c versus ΔG°′_binding in JdD35N

To test our hypothesis that the increased perturbations were caused by a Kase‐dependent conformational change, experiments were conducted using a double mutant, KaseC15AK70A. The K70A mutation has been documented to disrupt the ATP‐dependent conformational change in DnaK.25 The overall pattern of perturbations for J‐domain binding to KaseC15AK70A was comparable to that for binding to KaseC15A, suggesting that the substitution did not affect the binding site [Fig. 4(A)]. However, the slope of Δδ_c versus –ΔG°′_binding for most residues changed from negative to zero or positive [Fig. 4(A), insets, 4(B)]. In addition, the modulation of Jd‐KaseC15AK70A binding with KCl was observed to disproportionally affect residues in helix II of the J‐domain structure [Fig. 4(B)]. This coincides with the site for Jd‐Kase interaction and reflects the bending of helix II.19 A comparison of ∑Δδ_c versus –ΔG°′_binding for KaseC15A and KaseC15AK70A shows that conformational strain was unchanged (Jd) or increased (JdD35N) with increasing Kase‐binding affinity in the absence of the Kase conformational change [Fig. 5(B)].

Figure 4.

Structural distribution of Δδ_c for the [¹⁵N]J‐domain – KaseC15AK70A interaction. (A) Bars indicate Δδ_c for the individual backbone amides. Inset graphs indicate Δδ_c versus ΔG°′_binding at representative residues for binding at various [KCl]. (B) Bars indicate slopes of Δδ_c versus ΔG°′_binding for each amide. Slopes for most amides in both Jd and JD35N were positive, as expected if the K70A substitution blocked the potassium‐dependent Kase conformational change. The nonuniform distribution of mostly positive slopes is consistent with incrementally greater conformational strain in Jd at increasing J‐domain‐KaseC15AK70A binding affinity. More strain was observed for JdD35N than for Jd, suggesting that JdD35N is less rigid.

The nucleotide state (ATP/ADP) of Kase affected its binding interactions with Jd/JdD35N

Hsp70/DnaK function involves the transition between ATP‐ and ADP‐bound conformations. The binding of these two nucleotides has been associated with distinct conformations for Kase.28 The ability of Jd to stimulate ATP hydrolysis in Kase suggests its involvement in the transition between the ATP‐ and ADP‐bound Kase conformations. In order to observe the ATP‐bound state, we used a double‐mutant, KaseC15AT199A, in which the T199A mutation blocks ATP hydrolysis without impairing the conformational change induced by ATP binding.25

Differences in the bound nucleotide were observed to translate into changes in the Kase binding interactions with Jd and JdD35N. The changes were observed in terms of binding affinity, range of affected residues, and intensities of chemical shift perturbation. In terms of binding affinity, JdD35N was observed to have higher binding affinity for the ATP‐bound KaseC15AT199A compared to the ADP‐bound form. In contrast, Jd was observed to have similar binding affinities for KaseC15AT199A with either nucleotide‐bound (Supporting Information Table S1). A wider range of JdD35N residues was affected with the ATP‐ADP transition, and greater differences in chemical shift perturbations were observed with the nucleotide‐state transition (ΔΔδ_c:ADP‐ATP°) for JdD35N compared to Jd [Fig. 6(A)]. The greatest differences in chemical shift perturbations were observed near helix II. The summed chemical shift perturbations in helix II for JdD35N were significantly larger for ADP‐bound KaseC15AT199A, as compared to the ATP‐bound KaseC15AT199A [Fig. 6(B)].

Figure 6.

The nucleotide state (ATP/ADP) of Kase affects its binding interactions with Jd or JdD35N. (A) Composite chemical shift perturbations for interaction with ATP‐bound or ADP‐bound KaseC15AT199A. Greater chemical shift perturbations were observed for JdD35N. The largest perturbations were observed in helix II. (B) The integrated chemical shift perturbations for JdD35N were significantly larger for binding to the ADP‐bound KaseC15AT199A than for binding to the ATP‐bound form (P < 0.005). (C) These results suggest a model for Jd‐Kase interaction wherein the J‐domain experiences greater stress when binding to the ADP‐bound form of Kase and the less‐rigid JdD35N displays greater conformational strain.

