Analysis of three DnaK mutant proteins suggests that progression through the ATPase cycle requires conformational changes.

DnaK, the bacterial homolog of the eukaryotic hsp70 proteins, is an ATP-dependent chaperone whose basal ATPase is stimulated by synthetic peptides and its cohort heat shock proteins, DnaJ and GrpE. We have used three mutant DnaK proteins, E171K, D201N, and A174T (corresponding to Glu175, Asp206, and Ala179, respectively, in bovine heat stable cognate 70) to probe the ATPase cycle. All of the mutant proteins exhibit some alteration in basal ATP hydrolysis. However, they all exhibit more severe defects in the regulated activities. D201N and E171K are completely defective in all regulated activities of the protein and also in making the conformational change exhibited by the wt protein upon binding ATP. We suggest that the inability of D201N and E171K to achieve the ATP activated conformation prevents both stimulation by all effectors and the ATP-mediated release of GrpE. In contrast, the defect of A174T is much more specific. It exhibits normal binding and release of GrpE and normal stimulation of ATPase activity by DnaJ. However, it is defective in the synergistic activation of its ATPase by DnaJ and GrpE. We suggest that this mutant protein is specifically defective in a DnaJ/GrpE mediated conformational change in DnaK necessary for the synergistic action of DnaJ+GrpE.

The 70-kDa heat shock proteins (hsp70s) 1 comprise a ubiquitous family of essential proteins whose synthesis is induced by increases in temperature and other forms of stress (1)(2)(3)(4)(5). A vast collection of literature now supports the notion that a major function of the hsp70 proteins is to interact with substrate proteins to alter or maintain their conformation (6 -9). Binding and hydrolysis of ATP by hsp70s is central to their function as chaperones. The intrinsic ATPase of the hsp70s is modulated by binding substrate proteins and, at least in bacteria, by interacting with their cohort chaperones, DnaJ and GrpE (10).
DnaK is one bacterial homolog of this conserved protein class. It shares 50% amino acid identity with its hsp70 family members (11) and exhibits a number of similar biochemical properties, including a high binding affinity for ATP, a weak ATPase activity (12), and chaperone function (13)(14)(15). The chaperone activity of DnaK presumably mediates its participation in a diverse spectrum of cellular processes including host, phage, and plasmid DNA replication, cell division, proteolysis, flagellar biosynthesis, translocation of secretory proteins, and regulation of the heat shock response (16 -29).
In order to understand the mechanism of action of DnaK, we isolated two classes of partially defective dnaK mutants: recessive mutants selectively defective in regulation of the heat shock response, and dominant negative mutants (30). Because these mutants retain partial function, they are unlikely to result from gross structural alterations and are suitable for a structure-function analysis of DnaK. Three of the mutant proteins were of particular interest because they were defective in a variety of cellular processes even though they retained the ability to bind ATP. The two dominant mutations, E171K (Glu 175 in hsc70) and D201N (Asp 206 in hsc70), affect residues in the ATPase domain predicted to be in the vicinity of the Mg 2ϩ ion bound to ATP. Both were suggested to be candidate residues participating directly in catalysis (31,32). The recessive mutation, A174T (Ala 178 in hsc70), lies in proximity (in the linear sequence) to amino acids expected to participate directly in ATP hydrolysis; no inferences were made concerning a role of Ala 174 in catalysis.
The availability of these three DnaK mutant proteins that are able to bind ATP but are altered at amino acids suggested to be critical for ATP hydrolysis led us to perform a detailed characterization of their intrinsic and regulated ATPase activities. Interestingly, the same residues identified by our dominant negative mutations were chosen by two different groups for site-directed mutagenesis. The work reported here complements recent studies on the behavior of a series of mutant proteins altered at Glu 171 in DnaK (33), as well as Glu 175 and Asp 206 in the N-terminal domain of hsc70 (34). On the basis of our analysis, we suggest a pathway for the regulated ATPase activity of DnaK.

EXPERIMENTAL PROCEDURES
Materials-[␥-32 P]ATP at 3000 Ci/mmol was purchased from DuPont NEN, and NaB[ 3 H] 4 (12 Ci/mmol) was obtained from Amersham Corp. TPCK-trypsin used in partial proteolysis studies was bought from Sigma. All other chemicals were of analytical grade and purchased from standard sources.
Plasmid pNRK416 expressing wild type dnaK from the isopropyl-␤-D-thiogalactopyranoside-inducible lacUV5 promoter was from N. Kusukawa and T. Yura, while synthetic vsv-peptides A and C (Ͼ95% pure) (35) were synthesized by the Biomolecular Resource Center at University of California, San Francisco.
Growth and Lysis of Wild Type and Mutant DnaK Strains-Wild type (wt) cells transformed with pNRK416 were grown at 30°C in Luria broth containing 0.2% glucose and 50 g/ml ampicillin. When the A 600 reached 0.1, the expression of DnaK from the plasmid was induced by the addition of 0.25 mM isopropyl-␤-D-thiogalactopyranoside. Cells were harvested at an A 600 between 0.8 and 0.9. Mutant cells, expressing 5-fold higher levels of DnaK than untransformed wt cells, were grown in Luria broth ϩ 0.2% glucose at 30°C. To avoid the possibility of other bacteria contaminating the mutant cells, growth of mutant dnaK cells was selected for in the presence of 10 g/ml tetracycline.
