The Molecular Mechanism of Hsp100 Chaperone Inhibition by the Prion Curing Agent Guanidinium Chloride

Background: Guanidinium chloride (GdmCl) inhibits Hsp100 chaperones and leads to prion curing in yeast.

Results: The Gdm⁺ ion binds specifically to the N-terminal nucleotide binding domain, interacting primarily with a conserved glutamate and the bound nucleotide.

Conclusion: GdmCl affects the catalytic cycle by both interfering with the essential glutamate and modulating nucleotide binding affinities.

Significance: The study elucidates the mechanistic role of GdmCl in Hsp100 chaperone inhibition.

Abstract

The Hsp100 chaperones ClpB and Hsp104 utilize the energy from ATP hydrolysis to reactivate aggregated proteins in concert with the DnaK/Hsp70 chaperone system, thereby playing an important role in protein quality control. They belong to the family of AAA+ proteins (ATPases associated with various cellular activities), possess two nucleotide binding domains per monomer (NBD1 and NBD2), and oligomerize into hexameric ring complexes. Furthermore, Hsp104 is involved in yeast prion propagation and inheritance. It is well established that low concentrations of guanidinium chloride (GdmCl) inhibit the ATPase activity of Hsp104, leading to so called “prion curing,” the loss of prion-related phenotypes. Here, we present mechanistic details about the Hsp100 chaperone inhibition by GdmCl using the Hsp104 homolog ClpB from Thermus thermophilus. Initially, we demonstrate that NBD1 of ClpB, which was previously considered inactive as a separately expressed construct, is a fully active ATPase on its own. Next, we show that only NBD1, but not NBD2, is affected by GdmCl. We present a crystal structure of ClpB NBD1 in complex with GdmCl and ADP, showing that the Gdm⁺ ion binds specifically to the active site of NBD1. A conserved essential glutamate residue is involved in this interaction. Additionally, Gdm⁺ interacts directly with the nucleotide, thereby increasing the nucleotide binding affinity of NBD1. We propose that both the interference with the essential glutamate and the modulation of nucleotide binding properties in NBD1 is responsible for the GdmCl-specific inhibition of Hsp100 chaperones.

Introduction

Protein aggregation represents a severe threat for living organisms. It can occur under various cellular stress conditions, possibly most pronounced in the case of heat shock. The molecular chaperone Hsp104² and its bacterial homolog ClpB possess the unique ability to reactivate aggregated proteins, thereby providing increased thermotolerance for their respective organisms (1, 2). They belong to the family of AAA+ proteins (ATPases associated with various cellular activities) that generally use the energy from ATP hydrolysis to induce structural rearrangements in their substrates (3, 4). In particular, during the process of protein disaggregation, single polypeptide chains are translocated through the central pore of the hexameric Hsp104/ClpB ring, which is believed to be the active unit (5). Many mechanistic details including the interplay with the essential co-chaperones of the Hsp70/Hsp40 system (6–9) as well as allosteric regulation (10–12) within the hexameric Hsp104/ClpB complex are still under investigation. Additionally, Hsp104 is involved in the propagation and inheritance of yeast prions, such as [PSI⁺] and [URE3], which are amyloid structures formed from soluble precursor proteins (13–15). Similarly to protein disaggregation, the Hsp70/Hsp40 chaperone system is highly involved in these tasks (for a review, see Ref. 9).

It has been shown that low concentrations of guanidinium chloride (GdmCl) inhibit the activity of Hsp104, causing two effects in vivo: (a) impaired thermotolerance and (b) so-called “prion curing,” the loss of prion-related yeast phenotypes (16, 17). These effects have been associated with a GdmCl-specific inhibition of ATP hydrolysis in Hsp104 (18). A conserved aspartate (Asp-184 in Hsp104 from Saccharomyces cerevisiae) has been proposed to be essential for the Hsp104-related yeast prion propagation and the inhibition of Hsp104 by GdmCl (19, 20). Here, we present a detailed mechanistic study elucidating the role of the prion curing agent GdmCl in Hsp100 chaperone inhibition using the Hsp104 homolog ClpB from Thermus thermophilus, which is the only Hsp100 chaperone with high resolution structural data available.

The three-dimensional structure of the ClpB monomer from T. thermophilus is shown in Fig. 1 (21). It possesses two nucleotide binding domains, NBD1 and NBD2, bearing the characteristic Walker A and B motifs essential for ATP binding and hydrolysis. In between them, a long coiled coil domain is inserted. This so-called M domain is essential for the disaggregation activity (22, 23). The N-terminal domain, which is involved in substrate binding, however, is not essential (24). Like other AAA+ proteins, Hsp100 chaperones form hexameric ringlike structures, which were studied extensively by cryo-EM (25, 26). Previous studies utilized shorter ClpB constructs bearing just one of the two nucleotide binding domains (NBDs) each, allowing a more detailed mechanistic characterization (27). The construct NBD1-M(141–519), carrying the first NBD, was inactive in isolation, whereas NBD2(520–854), carrying the second NBD, could bind and hydrolyze ATP on its own. Both constructs mixed in solution were shown to reassemble into a functional oligomer with chaperone activity (28). Here, we show that the slightly longer construct NBD1-M(141–534) is nucleotide binding-competent and active in ATP hydrolysis, demonstrating that NBD1 also is a fully functional ATPase in isolation once the 15 additional C-terminal amino acids that complete the helical small domain are present. With this new construct in hand, we show that only NBD1, but not NBD2, is affected by GdmCl. Based on both structural and kinetic data, we derive a molecular mechanism of the inhibition of Hsp100 chaperones by the prion curing agent GdmCl.

FIGURE 1.

The domain architecture of ClpB from T. thermophilus. The ClpB monomer (Protein Data Bank code 1QVR) comprises two nucleotide binding domains, NBD1 (green) and NBD2 (red). The darker colors indicate the helical bundles (also called small domains) of both NBDs. Bound nucleotides are shown as space-filling models. The long coiled coil M domain (yellow) is an insertion into NBD1. The ClpB constructs used in this study in addition to the full-length protein are NBD1-M(141–534) and NBD2(520–854), each containing only one NBD. The N-terminal domain (blue) is excluded for the separate constructs.

EXPERIMENTAL PROCEDURES

Construct Design and Mutagenesis

Construct design for ClpB, DnaK, DnaJ, and GrpE from T. thermophilus as well as the ClpB variants NBD1-M(141–519) and NBD2(520–854) was described previously (27, 29, 30). The ClpB construct NBD1-M(141–534) was generated in the same manner by PCR using the pRS-ClpB plasmid, coding for full-length ClpB, as the template. The PCR product was purified and subcloned into an NdeI/EcoRI-digested pET28a vector that is used for proteins with an N-terminal, cleavable His tag. The mutation E209A was introduced by QuikChange PCR according to the QuikChange protocol (Agilent Technologies, Santa Clara, CA). Sequences were verified by DNA sequencing performed by Eurofins MWG Operon (Ebersberg, Germany).

