Visualization of a slow, ATP-induced structural transition in the bacterial molecular chaperone DnaK.

Recent reports have shown that the binding of ATP to a 70-kDa molecular chaperone induces a rapid global conformational transition from a "high affinity" state to a "low affinity" state, where these states are defined by tight and weak binding to (poly)peptides, respectively. To complete the activity cycle, a chaperone molecule must ultimately return to the high affinity state. In this report, this return to the high affinity state was studied using a chemical cross-linking assay in conjunction with SDS-polyacrylamide gel electrophoresis. The basis for this assay is that in the absence of nucleotide or in the presence of ADP, conditions that stabilize the high affinity state, cross-linking of the Escherichia coli molecular chaperone DnaK yielded two monomeric forms, with apparent molecular masses of 70 kDa (77%) and 90 kDa (23%), whereas cross-linking yielded only the 70-kDa monomeric form in the presence of ATP. This ATP-dependent difference in cross-linking was used to follow the kinetics of the low affinity to high affinity transition under single turnover conditions. The rate of this transition (kobs = 3.4 (+/-0.6) x 10(-4) s-1 at 25 degrees C) is almost identical to the reported rate of ATP hydrolysis (khy = 2.7 (+/-0.7) x 10(-4) s-1 at 22 degrees C). These results are consistent with a two-step sequential reaction where rate-limiting ATP hydrolysis precedes the conformational change. Models for the formation of two cross-linked DnaK monomers in the absence of ATP are discussed.

The highly conserved and ubiquitous 70-kDa family of molecular chaperones, which include stress induced (Hsp70) 1 and constitutively expressed (Hsc70) variants, promote protein-protein interactions via the of coupling ATP binding and hydrolysis to selective substrate binding and release. Molecular chaperones use this activity cycle to perform diverse biological processes such as the stabilization of partially unfolded and nascent proteins, protein translocation, and protein complex assembly and disassembly under normal growth and stress conditions (for reviews, see Refs. [1][2][3]. The underlying molecular events that enable the coupling of ATP binding and hydrolysis to selective substrate binding and release are poorly understood. The three-dimensional structure of the two separate 70-kDa chaperone functional domains have been determined. The NH 2 -terminal domain of Hsc70 binds nucleotide in the base of a cleft formed by two subdomains (4). Two K ϩ ions and a Mg 2ϩ ion are cofactors in the ATPase reaction by interacting with nucleotides in this cleft (5). The requirement for K ϩ is specific, because when K ϩ is replaced by Na ϩ , (i) the rate of Hsc70catalyzed ATP hydrolysis is five times slower (6), and (ii) DnaK-protein complexes do not dissociate in the presence of ATP (7). The COOH-terminal domain of DnaK (residues 389 -607) binds a peptide substrate in a channel formed by the loops from a ␤-sandwich (8). The molecular mechanism which couples the activities of the two domains is not understood. Several lines of evidence indicate that ATP induces a global structural transition in a 70-kDa chaperone molecule from a high affinity state to a low affinity state (9 -13). The high affinity state tightly binds (poly)peptides, whereas the low affinity state weakly binds (poly)peptides. Significantly, the induction of this structural transition is a consequence of ATP binding and not hydrolysis, because the rate of the transition (0.7 s Ϫ1 ) (14) is much faster than chaperone-catalyzed ATP hydrolysis (0.003-0.0005 s Ϫ1 ) (7,15,16). It appears that ATP hydrolysis is involved in the reverse transition, that is, the return from the low affinity state to the high affinity state (12,14,17). Small-angle x-ray scattering experiments conducted on recombinant bovine Hsc70 have indicated that the reverse reaction occurs in at least two steps, where the hydrolysis of ATP is followed by rate-limiting product release (14).
In this report, a chemical cross-linking/SDS-PAGE assay was used to visualize the slow transition from the low affinity state to the high affinity state of the Escherichia coli 70-kDa molecular chaperone DnaK under single turnover conditions. The results are consistent with a two-step sequential reaction where ATP hydrolysis is rate-limiting rather than product release, and ATP hydrolysis induces a second reaction, a structural transition in the chaperone molecule. This assay can also be used to probe the effect of other parameters such as salt, temperature, and even cochaperones, on the kinetics of the transition.