Cell pellets (typically 5 g) were resuspended in 115 ml of buffer A supplemented with 0.3 mg/ml lysozyme. Cell suspensions were lysed by repeated quick freezing and thawing cycles followed by sonication. Cell debris and unlysed cells were removed by centrifugation at 30,000 ϫ g for 45 min.
Purification of DnaK-The S-30 obtained following cell lysis was subjected to a 50% (NH 4 ) 2 SO 4 cut. The resulting precipitate was collected by centrifugation, resuspended in 30 ml of buffer B and dialyzed extensively against it. The dialysate was initially chromatographed over an 80 ml of DEAE fast flow Sepharose column (Pharmacia Biotech Inc.) equilibrated with buffer B. After washing the column with 100 ml of buffer B, bound proteins were eluted with 160 ml of a linear salt gradient ranging from 50 mM to 550 mM KCl in buffer B. The profile of DnaK eluted by the salt gradient was monitored by gel electrophoresis.
Peak fractions containing DnaK were pooled, dialyzed against buffer C and chromatographed through an ATP-agarose column (Sigma). The sample was allowed to interact with the column matrix and then washed with buffer C containing 2 M NaCl to eliminate proteins bound nonspecifically to the resin. The column was reequilibrated with buffer C and washed with 3 column volumes of buffer supplemented with 1 mM GTP to elute GTPases that may have bound to the column. To avoid the possibility of other lower affinity ATPases contaminating the preparation of DnaK, the column was next washed with buffer C containing 0.2 mM ATP. DnaK bound to the column was finally eluted with buffer C containing 10 mM ATP. Although DnaK was expected to elute predominantly in buffer containing 10 mM ATP, at least 50% of it was present in the 0.2 mM ATP eluate. A likely explanation for this behavior is that the 2 M salt wash may have weakened the interaction between DnaK and ATP-agarose allowing it to be eluted with lower ATP concentrations. Peak fractions of DnaK in the 0.2 mM and 10 mM ATP eluates were pooled separately. Each pool was concentrated using a Centricon 30 concentrator and free ATP was removed from the protein solution by extensive dialysis against buffer D.
The 0.2 mM eluate was further purified on a hydroxylapatite (Bio-Rad) column. This additional step also ensured removal of free ATP, left over after dialysis following ATP-Sepharose chromatography. Samples were diluted with buffer E to adjust the conductivity of the DnaK solution. DnaK was retained on the column under these conditions and was eluted with a linear gradient of sodium phosphate ranging from 10 mM to 120 mM in buffer E. Peak fractions containing DnaK were concentrated and dialyzed against buffer D containing 50 mM KCl. The dialyzed fractions were aliquotted, quick frozen in dry ice-ethanol, and stored at Ϫ80°C. Protein concentration was measured by the dye binding method using Coomassie Brilliant Blue G-250 (Bio-Rad protein assay kit), with bovine serum albumin as a standard. All preparations were greater than 95% homogeneous after elution from the ATP-Sepharose column.
Purification of DnaJ and GrpE-DnaJ was purified using the procedure of Zylicz et al. (36). GrpE was purified by a modification of the protocol used by Zylicz et al. (37). Briefly, GrpE was isolated by chromatography over a DEAE fast flow Sepharose column followed by separation on a hydroxylapatite column. This procedure was repeated a second time except that the pH of the buffer was increased from 7.5 to 8.5. Following hydroxylapatite column chromatography, the preparation was purified through a Q-Sepharose column. Finally, fractions containing GrpE were chromatographed on an ATP-agarose column to remove contaminating ATPases. GrpE was not retained by this column and was concentrated from the flow-through fractions. The GrpE prep-aration was Ͼ95% homogeneous and was devoid of endogenous ATPase activity.
ATPase Assays-Purified preparations of wt and mutant DnaK proteins, 0.6 M unless otherwise indicated, were typically incubated with 0.2 mM [␥-32 P]ATP (specific activity, ϳ300 cpm/pmol) at 30°C. In the initial experiments, the DnaK proteins were incubated in buffer F. Subsequently, the buffer system was altered to that used by the Bukau group (33), namely, buffer G. At indicated time periods, reactions were terminated by adding 25-l aliquots to 200 l of 0.1 N H 2 SO 4 , 1.5 mM carrier NaH 2 PO 4 . Free phosphate ( 32 P i ) released upon ATP hydrolysis was complexed with molybdenum by the addition of 50 l of 10% ammonium molybdate (38). The complex was extracted with 0.6 ml of water saturated 1-butanol and 200-l aliquots of the organic phase were counted in a liquid scintillation counter. Radioactivity in the organic phase was used to measure the extent of ATP hydrolysis; hydrolysis in the absence of exogenously added DnaK was routinely subtracted from data points. Each experiment was carried out at least two times; the error from one experiment to the next did not exceed 20%.