Analysis of J‐domain conformational strain

JdD35N exhibits faster amide‐group hydrogen/deuterium exchange rates than Jd

Protein conformational flexibility/rigidity can be analyzed using backbone amide‐group hydrogen/deuterium (H/D)‐exchange rates. The rate of H/D‐exchange for a given residue serves as a measure of solvent accessibility. Residues within ordered structures are generally less accessible to solvent and have slower H/D‐exchange rates than residues in unstructured parts of the protein.29 H/D‐exchange rates were determined for unbound Jd and JdD35N by monitoring the recovery of crosspeak intensities in successive HSQC‐NMR experiments. Exchange‐rate constants (k_HX) were calculated by fitting the crosspeak‐growth curves to single exponentials. The H/D exchange rates we observed for Jd are comparable to those reported by Huang et al.30 Seventy‐five percent of the amide groups with reliably‐fit regression lines (R ² ≥ 0.64) showed significantly different exchange rates between Jd and JdD35N (P < 0.05). Most of these amide groups reported faster H/D‐exchange rates for the JdD35N, and none reported significantly slower H/D‐exchange rates for the JdD35N (Fig. 7). These results provide further evidence for the decreased rigidity of JdD35N compared to Jd.

Figure 7.

Hydrogen‐Deuterium Exchange in Jd and JdD35N. (A) Representative graphs of time‐dependent changes in the intensities of backbone amide proton resonances. (B) Hydrogen‐deuterium exchange rates in Jd (k_HX‐Jd) and JdD35N (k_HX‐D35N). Data are presented as log k values. Recovery rates were measured in seconds. (C) Ratio of recovery times for individual amide protons observed in both Jd and JdD35N. At most positions, amide protons of wild‐type Jd exhibited longer recovery times.

Quantification of conformational strain through helix bending

Conformational strain in helix II of the J‐domain was quantified in terms of the bend in the helical axis, measured as an angle, theta (θ), which represents the deviation of the top part of a helix from its original axis. This deviation is due to the lengthening and shortening of opposite sides of the helix leading to a bend (Fig. 8).

Figure 8.

Angular quantification of helix bending (ΔΘ_ab). The bending of a helix may be described by the deviation from its original axis. The angular deviation (ΔΘ_ab) is dependent on the changes in the lengths of the helix sides (A and B; or C and D). Helix side lengths were estimated from the lengths of H‐bonds inferred from amide proton chemical shifts in helix II. Bend direction goes toward the helix side with largest decrease in H‐bond length. (A) Top and side views of the J‐domain structure (PDB ID: 1BQZ).^30. The molecular models were generated using Rastop (P. Valadon, 2000). (B)A representation of opposite helix sides (e.g., A and B, C and D) and the expected angle (θ) of helix bending along these axes. (C) The helix bend angles for the different Jd‐Kase pairings are represented as vectors showing the direction and magnitude of the bend with respect to the previously described helix side axes (i.e., A‐B; C‐D). Both Jd and JdD35N were observed to have bent conformations in their unbound forms. The initial bend directions were observed to differ between Jd and JdD35N. (D) Kase‐induced bending directions for Jd and JdD35N. Jd bending was in the same direction as the bend that corresponds to the difference in the structures of 1BQ0 and 1BQZ. These structures represent the unbound J‐domain, residues 2–78 (1BQZ), and the J‐domain in a larger molecule, residues 2–104 (1BQ0), which includes the G/F region of Hsp40 and was said to approximate a DnaK‐bound J‐domain structure.³⁰ The Kase‐induced bend direction for the dysfunctional mutant JdD35N differs from the bend direction of both Jd and 1BQ0 structures.