Protein Expression and Purification

All proteins were expressed recombinantly in Escherichia coli BL21(DE3) RIL. All ClpB variants as well as DnaK, DnaJ, and GrpE from T. thermophilus were purified as described previously (27, 29–31). The modified construct ClpB NBD1-M(141–534) was purified following the same protocol that was used for the shorter variant NBD1-M(141–519) with an additional alkaline phosphatase treatment for 3 h at room temperature as described for the NBD2(520–854) purification (31). The protein NBD1-M(141–534) was stored in buffer A (50 mm Tris/HCl, pH 7.5, 50 mm KCl, 5 mm MgCl₂, and 2 mm EDTA).

Crystallization and Structure Determination

Crystals of ClpB NBD1-M(141–534) were grown in a hanging drop vapor diffusion setup at 20 °C. The protein solution (10 mg/ml in buffer A) was supplemented with 3 mm ADP and 10 mm GdmCl and mixed 1:1 with the reservoir solution, which contained 0.1 m Tris/HCl, pH 7.5, 1.0 m LiCl, 18% PEG 6000, 10 mm MgCl₂, and 10 mm GdmCl. Needle-shaped crystals grew within 5 days. The cryoprotectant solution consisted of the reservoir solution supplemented with 20% ethylene glycol, 3 mm ADP, and 100 mm GdmCl. Crystals were passed quickly through the cryoprotectant solution and cryocooled in liquid nitrogen.

Diffraction data were collected at the synchrotron beamline X10SA at the Swiss Light Source (Villingen, Switzerland) with the crystals kept at 100 K. The program XDS was used for data processing (32). Phasing was done by molecular replacement using the program PHASER (33), and residues 141–534 of the ClpB full-length structure (Protein Data Bank code 1QVR) were used as the search model (21). Further model building and refinement were done in iterative cycles using the programs Coot (34) and REFMAC (35) including TLS refinement (36), respectively. The model quality was validated using MolProbity (37). Structure illustrations were prepared with PyMOL (38).

Stopped Flow Experiments

Transient kinetics experiments were performed with a BioLogic SFM-400 stopped flow instrument in single mixing configuration at 25 °C (BioLogic Science Instruments, Claix, France). The fluorescently labeled nucleotides MANT-ADP and MANT-dADP were purchased from BIOLOG (Bremen, Germany). The excitation wavelength was set to 296 nm, and the fluorescence signal was observed using a 400 nm long pass filter (400FG03-25, LOT Oriel Group). This setup allows a selective excitation of protein-bound MANT-nucleotides via fluorescence resonance energy transfer (FRET) from the initially excited tryptophan residues of the protein. Measurements were performed in buffer A. For measurements in the presence of GdmCl, 10 mm GdmCl was added to buffer A. Kinetic traces were recorded as triplicates and averaged. Data analysis was performed using the program GraphPad Prism 5.0.

Kinetic traces from direct binding experiments (2 μm ClpB NBD1-M(141–534) mixed with 10–50 μm MANT-nucleotide) were fitted to exponential functions. The extracted rate constants were plotted against the nucleotide concentration, leading to a linear function. The on-rates for MANT-nucleotide binding were obtained from the slope of this linear function.

Kinetic traces from dissociation experiments (2 μm ClpB NBD1-M(141–534) incubated with 20 μm MANT-nucleotide and subsequently chased with 10 mm ADP) were fitted to exponential functions. The extracted rate constant corresponds to the off-rate of MANT-nucleotide binding. The K_D was calculated from the ratio of off- and on-rates.

Dissociation experiments were also performed under high salt and varied pH conditions. The buffers contained 300 mm KCl (high salt) or 50 mm CHES, pH 9.5 (high pH) or sodium acetate, pH 5.5 (low pH), respectively.

Fluorescence Titrations

Fluorescence displacement titrations to determine K_D(ADP) and K_D(ATP) were performed at 25 °C in buffer A using a Jasco FP-8500 fluorescence spectrometer (Jasco Germany GmbH). The excitation wavelength was set to 296 nm. The MANT fluorescence signal was monitored at 441 nm. 2 μm ClpB NBD1-M(141–534) was incubated with 20 μm MANT-dADP and subsequently titrated with ADP or ATP. ATP titrations were performed in the presence of 2 mm phosphoenolpyruvate and 0.01 mg/ml pyruvate kinase (Roche Applied Science) as an ATP-regenerating system. For measurements in the presence of GdmCl, 10 mm GdmCl was added to the buffer. The data were corrected for dilution effects and analyzed with a cubic equation for competing ligands using the initial concentrations of protein and MANT-dADP as well as the K_D(MANT-dADP) from the stopped flow experiments as input values to determine K_D(ADP) and K_D(ATP) (39). The program GraFit 5.0 was used for data fitting.

To assess nucleotide binding affinities at higher protein concentration, 20 μm NBD1-M(141–534) was incubated with 40 or 50 μm MANT-dADP and titrated with ADP or ATP, respectively. In this case, the reference K_D(MANT-dADP) for the data analysis was determined in a separate titration against MANT-dADP.

Steady State ATPase Assay

Steady state ATPase activity was measured in a coupled colorimetric assay at 25 °C using a Jasco V-650 spectrophotometer (Jasco Germany GmbH). ClpB NBD1-M(141–534) was incubated at 25 °C in assay buffer (50 mm Tris/HCl, pH 7.5, 100 mm KCl, 2 mm EDTA, 2 mm dithioerythritol, 0.4 mm phosphoenolpyruvate, 0.4 mm NADH, 0.1 g/liter BSA, 4 units/ml pyruvate kinase, 6 units/ml lactate dehydrogenase, and 7 mm MgCl₂). Different protein concentrations ranging from 5 to 40 μm were used. The reaction was started by adding ATP (0.1–5 mm). The decreasing absorption at 340 nm was monitored over time, and the maximal slope was used to determine the ATPase turnover rate per molecule. The data were analyzed with the Hill equation (Equation 1) using the program GraphPad Prism 5.0.

Experiments to characterize the influence of GdmCl on the steady state ATPase activity were performed in the same manner using 20 μm ClpB NBD1-M(141–534), 10 μm NBD2(520–854), or 10 μm ClpB full-length. The assay buffer was supplemented with 0–15 mm GdmCl. 2 mm ATP or dATP was used to start the reaction. The ATPase rates were calculated as before from the slope of the decreasing A₃₄₀ signal. Control experiments were performed to verify that the coupled reactions catalyzed by pyruvate kinase and lactate dehydrogenase are not rate-limiting in the presence of GdmCl and/or deoxynucleotides.

ATPase activity assays were also performed under high salt and varied pH conditions. The buffers contained 300 mm KCl (high salt) or 50 mm CHES, pH 9.5 (high pH) or sodium acetate, pH 5.5 (low pH), respectively.

