The ATPase Cycle of the Endoplasmic Chaperone Grp94*

Grp94, the Hsp90 paralog of the endoplasmic reticulum, plays a crucial role in protein secretion. Like cytoplasmic Hsp90, Grp94 is regulated by nucleotide binding to its N-terminal domain. However, the question of whether Grp94 hydrolyzes ATP was controversial. This sets Grp94 apart from other members of the Hsp90 family where a slow but specific turnover of ATP has been unambiguously established. In this study we aimed at analyzing the nucleotide binding properties and the potential ATPase activity of Grp94. We show here that Grp94 has an ATPase activity comparable with that of yeast Hsp90 with a kcat of 0.36 min–1 at 25 °C. Kinetic and equilibrium constants of the partial reactions of the ATPase cycle were determined using transient kinetic methods. Nucleotide binding appears to be tighter compared with other Hsp90s investigated, with dissociation constants (KD) of ∼4 μm for ADP, ATP, and AMP-PCP. Interestingly, all nucleotides and inhibitors (radicicol, 5′-N-ethylcarboxamidoadenosine) studied here bind with similar rate constants for association (0.2–0.3 × 106 m–1 s–1). Furthermore, there is a marked difference from cytosolic Hsp90s in that after binding, the ATP molecule does not seem to become trapped by conformational changes in Grp94. Grp94 stays predominantly in the open state concerning the nucleotide-binding pocket as evidenced by kinetic analyses. Thus, Grp94 shows mechanistically important differences in the interaction with adenosine nucleotides, but the basic hydrolysis reaction seems to be conserved between cytosolic and endoplasmic members of the Hsp90 family.

The Hsp90 3 chaperone family is a highly conserved and ubiquitously expressed protein family in both eubacteria and eukaryotes (1)(2)(3). Cytosolic Hsp90 is involved in regulating the activity, turnover, and trafficking of numerous proteins involved in signal transduction and cell cycle control (www.hsp90.org). In eukaryotes, cytosolic Hsp90 exerts its functions in concert with a large number of co-chaperones (1, 2, 4 -6). Together, they seem to maintain their client proteins in specific conformational states required for subsequent activation. The ATPase activity of Hsp90 is essential for its in vivo functions in Saccharomyces cerevisiae (7,8). ATP is bound by the N-terminal domains of the dimeric protein (9). Inhibitors of Hsp90, such as geldanamycin (GA) and radicicol (RA), compete with ATP for binding to the N-terminal ATP-binding site (10). As a consequence of ATP-binding, the two N-terminal domains dimerize and associate with the middle domains (11,12). These conformational changes trap the ATP molecule and commit it to hydrolysis (13). The recently solved structure of full-length Hsp90 shows that nucleotide binding leads to large scale conformational changes in the entire molecule (14,15), which seem to be important to drive the chaperone cycle.
The endoplasmic reticulum paralog of Hsp90 is the glucoseregulated protein Grp94 (16 -18), also known as GP96 or endoplasmin. Grp94 is only found in higher eukaryotes. It must have arisen by duplication of the cytosolic gene (2,19). Like Hsp90, Grp94 is a dimeric protein (16,20,21). Grp94 seems important for the maturation of a number of client proteins, such as integrins, immunoglobulins, and Toll-like receptors (22,23). In addition, Grp94 also bind to peptides in vivo and in vitro (24 -26). Mouse knockouts of Grp94 are embryonically lethal at mesoderm formation, suggesting an important role for this protein in early development (27). In Leishmania, parasite virulence seems to depend on Grp94 (28). In addition, inhibitor studies suggest a central role for Grp94 in the maturation of kinases and other signal transduction proteins (29,30). Finally, Grp94⅐protein complexes can be exploited in the context of anti-tumor vaccines (31).
Interestingly, Grp94 is set apart in several aspects from cytosolic Hsp90. First, no co-chaperones have been detected for Grp94 yet. Second, Grp94 seems to bind but not hydrolyze ATP, and third, there is a specific ATP-binding inhibitor for Grp94, 5Ј-N-ethylcarboxamidoadenosine (NECA) (32,33). Furthermore, although structures of the N-terminal domain of Hsp90 in the absence or presence of ligands (ADP, GA, and RA) are almost identical (1), the structures of the N-terminal domain of Grp94 solved by the Gewirth laboratory show pronounced changes upon ligand binding (32)(33)(34). As expected from the high sequence conservation between cytosolic and endoplasmic Hsp90, the overall features of the N-terminal domain of Grp94 correspond to those of cytosolic Hsp90s. However, a comparison of the ligand-free structure of the N-terminal domain of Grp94 (34) with that in complex with inhibitors (33) or ATP/ADP (32) shows that, in contrast to cytosolic Hsp90, both the apo-form and the inhibitor-bound form of the domain are in a closed conformation, whereas the ADP-and ATP-bound forms are in a more open conformation concerning the accessibility of the nucleotide-binding site. Furthermore, it became evident that the specificity of the inhibitor NECA, which does not inhibit cytosolic Hsp90, is due to a fiveamino acid insertion in Grp94 that creates an additional binding pocket (33,34). This insertion shifts the position of the subdomain, consisting of helices 1-4-5, to the closed state in the apo-form. Although the crystallographic analysis clearly established the specific binding of ADP and ATP to the N-terminal domain of Grp94, the question of whether Grp94 is an ATP hydrolyzing enzyme is a highly disputed issue (35)(36)(37)(38)(39), and the prevailing notion is that Grp94 does not hydrolyze ATP. Indeed, the differences in the structures of the ATP-bound form of Grp94 compared with those for cytosolic Hsp90 together with biochemical evidence suggested that in the case of Grp94, ATP binding without hydrolysis may regulate its activity.
We have revisited this issue because it is important for understanding the mechanism of the entire Hsp90 family. We show that, unlike reported previously, Grp94 exhibits a specific ATPase activity with turnover numbers comparable with that of other Hsp90 members investigated so far. However, the ATP hydrolysis cycle of Grp94 differs substantially from that of cytosolic Hsp90 in strongly favoring the open form of the ATPbinding site even in the presence of ATP.