Changes in protein helix dimensions have been associated with changes in amide H‐bond lengths. Amide H‐bond lengths, R, were computed from observed proton chemical shifts based on the following formula adapted from Wishart et al.:31

$$R_{\text{observed}}=\frac{19}{\left(\delta \mathrm{H}-\text{observed}°\hspace{0.17em}-\hspace{0.17em}\delta {\mathrm{H}-}_{\text{random}\hspace{0.17em}\text{coil}}+\hspace{0.17em}9.7\right)}$$ (1)

The factor δH‐observed° is the normalized proton chemical shift at a given condition. This corresponds to the observed change in proton chemical shifts at one condition (e.g., 4 mM KCl) normalized against the total change observed within the range (2–10 mM KCl).

The effects of different factors on the amide H‐bond lengths (ΔR _X) of bound proteins are given by the following formula.

$$\begin{array}{l}\Delta R_{\mathrm{X}}=\left(R_{\text{bound};\mathrm{X}-\text{final}}– R_{\text{bound};\mathrm{X}-\text{initial}}\right)\\ -\left(R_{\text{unbound};\mathrm{X}-\text{final}}– R_{\text{unbound};\mathrm{X}-\text{initial}}\right)\end{array}$$ (2)

$${\theta }_{\text{AB}}=\hspace{0.17em}\left(a\hspace{0.17em}+\hspace{0.17em}b\right)/w$$ (3)

$${\theta }_{\text{CD}}=\hspace{0.17em}\left(c\hspace{0.17em}+\hspace{0.17em}d\right)/w$$ (4)

Equations (3) and (4) present how the angles of helix II bending, θ _AB and θ _CD, for Kase‐bound Jd and JdD35N were computed. The derivation for Equation (3) is presented in the Supporting Information.

Values for a and b correspond to the changes in length for opposite sides (e.g., sides A and B) of the helix [Fig. 8(A,B)]. The residues representing the four sides of the helix are listed in Supporting Information Table S2. Deviations between opposite sides A and B, or C and D are stated to occur in the A‐B, or C‐D axes, respectively. The direction of the bend is toward the side with greater decrease in H‐bond lengths. The parameter w corresponds to the width of the helix measured from available crystal structures (2.98Å).

KaseC15AK70A binding (ΔG°′_binding = ‐ 7 kcal) induced modest but significant bending of helix II along the A‐B axis for Jd (1.20 ± 1.2°) and JdD35N (1.6± 1.2°). At higher binding affinity (ΔG°′_binding = −8 kcal), the A‐B bend increased to 1.4 ± 1.2° for Jd and to 2.0 ± 1.3° for JdD35N.

Greater helix II bending was observed for Jd binding to the ATP‐hydrolysis deficient mutant, KaseC15AT199A. Binding to ATP‐bound KaseC15AT199A resulted in the bending of Jd by 1.9 ± 1.2° and JdD35N by 2.5 ± 1.2° along the A‐B axis. Binding to the ADP‐bound KaseC15AT199A resulted in greater bending of both Jd and JdD35N. Jd was bent by 2.1 ± 1.2° and JdD35N by 2.8 ± 1.2°.

The difference in Jd bending when bound to ATP‐bound and ADP‐bound Kase is hypothesized to approximate the J‐domain's response to Kase conformational change due to ATP‐hydrolysis. The functionally relevant change in nucleotide state from ATP to ADP was observed to induce greater bending in JdD35N compared to Jd. Results for the deviations in the J‐domain helix sides are presented in Supporting Information Tables S3 and S4.

J‐domain conformation in the unbound state

To further investigate the effect of Kase‐binding on the conformation of the J‐domain, the extent of helix II bending was quantified for both its unbound and bound structures. Similar to the bound structures, the direction and extent of helix bending were computed based on proton chemical shifts that translated to different helix side lengths (data not shown).