EXPERIMENTAL PROCEDURES
DnaK Purification-DnaK was isolated from the E. coli strain RLM893 (18) as described previously (19). The protein was maintained in a K ϩ /HEPES sample buffer (25 mM HEPES, 50 mM KCl, 5 mM MgCl 2 , 5 mM 2-mercaptoethanol at pH ϭ 7.0) and stored at 4°C prior to use. Protein concentration was determined by the Bio-Rad assay using bovine serum albumin as a standard and verified spectrophotometrically (⑀ 280 ϭ 15.8 ϫ 10 3 M Ϫ1 cm Ϫ1 ) (20). SDS-PAGE analysis demonstrated that the DnaK preparations were Ͼ95% pure. For some of the crosslinking and fluorescence experiments DnaK was dialyzed into a Na ϩ / HEPES sample buffer (25 mM HEPES, 50 mM NaCl, 5 mM MgCl 2 , 5 mM 2-mercaptoethanol at pH ϭ 7.0). All reagents were purchased from Sigma.
Cross-linking/SDS-PAGE-Samples of DnaK were cross-linked for 2 min at 25°C with glutaraldehyde (final concentration: 11 mM), and the reaction was stopped by the addition of an excess of glycine (final concentration: 200 mM). Cross-linked samples were loaded onto either a 4 -12% linear gradient denaturing polyacrylamide gel or a 10% denaturing polyacrylamide gel, electrophoresed, and stained with Coomassie Brilliant Blue G-250. Gel lanes were scanned with a Molecular Dynamics densitometer to determine the relative amount of species in each lane.
Fluorescence Measurements-A Photon Technology Inc. (South Brunswick, NJ) StrobeMaster lifetime spectrometer with a SE-900 steady state fluorescence option was used to monitor the slow increase in tryptophan fluorescence in the single turnover experiments. In these experiments the instrument was used in a time base mode, with ex ϭ 295 (3-nm bandwidth) and em ϭ 340 nm (5-nm bandwidth). Samples were maintained in a quartz cuvette (1-cm path length) with constant stirring and with temperature control via an external circulating heating/cooling bath (⌬T ϭ Ϯ 0.2°C). Sample temperature was verified using a hand held thermocouple, which was placed directly into the sample.
Data Analysis and Reproducibility of Measurements-The equilibrium and kinetic data were fit to the equations in the text using the program KaleidaGraph (Synergy, Reading, PA). All experiments were repeated three to four times.
The value for the equilibrium dissociation constant for ATP binding to DnaK has an error of 20%, while the values for the kinetic constants have 5 Ϫ 10% error.

RESULTS
In the course of conducting cross-linking experiments to determine the conditions that promote the oligomerization of DnaK (21), we have noticed that the treatment of DnaK (20 M) with glutaraldehyde produces two distinct monomer bands as well as dimer and trimer bands (Fig. 1A). On a 4 -12% polyacrylamide denaturing gel, one monomer band is centered at a molecular mass of 70-kDa, as expected, while the upper monomer band is centered at an apparent molecular mass of 90-kDa. The broad protein bands are probably due to variations in the amount of intramolecular cross-linking (22,23). Previously, two or more chaperone monomer bands have been reported and were attributed to heterogeneous cross-linking (22,24). In this report, the occurrence of these two monomer bands was exploited to visualize ATP-induced conformational transitions in DnaK.