Peptide Binding Measurements-(a) Radiolabeling of synthetic vsvpeptide C was carried out by a modification of the reductive methylation procedure of Tack 4 by chromatography over a gel filtration column (G10, Pharmacia). The amount of radioactivity in each fraction (volume ϭ 0.8 ml) was measured; those in the void volume containing peak radioactivity were pooled and stored at Ϫ70°C. (b) Binding of peptide C to DnaK (wt and mutant) was monitored by equilibrium dialysis, using microdialysis chambers separated by membranes with a molecular mass cut-off of 12-14 kDa. Each half chamber of the microdialyzer was filled with buffer G containing 0.1 mg/ml bovine serum albumin. DnaK (5 M) was added to one half of the chamber while [ 3 H]peptide C (specific activity, ϳ120 cpm/pmol) at concentrations ranging from 0.5 to 100 M was added to the other half chamber. Dialysis was carried out at room temperature for 12-16 h. Thereafter, 10-l aliquots were removed from each half chamber and the amount of [ 3 H]peptide present was determined by liquid scintillation counting. The extent of [ 3 H]peptide bound to DnaK and the K D values for binding were calculated using the "Kaleidograph" curve fitting program.
Construction and Use of His-GrpE-N-terminally hexa-His-tagged GrpE was constructed by cloning the BspHI-SphI grpE fragment from pPJR16 into pQE-30 (QIAGEN). His-tagged GrpE was expressed in strain M15 (pREP4) (QIAGEN), and purified to homogeneity (Ͼ95%) by a single step using Ni 2ϩ -NTA-agarose. 0.8 mg His-GrpE (in buffer H) was coupled to 0.2 ml (settled volume) of Ni 2ϩ -NTA resin (QIAGEN) by incubation at room temperature for 2 h. Coupling was monitored by doing a Bradford measurement of the starting solution and comparing it with that of the supernatant after incubation (Ͼ90% of the protein was bound to the resin under these conditions). The resin was equilibrated in buffer H and 10 g of each of the DnaK proteins, dialyzed against the same buffer, were batch adsorbed with 10 l (settled volume) of either Ni 2ϩ -NTA or His-GrpE-Ni 2ϩ -NTA resin (corresponding to ϳ40 g of bound GrpE). After a 1-2 h incubation at room temperature, the resin was centrifuged and the supernatant, corresponding to the flow-through fraction, was saved in a separate tube. The resin was washed twice in buffer D and then incubated in the presence of either 0.2 mM or 1 mM ATP in buffer D at 30°C for 30 min. The resin was respun, and the supernatant (or the eluate) was transferred to a new vial. Aliquots of the onput, flow-through, and eluate fractions were electrophoresed on 10% SDS-polyacrylamide gels and stained with Coomassie Brilliant Blue to detect the protein pattern.
Partial Proteolysis Studies-Partial proteolysis of wild type and mutant DnaK proteins was carried out using TPCK-treated trypsin. 3 g of DnaK were incubated with 1.6 units of trypsin in buffer I in the absence or presence of 5 mM ATP. The reactions were incubated at 30°C, and the rate of generation of tryptic peptides was followed for up to 40 min. Reaction aliquots were sampled directly into Laemmli (40) sample loading buffer and heated at 95°C for 3 min prior to electrophoresis on 12% SDS-polyacrylamide gels. The gels were acid-fixed and stained with Coomassie Brilliant Blue R-250 to visualize the proteolytic products.
Computer Modeling-Modeling was carried out using LOOK (Molecular Applications Group, Palo Alto, CA) (41) and MIDASPlus (University of California, San Francisco Computer Graphics Laboratory) (42,43). The Thr 179 side chain was subjected to energy minimization with the rest of the molecule held rigid. The resulting steric conflicts were so small (ϳ0.1-0.2 Å) that the rest of the molecule was not optimized. The ␤-carbon of Thr 174 is within 0.1 Å of the Ala 174 ␤-carbon. Note that the enzymology and modeling results for Ala 174 3 Thr both imply that the structure in the ATP-bound form is very similar to wild type. RESULTS We purified wild type and three mutant DnaK proteins, E171K, A174T and D201N, to apparent homogeneity by a three-step process involving chromatography over DEAE-cellulose, ATP-Sepharose, and hydroxylapatite column matrices. Because these mutations were expected to affect ATP hydrolysis by DnaK, we compared their basal and regulated ATPase activities with those of wt DnaK. Initially, two different techniques were used to quantitate the ATPase activity. The amount of either the free phosphate (P i ) released or ADP formed was measured by scanning polyethyleneimine cellulose sheets following thin layer chromatography. Alternatively, the free phosphate was complexed with molybdenum and extracted using H 2 O-saturated 1-butanol (see "Experimental Procedures"). All measurements gave identical results (data not shown). However, due to the sensitivity and rapidity of manipulations, the molybdenum assay was used in all experiments reported in this study.
Mutant DnaK Proteins Have Altered Intrinsic ATPase Activities-We examined the kinetics of ATP hydrolysis by wild type and mutant DnaK proteins using an ATP concentration of 0.2 mM. This concentration of nucleotide is saturating for wt DnaK which has a K m of 1 M for ATP. As observed by other investigators, the purified wt protein displayed extremely slow kinetics of hydrolysis at 30°C with a turnover number of ϳ0.03 min Ϫ1 . All three mutant proteins retained linear kinetics of ATP hydrolysis (not shown) but they exhibited rates that were either lower or higher than wt DnaK (Table I).
