JD induces conformational changes in Hsp70

To characterize the conformational changes in Hsp70 upon JD binding, we conducted unbiased all-atom molecular dynamics (MD) simulations based on crystal structures of DnaK T199A (PDB: 4JNE) in isolation and in complex with the JD of DnaJ (PDB: 5NRO), in which the T199 residue was restored. To identify the essential motions of Hsp70, we performed principal component analysis (PCA) on structural ensembles derived from the MD simulations. The first principal component, accounting for 28% of the conformational variability of the NBD (S2 Fig), was sufficient to distinguish between the NBD states of Hsp70 alone and in complex with JD (S1 Fig). Notably, the values of the first principal component were highly correlated with the torsion angle between the NBD lobes (R² = 0.95), indicating that JD binding decreases the inter-lobe angle by approximately 7° relative to Hsp70 alone, resulting in a more compact NBD conformation (Fig 2A, 2B).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. (A) Structure of the Hsp70’s NBD (DnaK) coloured by the subdomains (PDB: 5NRO).

The subdomain IA of NBD is depicted in blue, IB– in cyan, IIA– in red and IIB– in orange. Dots indicate subdomain centres of mass, used to define torsion angle describing subdomain dynamics. (B) Probability distribution of the torsion angle between the NBD subdomains during MD simulations of the Hsp70 (red) and Hsp70-JD complex (green). Black dashed line indicates an angle value adopted by the Hsp70 in the ADP state (PDB: 2KHO). (C) Chord diagram showing changes in NBD residues contact frequency associated with Hsp70-JD complex. Line widths are proportional to the contact frequency in Hsp70-JD complex and are coloured based on the contact frequency difference with respect to Hsp70. Only contacts displaying frequency difference of 40 percentage points or greater are shown. Dots positioned next to the NBD residues indicate ATP-binding residues (violet) or residues positioned at the interlobular interface (grey) in the Hsp70-JD complex. Coloured bars along the circle’s rim indicate NBD subdomain.

https://doi.org/10.1371/journal.pcbi.1014094.g002

To further characterize the conformational changes in Hsp70 resulting from JD binding, we compared the frequency of inter-residue interactions - including hydrogen bonds, ionic interactions, π-stacking, π-cation interactions and hydrophobic contacts - in Hsp70 alone and in the Hsp70-JD complex, with a focus on the NBD (Fig 2C). Notably, the most significant differences in contact frequency were observed in: (i) the JD binding site, (ii) the ATP-binding site, and (iii) lobe IIA (see also S3-S4 Figs). Strikingly, in the Hsp70-JD complex we observed an interaction between the hydroxyl group of T199 and the γ-phosphate of ATP that is absent in Hsp70 alone (S5 Fig). Analogous interaction has been observed in many ATPases, where it contributes to stabilizing of the metaphosphate intermediate during ATP hydrolysis [19,20].

Additionally, our simulations of the Hsp70-JD complex revealed further local rearrangements within the nucleotide-binding pocket. Notably, displacement of the Y145 sidechain from the SBDβ interface towards the ATPase catalytic site occurs more frequently in the presence of the JD, suggesting that this rearrangement may affect the transition towards the T199-close state (S6 Fig). Our observations are consistent with recent studies describing the client-binding-induced transition of DnaK T199A between the restraining (R) and stimulating (S) states [9,25]. This transition involves rotation of the NBD lobes and displacement of residue Y145 towards the γ-phosphate of ATP [9] (S20 Fig). Our results also capture the displacement of the G-loop (residues 196–199) observed in the crystal structure of the DnaK T199A S-state: following JD binding, G198 no longer interacts with G229, whereas G196 forms an interaction with T199 (Figs 2C, S4).

In parallel, several residues previously implicated in allosteric communication between the NBD and SBD - including K70, D148, R151, E171 - exhibit substantial changes in their interaction patterns in the Hsp70-JD complex compared with Hsp70 alone (S4 Fig) [26–28]. Specifically, the loss of contact between E171 and K70 in Hsp70-JD complex may allow E171 to coordinate the attacking water molecule during ATP hydrolysis [15]. Finally, our results suggest that the compact NBD conformation induced by JD binding is stabilized by the formation of inter-lobe salt bridge involving residues R71, R84, E230 and switching the R188 interaction partner from E209 to D336.