J‐domain helix II has a bent conformation in the unbound state. The extents and directions of the bends were different for Jd and JdD35N. Unbound JdD35N helix II bends more toward the B‐side of the helix, while unbound Jd helix II bends more toward the D‐side. Binding to KaseC15AK70A changes the direction of the bend for both J‐domain types, which has been represented as vectors in Figure 8(C).

In order to focus the comparison on the difference in bend direction and magnitude, vectors for the change in Jd and JdD35N are shown originating from a common “unbound” J‐domain position [Fig. 8(D)]. The bend observed for the dysfunctional JdD35N was in the opposite direction, with respect to the D‐C axis.

In previous work by Huang et al.,30 the structures of DnaJ(1–78) and DnaJ(1–104) were described as mimics of the free and DnaK‐bound structures of the J‐domain, respectively. The putative DnaK‐bound conformation included an increase in bending of helix II, and the bend was in approximately the same direction as observed in our actual DnaK‐bound Jd helix II [Fig. 8(D)]. The similarity in the bend directions of DnaK‐bound Jd and DnaJ(1–104) suggests the attainment of similar conformations in these functional J‐domain forms. In contrast the bend in the nonfunctional JdD35N was in a different direction from that of DnaJ(1–104).

Discussion

A molecular machine's moving parts often must balance conflicting needs for specificity, affinity, and activity. The interaction between the Hsp40 J‐domain and Hsp70 ATPase domain is characterized by low affinity and conformational polymorphism. The present studies reveal that the conformations of both partners are coupled to the binding affinity. The charge complementarity of the interaction surfaces of Jd and KaseC15A suggested the involvement of electrostatics in the binding interaction,18 and the thermodynamic signature from calorimetry was similar to that of nonspecific protein‐DNA binding.1 Indeed, we find that the binding affinity is reduced at elevated ionic strength, suggesting that electrostatic interactions were reduced by Debye‐Hückel screening. However, the higher degree to which potassium chloride modulates binding affinity compared to sodium chloride indicates that potassium has an ion‐specific interaction with the Jd‐Kase complex. The potassium ions most likely interact with the ATP‐binding site within Kase.32, 33 Binding of potassium in the ATP binding cleft of Kase has been found necessary for orienting ATP,32, 33 and for the ATP‐binding‐driven conformational change.6, 25 We conclude that increased potassium chloride alters the KaseC15A conformation, resulting in increased Jd conformational strain when it binds to KaseC15A. Control experiments observing the NMR chemical shifts of either Jd alone or JdD35N alone in the presence of increasing potassium ion concentrations confirmed that the potassium‐dependent increase in Jd conformational strain requires binding to Kase (data not shown).

Jd conformational strain is coupled to changes in the Kase conformational state. A Kase conformational change was previously hypothesized to occur simultaneously with the induced‐fit conformational strain in Jd.19 Here, we propose that the Kase‐conformational change promoted by KCl imposes conformational strain on the bound J‐domain [either Jd or JdD35N; Fig. 5(A)]. Thus, larger KaseC15A‐induced chemical shift perturbations were observed at higher [KCl] (Fig. 3). This result was initially unexpected because the Jd‐KaseC15A affinity is reduced at higher [KCl], but it makes sense in the context of a KCl‐dependent KaseC15A conformational change that causes more strain on Jd. The notion that the effect of KCl is exerted through the KaseC15A conformation rather than Jd‐KaseC15A binding affinity is also supported by the nearly uniform slopes of Δδ_c versus ‐ΔG°′_binding for individual residues across the J‐domains and the similarity of the whole‐protein ∑Δδ_c versus ‐ΔG°′_binding for Jd and JdD35N in these experiments [Fig. 5(A)]. The reversal of negative slopes in ∑Δδ_c versus ‐ΔG°′_binding for Jd and JdD35N binding to the conformational‐change‐deficient KaseC15AK70A further supports the proposal that the KCl‐dependent Kase conformational change imposes strain on Jd. Disruption of this conformational change with the K70A mutation converted the negative slope of ∑Δδ_c versus ‐ΔG°′_binding for Jd to essentially zero (although many residue‐specific slopes became positive), and converted the negative slope of ∑Δδ_c versus ‐ΔG°′_binding for JdD35N to positive [Figs. 4 and 5(B)].