Gel Filtration Experiments with Static Light Scattering (SLS) Analysis

Gel filtration experiments were performed on a Superdex 200 10/300 GL column connected to a refractive index detector (2414 from Waters, Milford, MA), a photodiode array detector (2996 from Waters), and a multiangle light scattering (MALS) detector (Dawn Heleos, Wyatt Technology, Santa Barbara, CA) in buffer A. The buffer was either nucleotide-free or supplemented with 2 mm ADP or 2 mm ATP. In the case of ATP, 2 mm phosphoenolpyruvate and 0.01 mg/ml pyruvate kinase (Roche Applied Science) were present as an ATP-regenerating system. The experiments were performed both in the absence or presence of 10 mm GdmCl. 40 μl of 100 μm NBD1-M(141–534) were injected, resulting in a final concentration of about 2 μm at the detector due to a 1:50 dilution during the experiment. Molecular weight values were extracted from the MALS data using the ASTRA software (Wyatt Technology).

Dynamic Light Scattering (DLS) Analysis

DLS experiments were performed at 10–50 μm protein concentration in buffer A that was either nucleotide-free or supplemented with 2 mm ADP or 2 mm ATP, respectively. In the case of ATP, 2 mm phosphoenolpyruvate and 0.01 mg/ml pyruvate kinase (Roche Applied Science) were present as an ATP-regenerating system. The measurements were performed with a Viscotek 802DAT DLS instrument (Viscotek, Waghäusel, Germany). 40 scans with a measuring time of 5 s per scan were recorded. The OmniSIZE 3.0 software package was used for data analysis.

Disaggregation Assay (Chaperone-assisted Reactivation of Heat-aggregated α-Glucosidase)

The assay was used to characterize the influence of GdmCl on the chaperone activity of ClpB. 0.2 μm α-glucosidase from Bacillus stearothermophilus was denatured for 8 min at 75 °C in reaction buffer containing 50 mm MOPS, pH 7.5, 150 mm KCl, 10 mm MgCl₂, 5 mm ATP, 1 mm dithioerythritol, and 0–12 mm GdmCl. Chaperones were added prior to refolding at 55 °C. The total chaperone concentrations were [ClpB] = 1.0 μm, [DnaK] = 1.6 μm, [DnaJ] = 0.5 μm, and [GrpE] = 0.2 μm. Samples were taken after 30, 60, and 120 min, and the α-glucosidase activity was measured at 40 °C using a microplate spectrophotometer (Varioskan, Thermo Electron, Vantaa, Finland). The samples were diluted into assay buffer containing 50 mm KP_i, pH 6.8, 2 mm para-nitrophenyl α-d-glucopyranoside, 0.1 mg/ml BSA, and 15 mm GdmCl. The measurement was performed in transparent 96-well plates (3641 from Corning). The average rate of absorption increase at 405 nm was monitored and normalized against a positive control containing α-glucosidase that was not heat-aggregated. Samples supplemented with either only ClpB or only DnaK/DnaJ/GrpE were used as negative controls.

Circular Dichroism Measurements

Protein unfolding experiments with GdmCl were performed as a control to show that the low concentrations of GdmCl used in the activity assays and nucleotide binding measurements did not influence the stability of ClpB NBD1-M(141–534). 5 μm ClpB NBD1-M(141–534) was incubated with various concentrations of GdmCl (0–6 m) in buffer containing 50 mm Tris/HCl, pH 7.5, 50 mm KCl, 5 mm MgCl₂, and 2 mm EDTA. Circular dichroism spectra were measured with a Jasco J-810 spectropolarimeter (Jasco Germany GmbH) using quartz cuvettes of 0.1-cm path length. Five spectra were averaged. The measured ellipticity at 222 nm was used to generate the denaturation curve, which was fitted using the equation for a two-state transition (40).

RESULTS

The Isolated NBD1 of ClpB Is Nucleotide Binding-competent and an Active ATPase

Recent studies showed that the two NBDs of ClpB from T. thermophilus can be expressed and purified separately (27). The construct NBD1-M(141–519) lacks nucleotide binding competence and does not possess ATPase activity in isolation, whereas the construct NBD2(520–854) binds and hydrolyzes ATP. Remarkably, both constructs mixed together in solution were shown to reassemble into a functional, oligomeric ClpB chaperone in which NBD1 retrieves its nucleotide binding competence and activity (28). Here, we use a slightly modified construct, NBD1-M(141–534), containing 15 additional C-terminal amino acids, which comprise the last helix of the NBD1 small domain located after the M domain insertion (Fig. 1). Interestingly, NBD1-M(141–534) is now nucleotide binding-competent, forms oligomers upon nucleotide binding, and shows ATPase activity in isolation.

To assess the relevant nucleotide binding parameters, binding of the fluorescently labeled nucleotide MANT-dADP to NBD1-M(141–534) was characterized by stopped flow experiments. Fig. 2A shows binding traces upon direct mixing of 2 μm NBD1-M(141–534) with different concentrations of MANT-dADP that could be fitted to single exponential functions. The extracted rate constants were plotted against the nucleotide concentration, leading to a linear function in which the slope represents the on-rate and the y axis intercept corresponds to the off-rate of MANT-dADP binding (assuming pseudo first order conditions and a simple one-step binding mechanism). The slope could be determined very accurately and was therefore used to extract the on-rate for MANT-dADP binding with k_on = 0.34 ± 0.02 μm⁻¹ s⁻¹.

FIGURE 2.

Nucleotide binding parameters and ATPase activity of NBD1-M(141–534). A, kinetic fluorescence traces upon direct mixing of NBD1-M(141–534) and MANT-dADP. The final concentration of protein is 1 μm in all cases. The final MANT-dADP concentrations are 5 (blue), 8.3 (green), 12.5 (yellow), 18.75 (orange), and 25 μm (red). Single exponential fits are shown as colored lines (upper panel). The rate constants extracted from the kinetic traces are plotted against the MANT-dADP concentration to obtain the on-rate for MANT-dADP binding from the slope of the linear function (lower panel). The off-rate for MANT-dADP binding can be estimated from the y axis intercept but was determined separately by a dissociation experiment as described in the text. All nucleotide binding parameters obtained from these experiments are given in Table 2. B, fluorescence titrations to obtain K_D(ADP) and K_D(ATP). 2 μm NBD1-M(141–534) was incubated with 20 μm MANT-dADP and subsequently titrated with ADP (upper panel) or ATP (lower panel), respectively. For the ATP titration, phosphoenolpyruvate and pyruvate kinase were present as an ATP-regenerating system. The data were fitted with the cubic equation for competing ligands using K_D(MANT-dADP) obtained from the stopped flow experiments as an input value. All nucleotide binding parameters obtained from these experiments are given in Table 2. C, steady state ATPase turnover rates per molecule plotted against the ATP concentration (upper panel). The ATPase activity per molecule strongly depends on the protein concentration with [NBD1-M] = 5 (blue), 10 (green), 18 (yellow), 25 (orange), and 40 μm (red). The data were fitted with the Hill equation. The obtained Hill equation parameters k_cat, K_m, and the Hill coefficient are plotted against the protein concentration (lower panel). The low k_cat and high K_m values at low protein concentration and Hill coefficients significantly higher than 1.0 indicate that NBD1-M(141–534) oligomers represent the active form. a.u., arbitrary units.