Taken together, our MD simulations of the ATP bound Hsp70-JD complex with restored T199 reveal the allosteric effects of JD binding, which involves conformational transitions within the NBD. These transitions include rotation of the NBD lobes and reorganization of the NBD contact network, notably the displacement of T199 towards the γ-phosphate of ATP.

JD facilitates the positioning of T199 next to the γ-phosphate of ATP

Our unbiased MD simulations revealed that JD binding favours displacement of T199 towards the γ-phosphate of ATP. To thermodynamically characterize this conformational transition, we determined free-energy profiles for T199 movement towards the γ-phosphate of ATP in both Hsp70 alone and in complex with JD, using enhanced sampling methods (Fig 3). We employed a well-established umbrella sampling approach, performing simulations until convergence was achieved.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. (A) Schematic representation of the reaction coordinate used in the umbrella sampling simulations.

The atoms shown in licoricey representation are coloured according to the elements – carbon in dark grey, oxygen in red, hydrogen in light grey, phosphorus in orange, nitrogen in blue. (B) Probability distributions (top) and free energy profiles (bottom) for the distance between hydroxyl oxygen of T199 and the phosphorus of ATP gamma phosphate in the absence (red) and presence (green) of the JD. The purple shade indicates the T199-close state (distance < 0.4 nm), whereas orange shade indicates the T199-far state (distance > 0.65 nm).

https://doi.org/10.1371/journal.pcbi.1014094.g003

Comparison of the resulting energy profiles indicates that the positioning of T199 strongly depends on the presence of JD. For Hsp70 alone, the free energy profile shows a well-defined minimum corresponding to a state in which the hydroxyl group of T199 is displaced from the γ-phosphate of ATP by ∼0.8 nm (T199-far state). Formation of the T199 - γ-phosphate contact in this state incurs an energy cost of ~6 kcal/mol, indicating that, in the absence of the JD, the T199 sidechain is unlikely to participate in stabilization of the ATP-to-ADP transition state.

In contrast, in the Hsp70-JD complex, the free energy profile exhibits a pronounced global minimum at ∼0.35 nm (T199-close state). In this state, formation of the T199 - γ-phosphate contact is energetically favourable, associated with a free-energy gain of ~3 kcal/mol. These results indicate that the T199-close state is predominantly adopted in the presence of JD, as reflected in the probability distribution along the T199 - γ-phosphate distance (Fig 3B). By contrast, in the absence of JD, the T199-close state is rarely accessible, being 6 kcal/mol less favourable than the T199-far state.

To further assess the role of Y145 in the transition to the T199-close state, we performed additional umbrella sampling simulations for the Hsp70 Y145A variant, both in isolation and in complex with JD (S21-S22 Figs). In the presence of JD, the free energy profiles showed only minor changes in the depth or position of the free-energy minima compared to the wild-type Hsp70-JD complex, indicating a limited role of Y145 in the transition to the T199-close state. Interestingly, the Y145A substitution reduced the free energy barrier for this transition (at ~0.6 nm of the T199- γ-phosphate of ATP distance) by ~2–3 kcal/mol. In contrast, the free-energy profile for isolated Hsp70 Y145A was flattened for the T199-far and intermediate distances. These findings are consistent with previous biochemical data showing that DnaK Y145A¹⁰, exhibits higher ATPase activity than the wild-type, both in the presence and absence of DnaJ, highlighting the functional relevance of the T199-close state. Collectively, our results support a model in which the JD binding induces rearrangements in the nucleotide binding pocket, promoting T199 interaction with the ATP γ-phosphate and thereby explaining ATPase stimulation.