Chemicals
All of the chemicals (including nucleotides) were purchased from Roche Applied Science, Merck, and Pharma Waldhof (Mannheim, Germany) at the highest quality available.

Construction of the Plasmid pSF94
For the expression of Grp94 in the cytoplasm of Escherichia coli, the plasmid pSF94 containing the cDNA sequence for canine Grp94 (41) was constructed by inserting a PCR fragment lacking the N-terminal signal sequence (amino acids 1-22) into a pET21 vector (Novagen, Madison, WI). The identity of the construct was confirmed by DNA sequencing.

Purification of Recombinant Grp94
E. coli BL21 (DE3) Codon Plus cells (Stratagene, La Jolla, CA) were transformed with pSF94. The cells were grown in LB medium at 37°C, and protein synthesis was induced with isopropyl ␤-D-thiogalactopyranoside (final concentration, 1 mM) at an optical density of 0.8. After 3 h of induction, the cells were harvested by centrifugation. The bacterial pellet was resuspended in 40 mM Hepes, pH 7.5. Subsequently, the cells were lysed, and KCl and CaCl 2 were added to final concentrations of 50 and 10 mM, respectively. The crude extract was cleared by centrifugation, and the supernatant was supplemented with (NH 4 ) 2 SO 4 to a final concentration of 1.2 M. Precipitated Grp94 was dialyzed against 40 mM Hepes, 50 mM KCl, and 10 mM CaCl 2 , pH 7.5 (buffer A), and then applied to a ResQ column (GE Healthcare, Munich, Germany). After washing the column with buffer A, Grp94 was eluted with buffer A supplemented with 0.5 M KCl. The eluted Grp94 was applied to a Superdex 200-pg gel filtration column (GE Healthcare), which was operated in 40 mM Hepes, 150 mM KCl, and 5 mM MgCl 2 , pH 7.5. The fractions of pure Grp94 were concentrated and stored in 40 mM Hepes, pH 7.5, containing 100 mM KCl and 10 mM CaCl 2 . The concentration of Grp94 was determined using a calculated extinction coefficient of E 280 ϭ 0.884 for a 1-mg/ml solution and a path length of 1 cm.