To obtain the off-rate for MANT-dADP binding with higher confidence, 2 μm NBD1-M(141–534) was incubated with 20 μm MANT-dADP and subsequently mixed with excess amounts of ADP. The stopped flow trace for this dissociation experiment could be fitted to a single exponential function where the rate constant corresponds directly to the off-rate of MANT-dADP binding with k_off = 3.3 ± 0.03 s⁻¹. The binding affinity for MANT-dADP binding was calculated from the ratio of off- and on-rate, resulting in K_D ≈ 10 μm. Fitting of the amplitudes obtained from the direct mixing experiment shown in Fig. 2A to a hyperbolic binding curve resulted in a similar K_D(MANT-dADP) of 8.3 ± 0.9 μm, which further legitimates the use of a one-step binding model.

Furthermore, we used fluorescence equilibrium titrations to obtain the binding affinities for the unlabeled nucleotides ADP and ATP. 2 μm NBD1-M(141–534) was incubated with 20 μm MANT-dADP and subsequently titrated with ADP or ATP, displacing the MANT-labeled nucleotide (Fig. 2B). ATP titrations were performed in the presence of phosphoenolpyruvate and pyruvate kinase as an ATP-regenerating system. The displacement titration curves were analyzed using the cubic equation for competing ligands (39), resulting in K_D(ADP) = 13.7 ± 1.1 μm and K_D(ATP) = 1520 ± 89 μm. The reference K_D for MANT-dADP was taken from the stopped flow experiment. Binding affinities for ADP and MANT-dADP were found to be very similar, demonstrating that the MANT-labeled nucleotide serves as an appropriate analog. Remarkably, ClpB NBD1 binds ADP 2 orders of magnitude better than ATP. Previous studies have already shown such a strong discrimination between di- and triphosphate for ClpB NBD2 albeit with about 20-fold higher affinities (31).

To characterize the catalytic activity of NBD1-M(141–534), the steady state ATPase turnover per molecule was measured for different concentrations of protein and ATP (Fig. 2C). NBD1-M(141–534) hydrolyzes ATP efficiently with turnover rates in the same range as measured before for isolated NBD2 and full-length ClpB. The data were analyzed using the Hill equation; fitting parameters are plotted in the lower panel of Fig. 2C. The ATPase activity per molecule strongly depends on the protein concentration. Hill coefficients between 1.7 and 2.7 indicate that NBD1-M(141–534) is active as an oligomeric species. The K_m value is also influenced by the protein concentration, indicating that oligomerization and nucleotide binding are coupled.

To characterize the oligomerization behavior of NBD1-M(141–534) upon nucleotide binding, we performed both dynamic light scattering (DLS) and static light scattering (SLS) experiments. Average molecular masses of 70–100 kDa, corresponding to dimeric species, and 120–160 kDa, corresponding to trimeric species, can be observed in DLS experiments in the presence of ADP or ATP, respectively, whereas mainly monomers are present under nucleotide-free conditions. This clearly indicates nucleotide-induced oligomerization of NBD1-M(141–534). Gel filtration runs in the presence of ADP or ATP and under nucleotide-free conditions, respectively, are shown in Fig. 3A. Average molecular weights of the eluted species were determined using a MALS detector. In the presence of nucleotide, the peaks were shifted to higher molecular weight, indicating a monomer/dimer equilibrium. Because of substantial dilution during gel filtration, the molecular weights obtained from these SLS data are not as high as in the DLS measurements performed at about 10-fold higher protein concentration.

FIGURE 3.

Nucleotide-dependent oligomerization of NBD1-M(141–534) in the presence and absence of GdmCl characterized by gel filtration and SLS. A, gel filtration profiles of NBD1-M(141–534) in nucleotide-free buffer (solid line) and with 2 mm ADP (dotted line) and 2 mm ATP (dashed line) present in the running buffer. The ATP-containing buffer was supplemented with phosphoenolpyruvate and pyruvate kinase as an ATP-regenerating system. In the presence of nucleotide, the peaks are broader and elute earlier, indicating a shift to higher molecular weight. Subsequent to the separation on the gel filtration column, a MALS detector was used to determine the average molecular mass of the eluted species, resulting in 44 (nucleotide-free), 58 (ADP), and 56 kDa (ATP). The actual molecular mass of the NBD1-M monomer is 45 kDa. B, the gel filtration runs were performed as described in A with 10 mm GdmCl present in the running buffer. The shift toward earlier elution in the presence of nucleotide is more pronounced than in A. The molecular masses determined by MALS data analysis are 44 (nucleotide-free) and 64 kDa (ADP). In the presence of ATP, two distinct masses were determined (86 and 48 kDa). The actual molecular mass of the NBD1-M monomer is 45 kDa. a.u., arbitrary units.

Next, we tested whether the nucleotide binding affinities are protein concentration-dependent and performed fluorescence titrations as described above at higher protein concentration (20 instead of 2 μm). Indeed, K_D(ADP) and K_D(MANT-dADP) decreased 2-fold, and K_D(ATP) decreased 3-fold at higher protein concentration (Table 2).

TABLE 2.

Nucleotide binding parameters of NBD1-M(141–534) and NBD2(520–854) in the presence or absence of GdmCl

	NBD1-M WT		NBD1-M E209A		NBD2 WT,a no GdmCl
	No GdmCl	10 mm GdmCl	No GdmCl	10 mm GdmCl
MANT-ADPb
kon (μm−1 s−1)	0.12	0.15	NDc	ND	ND
koff (s−1)	4.0	1.2	ND	ND	ND
KD (μm)	33	8	ND	ND	ND
MANT-dADPb
kon (μm−1 s−1)	0.34	0.36	1.00	1.09	4.2
koff (s−1)	3.3	2.8	4.1	4.2	0.015
KD (μm)	10 (5)d	8	4	4	0.004
Unlabeled nucleotides
KD(ADP) (μm)	13.7 (6.7)d	2.4	6.5	6.9	0.5
KD(ATP) (μm)	1520 (510)d	122	1190	1320	89

^a Nucleotide binding parameters of NBD2(520–854) as published in Ref. 31.

^b The differences in k_on between MANT-ADP and MANT-dADP binding in the absence of GdmCl can be explained when taking into consideration that only one of the two different MANT-ADP isomers binds efficiently, and its concentration in the mixture is lower than the overall concentration of MANT-ADP.

^c ND, not determined.

^d Values in parentheses refer to measurements at higher protein concentration of [NBD1-M] = 20 μm, whereas all other binding parameters were determined at low protein concentration of [NBD1-M] = 2 μm.

In contrast to the shorter variant NBD1-M(141–519), which does not assemble in isolation, NBD1-M(141–534) is active in an oligomeric context in which nucleotide binding, ATPase activity, and oligomerization are strongly coupled. This is in agreement with previous experiments studying the reconstituted ClpB complex of NBD1-M(141–519) and NBD2(520–854) in which the NBD1 helical small domain is completed by adding the NBD2 construct, leading to a retrieval of nucleotide binding competence and activity in NBD1 (28).