Allosteric signalling between JD and nucleotide-binding pocket of Hsp70

To elucidate how JD interaction at the surface of Hsp70 induces conformational changes within the nucleotide-binding pocket on the opposite side of the NBD [15,18], we first investigated the contribution of HPD motif residues to allosteric signalling. We focused on H33 and D35, which are known to form direct contacts with L391 and R167 of the NBD, respectively [15]. MD simulations of the Hsp70-JD H33A complex revealed an equilibrium conformation that closely resembles the wild-type complex. Consistently, the free energy profile for JD H33A binding to Hsp70 was nearly indistinguishable from that of the wild-type JD (S14 Fig;left), indicating that the H33A substitution does not impair JD-Hsp70 interaction. However the assessment of H33A effect on the transition to the T199-close state showed that the T199-far state is 3 kcal/mol more stable than the T199-close state (S14 Fig;right). These results demonstrate that although the H33A substitution preserves JD binding, it shifts conformational equilibrium towards the T199-far state, thereby preventing Hsp70 from adopting the stimulating (S) state.

Next, we examined the effects of the D35A substitution. MD simulations of the Hsp70-JD D35A complex showed that only helix II of JD remains in direct contact with the NBD, whereas helix III and the interhelical loop are displaced further from the Hsp70 binding site compared to the wild-type complex (S12, S13 Figs). Consistent with this observation, the free energy profile along the JD-Hsp70 distance (S14 Fig) indicates that D35A binding is ∼2 kcal/mol less favourable and leads to a relaxation of the complex, as evidenced by ∼0.2 nm shift in the free energy minimum. We further investigated the impact of the D35A substitution on the stability of the SBDβ-NBD interface. In previous studies, JD binding was shown to partially displace SBDβ from NBD [18]. However, umbrella sampling simulations of the Hsp70-JD D35A complex revealed that this displacement is abolished (S18 Fig). These findings underscore the critical role of the salt bridge between D35 of HPD motif and R167 of the NBD [18] in driving the conformational transition required for ATP hydrolysis. Thus, although the D35A and H33A substitutions differently affect JD binding, both prevent Hsp70 from undergoing the conformational transition to the stimulating (S) state, albeit through distinct mechanisms.

Because H33 and D35 within the HPD motif directly interact with L391 of the interdomain linker and R167 of the NBD, respectively, we hypothesised that local α-helix disruption by P34 enables H33 and D35 to adopt orientations that favour these interactions [29]. To test this hypothesis, we performed MD simulations of the isolated JD and its P34A variant. As expected, the P34A substitution extends the α-helical structure of helix II into the HPD motif. Therefore, a dihedral angle describing the relative orientation of H33 and D35 displays a broad distribution in the P34A variant, whereas in the wild-type JD this angle is narrowly distributed around values close to 0°, indicating that both residues face the same direction (S11 Fig). Together, these results demonstrate that P34 plays a key role in constraining orientation of H33 and D35, thereby enabling their productive interactions with L391 and R167.

Finally, to assess the combined effects of the HPD substitutions, we performed umbrella sampling simulations of Hsp70 in complex with JD HPD/AAA variant. The resulting free energy profile along the Hsp70-JD distance lacks an energetically favourable minimum (S15 Fig), indicating that the triple HPD/AAA substitution abolishes formation of a stable Hsp70-JD complex. Consistent with the effects observed for individual HPD substitutions, this triple JD mutant prevents formation of the stable Hsp70-JD complex required for Hsp70 to transition into the stimulating (S) state.

To elucidate the allosteric signal transmission routes connecting the HPD motif to T199, we performed additional analyses using graph theory, an approach that has proven effective in revealing allosteric communication in biomolecules [30,31]. We constructed a dynamic network model, in which the protein is represented as a network of Cα atoms (residues) interconnected by edges whose lengths are determined by the strength of correlated residue motions (connections). Analysis of our multi-µs MD trajectories revealed that all identified allosteric pathways follow a similar pattern: the signal originating at D35 and is relayed via H33, within the HPD motif, propagates through the interdomain linker and NBD β-strands 13 and 14 (residues F200–T215), and ultimately converges on T199 (Figs 4, S7). Notably, 7 of the 14 residues involved in these pathways have previously been implicated in allosteric communication by mutational and structural studies [22,27,32–34].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Allosteric signalling between JD and Hsp70.