The conversion of slopes for Δδ_c versus ‐ΔG°′_binding to positive values upon Jd binding to KaseC15AK70A produced the behavior that we anticipated from experiments with various mutants of the J‐domain and Kase (Fig. 1). When the Kase conformation is locked‐down by K70A, then it is possible to observe the gradual increase in Jd conformational strain (Δδ_c) with increasing stress of binding to Kase (‐ΔG°′_binding). The shift from uniformly negative slopes of Δδ_c versus ‐ΔG°′_binding [Fig. 3(B)] to a broad peak of positive slopes centered on helix II [Fig. 4(B)] indicates that helix II experiences most of this conformational strain, which we propose takes the form of helix II bending. It is unclear how the gradual strain‐stress relationship contributes to DnaJ‐DnaK function. It may have arisen as a necessary consequence of adopting a low‐affinity, flexible (with regard to orientation) interaction that depends on the formation of heterogeneous ternary complexes with non‐native client proteins. In the ternary complex, the J‐domain may approach Kase from a sub‐optimal orientation. It would be advantageous if the J‐domain could still deliver the allosteric signal to the ATPase domain, albeit less strongly, from a sub‐optimal orientation.

The proposed bend in helix II is compatible with the broad envelope of binding modes that have been proposed for J‐domain binding to the DnaK ATPase domain. Ahmad et al. proposed that the J‐domain engages DnaK through a “tethered” binding in which the J‐domain retains a great deal of rotational freedom.34 The authors present evidence for a Jd‐Kase binding site and orientation based on NMR and molecular dynamics simulations. The authors propose a binding interaction that involves residues of Jd (K26, and M30) facing the E206‐T221 loop in Kase. This corresponds to Jd helix II side B according to our notation (Fig. 8). Our current results show that Kase binding causes the ends of helix II to bend toward Kase, as if to wrap the Jd helix around the protruding E206‐T221 loop in Kase, although the actual amount of helix bending is small (∼2 degrees). The suggestion that the Jd retains some rotational freedom while bound to Kase would accommodate the J‐domain conformational plasticity that we describe here in the form of gradual helix II bending with increasing affinity for Kase.

Observations on the extent and direction of bending in J‐domain helix II show differences in the conformations attained by Kase‐bound Jd and JdD35N. Binding to Kase results in Jd helix II bending toward sides B and C. In contrast, JdD35N helix II bending goes toward helix sides B and D. The difference in bending direction indicates a difference in the conformations attained with Kase interaction. The altered conformation in the less rigid and dysfunctional JdD35N represents an interaction between a J‐domain and Kase that does not stimulate ATP hydrolysis. The dysfunction may lie in the reduced ability of JdD35N to resist the Kase conformational change.

Model for the Hsp40‐Hsp70 functional cycle

Jd binding may promote the active kase conformation by restricting subdomain movement

Our results support the hypothesis that the J‐domain restricts Kase subdomain movement in order to stabilize a conformation that catalyzes ATP hydrolysis. Residues in the interface between Kase subdomains IA and IB have been documented to experience the greatest changes in chemical shift between ADP and ATP bound Kase conformations.35 The rigidity of the bound J‐domain would determine the degree to which Kase subdomain movement may be restricted by J‐domain binding [Fig. 6(C)]. The rigidity of the bound J‐domain thus dictates the amount of conformational freedom for the Kase subdomains. A more rigid helix would be more efficient at restricting subdomain movement, promoting the active Kase conformation associated with ATP hydrolysis.36 The greater chemical shift change observed with the more flexible D35N mutant suggests a decreased capability for the restriction of subdomain movement, leading to an inability to stimulate DnaK function.