As we were finishing our studies, we became aware that our data with E171K differed significantly from those of the Bukau group (33). Whereas our preparation of the E171K mutant protein exhibited lower ATPase activity than the wt protein (Table I), theirs exhibited higher than normal activity. Differing purification procedures or assay conditions could have accounted for this discrepancy. Systematic analysis of these variables indicated that the assay conditions were responsible. The ATPase of E171K was 44-fold higher in buffer G used by Bukau than in our buffer, buffer F (Table II). This dramatic effect was specific for the E171K mutant. The activity of wt DnaK was identical in both buffer systems and those of A174T and D201N increased only ϳ2-fold in buffer G (Table II).
To identify the component that contributed to the increased activity of E171K, we switched individual components of both buffer systems. These experiments indicated that acetate was the key difference in the two buffers (Table III). Replacing Mg(OAc) 2 in buffer G with MgCl 2 reduced the activity of E171K to near background. Supplementing buffer G containing MgCl 2 with KOAc restored the ATPase of E171K activity to a level comparable to that observed in buffer G only, indicating that Cl Ϫ was not exerting a deleterious effect. Conversely, replacing MgCl 2 in buffer F with Mg(OAc) 2 raised the activity of E171K almost to the level observed in buffer G. The effect was specific to acetate since glutamate did not affect the activity of E171K (data not shown). These results indicate that acetate specifically alleviates the adverse effect of the lysine substitution in the E171K mutant protein. All the experiments reported in this study were performed in buffer G, unless otherwise indicated.
Mutant DnaK Proteins Have Altered Values of K m for ATP-To determine whether the mutant proteins were altered in their values of K m for ATP, we measured their rates of ATP hydrolysis as a function of substrate concentration and represented the results using the double reciprocal Lineweaver-Burke plot (Fig. 1). wt DnaK and D201N generated straight lines characteristic of Michaelis-Menten kinetics. The K m of wt DnaK for ATP was 1 M (Fig. 1A), consistent with the values reported for DnaK and eukaryotic hsp70 s (44,45), whereas the K m of D201N was about 8 M (Fig. 1B). On the other hand, the double reciprocal plot with A174T was distinctly biphasic and generated two K m values: a low K m of about 1 M and a high K m of about 20 M (Fig. 1C). E171K also had a slight tendency toward biphasicity, with a low K m of about 30 M and a high K m of about 175 M (Fig. 1D). Buchberger et al. (33) report a single K m for E171K, most likely because they used a more restricted range of substrate concentrations. The mixed K m values do not reflect a contaminating ATPase since all of the ATPase activity is removed by chromatography over immobilized GrpE (see Fig.  3). Nor do they reflect a mixed population of monomers and oligomers in which the oligomers exhibit a different affinity for ATP than the monomers. Analysis of the distribution of species on a native polyacrylamide gel indicated a rather homogeneous population of monomers, similar to that seen for the wt protein (data not shown). We favor the idea that the heterogeneity reflects a conformationally labile ATP binding site that results in a population with mixed affinities for ATP.
Effect of Substrate Proteins-The ATPase activity of hsp70 proteins is stimulated by the binding of substrate proteins and also by a variety of short synthetic peptides possibly because they mimic unfolded regions of substrate proteins (35). We tested the effects of saturating concentrations (0.5 mM) of two of these peptides, called A and C, on the ATPase activities of wild type and mutant DnaK proteins. The ATPase of wt DnaK was stimulated 4 -5-fold by peptide A and 7-8-fold by peptide C (Fig. 2A), similar to the previous observations with hsc70/BiP   1.9 Ϯ 0.5 a Indicates the ratio of picomoles of ATP hydrolyzed in buffer G (containing acetate) relative to that hydrolyzed in buffer F (lacking acetate) at an ATP concentration of 0.2 mM (see "Experimental Procedures" for buffer composition). (35) and DnaK (46,47). Peptide stimulation of the A174T ATPase was similar to but somewhat less than that of the wild type, with a 2-fold stimulation by peptide A and a 5-6-fold stimulation by peptide C (Fig. 2B). On the other hand, neither E171K nor D201N exhibited increased activity in response to peptides. The ATPase activity of E171K was unaffected by either peptide (Fig. 2C) and that of D201N was reproducibly inhibited 20 -25% by both peptides (Fig. 2D).
Peptide Binding to DnaK-The lack of a stimulatory effect of peptides on the ATPase activities of E171K and D201N could result from the inability of the mutant proteins to bind peptides. To investigate this, we followed the binding of [ 3 H]peptide C to wild type and mutant DnaK proteins by equilibrium dialysis. The mutant proteins retained the ability to bind peptide C. In fact, their K D values for binding were either identical to wild type (7-8 M with wt DnaK, A174T, and D201N) or differed at most by 1.5-fold (12 M with E171K) (Table IV). These results suggest that E171K and D201N are defective in responding to peptides following binding. Our data with peptide C are consistent with the reports of Buchberger et al. (33) using three different DnaK proteins altered at Glu 171 .
Effect of DnaJ and GrpE on ATP Hydrolysis-DnaJ and GrpE stimulate the intrinsic ATPase activity of DnaK. DnaJ stimulates the rate of ATP hydrolysis, whereas GrpE increases the rate of release of ADP bound to DnaK upon hydrolysis. Together, these proteins act synergistically to increase the ATPase activity of DnaK by 50-fold (44). We investigated whether DnaJ or GrpE had similar effects on the mutant proteins. We found that each of the mutant proteins differed from wt in their responses toward these heat shock proteins (Table  IV).