(A) Set of 5 optimal pathways between the D35 of the JD HPD motif and the T199 of Hsp70 (magenta spheres) calculated for 5 Hsp70-JD MD trajectories and projected onto the Hsp70-JD complex. C_α atoms of residues present in the set of optimal and suboptimal pathways shown as spheres and coloured according to the path occurrence frequency (for more details see S7 Fig). (B) Graph depicting connectivity of pathway forming residues. Arrows represent connections present in the set of 5 optimal pathways, residue colouring as in (A).

https://doi.org/10.1371/journal.pcbi.1014094.g004

Interestingly, none of the identified pathways directly involve the well-established interaction between D35 of the HPD motif and R167 of the NBD, which has been shown to be critical for stimulation of ATP hydrolysis [15]. Given that the otherwise non-functional DnaK R167A mutant is suppressed by the compensatory DnaJ D35N substitution both in vivo and in vitro [23,35], we next examined how these mutations affect the allosteric pathways described above. In the mutant complex, all identified pathways follow trajectories similar to those observed in the wild-type system (S24 Fig). Consistently, free-energy calculations comparing the T199-far and T199-close states for the Hsp70 R167A-JD D35N variant reveal a flattened free energy landscape, rendering the T199-close state more accessible than in Hsp70 alone (S23 Fig). Together, these results indicate that suppression of the R167A mutation by the D35N substitution does not reroute the allosteric communication network.

To further evaluate the proximity of the JD to T199 of the NBD in correlation space, we employed community network analysis [30,36], which simplifies the dynamic network by clustering strongly connected nodes into communities. The resulting community structure broadly reflects the division of the NBD into its constituent subdomains (S8–S9 Figs). T199 was assigned to the zeroth community, which is directly connected to the first community containing the JD. This organization provides a structural rationale for the observed routing of allosteric pathways between D35 of the JD and T199, as minimizing community crossings likely reduces path length in correlation space.

In addition, three leucine residues in the conserved VLLL motif of the interdomain linker were assigned to the same community as the JD, highlighting the tight coupling between these regions. We next quantified the strength of inter-community connections using the betweenness centrality (BC) metric (see Methods) and identified key interactions involved in allosteric signal transmission based on connections with the highest BC values. Both IA and IIA subdomains of the NBD establish strong connections with the JD. Specifically, JD residues D35, K26, Y25, H33 and F47 interact with NBD residues R167, E217, L391 and L392. Notably, D35 and H33 of the HPD motif form direct contacts with R167 of the NBD and L391 of the interdomain linker, respectively (S10 Fig). Interactions of H33 with L391 or L392 facilitate transmission of the allosteric signal towards the catalytic site (Figs 4, S10) [15].

JD-induced allosteric signalling is associated with destabilization of Hsp70’s β-strand 14 induced by steric repulsion

Analysis of our unbiased MD simulations revealed multiple events in which the T199-close state coincided with disruption of the interaction between β-strand 13 (containing T199) and β-strand 14 (residues 198–206 and 219–227; S19A Fig), accompanied by destabilization of the β-strand 14 itself. Probability distributions indicate that these disruption events are more frequent in the Hsp70-JD complex than in Hsp70 alone (S19B Fig). Notably, conformational changes in the β13-β14 region have previously been identified as structural signatures distinguishing the restraining (R) and stimulating (S) states of Hsp70 [9].

To examine whether these conformational changes are sufficient to drive the transition towards the T199-close state, we performed free energy simulations along the T199–ATP distance while enforcing two distinct conformations of the β13-β14 region. First, we enforced a disrupted conformation in Hsp70 alone (see Methods for more details). The resulting free-energy profile (S17 Fig; left) closely resembles that of native Hsp70, indicating that disruption of the β-strands is not sufficient to promote the T199-close state. Second, we enforced an undisrupted conformation in the Hsp70-JD complex. The corresponding profile shows a clear preference for the T199-far state (S17 Fig; right), demonstrating that the maintenance of an intact β-strands conformation prevents the transition to the T199-close state even in the presence of the JD. Taken together, these results indicate that disruption of the of the β13-β14 region is necessary, but not sufficient, for the transition to the T199-close state.