The Prion Curing Agent GdmCl Inhibits ATP Hydrolysis in ClpB NBD1 but Not NBD2

It has been established by Walter and co-workers (18) that small amounts of GdmCl inhibit ATP hydrolysis in Hsp104, the ClpB homolog in yeast. With our system of the isolated domains in hand, we aimed to clarify whether the ATPase activity of ClpB NBD1 or NBD2 is influenced by GdmCl and performed steady state ATPase assays in the presence of different concentrations of GdmCl separately for NBD1-M(141–534) and NBD2(520–854). The results clearly show that only NBD1, but not NBD2, is affected by GdmCl (Fig. 4A). The ATPase turnover in NBD1 shows a strong dependence on the GdmCl concentration and is 5-fold reduced in the presence of 15 mm GdmCl, whereas ATP hydrolysis in NBD2 is not altered significantly. ATP hydrolysis in full-length ClpB is also inhibited by GdmCl; however, a significant fraction of the ATPase activity remains most likely due to the contribution of NBD2.

FIGURE 4.

Influence of GdmCl on the ATPase and disaggregation activity of ClpB. A, steady state (d)ATPase turnover rates of NBD1-M(141–534) wild type (green) and E209A (blue), NBD2(520–854) (red), and full-length ClpB (gray) in the presence of 0–15 mm GdmCl. Filled circles represent ATP hydrolysis data, and empty circles show dATP hydrolysis data (dATP lacks the 2′-OH group). ATP hydrolysis in NBD1, but not NBD2, is inhibited by low concentrations of GdmCl. NBD1-M(141–534) E209A is less active than the wild type; however, it is no longer affected by GdmCl. Hydrolysis of dATP is slower than for ATP; however, the inhibiting influence of GdmCl is decreased when the 2′-OH group is missing. From the hyperbolic fit of the (d)ATPase data, the K_D for GdmCl binding to NBD1-M(141–534) could be determined: K_D(GdmCl) = 2.3 ± 0.1 mm in the presence of ATP (green; filled) and K_D(GdmCl) = 8.3 ± 1.9 mm in the presence of dATP (green; empty). Fitting the ATPase data of full-length ClpB yielded K_D(GdmCl) = 2.0 ± 0.2 mm (gray). B, steady state ATPase turnover of NBD1-M(141–534) wild type for various concentrations of ATP in the presence of 15 mm GdmCl (green). Data fitting using the Hill equation yielded k_cat = 1.03 ± 0.03 min⁻¹, K_m = 0.30 ± 0.02 μm, and n = 1.9 ± 0.3. Steady state ATPase turnover of NBD1-M(141–534) E209A for various concentrations of ATP in the absence of GdmCl (blue) is also shown. Data fitting using the Hill equation yielded k_cat = 0.46 ± 0.01 min⁻¹, K_m = 0.94 ± 0.04 μm, and n = 2.9 ± 0.3. C, unfolding of NBD1-M(141–534) by GdmCl treatment. The change in molar ellipticity at 222 nm is plotted against the GdmCl concentration. The data were fitted assuming a two-state transition. The m value (m_d-n) is 3.63 ± 0.56 kJ mol⁻¹ m⁻¹. The obtained midpoint lies at [D]^50% = 3.57 ± 0.15 m, which is more than 200 times more GdmCl than used in the activity assays. D, chaperone-assisted reactivation of heat-aggregated α-glucosidase. The assay was performed as described under “Experimental Procedures.” The relative α-glucosidase activity (normalized against the positive control) is shown for different time points during the ClpB/DnaK/DnaJ/GrpE-assisted disaggregation reaction in the absence of GdmCl (red) and in the presence of 3 (orange), 6 (yellow), 9 (green), and 12 mm GdmCl (blue). Negative controls included only ClpB present (dark blue) and only DnaK/DnaJ/GrpE present (brown). E, the relative α-glucosidase activity (normalized against the positive control) at t = 120 min is plotted against the GdmCl concentration, showing that the disaggregation reaction is significantly impaired in the presence of GdmCl. The colors refer to the GdmCl concentrations in D. deg, degrees.

From the hyperbolic fit of the ATPase data, we determined the K_D for GdmCl binding to NBD1-M(141–534) in the presence of nucleotide: K_D(GdmCl) = 2.3 ± 0.1 mm. Experiments performed at higher salt concentration (300 mm KCl) and varied pH conditions (pH 5.5 and 9.5) showed similar results with K_D(GdmCl) values ranging from 1.2 mm at pH 5.5 to 7.1 mm in the presence of 300 mm KCl. This indicates that the GdmCl interaction is not abolished under different salt and pH conditions.

Furthermore, we tested the influence of GdmCl on the chaperone activity of ClpB by performing disaggregation assays in which heat-aggregated α-glucosidase is reactivated by ClpB and the co-chaperones DnaK, DnaJ, and GrpE (Fig. 4, D and E). In the presence of GdmCl, the disaggregation activity of ClpB is severely impaired, which is in agreement with both NBDs being required to be active for efficient chaperone activity (41).

GdmCl as a chaotropic agent is widely used to induce protein unfolding. To verify that the low concentrations of GdmCl used in our activity assays do not influence the structural stability of NBD1-M(141–534), we performed protein unfolding experiments with GdmCl observing the change in circular dichroism (Fig. 4C). Data analysis assuming simple two-state unfolding shows that the midpoint of the unfolding transition lies at [GdmCl] = 3.6 m, which is more than 200 times more GdmCl than the highest concentration used in the activity assays.

The Crystal Structure of NBD1-M(141–534) in Complex with ADP and GdmCl

To understand the molecular mechanism of ATPase inhibition by GdmCl, which appears to be specific for NBD1, structural information is essential. We therefore determined the crystal structure of ClpB NBD1-M(141–534) in complex with nucleotide and GdmCl. Crystals of NBD1-M(141–534) in the presence of 3 mm ADP and 10 mm GdmCl diffracted to 2.2 Å resolution. Structure determination was done by molecular replacement using residues 141–534 of the full-length ClpB structure (Protein Data Bank code 1QVR) by Lee et al. (21) as the search model (Table 1).

TABLE 1.

Data collection and refinement statistics

Data collection
Wavelength (Å)	0.9785
Space group	P3121
Unit cell dimensions
a, b, c (Å)	61.0, 61.0, 213.9
α, β, γ (°)	90, 90, 120
No. of molecules	1
Resolution (Å)	50–2.2
No. of unique reflectionsa	24,236 (2,926)
Rmeas (%)a,b	5.4 (39.1)
I/σ(I)a	19.5 (4.8)
Completeness (%)a	99.4 (99.1)
Redundancya	4.9 (5.2)
Wilson B-factor (Å2)	39.9
Refinement
Rworkc/Rfreed	22.5/26.6
No. of atoms
Protein	2,976
ADP	27
Gdm+	4
Cl−	1
Water	86
Average B-factors (Å2)
Protein	22.2
ADP	28.7
Gdm+	25.0
Cl−	37.5
Water	25.0
r.m.s.d.e
Bond length (Å)	0.0094
Bond angles (°)	1.1892
Ramachandran favored (%)	98.9
Ramachandran disallowed (%)	0
MolProbity score [percentile]	1.56 [98th]
Protein Data Bank entry code	4HSE

^a Values in parentheses are for highest resolution shell (2.3–2.2 Å).