ATPase Measurement
Radioactive Assay-The ATPase activity of Grp94 was determined as described previously (13). In short, the assay was carried out in a buffer containing 40 mM Hepes, pH 8.0, 150 mM KCl, and 2 mM MgCl 2 . 2 M Grp94 were incubated at the temperatures indicated with 500 M (final concentration) unlabeled ATP and 10 Ci of [␥-32 P]ATP in a total volume of 30 l for steady state measurements.
In the case of single turnover measurements, 200 M Grp94 or yeast Hsp90 and 160 M (final concentration) ATP were used. ATP hydrolysis was stopped after different times by adding EDTA (final concentration, 24 mM). It was assured that this concentration of EDTA is sufficient to stop the reaction. Control measurements with trichloroacetic acid as stopping reagent showed the same results. After thin layer chromatography, the ATP to ADP ratio was quantified with a Typhoon 9200 PhosphorImager (Amersham Biosciences). Hydrolysis rates were corrected for uncatalyzed, spontaneous ATP hydrolysis at the respective conditions.
The temperature dependence of the yeast Hsp90 or Grp94 ATPase was determined by incubating the protein in the presence of 10 M ATP at the respective temperature for 10 min. ATP hydrolysis was then monitored using the radioactive ATPase assay at the same temperature.
Coupled Colorimetric Assay-ATPase assays were performed as described earlier using an ATP-regenerating system (8,42). The assays were measured in 40 mM Hepes, 150 mM KCl, 5 mM MgCl 2 , 1 mM ATP, pH 7.5, at the temperatures indicated in the presence of 3 M Grp94, if not indicated otherwise. For determination of the K m value, different concentrations of ATP were added, and the final buffer system of the assay was kept constant. The data were recorded for 40 min, and the assay was evaluated using the SigmaPlot software package (Systat Software, Erkrath, Germany).

Analytical Size Exclusion Chromatography
A SEC200 (GE Healthcare) gel filtration column was used. All of the experiments were performed at room temperature in 40 mM Hepes, pH 7.5, 150 mM KCl, and 5 mM MgCl 2 using a flow rate of 0.5 ml/min. The protein was observed with a Jasco FP 920 fluorescence detector (Jasco, Grossumstadt, Germany) using an excitation wavelength of 275 nm and an emission wavelength of 340 nm.

Circular Dichroism Spectroscopy
Far UV CD measurements were carried out in a Jasco J-715 spectropolarimeter (Jasco, Gross-Umstadt, Germany) equipped with a PTC343 peltier unit at 20°C. The proteins were dialyzed against 10 mM potassium phosphate, pH 7.5, overnight at 4°C. CD signals were accumulated ten times from 250 to 195 nm using a scanning rate of 20 nm/min.

Analytical Ultracentrifugation
To determine the sedimentation coefficients of Grp94, analytical ultracentrifugation was carried out using a Beckman XL-A ultracentrifuge equipped with a UV/Visible and interference detection unit (Beckman Coulter, Krefeld, Germany). The sample concentration was 1 mg/ml in 0.1 M Tris, pH 7.0, 50 mM KCl, and 2 mM MgCl 2 . Centrifugation was carried out at 25°C with a final rotation speed of 8.500 rpm in a TI-60 rotor. Detection was at 280 nm with UV scans recorded every 6 min. Evaluation of the data was carried out with the UltraScan software.

