Direct observation of chemo-mechanical coupling in DnaK by single-molecule force experiments

Abstract

Protein allostery requires a communication channel for functional regulation between distal sites within a protein. In the molecular chaperone Hsp70, a two-domain enzyme, the ATP/ADP status of an N-terminal nucleotide-binding domain regulates the substrate affinity of a C-terminal substrate-binding domain. Recently available three-dimensional structures of Hsp70 in ATP/ADP states have provided deep insights into molecular pathways of allosteric signals. However, direct mechanical probing of long-range allosteric coupling between the ATP hydrolysis step and domain states is missing. Using laser optical tweezers, we examined the mechanical properties of a truncated two-domain DnaK(1–552ye) in apo/ADP/ATP- and peptide-bound states. We find that in the apo and ADP states, DnaK domains are mechanically stable and rigid. However, in the ATP state, substrate-binding domain (SBD)^∗ye is mechanically destabilized as the result of interdomain docking followed by the unfolding of the α-helical lid. By observing the folding state of the SBD, we could observe the continuous ATP/ADP cycling of the enzyme in real time with a single molecule. The SBD lid closure is strictly coupled to the chemical steps of the ATP hydrolysis cycle even in the presence of peptide substrate.

Significance

Structural changes in proteins are essential for the regulation of their function. Hsp70 is a two-domain protein that binds ATP nucleotide in its nucleotide-binding domain. Protein substrates are bound in the second so-called substrate-binding domain (SBD). Hsp70 hydrolyses ATP to ADP; by doing this, Hsp70 undergoes significant changes in the SBD. Using a single Hsp70 molecule, we showed how the mechanics of Hsp70 is changed during ATP hydrolysis by using laser optical tweezers. We find that DnaK is mechanically stable and rigid in an ADP-bound state. However, in the ATP state, one domain is mechanically destabilized as the result of interdomain docking, followed by the partial unfolding of the SBD. By observing the folding state of the SBD, we observe the enzyme’s continuous ATP/ADP cycling in real time with a single molecule. The SBD lid closure is strictly coupled to the chemical steps of the ATP hydrolysis cycle even in the presence of peptide substrate.

Introduction

Large proteins are made of multiple domains that are organized like beads on a string. In some cases, domains can interact with each other in a ligand-dependent manner, which results in the regulation of the function of the distal domain. The result of such long-range domain-domain interaction can be, for example, a change in the affinity of a ligand, binding site accessibility, and activation/suppression of a protein activity (1,2,3,4,5). These examples are well described by the concept of allostery (reviewed and see references herein (6,7,8)). The existence of allosteric control of functional properties of proteins is long well known from a large number of biochemical and biophysical studies. For some intensively studied proteins, molecular mechanisms and pathways were elucidated using mutagenesis (9,10). However, revealing fundamental mechanical principles and energetics of long-range allosteric ligand-induced coupling between subdomain interfaces is challenging.

The scale of conformational changes in proteins varies significantly; conformational changes can be local at residue level (e.g., side-chain rotations (11)), native-state rearrangements (e.g., monomer/dimer equilibria (12) and opening/closing of binding sites (13,14)), and extensive conformational transitions (e.g., unfolding of polypeptide chain (15)). Under favorable circumstances, single-molecule force spectroscopy can provide an efficient tool to observe mechanical and structural changes in proteins at the subnanometer scale (16,17). The development of highly sensitive differential detection dual-beam laser tweezers enabled observing folding/unfolding transitions of many proteins and observing subnanometer-sized opening and closing of the active site of adenylate kinase. The primary benefit of the force technique is the possibility of controlling the state of the protein structure and determining its propensity to undergo structural rearrangements (18). The immediate outcomes of single-molecule assays using optical tweezers measurements are two parameters describing the observed conformational changes: force and distance. Under equilibrium conditions, force and distance fluctuations completely specify the underlying energy landscape of the observed conformational changes, their minima, barriers, and transition path times (19,20). Here, we employ laser optical tweezers to observe and monitor ATP-induced allosteric changes in the domain conformations of a well-studied, two-domain protein, DnaK.