Knowing that disruption of the β13-β14 region alone is insufficient to reach the T199-close state and that tight JD binding is required for allosteric signal transmission, we hypothesised that in the Hsp70-JD complex, steric repulsion from the JD exerts a mechanical push that drives both displacement of the SBDβ from the NBD and disruption of the β-strands. To test this, we constructed an Hsp70 complex with a JD variant lacking helix III and IV (residues 41–65), reasoning that the JD ∆H3-4 variant would bind Hsp70 but be unable to exert the mechanical push. Indeed, free energy profiles along the SBDβ-NBD and T199-ATP distances for the Hsp70-JD ∆H3-4 variant are similar to those of Hsp70 alone, indicating that the truncated JD prevents SBDβ displacement from the NBD (Fig 5A) and maintains the T199-far state (Fig 5B). These findings support our hypothesis that steric repulsion exerted by the intact JD is necessary for allosteric signal transmission.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Role of the helix III of JD.

(A) Free energy profiles for the dissociation of the NBD and SBDβ. The profile for the Hsp70 without JD is depicted in red, for the complex of Hsp70 and wild-type JD– in green, whereas for the complex of Hsp70 with the JD lacking helix III and IV– in blue. The definition of the distance between centres of mass of the SBDβ and the part of NBD interfacing SBDβ is shown above. (B) Differences of the free energy associated with the transition from the T199-far to T199-close state for the same systems as in 4A (for the free energy profiles see S16 Fig).

https://doi.org/10.1371/journal.pcbi.1014094.g005

Simulation protocol

All simulations were performed using Gromacs 5.0.4 [39] with the Plumed 2.1 plugin [40]. CHARMM36m force field [41] was used for proteins, ions and Mg-ATP, and the TIP3P model [42] was used for water. In each of the simulation boxes, the numbers of Na⁺ and Cl⁻ ions were adjusted to 0.15 M. Temperature was kept at 310 K with the v-rescale algorithm [43] using a coupling constant of 0.1 ps. Pressure was kept at 1 bar using the Parrinello-Rahman algorithm [44] with a coupling time of 5 ps. Periodic boundary conditions were applied and the Particle Mesh Ewald summation [45] was used to calculate long-range electrostatic interactions with a cut-off radius of 1 nm and a Fourier grid spacing of 0.12 nm. Van der Waals interactions were calculated with Lennard-Jones potential with a cut-off radius of 1 nm. All bonds involving hydrogen were constrained using the LINCS algorithm [46]. Leap-frog Verlet algorithm [47] was used to integrate equations of motion with a time step of 2 fs.

The energy of the systems was minimized in two stages using steepest descent algorithm: at first, all heavy atoms of the protein were kept fixed; in the second step only the position of the protein backbone was restrained, allowing protein side chains to relax [48,49]. After that, the systems were relaxed with short MD simulation (10 ns) with constrained protein backbone.

Conventional MD simulations

The Hsp70 system was prepared using 4JNE PDB structure [37] placed in a ∼13.5 nm x 13.5 nm x 13.5 nm box solvated with ∼55000 water molecules. The system was simulated in two copies and for each copy the 2.5 µs-long trajectory was obtained. The Hsp70-JD system was prepared using 5NRO PDB structure [15] placed in a ∼13 nm x 13 nm x 13 nm box solvated with ∼52000 water molecules. The system was simulated in five copies and for each copy the 1.8 µs-long trajectory was obtained. In each system, the crystal water molecules coordinated to Mg²⁺ ion in Mg-ATP were kept. As the terminal region of SBDα is dispensable for the transmission of the allosteric signal [9,50] to reduce the size of the system and thus, to allow for the better convergence of the free energy profiles, the C-terminal region of the lid (residues 534–609) was removed. The final structures from these simulations were used as the initial configurations for the subsequent free energy calculations.

The wild-type and P34A JD systems (residues 3–65) were prepared using JD from the 5NRO PDB structure [15]. Each system was placed in the ∼6 nm x 6 nm x 6 nm box, solvated with ∼7000 water molecules and simulated for 1 µs.

Trajectories analysis

Principal Component Analysis was performed for the backbone of the Hsp70’s NBD (residues 2–384) with gmx covar and gmx anaeig tools [39]. The eigenvectors for JD-induced NBD conformational changes were assigned based on the merged NBD trajectories from Hsp70 and Hsp70-JD systems. The merged trajectory was then projected onto the first eigenvector.