A174T responded similarly to wt DnaK to the separate ad-dition of either DnaJ or GrpE. However, DnaJ and GrpE did not function synergistically to increase the ATPase activity of A174T. Together, DnaJ and GrpE stimulated the A174T ATPase with less than twice the effectiveness of DnaJ alone. In contrast, for wt DnaK, DnaJ and GrpE together were approximately 25-fold more effective in stimulating the ATPase activity than DnaJ alone. The ATPase activity of E171K was not stimulated either by the separate or simultaneous addition of DnaJ or GrpE. In addition, there was no stimulation by DnaJ Ϯ GrpE when the experiments were carried out in buffer containing MgCl 2 rather than Mg(OAc) 2 where the intrinsic ATPase activity of E171K was reduced by more than 40-fold (data not shown). As was true with E171K, neither DnaJ, alone, nor DnaJ ϩ GrpE stimulated the ATPase activity of D201N. Moreover, GrpE reduced the ATPase of D201N by as much as 40% (Table IV). Though small, the inhibitory effect of GrpE was reproducible.
Binding of GrpE to DnaK-Each of the mutant proteins failed to show a synergistic increase in ATPase activity following the simultaneous addition of DnaJ and GrpE, leading us to wonder whether the DnaK mutants interacted normally with these proteins. We took advantage of the availability of His-tag GrpE protein to study the interaction between the mutant DnaK proteins and GrpE. GrpE binds tightly to DnaK in the absence of ATP and weakly in its presence (37). This fact serves as the basis for a prevalent purification of GrpE in which GrpE is chromatographed through a DnaK affinity column and eluted with ATP. We performed similar experiments in reverse. His-GrpE was coupled to Ni 2ϩ -NTA resin, the DnaK proteins were adsorbed to the resin and then eluted with ATP.
wt DnaK and all three mutant proteins bound specifically to the his-GrpE column. Whereas Ͼ90% of the DnaK was present in the flow through fraction when chromatographed over the Ni 2ϩ -NTA resin alone (Fig. 3, Panel A), only trace amounts were detectable in the flow through fractions of the His-GrpE-Ni 2ϩ -NTA resin (Fig. 3, Panel B). Thus, the mutants were not grossly defective in their interaction with GrpE. We note in passing that the ATPase activity in the flow-through of the His-GrpE-Ni 2ϩ -NTA resin was proportional to the small amount of DnaK present in that fraction, indicating that the activities we have measured throughout this report are attrib-utable to DnaK, rather than to a contaminating ATPase.
In contrast to the binding phase, the ATP elution phase revealed differences between wt and mutant proteins (Fig. 3,  Panel B). As expected, wt DnaK was efficiently eluted by 0.2 mM ATP; A174T behaved similarly to the wt protein. In contrast, only 10 -20% of E171K and D201N were recovered. ATP concentrations as high as 1 mM were unable to further elute E171K from the bound resin. 2 These data suggest that both E171K and D201N are defective in the ATP mediated conformational change in DnaK necessary for release of GrpE.
Partial Proteolysis of wt and Mutant DnaK Proteins-Partial proteolysis is a convenient way to monitor the conformational change induced in DnaK upon ATP binding. We have used this assay to compare the conformational changes exhibited by mutant proteins with those exhibited by wild type DnaK. As demonstrated in Fig. 4, in the absence of protease, both mutant and wild type proteins are stable over the time course of the assay.
The differences in digestion of wt DnaK with and without nucleotide have been carefully documented (48,49) and were reproduced here (Fig. 5A). In the absence of ATP, wt DnaK exhibited five predominant proteolytic products of about 55, 46, 44, 31, and 17 kDa in size. In the presence of ATP, proteolysis was considerably accelerated and the predominant digestion products were altered. The 55-, 44-, and 17-kDa fragments remained, the 31-kDa fragment disappeared, and fragments of 53 and 45 kDa became prominent. The 45-kDa fragment, which is the major band generated upon prolonged digestion in the presence of ATP, corresponds to the N-terminal ATPase domain of DnaK and may be distinct from the 46-kDa band observed in the absence of ATP (33,34).
The pattern of proteolytic products generated by A174T was almost identical to that of wild type DnaK both in the absence and presence of ATP (Fig. 5B). Thus, at the level of sensitivity of this assay, both the native structure of A174T and its conformational change upon binding ATP are virtually identical to that of the wt protein.
For D201N, the tryptic peptides in the absence of ATP were almost identical to wt (Fig. 5C). However, addition of ATP induced very little change in either the rate of appearance or pattern of tryptic peptides. These results suggest that although the conformation of D201N in the absence of ATP is similar to that of the wt, it does not exhibit the conformational change characteristic of wt DnaK upon nucleotide binding. 2 A. S. Kamath-Loeb, unpublished data. e Refers to the percentage of DnaK dissociated from a complex with GrpE by ATP. The numbers were calculated from the intensities of the Coomassie-stained bands of the ATP eluates and from ATPase assays of these fractions and represent the average of two determinations.
f Acetate effect: sensitivity to acetate ions in reaction buffer. Numbers refer to the rate of ATP hydrolysis in buffer G relative to that in buffer F.

FIG. 3. Interaction of wt and mutant DnaK proteins with GrpE.