^b R_meas = Σ_h[n_h/(n_h − 1)]^1/2Σ_i|Î_h − I_h,i|/Σ_hΣ_iI_h,i where Î_h is the mean intensity of symmetry-equivalent reflections and n_h is the redundancy.

^c R_work = Σ‖F_o| − |F_c‖/Σ|F_o|.

^d R_free is calculated like R_work using 5% of the data that are excluded from refinement.

^e Root mean square deviation.

The overall structure of NBD1-M(141–534) is very similar to the equivalent part in the published structure of full-length ClpB. The main deviation is caused by a slight bending of the coiled coil toward the tip of the M domain. This area (residues 421–427) showed weaker electron density and could not be fully modeled most likely because of the presence of several alternative conformations. The orientation of the nucleotide in the active site is almost identical to the full-length structure, which contains the non-hydrolyzable ATP analog AMPPNP instead of ADP. When the crystallization mixture was supplemented with GdmCl, additional F_o − F_c density of triangular, flat shape was observed in the active site that fits the size of a Gdm⁺ ion (Fig. 5A). A network of polar interactions with both the protein and the nucleotide stabilizes the Gdm⁺ ion. The main interaction between Gdm⁺ and the protein is formed with the negatively charged side chain of glutamate 209 (2.7 Å). Additionally, Gdm⁺ forms a hydrogen bond with the 2′-OH group of ADP (2.8 Å). Further hydrogen bonds are formed between Gdm⁺ and the backbone carbonyl groups of proline 171 (3.0 Å) and aspartate 170 (2.9 Å). This residue (corresponding to aspartate 184 in Hsp104 from S. cerevisiae) was proposed as the main interaction partner for Gdm⁺ in previous publications (19, 20). However, we do not observe any direct side chain interaction with aspartate 170.

FIGURE 5.

The crystal structure of NBD1-M(141–534) in complex with ADP and GdmCl. A, the Gdm⁺ ion binds in the active site of ClpB NBD1, forming contacts with the side chain carboxyl group of glutamate 209 and the backbone carbonyl groups of aspartate 170 and proline 171. The Gdm⁺ ion interacts directly with the nucleotide via a hydrogen bond with the 2′-OH group of ADP. The F_o − F_c electron density map (contoured at 3σ) was obtained after initial phasing prior to modeling Gdm⁺ and ADP. The side chain of lysine 204, the catalytically essential Walker A residue, contacts the β-phosphate of ADP and is not involved in the Gdm⁺ interaction. Dotted lines refer to hydrogen bonds or ionic interactions. Distances are given in Å. B, structure alignment of ClpB NBD1 (green) and NBD2 (red) active sites. The P-loop regions, residues 195–215 of the NBD1-M structure (Protein Data Bank code 4HSE) and 592–612 of the published full-length structure (Protein Data Bank code 1QVR), respectively, were chosen for superposition. The equivalent residue to glutamate 209 in NBD1 is lysine 606, an amino acid of opposite charge, in NBD2. This together with steric restrictions explains the specificity of GdmCl binding to NBD1.

GdmCl Binding Interferes with the Role of an Essential Glutamate (Glu-209)

According to the crystal structure, binding of Gdm⁺ to ClpB NBD1 is mediated primarily by glutamate 209. This amino acid is conserved throughout different homologs of ClpB from several species including Hsp104 from S. cerevisiae in which the prion curing effect upon GdmCl treatment has been observed (17). We generated an NBD1-M(141–534) variant in which the glutamate is replaced by an alanine (E209A) and tested for GdmCl dependence of ATP hydrolysis. The steady state ATPase activity of NBD1-M(141–534) E209A is 10-fold reduced and no longer affected by increasing concentrations of GdmCl (Fig. 4, A and B), indicating that glutamate 209 is indeed important for the GdmCl interaction.

When introducing the E209A mutation into the full-length protein, we observed a 3–4-fold reduced disaggregation activity. This together with the 10-fold reduced ATPase activity of NBD1-M(141–534) E209A indicates that glutamate 209 is an essential residue possibly involved in positioning the nucleotide or stabilizing an intermediate state during the hydrolysis cycle. We therefore propose that the presence of Gdm⁺ in the active site interferes with the role of glutamate 209.

The crystal structure also revealed a direct interaction between Gdm⁺ and the nucleotide via the 2′-OH group of the ribose. Thus, we performed activity assays using dATP, which lacks the 2′-OH group, and tested for GdmCl dependence of the hydrolysis rates (Fig. 4A). In the absence of GdmCl, hydrolysis of dATP by NBD1-M(141–534) was 4 times slower compared with ATP. However, adding up to 15 mm GdmCl caused only a 2-fold decrease in dATPase activity, showing that dATP hydrolysis is influenced less by GdmCl. This is further supported by a 4-fold reduced GdmCl binding affinity of K_D(GdmCl) = 8.3 ± 1.9 mm in the presence of deoxynucleotides. In the presence of the mutation E209A, the dATPase activity was just 2-fold reduced compared with the ATPase activity with no further changes caused by GdmCl.

When comparing ClpB NBD1 with NBD2 by both amino acid sequence and structural alignments, it becomes apparent that the equivalent residue to glutamate 209 in NBD1 is lysine 606, an amino acid of opposite charge, in NBD2 (Fig. 5B). This together with the steric restrictions explains why NBD2 cannot bind the Gdm⁺ ion so that ATP hydrolysis in NBD2 is not affected by GdmCl.

GdmCl Modulates the Nucleotide Binding Properties of ClpB NBD1

The direct hydrogen bonding interaction of the Gdm⁺ ion with the nucleotide observed in the crystal structure led us to hypothesize that the nucleotide binding parameters of ClpB NBD1 might be influenced in the presence of GdmCl. All rate constants and nucleotide binding affinities obtained from the experiments described below are listed in Table 2.