DnaK is an Hsp70 chaperone that binds unfolded polypeptide chains as substrates and prevents unproductive misfolding and aggregation (1). Substrate binding activity is located in the so-called substrate-binding domain (SBD). Its activity is regulated by the ATP/ADP status of the N-terminal nucleotide-binding domain (NBD; Fig. 1 A). In the ADP state, the substrate binds at high affinity but with slow kinetics. In the ATP state, substrate binding is orders of magnitude faster but at low affinity. Three-dimensional structures of DnaK in different states provide clues for ATP-regulated substrate binding (21,22). In the ATP state (9,10,23), a set of structural rearrangements occurs: the interdomain linker binds to the NBD, the SBD docks to NBD while the helical lid opens, and β8 dislocates, which restructures the β-core and exposes the altered substrate-binding site. Our initial goal was to monitor interdomain docking and conformational motion of the lid after ATP binding. We find that observing ATP-induced allosteric action variant optimization was essential, which made us arrive at a minimal allosteric two-domain DnaK(1–552ye) variant (10,24). Because in this variant the ATP-induced conformational change was coupled to the complete unfolding of the helical lid, we were able to use force spectroscopy as a probe of the SBD mechanics in response to the status of a single ATP molecule in the NBD and detect conformational switching of the SBD after the hydrolysis. Under conditions used in the ATP- and peptide-bound substrate, SBD^∗ye remains open and highly flexible with the lid closure strictly coupled to ATP hydrolysis. We could observe equilibrium fluctuations of the force-induced interdomain docking/undocking process in the presence of the peptide substrate. By the transient trapping of the DnaK^∗ye ADP state, we could follow the structural events after the exchange of ADP to ATP during which the SBD^∗ye transits from the closed form to the open, domain docked/undocked form. To summarize, force spectroscopy provides insights into individual microscopic events during allosterically induced structural changes in a protein.

Figure 1.

Single-molecule force spectroscopy of a two-domain DnaK. (A) A three-dimensional structure of Hsp70 in the ATP state (PDB: 4JNF (22) and 4B9Q (21)) and ADP states (PDB: 2KHO (43)). (B) Color codes are according to subdomains: the nucleotide-binding domain lobe I (yellow) and lobe II (blue), the substrate-binding domain β-subdomain (green), and the α-helical subdomain (red). The largely disordered C-terminal part is shown in gray. Constructs used in the study are indicated as well. (C) The design of the optical tweezers assays with the two-domain DnaK (1–607). (D) Force-extension curves of the apo form of two-domain DnaK^∗ at pulling velocity 20 nm/s. Pulling traces show the hierarchical unfolding of DnaK domains, unfolding of the substrate-binding domain (SBD; two shorter unfolding events at ∼10 pN) and the nucleotide-binding domain (NBD; large unfolding event at ∼30 pN). The order of unfolding events for the SBD subdomains is different: (left) the unfolding of the β-subdomain followed by the α-helical subdomain, and (right) the unfolding of the α-helical subdomain followed by the β-subdomain. The blue circle shows a DnaK autoinhibited state. (E) Mechanical unfolding of the SBD in the absence (left) and the presence (right) of 250 μM NR peptide. Pulling velocity was 20 nm/s. (F) Model of the observed processes, during which the α-helical subdomain unfolds and the part of this unfolded domain binds to the folded β-subdomain (autoinhibited state). (G) DnaK variant, which is devoid of autoinhibition behavior, DnaK^∗ye. A force-extension curve of single DnaK^∗ye (1–552, L542Y/L543/E) (23) shows two unfolding events corresponding to the unfolding of the individual NBD (left; larger event at ∼30 pN) and SBDye (left; shorter event at ∼40 pN). Complete refolding and reappearance of a fully native two-domain DnaK^∗ye within the lifespan of an observation of a single molecule is a slow process. To see this figure in color, go online.

Materials and methods

Experimental procedures

Peptides (NRLLLTG and ¹³²QRKLFFNLRKTKQC derived from E. coli σ³² protein (25)) were purchased from GenScript. The DnaK wild-type and DnaK^∗ye gene were amplified by PCR from E. coli XL1 blue using primers containing additional codons for cysteine residues. Then, the PCR fragment was cloned into pET11a (Novagen, Madison, WI, USA) using the NdeI and BamHI restriction site. Protein truncations and L542Y/L543E, D393R, and T199A substitutions were introduced by QuickChange mutagenesis (Agilent, Santa Clara, CA, USA). Protein sequences and purification procedures are in the supporting material. The experiments were performed in 100 mM Tris-HCl, 150 mM KCl, 0.65% glucose, 0.15 M NaCl, and 2% glycerol (pH 7.8) in the absence or presence of 5 mM MgCl₂ and 1 mM ATP or ADP. The experimental setup used for optical trapping is a custom-built high-resolution dual-trap optical tweezers with back-focal plane detection as described previously (26,27).