Torsion angle of NBD subdomains was defined as the angle between centre of mass of the backbone atoms of lobes IB (residues 40–115), IA (residues 1–39, 116–188, 361–384), IIA (residues 189–228, 307–360) and IIB (residues 229–306) [51].

Contact analysis was performed using GetContacts tool [52] for the NBD and interdomain linker of Hsp70 (residues 2–394) using merged trajectories of Hsp70 and Hsp70-JD complex. The interactions were classified as differentiating if the difference in contact frequency between Hsp70 and Hsp70-JD systems was greater than 40 percentage points. For defining inter-residue contacts following types of interactions were considered: hydrogen bonds, ionic interactions, π-stacking, π-cation interactions and hydrophobic interactions (defined as Van der Waals interactions between purely hydrophobic residues). Default GetContacts criteria were used for all interaction types with exception of the criteria for the hydrogen bond detection were the α_DH···A angle value greater than or equal to 150 ° was required. Calculated contact frequencies were then visualized as chord diagrams using the Flareplot tool [52]. NBD residues were classified as interacting with ATP if they remained in contact for at least 50% of the trajectory. Residues were considered to be at the NBD interlobular interface if they remained in contact with residues from the opposite NBD lobe for at least 50% of the trajectory.

Dynamic network analysis was performed with dynetan 1.0.1 python package [31]. Merged trajectories of Hsp70-JD system were used to infer common network topology and perform community and betweenness centrality analyses. For tracing allosteric pathways between JD D35 and DnaK T199 weights of connections in the network were calculated using individual trajectories as separate analysis windows, while keeping previously obtained network topology. Nodes of the network were defined as the Cα atoms positions of the NBD and interdomain linker (residues 2–394) or JD (residues 3–65). Additional nodes were assigned for the magnesium ion associated with ATP molecule in the catalytic site of Hsp70, as well as for the ATP itself (two nodes, matching the positions of N1 atom of purine ring and terminal phosphorus atom). Nodes (residues) were interconnected by edge (connection) if any heavy atoms from two nodes remained within 0.45 nm from each other for at least 75% of the simulation length. In the case of ATP, the heavy atoms of the purine ring were assigned to the N1 atom node, whereas heavy atoms of phosphate groups and ribose were assigned to the γ-phosphate node. For each connection, the weight of the connection was calculated as the generalized correlation (GC) coefficient, which captures both linear and nonlinear correlations between residues.

Clustering of the densely interconnected nodes into communities was performed using the Louvain method [53], taking into the account the weight of the connection. Communities that made up for less than 1% of all the nodes were discarded in further analysis. The strength of the connection between the communities was quantified using the betweenness centrality measure, defined as the fraction of the shortest pathways, between all pairs of nodes, passing through a selected node, showing which nodes play a bridging role in the network.

For each of the simulated replicas, the allosteric pathways between D35 from JD conserved HPD motif (“source”) and T199 of NBD (“sink”) were calculated using Floyd-Warshall algorithm [54]. For given connection between atoms i and j, the distance parameter d_ij, defined as d_ij= − log(GC_ij), was calculated and the optimal path was assigned in such a way that the sum of distances along the pathway from the “source” to “sink” was minimized. Additionally, twenty sub-optimal pathways were determined. Based on the calculated pathway set, the residue pathway occurrence frequency was calculated. To confirm that the general conclusions from the pathway’s analysis are not dependent on the values of the parameters used to construct the network, we performed additional sensitivity analysis (S42-S44 Figs). The connections in new networks were assigned by taking into the account different thresholds, both for the distance cut-off between heavy atoms (4.2-4.8 Å), as well as for the contact persistence (65–85%). Then, pathway similarity was assessed using Jaccard index (overlap between sets of pathways, measured as residue overlap and connection overlap).