10 g of each DnaK protein was mixed with either Ni 2ϩ -NTA resin (Panel A) or with His-GrpE-Ni 2ϩ -NTA resin (Panel B). Following incubation at room temperature for 1-2 h, the resins were centrifuged, and the supernatants (or flow-through fractions, FT) were saved separately. The resins were then incubated with 0.2 mM ATP (for wt DnaK, A174T, and D201N) or 1 mM ATP (for E171K) at 30°C for 30 min. The resins were recentrifuged, and the ATP eluates were likewise transferred to a separate tube. Aliquots of the starting material, flow-through fractions, and ATP eluates were electrophoresed through 10% SDS-polyacrylamide gels; the profile of DnaK in these fractions was visualized by Coomassie Blue staining. The lower band reflects GrpE, which appears to show a slow elution from the column under these conditions. The partial proteolysis products of E171K differed most from those of wt DnaK (Fig. 5D). E171K was much more labile in the absence of ATP and generated peptides that were distinct from those of wt DnaK even after limited extents of proteolysis. The presence of ATP reduced the rate of trypsin digestion and in particular, stabilized the 46-kDa ATPase fragment from deg-radation. Since the ATPase activity of E171K was dramatically increased in the presence of acetate ions relative to that in its absence, we wondered whether the protein under went major structural changes in response to acetate. We tested this by carrying out partial proteolysis of E171K in buffer I containing MgCl 2 instead of Mg(OAc) 2 . Interestingly, the pattern of tryptic peptides generated without acetate was virtually identical to that with acetate suggesting that acetate may cause only local and/or subtle changes in the structure, if any (data not shown). These results suggest that the conformation of E17lK is discrepant from that of the wild type protein both in the absence and presence of ATP. DISCUSSION The ATPase activity of DnaK is central to its mode of action as a chaperone. The intrinsic ATPase activity of DnaK is very low (k cat ϳ 0.1-1.0 min Ϫ1 ), and physiologically important modulators, including denatured proteins and the cohort heat shock proteins DnaJ and GrpE increase this basal rate of hydrolysis. Both direct physical measurements and inferences from homologous crystal structures support the idea that hsp70s undergo conformational changes upon binding ATP. This conformational change is crucial to chaperone activity as it alters the kinetics with which hsp70 binds and releases substrate proteins. The availability of three mutant DnaK proteins that retain the ability to bind ATP but are defective in chaperone activity in vivo allowed us to explore the relationship between the ATPase cycle and chaperone function. The effects of each of these mutations on the intrinsic and modulated ATPase activities are summarized in Table IV. Below, we discuss how analysis of the effects of these mutational changes has altered our view of the function of hsp70.
The E171K Mutation-The E171K mutation is a severely defective allele of dnaK that allows growth of cells only up to 34°C. Cells containing this allele are unable to modulate the heat shock response or support the growth of bacteriophage , suggesting a general defect in chaperone function. Finally, the defective allele is dominant as the growth of bacteriophage is inhibited even in the presence of wild type DnaK (30). A dominant negative phenotype is also observed when the comparable glutamic acid residue in Drosophila hsc72 is altered to serine. 3 Glu 171 (Glu 175 in bovine hsc70) occupies a position in close proximity to the active site, forming a hydrogen bond with a water molecule in the first coordination shell of Mg 2ϩ . Proper positioning of the Mg 2ϩ ion, which interacts with the ␤-phosphate of ATP, is crucial for catalysis. The Glu 175 position has been the subject of a great deal of study because it has been hypothesized both to be a catalytic base in the ATPase reaction and essential for the subdomain movement that may couple ATPase activity with substrate binding. To pursue these ideas, the McKay and Bukau groups made site-directed changes at Glu 175 and examined various aspects of structure and function of the mutant proteins. Our studies add additional information concerning the function of this residue in both basal and regulated ATPase activities of DnaK.