Effect of Nucleotide on Cross-linking- Fig. 1B shows the effect of nucleotides on the cross-linking of DnaK in the monomer region. At the concentration of DnaK (6 M) used in these experiments, 85-90% of the protein is monomeric (21). Since 10% polyacrylamide gels gave a slightly better separation of the two monomer bands, 10% gels were used in these experi-ments. 2 Uncross-linked DnaK appeared as a single monomer band (lane 1), whereas DnaK that was cross-linked in the absence of nucleotide appeared as two monomer bands (lane 2). When DnaK was cross-linked in the presence of ADP (2.0 mM), the two monomer bands were still present (lane 3). In contrast, cross-linking in the presence of ATP (2.0 mM) abolished the upper monomer band (lane 4). Other nucleotide triphosphates, GTP, CTP, and UTP, also abolished the upper monomer band (data not shown). On the other hand, cross-linking in the presence of AMP-PNP (2.0 mM), a nonhydrolyzable analog of ATP, had no effect on the upper band (lane 5). On the basis of these results, we conclude that the abolition of the upper monomer band is a consequence of a highly specific conformational change in DnaK, induced by the action of hydrolyzable nucleotide triphosphates. The ␥-phosphate group -O-P ␥ O 3 2Ϫ of the nucleotide triphosphate is required for the induction of the conformational change since the nucleotide triphosphate AMP-PNP, where the ␥-phosphate group is -NH-P ␥ O 3 2Ϫ , did not induce the conformational change. Possibly the nitrogen atom linkage in AMP-PNP modifies the coordination of Mg 2ϩ , which in turn prevents the proper docking of AMP-PNP within the ATP binding site of DnaK.
Two models are proposed to explain the above results ( Fig.  2). In both models: (i) E and E** stand for high and low affinity chaperone states, respectively; (ii) reaction intermediates have been ignored; (iii) in the absence of ATP, all DnaK molecules populate the E state; (iv) and the binding of hydrolyzable nucleotide triphosphates to the E state drives the conversion to the E** state. Model 1 is a two-state mechanism defined by two distinct chaperone conformations, the E and E** states. It is postulated that certain structural features of the E state (and the E-ADP state) promote heterogeneous cross-linking. The upper monomer band is probably due to a population of DnaK monomers with a large number of intramolecular cross-links, which tend to inhibit SDS loading and therefore lead to an apparent molecular mass higher than 70 kDa; whereas, the lower band is probably due to a population of DnaK monomers with a relatively small number of intramolecular cross-links. ATP binding to the E state produces the E** state, which happens to cross-link homogeneously, yielding a single monomer band. Model 2 is a three-state mechanism defined by two distinct E state isomers (E a N E b ) and one E** state. Since there are two E state isomers, cross-linking produces two bands. Upon the addition of ATP, either both E states convert to the E** state directly, or one state converts to the E** state and the equilibrium readjusts. Both paths lead to the E** state, which yields one band on cross-linking. An alternative interpretation of the gel data that does not assume the validity of (iii) is discussed below.
Chromatography-If Model 2 is correct it should be possible to separate the E state isomers. In an attempt to separate the putative isomers, DnaK was electrophoresed on both a 10 -15% and a 15-20% linear gradient nondenaturing polyacrylamide gel, and, in each case, a single band was observed. Although unsuccessful in our attempts to separate these putative isomers, Model 2 should not be ruled out, as discussed below.
Equilibrium Experiments-The thermodynamics of ATP binding to DnaK were investigated by exploiting the different cross-linking patterns of DnaK in the absence and presence of ATP. Samples of DnaK (6.0 M) with varying amounts of ATP were incubated for 1 min at 25°C and then cross-linked. Fig.  3A shows the relative amounts of the two monomer bands before and after the addition of ATP. The upper monomer band was abolished when [ATP]/[DnaK] Ն 1.6, indicating that the population of the E state was also abolished.
The data in Fig. 3A were used to estimate the apparent equilibrium dissociation constant (K d ) for ATP binding to DnaK according to Model 1 (Fig. 2). Each lane in the gel was scanned to determine the fraction of the upper monomer band (f E ), which is an indicator of the amount of the E state, and then f E was plotted as a function of the total concentration of ATP (Fig. 3B). The apparent equilibrium dissociation constant for the reaction E**-ATP N ϩ ATP is related to f E according to Equation 1,  (17).