Our experiments with the KaseC15AT199A mutant identify changes in strain on the JdD35N structure associated with the functionally‐relevant nucleotide‐induced conformational changes. Larger chemical shift perturbations were observed with ADP‐bound JdD35N compared to the ATP‐bound forms. Thus, the ATP‐bound KaseC15AT199A imposes less strain on JdD35N. The fact that greater strain was observed for JdD35N compared to Jd when bound to KaseC15AT199A (with ATP or ADP) is consistent with Jd stiffness pushing back on the conformation of KaseC15AT199A. The rigid Jd resists the conformational strain and acts like a semi‐elliptical spring to push the domains of Kase into the active conformation. Since the dysfunctional JdD35N lacks rigidity, it cannot resist the conformational strain imposed by Kase, making it unable to promote the active Kase conformation [Fig. 6(C)]. We suppose that Kase‐ATP is more similar to the transition‐state than Kase‐ADP because Kase‐ATP requires less helix II bend than Kase‐ADP and JdD35N binding to Kase‐ATP is more favorable than binding to Kase‐ADP. After hydrolysis the affinity of JdD35N is reduced. The difference in binding affinity for the two Kase states contributes to the cyclic assembly/disassembly of DnaJ‐DnaK‐client complexes. [Here, we are using JdD35N as a probe of Kase conformational states. Presumably, similar observations with Jd were confounded by the ability of Jd to push back on the conformation of Kase. Nevertheless, we raise the caveat that JdD35N‐binding to Kase differs from Jd‐binding to Kase in strength and geometry (see above).]

These observations fit into the overall model for DnaJ‐DnaK operation previously described as follows1 (Fig. 9). The J‐domain effectively couples ATP hydrolysis in DnaK with binding to authentic DnaJ‐DnaK client proteins. Chaperone power depends on the ability of a client to attract both DnaJ and DnaK.37 Once both are assembled onto the client by virtue of their respective client‐binding activities, the proximal Jd stimulates the DnaK ATPase activity. Thus, Jd is expected to act as a pure catalyst to stabilize DnaK in an ATPase‐active conformation. Presumably, this is mediated by its direct interaction with the DnaK ATPase domain and synergizes with the tug‐of‐war signal11 from the DnaK peptide‐binding domain.

Figure 9.

Model for the Hsp40‐Hsp70 functional cycle. The Hsp40‐bound unfolded polypeptide binds to the Hsp70. The Hsp40 J‐domain binds to the Hsp70 ATPase domain. The J‐domain promotes ATP hydrolysis in Hsp70, which requires J‐domain rigidity. The dysfunctional mutant JdD35N has decreased rigidity and is unable to promote ATP hydrolysis. ATP hydrolysis promotes closure of the Hsp70 peptide‐binding domain and dissociation of the J‐domain.

This study demonstrates the importance of protein structural rigidity for the function of the Hsp40/DnaJ‐Hsp70/DnaK molecular machine. The identification of structural rigidity as a requirement for maintaining protein function provides another dimension through which the needs of specificity, affinity and activity of molecular machines and cellular complexes may be regulated and controlled.

Materials and Methods

Plasmid expression vectors

Plasmids encoding Kase and Jd were pRLM157 and pPLJ(1–75), respectively, which were generously provided by Dr. R. McMacken. The QuickChange site‐directed mutagenesis kit (Stratagene) was used to generate the following mutations on these plasmids. The mutation encoding C15A was generated on pRLM157, resulting in pRLM157C15A, which produces KaseC15A. An additional mutation was introduced in pRLM157C15A to generate pNAB09, which produces KaseC15AT199A, a DnaK ATPase domain with substitutions at residue 15 (C15A) and residue 199 (T199A). Similarly, an additional mutation was introduced in pRLM157C15A to generate pNAB11, which produces KaseC15AK70A, a DnaK ATPase domain with substitutions at residue 15 (C15A) and residue 70 (K70A). The mutation encoding D35N was introduced to pPLJ(1–75) as described.18