Results

The design of a minimal allosteric two-domain DnaK for mechanical interrogation

The mechanical properties of the individual domains of DnaK have been characterized in detail using single-molecule optical tweezers experiments (28,29,30,31). Wild-type DnaK from E. coli contains the intrinsically disordered C-terminal part (607–638), which was initially present also in our construct (Fig. 1 B). However, we found that the DnaK wild-type with the C-terminal was not very well amenable to single-molecule experiments. Therefore, we employed the shorter two-domain DnaK^∗ construct (residues 1–607) for optical trapping experiments (Fig. 1 C). DnaK^∗ presents a suitable model for Hsp70 proteins and their interdomain allostery. After introducing cysteine residues at the N and C termini of DnaK^∗, modification by DNA-maleimide conjugates, and hybridization with long DNA handles attached to functionalized beads (see for details (32)), single-molecule force spectroscopy experiments were performed. Force-extension constant velocity experiments at 20 nm/s of the apo form of Hsp70 showed three unfolding events (Fig. 1 D): two shorter unfolding events at ∼10 pN with contour-length increases (L_c) of ca. 22 and 41 nm and a third, larger, unfolding event observed at ∼30 pN with L_c of ca. 132 nm. All observed L_cs are in excellent agreement with the previously reported fingerprints for the SBD subdomains (29) and the NBD (28,30,31). NBD unfolding consists of a single event at ∼30 pN with a contour length increase of ca. 134 nm (28).

The SBD fingerprint consists of two unfolding events, reflecting the unfolding of the α-subdomain (22 nm) and the β-subdomain (40 nm) (29). Interestingly, in our first set of experiments at a slow pulling speed (20 nm/s), we observed that in the case where the α-subdomain unfolds first, this unfolding event displayed overshoot-like behavior (Fig. 1 D, blue circle), which was consistently observed for the isolated SBD domain as well (Fig. 1 E, left). We interpreted this unusual response as a first full unfolding of the α-helical lid of DnaK (Fig. 1 E) followed by ca. 6-nm-long back binding of the unfolded lid to the folded β-subdomain of the SBD (Figs. 1 E and 2.). Consistently, in cases where the β-subdomain unfolded first, such behavior was not observed (Fig. 1 D, left). To further test this hypothesis, we performed single-molecule experiments in the presence of a large excess of the NR peptide, a known substrate peptide for DnaK (33) (250 μM; Fig. 1 E, right). In this case, such overshoot-like behavior was not observed. We conclude that the unfolded α-helical lid serves as a substrate for the β-subdomain of the SBD, thus autoinhibiting its substrate binding capability. To avoid interference with our pulling experiments, we decided to use a DnaK variant, which did not display such autoinhibition. Such a variant where two hydrophobic amino acids in the lid were replaced (L542Y/L543E) and the α-helical lid domain was further truncated (1–552) (Fig. 1 B) was previously described by Swain et al. (24). This so-called DnaK^∗ye variant is an allosterically functional chaperone with proven in vitro and in vivo activities displaying that it could be substituted for wild-type DnaK (24). Force-extension traces from constant velocity experiments of DnaK^∗ye (Fig. 1 G, blue circle) confirmed that the protein does not display the autoinhibition anomaly observed for DnaK^∗ (1–607). Therefore, all our subsequent experiments were conducted using DnaK^∗ye (for amino acid sequence, see supporting material).

Figure 2.

Mechanical properties of DnaK^∗ye domains in the presence of ATP/ADP ligands. (A) Examples of force-extension pulling traces from constant velocity experiments of the DnaK^∗ye apo form performed at 20 nm/s pulling speed. (B) Force-extension pulling traces of the DnaK^∗ye were measured in the presence of 1 mM ADP and 5 mM MgCl₂ (blue; further refer as “ADP”). (C) Force-extension pulling traces of the DnaK^∗ye in the presence of 1 mM ATP and 5 mM MgCl₂ (red; further refer as “ATP”). Black dashed lines (A–C) correspond to worm-like chain fitting curves. (D and E) Shown are unfolding force histograms for the NBD (D; n_apo = 109, n_ADP = 43, n_ATP = 49) and the SBD (E; n_apo = 30, n_ADP = 72, n_ATP = 99), where the total number of unfolding events measured under specific condition is indicated by the value of n_i. (F and G) Shown are contour-length increase histograms for the NBD (F; n_apo = 109, n_ADP = 43, n_ATP = 49) and the SBD (G; n_apo = 30, n_ADP = 72, n_ATP = 99). For all measurements, the buffer was 100 mM Tris-HCl (pH 7.8; at 25°C) plus an oxygen scavenging system (17). To see this figure in color, go online.

Force spectroscopy of DnaK^∗ye in the apo- and ADP/ATP-bound state

Single-molecule force spectroscopy experiments with DnaK^∗ye were performed under various nucleotide conditions. In the absence of nucleotides, DnaK^∗ye apo (Fig. 2 A) force-extension curves show the well-known NBD unfolding (28) as well as a single cooperative unfolding peak of the SBD^∗ye. In the presence of ADP (Fig. 2 B), force-extension traces of DnaK^∗ye are very similar to the apo form. The presence of ADP was confirmed by the observation of characteristic pattern of intermediates the NBD unfolding (Fig. S1), as observed previously (28). Note that under these conditions, the sequence of unfolding events NBD/SBD^∗ye can vary (for an example, see Fig. 1 G). In the presence of ATP, force-extension traces of DnaK^∗ye showed two different unfolding patterns (Fig. 2 C). While the unfolding forces and contour-length increases for the NBD are similar to the apo/ADP form, we observed significant changes in the mechanical properties of the SBD^∗ye (Fig. 2 C, red traces). The vast majority of pulling traces showed SBD^∗ye unfolding with significantly decreased unfolding forces and shorter contour-length increases (Fig. 2 C, dark red traces). Occasionally (ca. 10%), however, we observed a DnaK^∗ye pattern (Fig. 2 C, light red trace) very similar to the apo/ADP form with higher unfolding forces and contour length increase.