Transient Kinetics
Stopped-flow measurements were performed with a TgK Scientific (Bradford on Avon, UK) SF-61 DX2 instrument in 40 mM Hepes/KOH, pH 8.0, 150 mM KCl, and 5 mM MgCl 2 at 25°C if not mentioned otherwise. The excitation slit was set to 1 nm. For measurements of intrinsic tryptophan fluorescence excitation was set to 296 nm with a long pass filter (cut-off) of 320 nm for emission. Measurements with the fluorescent ATP analog (P␥)-MABA-ATP where performed with fluorescence energy transfer using 296 nm for excitation and a cut-off filter of 420 nm (13). Concentrations are noted as conc1/conc2 (syringe/ cell). Typically each experiment was performed three to six times, and the resulting time traces were averaged with the software from TgK Scientific.

Data Analysis
Data analysis was performed with Origin (MicroCal, LLC Northampton, MA) or Grafit, version 5.0.13 (Erithacus Software Limited, Horley, Surrey, UK). For binding isotherms (amplitude information from stopped-flow titrations) the quadratic equation was used that also includes the case of tight binding processes (43).
where F is the signal, F o is the initial amplitude, F max is the maximal amplitude, A o is the concentration of Grp94, B o is the concentration of ligand, and K D is the dissociation constant of ligand.

Binding Kinetics and Pseudo First Order Kinetics
In the case of experiments measuring association of nucleotides or RA, the concentration of ligand was increased starting with an excess to approximate pseudo first order binding kinetics. Accordingly, the primary data were analyzed with single exponential equations for the binding of (P␥)-MABA-ATP or nucleotides with the software of TgK Scientific.
where F is the signal, A is the amplitude, F o is the offset, k is the observed rate constant, and t is time.
For secondary data analysis, the observed rate constants (k obs ) as obtained from a single exponential fit of each time trace versus concentration of ligand in excess were analyzed with a linear equation. This follows from the fact that under pseudo first order conditions (one ligand in excess) and for a simple one step binding mechanism, k obs ϭ k on ϫ L excess ϩ k off , with L excess being the concentration of ligand in excess, and therefore a plot of k obs versus L excess gives k off as the intercept and k on as the slope (44).
The data were measured using the Foerster energy transfer from intrinsic tryptophan to the fluorescent group of the nucleotide. Experimental conditions were: temperature, 25°C; puffer, 40 mM Hepes/NaOH, pH 7.5, 150 mM KCl, and 5 mM MgCl 2 .
The affinity of unlabeled nucleotides were either calculated with K D ϭ k off /k on or from the amplitude information as determined using the intrinsic fluorescence signal of Grp94 and ATP/ADP or with competition experiments where (P␥)-MABA-ATP was simultaneously added with various concentrations of ATP/ADP. These competition data were analyzed with a cubic equation as described (45) using knowledge about (P␥)-MABA-ATP affinity determined independently with a pseudo first order concentration series as described above.
In the case of RA, the binding constant was determined with the well defined rate constant for association (k on ϭ slope) from a series of pseudo first order kinetics as described above and a time trace with initial equal concentrations of Grp94 and RA according to the following equation.
This procedure is necessary because the intercept defining k off is ill defined for RA because of its very low value.