Protein expression and purification

Expression and purification of the Hsp40/DnaJ J‐domains (Jd and JdD35N) and Hsp70/DnaK ATPase domains (KaseC15A, KaseC15AK70A, and KaseC15AT199A) were as described by R. McMacken and coworkers.38, 39 Isotopically enriched [¹⁵N]Jd and [¹⁵N]JdD35N were prepared as described.18

Nucleotide removal from KaseC15AT199A

Multiple passes through a quaternary amine ion exchange (MonoQ) column resulted in the removal of the nucleotide from KaseC15AT199A. Successive MonoQ runs and repooling of the selected fractions was done while monitoring the changes in the elution readout. Intensities of peaks assigned to nucleotide, nucleotide‐bound protein, and nucleotide‐free protein are observed to vary with successive purification runs. The process was repeated until the peak assigned for the free nucleotide was no longer detectable. The nucleotide content of the elution fractions was also analyzed by UV absorbance at 260 nm and 280 nm. Samples having a 280/260 ratio of 1.5 or higher were regarded as nucleotide‐free.40

Nucleotide‐loading into nucleotide‐free KaseC15AT199A

Aliquots of nucleotide‐free KaseC15AT199A were dialyzed using dialysis cassettes with 10000‐MWCO membranes. Three milliliters of protein was dialyzed in a liter of 50 mM MOPS‐NaOH, pH 6.8, 2 mM KCl, 1 mM MgCl_2, and 1 mM nucleotide (ADP or ATP) for 4 hours. Proteins which were used in the same experiments as these KaseC15AT199A aliquots (i.e., Jd, JdD35N, ¹⁵N‐Jd and ¹⁵N‐JdD35N) were also dialyzed in the same buffer. The proteins were aliquoted post‐dialysis, flash‐frozen and stored at at −80° C until further use. Five‐milliliter aliquots of sterile‐filtered dialysis buffers were also flash‐frozen and stored at at −80°C until further use.

Isothermal titration calorimetry

Isothermal titration calorimetry experiments were performed on a MicroCal VP‐ITC unit. The sample cell was equilibrated to 25°C for a week prior to data acquisition. Proteins were dialyzed against 50mM morpholinoethanesulfonic acid (MOPS) – NaOH, pH 6.8. The sodium or potassium ion concentration was modified by addition of appropriate amounts of 1 M KCl or 1 M NaCl. Kase proteins (Kase, KaseC15A, KaseC15AK70A, KaseC15AT199A) were loaded in the sample cell at 0.05 mM. Jd proteins (Jd and JdD35N) were loaded into the syringe injector at either 2 mM or 1 mM concentration. Protein solutions were degassed with the ThermoVac vacuum system (MicroCal) for 5 min prior to each experiment. Jd proteins were titrated into the sample cell at 10‐μl steps with 240‐s intervals. The association binding constant (Ka) was determined by fitting the calorimetry data with a one‐site binding curve using the MicroCal – ITC software (MicroCal). Free energy of binding was calculated using the formula: ΔG = –RTln(1/Ka).