Histograms of unfolding forces and contour-length increases under the various conditions (apo, ADP, and ATP) are shown in Fig. 2 D–G. The apo form of NBD in DnaK^∗ye (Fig. 2 D) unfolds at a force of 32.6 ± 0.4 pN (average value mean ± SE) (n_apo = 109) and the ADP form at 34.2 ± 0.8 pN (n_ADP = 43). In the presence of ATP, the average unfolding force was 33.9 ± 1.1 pN (n_ATP = 49). The average contour-length increase for the NBD (Fig. 2 F) is 134.1 ± 0.1 nm for the apo form, 134.9 ± 0.1 nm for the ADP form, and 134.5 ± 0.1 nm in the presence of ATP. Fig. 2 E summarizes the data for the SBD^∗ye unfolding in DnaK^∗ye. The average unfolding force is 26.4 ± 1.3 pN for the apo form (n_apo = 72), 22.6 ± 2.1 pN in the ADP form (n_ADP = 30), and 16.8 ± 0.6 pN in the presence of ATP (n_ATP = 99). The average contour-length increase is 52.9 ± 0.1 nm for the apo form and 52.1 ± 0.1 nm for the ADP form. In the presence of ATP (Fig. 2 G, red), the contour-length increase histogram shows bimodality: a major population (87 of 99) of the SBD^∗ye unfolding events occurs an a shorter contour-length increase centered around 35.6 ± 0.1 nm, while a minor population (12 of 99) of the events exhibits contour-length increases centered around 52.5 ± 0.2 nm, which is identical to the value we find for the apo/ADP form.

A scatter plot of unfolding contour lengths versus unfolding forces shows the bimodality for the SBD^∗ye unfolding in the presence of ATP even more clearly (Fig. 3 A). The red data points fall into two clusters: one cluster (87 of 99) at low forces <20 pN and small contour length (35.6 nm), and a second cluster at higher contour length and with a broader force distribution (12 of 99). Corroborating our finding, also a k-means clustering approach (34) identified two clusters with the best elbow Silhouette score (35). Interestingly the minor cluster we identify coincides with the distributions of forces and contour lengths we find for ADP and apo conditions (Fig. 3 A, gray and light blue ellipses). The shorter contour-length increase of the major population for the SBD^∗ye in the presence of ATP indicates that part of the SBD is unstructured and hence in the open form. Significant loss of α-helical lid structure in the ATP state was also reported for DnaK^∗ye (10).

Figure 3.

Single-molecule force spectroscopy of ATP-deficient and non-allosteric DnaK variant. (A) Unfolding forces plotted versus contour-length increases for the SBD^∗ye in DnaK^∗ye in the presence of ATP (n_ATP = 99). For ADP and apo forms, DnaK^∗ye data are indicated by the semi-transparent ellipsoid (apo: gray, ADP: blue). (B) A simple scheme for the ATP cycle of DnaK. Time constants between individual steps of the cycle are published (τ = 1/k_i) (40,49,52). (C) Force-extension pulling cycles of a single DnaK^∗ye in the presence of ATP. During cycling, the mechanical fingerprint of the SBD^∗ye domain switches between ATP-induced open SBD^∗ye (red traces) and ADP/apo-like closed SBD^∗ye (blue trace). (D) Design of DnaK^∗ye variants. The T199A variant (brown) locks DnaK^∗ye in the ATP state because the hydrolysis is blocked. The D393 variant (green) abolished allosteric communication between NBD and SBD. (E) Force-extension traces of DnaK^∗ye T199A (brown traces) and DnaK^∗ye D393R (green traces) in the presence of ATP. The pulling speed was 20 nm/s. (F) Scatter plots of the unfolding traces and L_c for DnaK^∗ye (red squares in the background; n = 99), DnaK^∗ye D393R (green squares; n = 35), and DnaK^∗ye T199A (brown squares; n = 39). Experiments were performed in the presence of ATP and pulling speed 20 nm/s. For experimental conditions, see the caption for Fig. 2. To see this figure in color, go online.