Characterization of Recombinantly Expressed Grp94-To
analyze the ATPase activity of Grp94, we overexpressed fulllength canine Grp94 lacking the N-terminal signal sequence in the cytosol of E. coli and purified it to homogeneity. Mass spectrometric analysis revealed that the purified protein was without any detectable modifications or truncations (data not shown). SEC and analytical ultracentrifugation showed that the protein is a dimer in solution (data not shown). CD spectroscopy further confirmed that the protein exhibits the expected secondary and tertiary structure (data not shown).
Steady State ATPase Activity-The intrinsic ATPase activity of Grp94 was determined either by monitoring the hydrolysis of radiolabeled ATP or by a coupled colorimetric assay. To exclude the possibility that our Grp94 preparations contain contaminations that contribute to the overall ATPase activity, we performed steady state ATPase assays under standard conditions (see "Experimental Procedures") in the absence and in the presence of the anti-tumor drugs GA, RA, and the Grp94specific ATP-binding inhibitor NECA. GA binds with high affinity to the Hsp90 nucleotide pocket (9, 10, 46) and acts as a competitive inhibitor of ATP (46). A 5-fold molar excess of GA over Grp94 was sufficient to inhibit almost completely ATP hydrolysis in the presence of a 200-fold excess of ATP over Grp94 (Fig. 1A). Similar results were obtained in the presence of RA or NECA. Because NECA specifically inhibits Grp94, but not Hsp90 (33,36) and because GA and RA are specific for Hsp90s, these results prove that the ATPase activity observed in our experiments can be exclusively assigned to Grp94.
The calculated k cat of 0.36 min Ϫ1 at 25°C is relatively low compared with other ATP hydrolyzing enzymes but is in good agreement with turnover numbers that had been published for Hsp90 (8,47) or Hsc70/DnaK (48,49).
Next, we were interested in determining the temperature dependence of the Grp94 ATPase. We incubated Grp94 at the respective temperatures and followed the hydrolysis of radiolabeled ATP at the same temperature. As can be seen in Fig. 1B, the k cat increases with temperature. It reaches a plateau between 25 and 37°C. At higher temperatures, the turnover number drops significantly, possibly because of protein denaturation. To ascertain that the catalytic activity we monitored is exclusively that of Grp94, we performed experiments in parallel in which we preincubated the sample with NECA. Here, we did not observe significant ATPase activity at all of the temperatures tested. In contrast to the results obtained for Grp94, the activity of yeast Hsp90 increased with temperature up to 60°C (Fig. 1C). In the presence of GA, the observed ATPase activity could be completely suppressed.
The K m value and true k cat value at saturating ATP were determined with a coupled colorimetric assay as shown in Fig. 1D. The determined K m value for ATP of 92 M was very high. Furthermore, we determined an ATP turnover number of 0.36 min Ϫ1 at 25°C, which is nearly three times higher than that of Hsp90 from yeast (13). Although such a high K m value with a moderate ATPase activity in principle indicates the presence of contaminating ATPases, we could exclude that by the nearly complete inhibition of ATPase activity by NECA or RA as described above (Fig. 1A).
Single Turnover Experiments to Determine the Partitioning of the Grp94⅐ATP Complex-To identify the rate-limiting step of the ATPase cycle, we performed experiments under single turnover conditions in which we directly compared Grp94 to yeast Hsp90. In this type of experiment, Grp94 or yeast Hsp90 is nearly saturated with radiolabeled ATP and consequently only undergoes one round of hydrolysis. This allows neglect of nucleotide release and detection of a potential rate-limiting step prior to or concomitant with hydrolysis.
The ratio of radioactive ATP to Grp94 or yeast Hsp90 in these experiments was set at 0.8:1. As shown in Fig. 2A, the k cat of the hydrolysis reaction could be determined to be 0.36 min Ϫ1 under single turnover conditions at 30°C. Because this value is 80% of the respective value obtained under steady state conditions at 30°C (0.41 min Ϫ1 ), this experiment suggests that nucleotide release is not the rate-limiting step, consistent with the rate constant for dissociation of ADP as shown below. Consistent with previous results (13), nucleotide hydrolysis was also determined to be the rate-limiting step for yeast Hsp90 (Fig.  2B). Single turnover experiments can be further exploited to give valuable information about the commitment of the bound ATP to hydrolysis (13). To test this, an excess of cold ATP is  DECEMBER 7, 2007 • VOLUME 282 • NUMBER 49 added within the first minutes of the reaction. If the bound nucleotide is not trapped by the protein, it will readily exchange with unlabeled ATP and as a result the observed hydrolysis of radiolabeled ATP will stop immediately. On the other hand, if the bound ATP is trapped in the protein, its ability to exchange with cold ATP is decreased, resulting in the continued hydrolysis of radiolabeled ATP. In the case of Grp94, when an excess of cold ATP was added during the single turnover reaction, the signal intensity of the product of hydrolysis did not increase any longer, implying that the bound nucleotide is not protected against exchange during the ATPase reaction. However, when the experiment was performed under identical conditions with yeast Hsp90, the addition of a large excess of cold ATP did not result in an immediate stop of the hydrolysis reaction (Fig.  2B). A significant fraction of the bound ATP was still hyrdolyzed. This behavior is in marked contrast to Grp94 but consistent with previous results obtained for yeast Hsp90 (13). Thus, the partitioning experiments clearly show that Grp94bound ATP is not committed to hydrolysis in a closed nucleotidebinding pocket as it was observed for Hsp90 from yeast.