Single Turnover Experiments-Since the upper monomer band is abolished at near stoichiometric concentrations of ATP, the assay is an ideal way to monitor the kinetics of structural transitions in a DnaK monomer that occur during a single turnover. Single turnover experiments at 25°C were conducted by adding a stoichiometric amount of ATP (6.0 M) to a solution of DnaK (6.0 M) in a K ϩ /HEPES buffer (Fig. 4A). Aliquots were removed at the indicated times and cross-linked as out- The lanes were scanned with a densitometer to determine the relative amounts of the two DnaK monomers, and the results are plotted in Fig. 4B. The reappearance of the signal from the upper monomer band followed single exponential kinetics according to S(t) ϭ A(1 Ϫ e Ϫkobst ) ϩ B, with k obs ϭ 3.4 (Ϯ0.6) ϫ 10 Ϫ4 s Ϫ1 . Significantly, this rate is identical within experimental error to the reported steady state rate of DnaK-catalyzed ATP hydrolysis (k hy ϭ 2.7 (Ϯ0.7) ϫ 10 Ϫ4 s Ϫ1 at 22°C) (25). On the basis of these gel kinetic results, we conclude that (i) the ATP-induced transition from the high affinity state to the low affinity state is fast (t1 ⁄2 Ͻ 2 min) in the presence of K ϩ ions, therefore this transition is due to ATP binding and not hydrolysis; and (ii) the reverse transition (E** 3 E) occurs at the same rate as DnaK-catalyzed ATP hydrolysis.
Complementary single turnover experiments were conducted to see whether the reported slow change in tryptophan fluorescence (14,17) matched the slow conformational transition detected in these gel experiments. ATP induced changes in the tryptophan fluorescence of DnaK on two distinctly different time scales: upon the addition of ATP, a 15% decrease in fluorescence occurred within 1 min and then the decrease was slowly reversed over the course of an hour (Fig. 5). Specifically, the slow increase in fluorescence followed single-exponential kinetics, with k obs ϭ 3.0 (Ϯ0.3) ϫ 10 Ϫ4 s Ϫ1 . For comparison, the gel kinetic data from Fig. 4B are superimposed on the fluorescence data. On the basis of the similarity of the two kinetic curves, we conclude that both methods monitor the low affinity state to high affinity state transition (E** 3 E) and that the change in tryptophan fluorescence occurs in concert with the conformational transition.
Since Na ϩ ions inhibit the high affinity state to low affinity state transition (7), Na ϩ ions should also inhibit the ATPinduced elimination of the upper monomer band. To test this idea, single turnover gel-and fluorescence-detected experiments were also conducted in a Na ϩ /HEPES buffer. Using the gel assay we found that Na ϩ ions almost completely prevented the ATP-induced elimination of the upper monomer band (Fig.  5, inset). Similarly, Na ϩ ions both attenuated the initial reduction in fluorescence upon the addition of ATP and retarded the rate of reappearance of the fluorescence. That Na ϩ ions inhibit the ATP-induced abolition of the upper monomer band is consistent with our claim that the abolition of the upper monomer band is an indicator of the high affinity state to low affinity state transition.

DISCUSSION
The gel-and fluorescence-detected single turnover experiments showed that ATP induces a rapid conformational change in a molecule of DnaK. For example, the addition of a stoichiometric amount of ATP to DnaK lead to the elimination of the upper monomer band (Fig. 4A) within 2 min or less. Similarly, the addition of a stoichiometric amount of ATP to DnaK resulted in a rapid reduction in fluorescence (Fig. 5). Since the rapid conformational change and the rapid spectral change occur much faster than the steady state rate of ATP hydrolysis (25), and because no rapid burst phase of ATP hydrolysis has been detected in any single turnover experiments (7), we conclude that the conformational change and the attending spectral change are due to ATP binding to DnaK. This conformational change is consistent with the high affinity state to low affinity state transition (E 3 E**) (7,15,16).
The major finding in this study was that the gel-detected conformational change and the change in tryptophan fluorescence occur at exactly the same rate (3 ϫ 10 Ϫ4 s Ϫ1 ), the rate of DnaK-catalyzed ATP hydrolysis. The simplest explanation of these results is that ATP hydrolysis and the conformational change in the chaperone occur simultaneously, according to Reaction 1.
However, it is unlikely that the cleavage of the ␥-phosphate group of ATP and a global conformational change in DnaK occur at precisely the same instant. A more likely explanation is that ATP hydrolysis precedes the conformational change, according to Reaction 2.