In summary, in the apo and ADP states, the mechanical properties of the NBD and the SBD are similar, and they correspond to the properties of the undocked individual domains. When ATP is bound, the SBD^∗ye is destabilized but can transition back to its fully folded form when ATP is hydrolyzed in the NBD.

ATP-deficient and non-allosteric DnaK variants disrupt ATP/ADP cycling of the SBD^∗ye

To investigate SBD^∗ye cycling between the stable (apo/ADP) and the less stable, partially open form (ATP) in more detail, we prepared two variants interfering with different molecular events (Fig. 3 D). In the first variant, DnaK^∗ye D393R, the allosteric coupling between ATP hydrolysis and ATP sensing of SBD^∗ye is abolished. The conserved aspartic acid residue at position 393 is located in the linker that connects the NBD-SBD domains. As shown previously (36), the D393 variant retains both domains folded and individually functional; however, the allosteric communication between the domains is completely abolished (36). Force-extension traces of the DnaK^∗ye D393R variant in the presence of ATP showed a mechanically highly stable and closed form of the SBD^∗ye similar to the ADP/apo form of the DnaK^∗ye (Fig. 3 E, green traces). The second variant, DnaK^∗ye T199A, contains an amino acid substitution in the NBD catalytic site and disrupts the ATPase activity (37). Force-extension traces of the DnaK^∗ye T199A in the presence of ATP (Fig. 3 E, brown traces) displayed only the partially open SBD^∗ye form as becomes evident from the shorter contour-length increase. The scatter plot of the unfolding forces versus contour-length increases confirms that these two variants exist in only one form, which is reflected by the unimodality of the L_c distribution. For the D393R variant, the average unfolding force is 25.6 ± 1.9 pN (n_ATP = 35; Fig. 3 F, green squares), and the average contour-length increase is 52.8 ± 0.2 nm for the ATP form (Fig. 3 F, green squares). For the T199A variant, the average unfolding force is 20.1 ± 0.9 pN (n_ATP = 39; Fig. 3 F, brown squares), and the average contour-length increase is 33.6 ± 0.2 nm in the presence of ATP (Fig. 3 F, brown squares). Hence, the D393R variant only exists in the mechanically stable, closed SBD^∗ye form even in the ATP state, while the T199A variant only exists in the partially open form. Both variants disrupt the ATP/ADP cycling observed for DnaK^∗ye and support the proposed molecular model.

Internal mechanics of SBD^∗ye in the ATP- and peptide-bound form DnaK^∗ye

Having characterized the individual apo/ADP/ATP states of DnaK^∗ye, we examined how peptide substrate binding to the SBD^∗ye affects the internal mechanics of the subdomains. The binding of the peptide substrate to the SBD stimulates ATP hydrolysis (38) (Fig. 4 A), which turns DnaK into a high-affinity ADP state. Once the ADP dissociates, ATP is bound rapidly (1 mM solution concentration) to DnaK, which results in a low-affinity state that, however, has faster peptide binding/unbinding kinetics (39,40). All these processes are part of the DnaK chaperone cycle during its interaction with a substrate.

Figure 4.

DnaK^∗ye SBD cycling between open/closed forms in the presence of ATP and peptide substrate. (A) A simple scheme for the ATP cycle of DnaK and stimulating effect of the peptide substrate. (B) Examples of three types of force-extension DnaK^∗ye traces in the presence of ATP and peptide ligands. Traces report on different SBD^∗ye states: closed state (ADP, blue), lid open state (ATP, red), and unfolded state (black trace, no background color). Experiments were performed in the presence of 1 mM ATP, 5 mM MgCl₂, and 250 μM NR peptide in the buffer (see caption for Fig. 2), and a pulling speed of 20 nm/s. (C) Analysis of 119 force-extension traces (1 trace = 1 black square) of DnaK^∗ye (+ATP, +peptide) with folded NBD while SBD^∗ye is probed by stretching cycle. Background color indicates a particular state (ATP: red, ADP: blue, unfolded SBD^∗ye: no color). (D) The force-extension trace of DnaK^∗ye with folded NBD (with transiently bound ADP); at ∼13 pN, the stretching process was stopped and continued as constant trap-trap distance shown in (E). Monitoring forces over time with the folded NBD (ADP) and initially the SBD^∗ye closed form. (F) Force fluctuations originating from the partially open form of the SBD^∗ye observed in the ATP state (folding rates: open circles, unfolding rates: filled circles). Solid lines are theoretical fits of the force-dependent rate model, which includes tether/polypeptide chain entropic elasticity (53). (G) A comparison of force-extension traces for DnaK^∗ye + ATP + peptide and isolated SBDΔα + peptide as a mimic for the undocked state. (H) Analysis of 42 force-extension traces (1 trace = 1 black square) of the DnaK^∗ye T199A variant (+ATP, +peptide) with folded T199A NBD while SBD^∗ye is probed by stretching cycle. (I) Model of the coupling between ATP/ADP state of the NBD and mechanical properties of SBD^∗ye. To see this figure in color, go online.