The antiparallel beta sheet content (s) of residues 198–206 and 219–227 of Hsp70 was calculated using ANTIBETARMSD method [55] from Plumed plugin with default parameter values and strands cutoff equal to 1. In the wild-type and P34A JD systems the torsion angle between H33 and D35 residues was defined as the angle between H33 N∊, H33 Cβ, D35 Cα and D35 Cγ. The probability distributions for distance between Helix III of JD (residues 41–58) and SBDβ (residues 395–505) and the distance between HPD motif and R167 were calculated using gmx distance and gmx mindist tools [39], respectively.

All molecular images were created using VMD and Blender [56,57].

Free energy calculations

All of the free energy profiles were computed with the replica-exchange umbrella sampling (REUS) method [58] and determined using the weighted histogram analysis method [59]. Uncertainties of the free energy profiles were estimated using bootstrap error analysis taking into account the autocorrelation in the analysed time series. (for the convergence of the calculated profiles see S25-S41 Figs).

Conformational change in the catalytic site of NBD

The free energy profiles associated with the T199 conformational change in the catalytic site of the NBD were calculated along the reaction coordinate defined as the distance between the T199 hydroxyl oxygen and the terminal phosphorus atom of ATP. The initial configurations for REUS simulations were obtained with 80 ns-long steered-MD simulation, during which the distance between T199 and ATP was gradually increased using a moving one-sided potential with a force constant of 2000 kJ· mol⁻¹· nm⁻². Ten equally spaced US windows were used, spanning the range of 0.3 to 1.2 nm of the reaction coordinate. The spring constant of the harmonic biasing potential was set to 1500 kJ· mol⁻¹· nm⁻² and each of the REUS windows was simulated for 1 µs. The same procedure was applied for the substitution variants.

During the investigation of the beta-sheet disruption influence on the T199-far to T199-close transition additional simulations were performed. In the Hsp70-JD system, the beta-sheet was restrained by setting the parameter s to 2.73, whereas in the Hsp70 system the beta-sheet was disrupted by keeping the s parameter at the value of 1.8. In both cases the one-sided harmonic potential was used, with the spring constant of 4000 kJ· mol⁻¹· nm⁻².

In the case of the JD ∆H3-4 system, during the free energy calculations the conformation of the remaining JD (residues 3–40) was restrained with the additional harmonic biasing potential (with the spring constant set to 5000 kJ· mol⁻¹· nm⁻²), which kept the 0.0 nm RMSD value with respect to the wild-type helix II conformation.

JD positioning on Hsp70

The free energy profiles associated with the JD binding on Hsp70 were calculated along the reaction coordinate defined as the distance between the centre of mass of backbone atoms of the JD (residues 25–31 and 48–54) and the centre of mass of Cα atoms of Hsp70 binding site (residues 208–216 and 415–422). The initial configurations for REUS simulations were obtained with 100 ns-long steered-MD simulation, during which the distance between Hsp70 and JD was gradually increased using a moving one-sided harmonic potential with a force constant of 1500 kJ· mol⁻¹· nm⁻². Six equally spaced US windows were used, spanning the range of 0.8 to 1.3 nm of the reaction coordinate. The spring constant of the harmonic biasing potential was set to 3500 kJ· mol⁻¹· nm⁻² and each of the REUS windows was simulated for 800 ns.

SBD and NBD dissociation

In the studies of SBD and NBD dissociation the reaction coordinate was defined as the distance between the centre of mass of Cα atoms of the SBDβ domain (residues 395–505) and the centre of mass of Cα atoms of the NBD fragments, which are in direct contact with SBDβ (residues 146–226). To make the system significantly smaller and thus, to achieve better convergence of the free energy profiles, the lid sub-domain, SBDα, was truncated by removing residues 506–602 of Hsp70.

The initial configurations for REUS simulations were obtained with 500 ns-long steered MD simulations, during which the distance between NBD and SBDβ was gradually increased using a moving one-sided potential with a spring constant of 2500 kJ· mol⁻¹· nm⁻², from the initial value of 2.58 nm up to 3.58 nm. Five equally spaced US windows were used, spanning the range of 2.5 to 2.9 nm of the reaction coordinate. The spring constant of the harmonic biasing potential was set to 3500 kJ· mol⁻¹· nm⁻² and each of the REUS windows was simulated for 1 µs.