We therefore attempted to fit the single turnover kinetic data to Equation 2, which describes product formation according to the above two-step mechanism, where S(t) is the signal due to the product, A is the amplitude, k hy and k c are the first-order rate constants defined in Reaction 2, and B is the offset. A similar function has been used to fit single turnover data from fluorescence experiments (14,17). Unfortunately, due to the small number of data points, the gel kinetic data (Fig. 4B) could not be reasonably fit to Equation 2. On the other hand, the large number of data points in the fluorescence traces allowed those traces (Fig. 5) to be fit to Equation 2. The best fit curves were nearly superimposable on the data, and the best fit values for k hy and k c were 3.0 (Ϯ0.2) ϫ 10 Ϫ4 s Ϫ1 and 1.9 (Ϯ0.5) ϫ 10 Ϫ3 s Ϫ1 , respectively. It is noteworthy that when ATP hydrolysis is rate-limiting (k hy Ͻ k c ), Equation 2 simplifies to S(t) ϭ A(1 Ϫ e Ϫkhyt ) ϩ B, which explains why the single turnover data followed single exponential kinetics and why k obs ϭ k hy . In conclusion, the single turnover kinetic data from both assays are consistent with Reaction 2, where ATP hydrolysis occurs first and is rate-limiting. 3 On the basis of this work, we conclude that the mechanistic details of how the bacterial Hsp70 and the eukaryotic cytosolic Hsc70 couple ATP hydrolysis to structural transitions are different. Single turnover experiments on bovine Hsc70 have revealed the same two-step mechanism as shown in Reaction 2 (14); however, the rate-limiting step for this eukaryotic cytosolic Hsc70 is not the first step, ATP hydrolysis, it is the second step, product release. This mechanistic difference is probably related in some way to the fact that the bacterial Hsp70 chaperone DnaK depends on the cochaperones DnaJ and GrpE to accelerate ATP hydrolysis and ADP release, respectively, whereas the eukaryotic cytosolic Hsc70 chaperone is GrpEindependent (26,27).
Although the cross-linking results in this report were interpreted in terms of Model 1, there are two reasons why Model 2 should not be discarded. First, a genetically engineered COOHterminal fragment of DnaK, with a peptide in the binding site, crystallizes in two different forms (8). In one crystal form the ␣-helical lid blocks the deep peptide binding channel, while in the other crystal form the ␣-helical lid is rotated away from the channel, making the channel more accessible. Possibly, these two different crystals represent the two different E state isomers. Second, hexokinase, which has an ATP and glucose binding core that is identical in tertiary structure to the ATPbinding core of 70-kDa molecular chaperones (4), equilibrates between closed and open forms in the absence of glucose (28,29). Given the similar tertiary structures of hexokinase and the 70-kDa chaperones, it is reasonable to postulate that the ATPase domain of 70-kDa chaperones also equilibrates between closed and open forms in the absence of substrates, as indicated in Model 2. Clearly, more experiments are required before the presence of the two monomers bands can be definitively assigned to a specific model.
Last, we address one other interpretation of the gel results. Suppose that the high affinity species, which migrates as the apparent molecular mass 90-kDa species after cross-linking, and the low affinity species, which migrates as the 70-kDa species after cross-linking, are in equilibrium in the absence of ATP, the equilibrium lies to the low affinity state, and ATP binding shifts the equilibrium completely to the low affinity state. Specifically, suppose that 30% of the DnaK molecules populate the high affinity state and 70% of the DnaK molecules populate the low affinity state in the absence of ATP; crosslinking such a sample would yield a doublet, exactly as shown in Fig. 2. On the other hand, because ATP binding shifts the equilibrium to the low affinity state, cross-linking would yield a singlet, exactly as shown in Fig. 2. In this interpretation of the gel data, polypeptide binding should shift the population of molecules from the low affinity to the high affinity state. But we have been unable to cause such a shift by incubating DnaK with a large excess of peptide and then cross-linking. In addition, the idea that DnaK predominantly populates the low affinity state in the absence of ATP also contradicts results from structural (12,13) and kinetic (15) studies, which have shown that in the absence of ATP 70-kDa chaperone molecules populate the high affinity state. We believe the best explanation of the gel results is Model 1, where in the absence of ATP all DnaK molecules populate the high affinity state, which happens to cross-link heterogeneously. Whichever model turns out to be correct, the low affinity to high affinity structural transition in DnaK molecules occurs at exactly the same rate as DnaK-catalyzed ATP hydrolysis, consistent with Reaction 2.