Here, we analyze how the peptide binding to the SBD affects the switching between closed and open forms in the ATP-bound form of DnaK^∗ye. The stimulation of ATP hydrolysis depends on the substrate affinity (41). In our assay, we used 250 μM NR peptide, which stimulates the ATP hydrolysis ca. threefold (see Fig. S5 in (10)). To study the conformational cycling, a suitable readout for the actual form of the SBD^∗ye can be the profile of the native force-extension trace pulled to forces, which are below the unfolding force (Fig. 4 B, blue trace). Thus, mechanical stretching experiments can be used to analyze the open/closed SBD^∗ye form. During mechanical probing in the ATP state, however, SBD^∗ye without peptide substrate can unfold even at low forces (Fig. 4 B, red trace), and it may need a variable amount of time for refolding. During this variable time, SBD^∗ye stays unfolded (Fig. 4 B, black trace), and, hence, no information can be obtained. The force-extension fingerprint is unique for all three situations, and hence during each stretching cycle, we get the information on the SBD^∗ye form (closed, open, unfolded). We performed a sequence of stretch/relax cycles of a single DnaKye^∗ in the presence of ATP and NR peptide (Fig. 4 C). In the experiment, applied forces are significantly below the unfolding forces for the NBD, and hence the NBD retains a fully folded state during the whole sequence of probing experiments. The state of SBD^∗ye was determined for each cycle by the classification of Fig. 4 B and was monitored over 400 s (Fig. 4 C). After each stretching cycle, the protein was rested at zero force for 7 s. The data show that SBD^∗ye in DnaK^∗ye can switch between open and closed forms (Fig. 4 C, red to blue), thus providing direct information on ATP hydrolysis events. We find the average lifetime is 60 ± 20 s for the ATP SBD^∗ye open state and 28 ± 8 s for the closed ADP state. To improve the time resolution for capturing the closed-to-open SBD^∗ye transition triggered by the binding of a new ATP molecule after ADP release has occurred, we combine constant velocity experiments with a constant trap-trap distance experiment (Fig. 4 D and E). After confirming that SBD^∗ye is trapped in the closed ADP form (Fig. 4 D), the stretching ramp was stopped at 13 pN, and the two traps were held at constant separation (Fig. 4 E). In this data trace, we observed the switching between the closed SBD^∗ye form to the open ATP-bound form in real time (Fig. 4 E, from blue to red). In the open state, characteristic force fluctuations can be observed (Fig. 4 F) indicative of rapid folding/unfolding of parts of the β-subdomain. These structural changes in the SBD can be attributed to detachment of β7-8 from the β-core as observed for isolated SBD (i.e., β^F fluctuations in Mandal et al. (29)). Next, we wanted to examine whether the structural fluctuations in the open state of SBD^∗ye in the DnaK⋅ATP⋅peptide complex are affected by the tight docking of the SBD to the NBD. We therefore prepared a construct of only the β-subdomain of SBD^∗ye (SBD^∗yeΔα; supporting material) and measured force-extension curves in the presence of the peptide (Fig. 4 G). These traces show that the observed fluctuations are identical in length and force range to those observed in the full-length construct. This result suggests that SBD^∗ye in the DnaK⋅ATP⋅peptide complex is mechanically not affected by the contacts to the NBD.

In our experiments, we monitored the open/closed forms of the SBD^∗ye, which may or may not be strictly coupled to the ATP/ADP state of DnaK^∗ye. To examine the extent of coupling between SBD^∗ye form and nucleotide status of DnaK^∗ye, we investigated the T199A DnaK^∗ye variant, where ATPase activity is abolished and hence permanently stays in the ATP state. Only the open SBD^∗ye form was observed in the presence of ATP and NR substrate peptide (Fig. 4 H), and we can rule out that the lid closes, which indicates strong coupling lid opening/closing to the ATP/ADP status of DnaK^∗ye.

Discussion

Opening and closing an enzyme’s active site is a crucial strategy to control the closest environment of a substrate. For example, for productive catalysis of kinases, the closing of the active site isolates substrates and expels water molecules to suppress unproductive hydrolysis (adenylate kinase, for example, (42)). DnaK, the Hsp70 chaperone, operates under a similar principle: lid closing after binding of the hydrophobic peptide stretch isolates the aggregation-prone substrate and suppresses its unproductive misfolding and aggregation. How quickly and strongly the substrate is bound to DnaK are allosterically regulated by the ATP/ADP status of NBD.