N‐H HSQC NMR

NMR chemical shift perturbations in [¹⁵N]Jd with Kase variants were analyzed as described previously.19 Samples were prepared initially in 550 μL of 50 mM MOPS‐NaOH, pH 6.7, and then 48 μL D₂O and 1.2 μL of 80 mM NaN₃ were added. The final concentration of [¹⁵N]Jd prior to addition of a Kase variant was 280 μM. The Kase variants dissolved in 50 mM MOPS‐NaOH, pH 6.7 were added in increments that produced final concentrations as follows: KaseYND, 126, 185, and 242 μM; KaseEV, 157, 206, and 385 μM; and KaseDRQ, 142, 207 and 268 μM. The fraction of bound [¹⁵N]Jd ranged from 28–52%. All NMR spectra were recorded at 25°C on a Bruker DRX 500 spectrometer equipped with a double‐resonance broadband (BBI) probe. Proton chemical shifts were referenced to external 3, 3, 3‐trimethylsilyl propionate (TSP). ¹⁵N chemical shifts were referenced indirectly relative to the TSP ¹H frequency. Two‐dimensional ¹⁵N‐¹H HSQC NMR spectra employed the standard Bruker pulse sequence phase sensitive Echo/Antiecho‐TPPI gradient selection41, 42 (invietgp) with 1000 acquisition points for spectral width 2003 Hz in the ¹H dimension and 256 increments in t ₁ for 1800 Hz in the ¹⁵N dimension.

Prior to addition of salt solutions, the NMR samples were comprised of 100 μM [¹⁵N]Jd or [¹⁵N]JdD35N and Kase proteins in 600 μl of a 50 mM MOPS pH 6.8, 10% D₂O solution. Potassium chloride concentration was varied from 2 to 10 mM in 2‐mM steps by the addition of 2.4 μl of a 0.5‐M KCl solution to the NMR sample. Acquisition of spectra was performed at 25°C in a Bruker DRX 500 spectrometer equipped with a 5 mm Bruker inverse triple‐resonance (TXI) probe. Pulses were calibrated using standard methods. Two‐dimensional ¹⁵N‐¹H HSQC spectra were acquired using the standard Bruker pulse sequence with sensitivity improvement and phase‐sensitive echo/antiecho‐TPPI gradient selection (invif3gpsi). One thousand twenty four (1024) acquisition points for spectral width 2003.205 Hz were acquired for the ¹H dimension. Two hundred fifty six (256) increments were done in t1 for 2027.37 Hz in the ¹⁵N dimension.

NMR HSQC spectra were analyzed using the Felix software program (Accelrys). Observable crosspeaks at similar contour depths were selected for each dataset using the automated Peak Picker function of the software. Coordinate measurements for each individual residue crosspeak center was manually validated. Residue crosspeak coordinates were exported to Excel (Microsoft) for further processing. Values for composite chemical shift (δ_c) were computed based on the following formula: (δ_c)² = (δH)² + [0.17(δN)]². Changes in proton and nitrogen chemical shifts (δH and δN, respectively) were based on observed crosspeak coordinate differences from the appropriate unbound J‐domain protein (Jd or JdD35N). A 0.17 correction factor was used to compensate for the observed greater sensitivity of N compared to H for chemical shift perturbation.19 Computed composite chemical shifts were normalized to reflect 100 percent binding to account for variabilities in interaction due to differences in the percentage of the Jd‐protein bound.19 Integrated composite chemical shifts (∑δ_c) were computed as sums of acquired δ_c for identical residues from the compared sets. Graphs relating composite chemical shifts and free energy of binding were created using Prizm (Graphpad). Regression line R² and slope values were computed for the effect of binding affinity on composite chemical shifts for individual residues (δ_c) and the whole protein (Σδ_c)

H¹‐H²Exchange HSQC NMR

H¹‐H² Exchange HSQC NMR experiments were conducted as another measure of J‐domain rigidity. Lyophilized [¹⁵N]Jd or [¹⁵N]JdD35N was reconstituted in 99.96% D₂O (Cambridge Isotope Laboratories) and incubated at 37°C for 16 h to facilitate exchange of native hydrogen (¹H) for deuterium (²H). A 1 mM [²H]Jd protein sample was made with 50 mM MOPS pH 6.8 just prior to data acquisition and recovery of ¹H was monitored over time. HSQC spectra were collected in a Bruker DRX 500, using the Bruker Topspin hsqcetgp pulse program. Each experiment involved 40 sequential aqcuisition sets following the deuterium exhange reaction in 20‐min steps (Total acquisition time: 800 min).

Supporting information

Supporting Information