We performed stopped flow experiments to characterize binding of MANT-ADP to NBD1-M(141–534) in the presence and absence of GdmCl. MANT-ADP is a mixture of two isomers carrying the fluorescent MANT group either on the 2′- or 3′-position of ADP. It was shown previously that slow kinetic phases due to 2′/3′-MANT isomerization could occur when MANT-ADP was used in stopped flow experiments and that this could be avoided by using MANT-dADP (31). However, for studying the GdmCl interaction, the presence of the 2′-OH group was essential. MANT-dADP that lacks the 2′-OH group was used as a control. 2 μm NBD1-M(141–534) was mixed with different concentrations of MANT-nucleotide in the presence and absence of GdmCl. The data were analyzed as described in Fig. 2A. The fitted rate constants plotted against the nucleotide concentration are shown in Fig. 6, A and B. The slopes of the linear functions representing the on-rate of nucleotide binding are only slightly increased in the presence of GdmCl. However, the y axis intercept representing the off-rate changes significantly for MANT-ADP binding in the presence of GdmCl. This indicates that MANT-ADP binding in the presence of GdmCl is stronger due to a lower off-rate. The off-rate for MANT-dADP, which lacks the 2′-OH group, is not changed significantly by GdmCl. To verify these results, we also performed dissociation experiments to determine the off-rates more accurately. As described above, 2 μm NBD1-M(141–534) was incubated with 20 μm MANT-nucleotide and subsequently mixed with an excess of unlabeled ADP. The extracted off-rate for MANT-ADP binding decreases from 4.0 s⁻¹ in the absence of GdmCl to 1.2 s⁻¹ in the presence of 10 mm GdmCl, leading to a 4-fold improved binding affinity for MANT-ADP in the presence of GdmCl. Dissociation experiments performed at higher salt concentration (300 mm KCl) and varied pH conditions (pH 5.5 and 9.5) gave similar results. The decrease in MANT-ADP off-rate due to GdmCl was observable under all conditions, indicating again that GdmCl binding is not impaired by different salt and pH conditions. The strongest effect was observed for pH 9.5 with k_off = 8.6 s⁻¹ in the absence of GdmCl and k_off = 1.5 s⁻¹ in the presence of 10 mm GdmCl.

FIGURE 6.

Nucleotide binding behavior of NBD1-M(141–534) in the presence of GdmCl. A, MANT-ADP binding to NBD1. NBD1-M(141–534) and MANT-ADP were mixed directly in the presence (filled circles) or absence (empty circles) of 10 mm GdmCl. The fitted rate constants of the kinetic traces were plotted against the nucleotide concentration. Although the slope of the linear function (corresponding to the on-rate) is only slightly increased, the y axis intercept (corresponding to the off-rate) is significantly lower in the presence of GdmCl. The off-rates were determined separately by a dissociation experiment as described in the text, confirming a 4-fold increased binding affinity for MANT-ADP in the presence of GdmCl due to a slower off-rate. B, MANT-dADP binding to NBD1. The stopped flow experiments were performed as described in A using MANT-dADP, which lacks the 2′-OH group, instead of MANT-ADP. Nucleotide binding in the presence of 10 mm GdmCl (filled circles) is only slightly increased compared with the absence of GdmCl (empty circles), indicating that the 2′-OH group is essential for the GdmCl interaction. C, fluorescence titrations to obtain K_D(ADP) in the absence of GdmCl (empty circles; data from Fig. 2B) and in the presence of 10 mm GdmCl (filled circles). ADP binding is enhanced 6-fold in the presence of GdmCl. D, fluorescence titrations to obtain K_D(ATP) in the absence of GdmCl (empty circles; data from Fig. 2B) and in the presence of 10 mm GdmCl (filled circles). ATP binding is enhanced 12-fold in the presence of GdmCl. The measured fluorescence was normalized for direct comparison of different titrations in one graph. All nucleotide binding parameters obtained from the experiments shown in this figure are given in Table 2. a.u., arbitrary units.

Next, we wanted to check how GdmCl alters the binding affinities of the unlabeled, physiologically relevant nucleotides ADP and ATP. Thus, we performed fluorescence equilibrium titrations in which 2 μm protein was incubated with 20 μm MANT-dADP and subsequently titrated with either ADP or ATP (ATP titrations were done in the presence of phosphoenolpyruvate and pyruvate kinase). The titration curves were analyzed using the cubic equation for competing ligands (39). The reference K_D values for MANT-dADP were taken from the stopped flow experiments described above. In the presence of 10 mm GdmCl, K_D(ADP) for NBD1-M(141–534) is 6 times lower than in the absence of GdmCl; K_D(ATP) is even 12 times lower (Fig. 6, C and D, and Table 2). This clearly shows that nucleotide binding to NBD1-M(141–534) is significantly enhanced in the presence of small amounts of GdmCl. This result is supported by steady state ATPase measurements at different ATP concentrations in the presence of 15 mm GdmCl, which yielded a significantly reduced K_m value (Fig. 4B).

Furthermore, we performed gel filtration experiments with subsequent SLS analysis with GdmCl present in the running buffer. The nucleotide-induced oligomerization of NBD1-M(141–534) was more pronounced than in the absence of GdmCl, showing that enhanced nucleotide binding due to GdmCl also triggers stronger oligomerization (Fig. 3B). This is in agreement with previous experiments performed with full-length Hsp104 by Walter and co-workers (18).

The oligomerization behavior of the mutant variant NBD1-M(141–534) E209A was also characterized by DLS and SLS measurements, and there were no significant differences compared with the wild type in the absence of GdmCl. However, adding GdmCl did not enhance the nucleotide-induced oligomerization as observed for the wild type, further supporting that Glu-209 is involved in GdmCl binding. Despite causing reduced ATPase activity, the mutation E209A does not impair nucleotide binding; K_D(MANT-dADP) is even lower than for the wild-type protein. However, the binding affinities for ADP and ATP are not modulated by GdmCl for NBD1-M(141–534) E209A in agreement with the hypothesis that Glu-209 is essential for the interaction of Gdm⁺ and the protein.

In sum, these experiments show that in the presence of GdmCl, which binds specifically to the active site of ClpB NBD1, nucleotide binding affinities are increased significantly mainly due to lower off-rates. This effect is mediated by a direct interaction of the Gdm⁺ ion with the 2′-OH group of the nucleotide.

DISCUSSION

It has been reported recently that low concentrations of GdmCl influence the catalytic activity of the Hsp100 chaperones Hsp104 and ClpB (18, 20). In vivo studies showed that both thermotolerance and prion propagation in yeast are impaired by GdmCl (17, 42). In vivo experiments with E. coli showed no significant effect on thermotolerance upon GdmCl treatment (20). However, the removal of intracellular aggregates accumulated after heat shock was indeed affected. Furthermore, it has been shown that E. coli cell growth is impaired at increasing GdmCl concentrations (43).

Here, we present mechanistic details explaining how the prion curing agent GdmCl inhibits Hsp100 chaperones. We made use of the Hsp104 homolog ClpB from T. thermophilus, the only Hsp100 chaperone for which high resolution structural information and well characterized constructs of the two separate NBDs are available (27). The basis for our study was the finding that the isolated NBD1 of ClpB also is a fully active ATPase once an additional helix (residues 520–534) located directly after the M domain insertion is present in the construct. Although not directly interacting with the nucleotide, these 15 residues complete the C-terminal helical bundle (also named small domain) of NBD1, which evidently has to be intact to ensure oligomerization and nucleotide binding competence. This ClpB construct, NBD1-M(141–534), allowed for the first time a separate measurement of nucleotide binding and ATPase properties of NBD1 as performed previously for NBD2 using the construct NBD2(520–854) (31). As already observed for ClpB NBD2, NBD1 discriminates strongly between di- and triphosphate, binding ADP about 100-fold more strongly than ATP, which seems to be a rather disadvantageous feature for an ATPase. However, nucleotide binding properties might be modified due to conformational changes during the different steps of the ATPase cycle. Compared with a previous study that estimated K_D(ADP) and K_D(ATP) of ClpB NBD1 from T. thermophilus to be around 40 and 50–150 μm, respectively, by analyzing different protein proteolysis patterns, our direct measurements showed a significantly stronger discrimination between ADP and ATP (11).