In this study, we observed allosteric ATP-induced mechanical and structural changes of a single two-domain DnaK molecule while undergoing ATP/ADP cycle in the presence of a peptide substrate. Capturing the ATP-induced changes in DnaK is technically challenging; based on available structures of the ATP-bound open and ADP-bound closed forms, conformational changes are in the range of ∼1 to 2 nm (21,43). Under certain circumstances, mechanical probing of such small conformational transitions is feasible (e.g., adenylate kinase (13)), mainly due to the technical improvements of laser optical tweezers (13,26). In the present study, successful observation of a single allosteric protein at work was achieved by several other improvements, including 1) usage of DnaK ye variant for single-molecule force spectroscopy, 2) extensive characterization of individual domains and comparison with two-domain DnaK, 3) comparison with non-allosteric and hydrolysis-deficient variants, and 4) developing assay protocols that monitor the hydrolysis state of a single ATP molecule bound to DnaK.

DnaK variant optimization is required for practical single-molecule force experiments

In our initial experiments, we tried to use wild-type DnaK (1–638). Even after several rounds of assay optimization, we could not obtain enough single-molecule tethers. Force-extension traces of four different DnaK molecules are shown in the supporting material. Removing the C-terminal intrinsically disordered tail of DnaK, corresponding to residues 607–638, improved the success rate in the assay. Even though the disordered, variable-length C-terminus was reported to enhance the overall fitness of a bacteria (44,45), it is not essential for the two-domain DnaK allostery. However, we found that at slow pulling speeds, when the α-helical SBD domain unfolds first, the unfolded part is rapidly rebound by the remaining folded β-subdomain (see Fig. 1 E). This rebinding process occurs even under load but can be completely suppressed by a large excess of the peptide substrate in the measurement buffer. Since the final product of our mechanical stretching assay is an unfolded polypeptide chain, a part of this polypeptide chain is recognized as a substrate for the still-folded β-subdomain of the SBD. According to the LIMBO prediction algorithm (46), there are four DnaK-binding motifs in the full-length DnaK sequence (supporting material). Two of these binding motifs are located in the NBD (⁷⁰KRLIGRR and ²³⁴SRLINYL), one motif is in the conserved connecting linker (³⁸⁹VLLLDVT), and one is at the end of α-subdomain (⁵⁹⁶QKLMEIA). The rebinding event we observe in our single-molecule experiments is not consistent with those predictions. First, the rebinding event is observed after 23.3-nm-long unfolding of the α-helical part of the SBD. Second, assuming the rebinding of the C-terminal ⁵⁹⁶QKLMEIA motif, the contour-length increase should match the observed unfolding event, i.e., 23.3 nm. Instead, the length changes we measure for the rebinding event (the total length of 6.5 nm) are consistent with the location of the ⁵⁴¹HLLH-binding motif.

Interestingly, previous studies have shown that under the circumstances of destabilizing the α-subdomain, DnaK binds its own C-terminal part (24,47,48). In particular, those studies showed that the peptide-binding site of the β-SBD recognizes the internal sequence ⁵⁴¹HLLH that is part of the long, lid-like helix B. We conclude that this HLLH motif is responsible for the rebinding events we find in our single-molecule experiments. Side chains of the two consecutive hydrophobic leucine residues are oriented toward helices C and D and create a highly hydrophobic core with residues L591, L569, 572, and 576 (PDB: 2KHO; (43)). Mechanical unfolding of the B-E α-helices destroys the hydrophobic core and exposes the HLLH motif to the folded β-SBD, which rapidly binds this motif. To suppress the intramolecular binding of the unfolded α-subdomain by the β-SBD, we prepared the following variant: DnaK^∗ye, originally designed by the Gierasch group (23,49). In DnaK^∗ye, helices C-E are removed, and the two consecutive bulky hydrophobic leucine side chains of the HLLH motif are replaced by an aromatic tyrosine side chain and a negatively charged glutamic acid, respectively. These structural changes yielded a highly cooperative SBD^∗ye domain with increased mechanical stability and abolished rebinding events. In the wild-type SBD, passive mechanics of the α- and β-subdomains are optimized for large allosteric ATP-induced conformational changes (29). Opening the α/β interface only requires 2 k_BT. In the apo form of DnaK^∗ye, the opening of the α/β interface is now highly cooperative and occurs at high unfolding forces. Apparently, the strong interaction introduced by the substituted residues overrules the subtle energetic balance between the subdomains, and they behave as one.

Nucleotide state of DnaK^∗ye NBD controls the mechanical properties of SBD^∗ye

Systematic investigation of force-extension traces of DnaK^∗ye in the absence and presence of ATP/ADP made it possible to reveal the allosteric action of the ATP-bound NBD on SBD^∗ye (Fig. 2). While the unfolding force distributions and contour-length gains of the NBD in DnaK^∗ye are similar to those observed for the individual NBD domains (28), differences can only be observed for the SBD. This indicates that even in the ATP state, during which the NBD interacts with SBD, there are no significant changes in the NBD mechanics, and NBD acts as a mechanically rigid part of DnaK^∗ye.