ATPase Cycle of Grp94
Analysis of Nucleotide Binding to Grp94-To assess the binding of ATP to Grp94 but also to determine the relevant kinetic constants for different nucleotides, binding transients were measured with a stopped-flow rapid mixing apparatus using the intrinsic tryptophan signal of Grp94. Upon the addition of nucleotide to Grp94, the fluorescence signal at 350 nm decreases by ϳ20% (corrected for dilution and inner filter effects) without a significant shift in the peak emission (data not shown). Using this signal, we measured the binding of various nucleotides under pseudo first order conditions with a long pass filter of 320 nm (see "Experimental Procedures"). Fig. 3A shows representative binding traces of ATP that could be well described with a single exponential equation without significant systematic deviations. Thus, ATP binding to Grp94 seems to obey a simple one step binding mechanism. Accordingly, the analysis of observed rate constants (k obs ) at different ATP concentrations with a linear equation gives the association and dissociation rate constants of ATP with k ϩ1 ϭ 0.24 Ϯ 0.02 M Ϫ1 s Ϫ1 and k Ϫ1 ϭ 0.8 Ϯ 0.3 s Ϫ1 , respectively (Fig. 3B). Analysis of the resulting amplitudes (Fig. 3C) of the kinetics performed at different ATP concentration with the appropriate binding equation (see "Experimental Procedures") results in a dissociation constant of K D ϭ 3.5 Ϯ 0.9 M that is consistent with the kinetically   Fig. 3), the rate constant for dissociation (k off ) and equilibrium dissociation constant (K D ) could be determined with high accuracy to K D ϭ 0.012 M and k off ϭ 0.0028 s Ϫ1 (see "Experimental Procedures" and Table 1). determined one (k Ϫ1 /k ϩ1 ϭ 3.4 M). This consistency further corroborates the assumption of a simple one-step binding model.
The binding of ADP and AMP-PCP was measured and analyzed with the same approach, assuming that there are no observable deviations in general binding properties. The resulting constants, as summarized in Table 1, show that the binding properties of ATP, ADP, and AMP-PCP are nearly identical with rate constants for association of 0.17-0.28 M Ϫ1 s Ϫ1 and rate constants for dissociation of 0.6 -0.8 s Ϫ1 . Accordingly, the calculated as well as measured equilibrium dissociation constants are ϳ2.4 -3.5 M.
The binding of the inhibitor RA could only partially be assessed by the approach described above. Although the slope (k ϩ1 ) was determined well with 0.24 M Ϫ1 s Ϫ1 , the intercept is too close to zero and thus is not determined sufficiently well. To nevertheless gain knowledge of the rate constant for dissociation of RA, we followed a different approach.
Here, Grp94 and RA were mixed in equimolar concentrations because the equation that describes a one-step binding process under these conditions allows determining k Ϫ1 accurately if k ϩ1 is known (see "Experimental Procedures"). The resulting rate constant for the dissociation of RA is 0.0028 s Ϫ1 (Fig. 3D) and thus appreciably slower than that observed for the nucleotides. Accordingly, the resulting dissociation constant is much lower (12 nM). This allows the conclusion that the high affinity of RA is solely based on a highly reduced rate constant for dissociation. Consistent with this observation is the rather closed conformation of the N-terminal domain of Grp94 in the presence of RA as opposed to the ATP state (32). The binding of GA and NECA could not be analyzed with stopped-flow measurements because they are either highly light-sensitive (GA) or produce no signal change upon binding (NECA) (data not shown). However, a K D of 200 nM for NECA was reported using isothermal microcalorimetry (36).
To further substantiate the results obtained using the intrinsic tryptophan fluorescence of Grp94, we additionally used the fluorescent nucleotide analog (P␥)-MABA-ATP as described previously (13). Displacement of (P␥)-MABA-ATP from a Grp94-(P␥)-MABA-ATP complex with excess unlabeled ATP allowed the direct measurement of the rate constant for dissociation of the labeled nucleotide with 0.11 s Ϫ1 (data not shown). The binding of (P␥)-MABA-ATP was measured by the fluorescence energy transfer signal (Trp/MABA) and variable (P␥)-MABA-ATP concentrations. A replot of the rate constants obtained from single exponential analyses is shown in Fig. 4. It gives 0.17 s Ϫ1 and 0.14 M Ϫ1 s Ϫ1 as rate constants for dissociation and association, respectively. Analysis of the amplitudes (Fig. 4, inset) using a quadratic binding equation results in a K D (P␥)-MABA-ATP ϭ 0.42 M compared with a calculated value of 0.82 M with the more accurately determined value of k Ϫ1 from the displacement experiment.