Glu 175 is in a structurally equivalent position to an aspartic acid proposed to be a catalytic base in hexokinase, leading the McKay group to suggest Glu 175 as one of four possible acidic residues that could participate in catalysis (31,32,34,49). The ATPase activities of N-terminal fragments of hsc70 containing either E175S or E175Q substitutions are defective; the k cat values are 5-20-fold lower, and the K m values are about two orders of magnitude greater than the wt (34,49). However, both mutants retain ATPase activity indicating that Glu 175 is 3 K. Palter, personal communication. not essential in catalysis. Similar studies with E171A, E171L, and E171K substitutions in DnaK also demonstrated that the mutant proteins exhibit significant catalytic activity (33). In fact, in contrast to the results with the hsc70 mutants, the DnaK mutants exhibit V max values 3-30-fold greater than wt. Our results with E171K, isolated using a different purification protocol, are in agreement with the results of the Bukau group. ATP hydrolysis by E171K is about 13-fold higher than the wt rate. However, we find that the high k cat is not an intrinsic property of E171K, but is dependent on having acetate in the buffer. The presence of acetate specifically enhances the rate of hydrolysis of E171K about 40-fold. Glutamate cannot substitute for acetate indicating that a specific interaction between E171K and acetate is involved in this effect. We currently do not know whether the acetate effect is specific for the E171K mutation or if it affects the k cat of other substitutions at residue Glu 171 . Moreover, we do not know whether acetate also enhances the k cat of hsc70. The McKay assays were performed in acetate; however, if slight structural differences prevented acetate enhancement of hsc70 catalysis, the k cat differences of mutant hsc70 and DnaK proteins would be explained. We consider three possible roles for acetate in enhancing ATP hydrolysis by E171K: 1) acetate may perform the function of the original glutamic acid residue, 2) binding of acetate may alter the position of the lysine residue permitting greater function, or 3) binding of acetate may specifically neutralize the deleterious charge effects of the lysine residue. Examination of the homologous crystal structure indicates that some possible conformations of lysine could accommodate an acetate ion. 4 Glu 171 is located in one of two crossed ␣-helices connecting the two N-terminal subdomains and is the only residue in these helices that interacts with Mg 2ϩ . In analogy to actin, Bukau and collaborators (33) suggest that the two subdomains move upon binding ATP using the crossed helices as a hinge and that this movement requires a Mg 2ϩ connection, thus implicating Glu 171 in this process. Further correlates are that this movement is essential for coupling ATPase activity to substrate binding which, in turn, is essential for chaperone function. Consistent with this idea, they find that E171A, E171L, and E171K mutant DnaK proteins 1) do not undergo the wt conformational change upon binding ATP as judged by partial proteolysis, 2) bind peptides normally but exhibit neither peptidestimulated ATPase nor ATP-stimulated release of peptides, and 3) are defective in chaperone activity. Our studies agree with these findings and indicate that E171K is also defective in its responses toward DnaJ and GrpE. Its ATPase can neither be stimulated by DnaJ, nor synergistically increased by DnaJ ϩ GrpE. Additionally, the mutant protein is defective in ATPmediated release from GrpE. Our view of the origin of the E171K phenotype is somewhat more general than that of Bukau. We believe that E171K is one example of a class of mutations that interfere with the normal conformational change of DnaK upon binding ATP and that the inability to make this conformational changes interferes with all of the regulated activities of the protein. This view is based primarily upon the phenotype of D201N and is explicated below.
The D201N Mutation-The in vivo phenotypes of D201N are similar to those of E171K. Cells containing the dominant D201N mutant protein grow only up to 34°C and are unable to modulate the heat shock response or support the growth of bacteriophage (30). A dominant negative phenotype is also observed when the comparable aspartic acid residues in Drosophila hsc70 and hsc72 are changed to serine. 3 Like Glu 171 , Asp 201 (Asp 206 in bovine hsc70) occupies a posi-tion in close proximity to the active site. Asp 201 is coordinated via one carbonyl oxygen to the K ϩ ion that lies at the interface between protein and P i in the ADP form of bovine hsc70 (50). In addition, Asp 201 is coordinated via a water molecule to Glu 171 (50). Finally, like Glu 171 , Asp 201 by analogy to the actin structure was suggested to act as a proton acceptor in catalysis of ATP (34). However, studies by the McKay group indicated that altering Asp 206 to either Asn 206 or Ser 206 did not eliminate catalytic activity of the N-terminal fragment of hsc70. In fact, of the four positions substituted by the McKay laboratory, mutations at Asp 206 are the least defective; the mutant proteins exhibit almost no alteration in K m and less than a 10-fold decrease in k cat (34). Our studies on D201N, performed on full-length DnaK, support those of the McKay group in that we find only a small increase in K m . However, reminiscent of the results with Glu 171 , D201N also exhibits an increase in k cat . In addition, crystallographic studies of D206N and D206S demonstrate that the structure of the mutant proteins bound to ADP is virtually identical to wt (49). In summary, D201N has only a very small effect on basal ATPase activity or on the structure of the N-terminal fragment bound to ADP. In stark contrast, D201N is severely defective in all of its regulated activities. Peptides bind normally to D201N, but do not stimulate its ATPase activity. On the contrary, they slightly, but consistently, inhibit the ATPase activity. Similarly, neither DnaJ nor DnaJ in combination with GrpE stimulate D201N ATPase. In fact, like peptide, GrpE also slightly inhibits the ATPase activity of D201N. Finally, as observed with E171K, D201N is defective in ATP-mediated release from GrpE.
Why does the D201N change completely abolish effector stimulation of the ATPase activity? The partial proteolysis phenotype of D201N may provide us with a possible explanation. Prior to binding ATP, the conformations of D201N and wt DnaK are indistinguishable as judged by partial proteolysis profiles. Following ATP binding, the wt protein exhibits significant qualitative and quantitative changes in partial proteolysis pattern, whereas D201N does not. This suggests that D201N is very defective in undergoing a conformational change after binding ATP. If the conformationally altered ATP bound form of DnaK (DnaK*) is the substrate with which DnaJ, GrpE, and peptide all interact productively, then D201N will be defective in all regulated activities because it cannot form DnaK*. Two studies, using completely different methodologies, indicate that ATP binding, without hydrolysis, alters DnaK so that it both binds and releases peptide more rapidly (9,51). These studies provide independent evidence for the idea that proper interaction with peptides requires DnaK*. Our explanation for the mutant phenotype suggests that productive interaction with DnaJ and GrpE requires DnaK* as well.
The A174T Mutation-The A174T mutation is a weakly defective, recessive allele of dnaK that permits growth of cells up to 40°C. This alanine residue is found in all 22 known bacterial DnaKs and 74 out of 78 eukaryotic hsc70 homologs. A characteristic feature of A174T is that it exhibits a selective deficiency in chaperone function. Both in vivo and in vitro studies indicate that the DnaK:DnaJ:GrpE chaperone team is required for phage replication and for regulation of the heat shock response. Yet, cells with the A174T allele are proficient for growth but defective in regulation of the heat shock response (30). This selective phenotype suggested that A174T was worthy of detailed characterization.