For the SBD of ADP/apo-DnaK^∗ye, we find that truncation and introducing two-point mutations resulted in a mechanically more stable subdomain than the wild-type. Several force-extension traces showed unfolding forces of the SBD^∗ye that were even higher than the unfolding forces of the NBD (∼30 pN at 20 nm/s). Close inspection of the unfolding force histogram shows a broad range of unfolding forces of the SBD in ADP/apo-DnaK^∗ye. Such broad force distribution might imply a very small transition state position value, indicating that SBD^∗ye is a brittle protein structure, and the native state cannot undergo large fluctuations without committing to cross the unfolding barrier (50). Another possible explanation would be that native SBD^∗ye is intrinsically heterogeneous and exists in several nearly isoenergetic structural states with different heights or locations of their unfolding barriers. The shape of the force distribution appears multimodal instead of supporting the latter interpretation. However, the statistics were not yet good enough to clearly distinguish between the two. The molecular origin of such heterogeneity cannot be assessed from our experiments; heterogeneous orientation of the B helix and a metastable SBD^∗ye α-helical region were reported (24,51) and may contribute to observed multimodality. In contrast to multimodal force distributions, contour-length gain distributions in the ADP/apo forms are well defined and unimodal.

When ATP is bound to the NBD in a two-domain DnaK^∗ye, structural changes occur that result in a dramatic change in the integrity and mechanical properties of SBD^∗ye. First, the average unfolding forces of the SBD^∗ye dramatically decrease from 26 (apo) to 16 pN (ATP). Second, the average contour-length gain decreases from 52.9 (apo) to 37.6 nm (ATP). An analysis of the scatter plot of unfolding forces versus contour-length gains performed by the k-means clustering method (Fig. 3 A) shows the existence of data clusters. A minor cluster (12%) is statistically indistinguishable from SBD^∗ye data measured under apo/ADP conditions (Fig. 3 A). A second cluster (88%) is unique for ATP conditions and well separated from the first cluster. The population of ATP-SBD^∗ye extracted from this cluster has shorter contour-length gains (35.5 ± 0.4 nm) and lower unfolding forces (16.8 ± 0.6 nm pN). Based on this data, we conclude that the observed bimodality is due to hydrolysis of ATP to ADP, which turns mechanical/structural properties of minor molecules of SBD^∗ye to ADP/apo like. ATPase-deficient T199A DnaK^∗ye shows the shorter contour-length gain (Fig. 3 E and F) and hence supports our conclusions. The D393R DnaK^∗ye variant confirmed the essential role of functional ATP-induced allosteric communication; only fully folded SBD^∗ye with a contour length gain of 53 nm could be found in this variant (Fig. 3 E and F). To summarize, single-molecule force experiments can detect allosterically induced mechanical destabilization of the distal SBD^∗ye while the NBD undergoes ATP/ADP cycling.

Internal mechanics of two-domain DnaK undergoing an ATP/ADP cycle in the presence of peptide substrate

Combining our results about nucleotide state-dependent contour-length changes within the SBD, we were now able to monitor the ATP cycling and associated mechanical and structural status of SBD^∗ye in real time. Consistent with the literature, the lifetimes in the ATP state become shorter when the peptide is added to the solution (Fig. 4 C). The average lifetime of the closed ADP-like form (30 s) is slightly shorter than the expected time for the dissociation of the nucleotide (ca. 50 s; supporting material). Opening and closing of the SBD^∗ye lid could, at least in theory, occur independently of the ADP/ATP status of DnaK^∗ye. The ATPase-deficient T199A DnaK^∗ye shows that only the lid-open form of the SBD^∗ye is populated. This finding indicates that, for DnaK^∗ye, there is a strong coupling between open and closed states of SBD^∗ye and the nucleotide status of the NBD. Apparently, strong interactions between the NBD and the SBD affect a high synchronization and timing of allosteric events.

The following model can be deduced from our single-molecule force spectroscopy study (Fig. 4 I). In the ADP state, DnaK can be viewed as consisting of two tightly folded domains; however, from a mechanical viewpoint, the two domains behave independently of each other like beads on a string, consistent with earlier studies (24). In the ATP form, the NBD remains rigid. However, the ATP-induced change in the NBD conformation triggers NBD-SBD interactions, resulting in stable docking of the SBD to the NBD. As a result of the interdomain docking, the internal reorganization of the SBD leads to the release and simultaneous unfolding of the α-helical lid, leaving a mechanically less stable β-SBD. Even in the presence of peptides, the strong domain-domain coupling is preserved.

Author contributions

G.Z. and M.R. designed research; G.Z. and A.S. performed research; G.Z. and A.S. analyzed data; and G.Z. and M.R. wrote the paper.

Editor: Mark Williams.