DISCUSSION
The Hsp90 family of molecular chaperones is characterized by an unusual ATP-binding site that is located in the N-terminal domain of the protein. Crystallographic analyses show that the ATP molecule is bound in a kinked conformation with the ␥-phosphate group sticking out of the binding pocket (9,10,33). ATP binding results in the repositioning of the so-called "lid" region in the ATP-binding domain, which consists of two helices and the loop connecting them. This lid domain closes over the ligand-binding site in response to nucleotide binding and at the same time allows dimerization of the N-terminal domains and the association of the N-terminal with the middle domain, which is required for efficient ATP hydrolysis (2,14). This sequence of events has been elucidated by a combination of biochemical approaches and x-ray crystallography, specifi-  a The rate constant for dissociation of radicicol under pseudo first order conditions could not be determined since it is too close to zero. Instead a combination of pseudo first order and second order kinetics with equal concentration initial conditions was used here (see "Experimental Procedures"). b These values were derived from complex kinetic competition experiments where either amplitudes were analyzed or replots of observed rate constants lead to rate constants for dissociation of (P␥ )-MABA-ATP, ADP, or ATP (40).
cally for cytosolic Hsp90 from S. cerevisiae (11,12,14,50). Although the similarity of the mechanism to that of DNA gyrase and MutL, which, together with Hsp90, form the GHKL ATPase family (51), suggests that it may be conserved within the Hsp90 family, there seem to be variations for different members of the Hsp90 family. Human cytosolic Hsp90 for example has an extremely low ATPase activity with 0.0054 min Ϫ1 (52). Furthermore, it has been suggested that for human Hsp90 N-terminal dimerization may not be part of the reaction cycle (53). For Grp94, the endoplasmic member of the Hsp90 family, the differences seem even more pronounced. Here, x-ray crystallography and biochemical experiments have shown that, whereas the overall architecture of the N-terminal domain is similar to that of cytosolic Hsp90 species, significant differences exist. A striking one is the specific binding of NECA, an adenosine derivative known to act as an adenosine A 2 receptor antagonist (54,55) and a specific Grp94-binding agent, to the ATP-binding site of Grp94 (36). This discriminates Grp94 from cytosolic Hsp90 members (36). Also, there are structural differences in the N-terminal domain that could be the basis for differences in the ATPase mechanism (33,34). Astonishingly, although binding of ATP and ADP to the N-terminal domain of Grp94 was observed (36), there are highly conflicting reports on the ATPase activity of Grp94, ranging from the complete absence of ATPase activity to significant turnover (35-37, 39, 56).
The results of this study show unambiguously that highly purified, recombinantly produced canine Grp94 has an ATPase with turnover numbers comparable with or even higher than those reported for cytosolic Hsp90 from yeast. This ATPase activity can be completely inhibited by NECA, which demonstrates that there were no contaminations of other ATPases in our preparations. Interestingly, canine Grp94 is a relatively labile protein. Already at temperatures above 30°C, the protein starts to loose activity. This is in contrast to cytosolic yeast Hsp90, which can be readily assayed at temperatures at least up to 40°C. At present it is not clear why in some previous studies on Grp94 ATPase activity was either very low or not detected at all. It may be due to the unusual properties of this ATPase including the relatively weak binding of nucleotides (compared with other ATPases), low turnover numbers and last but not least the instability of the protein. In addition, we cannot rule out that the ATPase activity of the authentic protein is influenced by posttranslational modifications.