Our analysis indicates that the mutation has a minimal effect on most of the activities we have examined. The basal ATPase activity of A174T is basically intact. The mutant protein exhibits a k cat about 50% that of the wt and a heterogene-ous K m for ATP. The lower K m of 1 M is identical to that of the wt, whereas the higher value is about 20-fold greater. However, given the high intracellular ATP concentration (2.7 mM) (52,53), this heterogeneity is unlikely to affect basal ATPase function. Likewise, the K D for peptide binding is unaltered relative to the normal value and the ATPase of A174T is increased by two different substrate proteins. Finally, as monitored by partial proteolysis, A174T and wt DnaK undergo similar conformational changes in the presence of ATP.
The outstanding characteristic of A174T is that the simultaneous addition of DnaJ and GrpE does not result in a synergistic increase in its ATPase activity. This is particularly surprising since assays for the effects of the individual proteins on the A174T ATPase did not reveal any defects. DnaJ stimulation of the A174T ATPase is equivalent to that of wt. Likewise, binding and release of A174T to/from GrpE are normal. The mutant phenotype could be explained if binding of DnaJ, GrpE, or both proteins normally induces a conformational change in DnaK (to give DnaK**) that is required for the observed synergistic action of DnaJ ϩ GrpE. The A174T mutation may prevent this conformational change and abolish synergy. Ala 174 is located in the same ␣-helix as Glu 171 (described above), one of the two crossed ␣-helices in the hinge region connecting the N-terminal subdomains. This region of DnaK is believed to be crucial in conformational changes in response to effectors. Whereas Glu 171 is at the N terminus of this helix and points into the active site, Ala 174 lies in the middle of the helix facing a large ␤-sheet in subdomain 2A (33). To examine the structural consequences of altering this alanine to threonine, we modeled the A174T change into the known structure of hsc70, where the corresponding residue is Ala 179 (Fig. 6). Substitution of threonine for alanine results in a small but significant steric overlap between the threonine methyl group and valines 335 and 337 of the adjacent ␤-sheet, which implies either distortion or displacement of the Ala 179 ␣-helix relative to the ␤-sheet. In addition, this threonine is now in close proximity to Val 369 in the other crossed ␣-helix connecting the two subdomains. The extra bulk in the vicinity of the hinge region is likely to interfere with a number of possible conformational rearrangements involving the two crossed helices and the associated ␤-sheet.
It has recently been suggested that DnaJ induces a conformational change in DnaK. Studies of partial DnaJ proteins indicate that the ability to stimulate ATPase and cause the conformational change reside in partially distinct regions of the DnaJ protein (54). This conformational change is required for the tight binding of DnaK to 32 which is a prerequisite for proper regulation of the heat shock response (54). If A174T were unable to make this conformational change, the defect of A174T in regulating the heat shock response could be explained. It is currently unknown whether a similar tight binding state is required for efficient participation of DnaK in replication.
Summary and Conclusions-Each of the three mutant DnaK proteins that we have examined carry out the basal ATPase cycle but are defective in some aspects of the regulated ATPase cycle of DnaK. The spectrum of activities carried out by the mutants lead us to suggest that two different conformational changes are required for progress through the regulated ATPase cycle. The first conformational change, leading to the DnaK*, is accomplished upon binding ATP. The DnaK* form of the protein is required for all subsequent regulated activities of DnaK. The second conformational change, leading to DnaK**, is accomplished after binding DnaJ/GrpE. The DnaK** form of the protein is required for synergistic activation by DnaJ and GrpE.
Our evidence that the DnaK* form of the protein is required for all subsequent regulated changes in the ATPase cycle comes from the study of two mutant proteins, E171K and D201N. This evidence is particularly compelling for D201N, which exhibits a normal conformation in the absence of ATP but fails to achieve the conformationally altered nucleotide bound form of DnaK. In addition, D201N fails to perform all regulated activities of the DnaK protein that we have tested. The simplest way to explain this mutant phenotype is to suggest that the conformationally altered DnaK* form of the protein is required for all regulated activities including: 1) stimulation of the ATPase activity of DnaK by peptides, DnaJ, or DnaJ ϩ GrpE and 2) ATP-mediated release of GrpE from DnaK. DnaK* could be required either for effector binding or for signal transmission. For peptides and GrpE, the DnaK* form must be required for signal transduction, since both ligands can bind to DnaK. For DnaJ, it is not known whether DnaK* is required for binding or signal transduction.
Our evidence that the DnaK** form of the protein is required for synergistic activation of DnaK by DnaJ and GrpE comes from the study of the third mutant protein A174T. A174T binds both DnaJ and GrpE normally but is specifically defective in the synergistic stimulation of ATPase caused by the simultaneous binding of these effector molecules. We suggest that this mutant protein is specifically defective in a DnaJ/GrpE mediated conformational change in DnaK (to give DnaK**), and that this conformational change is necessary for the synergistic action of DnaJ ϩ GrpE. The biochemical phenotypes of intragenic second site revertants of each mutation may test the validity of these ideas.