The steady state ATP hydrolysis rates measured for NBD1-M(141–534) were similar to those of NBD2(520–854), demonstrating again that both NBDs of ClpB are equally active and contribute to the overall ATPase activity. However, the activity of full-length ClpB is lower than the sum of both independent NBDs, indicating that allosteric regulation takes place.

The high Hill coefficients obtained from fitting the steady state ATPase data suggest that NBD1 is active as an oligomeric species, which is further supported by the observed coupling of nucleotide binding and oligomerization. These findings are in agreement with a previous study in which we proposed that NBD1 is regulated in a trans-acting fashion (28).

Next, we were interested in what changes occur in the presence of the prion curing agent GdmCl and whether NBD1, NBD2, or both NBDs are affected by GdmCl. ATP hydrolysis only in NBD1, but not NBD2, is inhibited by increasing amounts of GdmCl, indicating that GdmCl interacts specifically with NBD1. The steady state ATPase activity of full-length ClpB is also inhibited, which is in contrast to a recent study showing that ClpB from E. coli possesses severalfold increased ATPase activity in the presence of GdmCl (20). However, as discussed in detail further below, these differences can be explained by a slightly different modulation of nucleotide binding properties by GdmCl in ClpB from E. coli.

The high resolution crystal structure of NBD1-M(141–534) in complex with ADP and GdmCl shows that the Gdm⁺ ion binds in the active site of NBD1, forming contacts with both the protein and the nucleotide. The aspartate residue proposed previously to be essential for the GdmCl interaction (Asp-184 in Hsp104 from S. cerevisiae corresponds to Asp-170 in ClpB from T. thermophilus (19, 20)) is indeed part of the Gdm⁺ binding site providing a hydrogen bond to the Gdm⁺ ion via its backbone carbonyl group, but there is no direct side chain interaction observable for Asp-170. However, side chain variations can also affect the backbone conformation, which might have caused the observed effects in previous studies using mutational analysis. Furthermore, we cannot exclude that there are alternative conformations in which the side chain is involved in GdmCl binding.

We identified a conserved glutamate residue (Glu-209) as the main interaction partner of NBD1 and Gdm⁺ and analyzed the E209A mutant to verify this result. Nucleotide binding properties and ATPase rates of NBD1-M(141–534) E209A are not affected by GdmCl. The conserved glutamate appears to be essential because ATP hydrolysis in the case of the E209A mutation is 10-fold slower compared with the wild-type protein. However, unlike the residues of the highly conserved Walker A and Walker B motifs, Glu-209 is not directly involved in the catalysis because it is located too far away from the phosphates of the nucleotide. We therefore suggest that it rather stabilizes an intermediate state and positions the nucleotide and/or other essential residues properly to allow an efficient hydrolysis. We propose that the binding of Gdm⁺ impairs this function, thereby causing reduced ATPase activity in the presence of GdmCl.

Furthermore, the direct interaction between the Gdm⁺ ion and the 2′-OH group of the bound nucleotide causes a modulation of nucleotide binding properties of NBD1 that ultimately influences the ATP hydrolysis cycle. Possible consequences of enhanced nucleotide binding are illustrated in a simplified energy diagram describing an ATPase reaction (Fig. 7) where substrate binding is followed by the hydrolysis step and the final product release. The energy barriers, ΔG^≠, between these states are directly related to reaction rates. A stabilization of the substrate-bound state may therefore decelerate the hydrolysis step. A stabilization of the product-bound state may decelerate the product release step. Of course, this is just one possible scenario; also transition states might be stabilized in the presence of GdmCl, which would then reduce ΔG^≠ and accelerate certain steps. This simplified model can be expanded by taking into consideration that nucleotide binding and oligomerization are coupled. The GdmCl-induced increase in nucleotide binding affinities causes a stabilization of oligomeric NBD1-M species. Assuming that the interface between adjacent NBD1 subunit (where the catalytic action takes place) closes upon ATP binding and has to open again after hydrolysis to release ADP and phosphate (which does not necessarily involve a complete dissociation of subunits), it seems reasonable to hypothesize that GdmCl inhibits a step in the ATPase cycle that is associated with product release.

FIGURE 7.

Energy diagram of a simplified ATPase reaction. The energy diagram schematically depicts an ATPase reaction consisting of three steps: substrate binding, hydrolysis, and product release. The energy barriers, ΔG^≠, for the individual steps (represented by vertical arrows between ground and transition states) are directly related to the reaction rates. The rate-limiting step of the overall reaction is characterized by the highest ΔG^≠. A possible modulation of the energy landscape due to the influence of GdmCl is shown in red. The higher nucleotide binding affinities in the presence of GdmCl stabilize both the ATP- and ADP-bound state (or substrate- and product-bound state, respectively), thereby increasing ΔG^≠ for both the hydrolysis and product release steps (assuming that the transition states are not changed to the same extent). Depending on which step is initially rate-limiting, the overall reaction can be significantly inhibited by modulated nucleotide binding affinities.

There are other examples of ATPase inhibitors that form ternary complexes with enzyme and product leading to slower product release. Small molecule inhibitors of the human kinesin spindle protein (or Eg5), a mitotic motor protein and promising target for the development of new anticancer drugs, bind to an allosteric site 12 Å from the catalytic center (44). Kinetics studies utilizing MANT-labeled nucleotides showed that ADP release is inhibited in the presence of the inhibitors monastrol, S-trityl-l-cysteine, and ispinesib (45–47). However, in those cases, there is no direct interaction between nucleotide and inhibitor as observed in our crystal structure but rather an allosteric modulation of the catalytic site due to inhibitor binding.

The Hsp100 chaperones ClpB and Hsp104 share 45% overall sequence identity (calculated by LALIGN³) and possess highly conserved AAA+ domains and active sites. Using only the isolated N-terminal nucleotide binding domain of ClpB, we could observe the same GdmCl-induced effects on ATPase activity, nucleotide binding, and oligomerization described previously for Hsp104 by Walter and co-workers (18), allowing us to transfer the mechanistic insights obtained from studying ClpB to Hsp104 where the prion curing effect in yeast upon GdmCl treatment was observed (17). Based on structural and functional data, we conclude that the prion curing agent GdmCl specifically inhibits NBD1 of Hsp100 chaperones, but not NBD2, by binding in the active site and modifying the nucleotide binding affinities via a direct interaction with the 2′-OH group of the nucleotide. We propose that two mechanistic effects contribute to the observed GdmCl-specific inhibition of Hsp100 chaperones. First, GdmCl impairs the function of an essential glutamate. Second, GdmCl-induced changes in nucleotide binding properties, which go along with a modulation of the oligomerization behavior of NBD1, interfere with the ATPase cycle.