Except for cytosolic Hsp90 from S. cerevisiae and Homo sapiens, a detailed kinetic analysis is missing for other family members, such as organellar Hsp90. Especially in the light of the differences in the structure of the N-terminal domain of Grp94 (33) compared with cytosolic Hsp90 species, it seemed rewarding to dissect the complete ATPase cycle to learn about potential mechanistic reasons for the discrepancies.
Already the initial step of the ATPase cycle of Grp94, nucleotide binding, is different from that of cytosolic Hsp90. Nucleotides interact in general more tightly (summarized in Table 1) than observed for other Hsp90 proteins. However, the kinetics of binding of nucleotides is in the same order of magnitude as determined for Hsp90 from yeast and human Hsp90 (13,53), and the rate constants for association (k ϩ1 in Fig. 5) of all nucleotides, as well as that of the inhibitor RA, are comparable with 0.16 -0.28 M Ϫ1 s Ϫ1 . Obviously, the differences in affinities of the nucleotide interaction between Grp94 and other Hsp90 family members are determined by variations in the rate constants for dissociation. It should be noted that the binding constants measured for ATP with stopped-flow experiments represent the true binding properties because ATP hydrolysis is appreciably slower (0.006 s Ϫ1 at 25°C) and thus does not interfere with the rapid binding process of ATP as measured by transient kinetics. Interestingly, studies with authentic Grp94 isolated from porcine pancreas gave similar results concerning ATP binding (36). Although not quantitatively determined, the affinity of ATP and ADP for Grp94 was found to be comparable, which is in contrast to most other Hsp90 proteins.
The ATPase activity of Hsp90 proteins is believed to be strictly coupled to a closure of the N-terminal domains and the association with the middle domains (1,2). This reaction results in the trapping of the nucleotide in the binding site and seems to be the rate-limiting step of the cycle (13). We focused on this important step as a clue for potential differences in mechanism. As previously observed for Hsp90 from yeast (13), single turnover experiments with Grp94 gave values comparable with that observed in steady state experiments. This shows that, in agreement with the dissociation rates for nucleotides, product release is not rate-limiting. In contrast to experiments for yeast Hsp90 (13), however, a kinetic partitioning analysis of Grp94 showed no commitment of the bound ATP for hydrolysis. This could either mean that k Ϫ2 greatly exceeds k hyd (Fig. 5) or that actually the enzyme is predominantly in the open form (E o ) in the presence of ATP. A strong indication that the latter is indeed true comes from the comparison of the measured K D and K m values for ATP, where the K m value (92 M) is substantially higher than the K D value (3.4 M), measured under otherwise nearly identical conditions. In this case, the K m value would be the product of K D1 (ATP binding) times K D2 , the value that reflects the equilibrium of open to closed form with ATP bound. Consistent with this interpretation is the large change in the tryptophan signal intensity of Grp94 mediated by RA as opposed to other, weaker binding ligands (nucleotides). This correlation is indicative of conformational changes in the presence of RA that induce the closed form. It should be noted that our experiments do not reveal the structural changes associated with the binding of the nucleotide. The terms "open" and "closed" thus refer in general terms to the closure of the nucleotide-binding site as evidenced by kinetic methods.
The structural analysis of the N-terminal domain of Grp94 in the presence or absence of ligands complements this picture. In the crystal structure, the N-terminal domain of Grp94 is in the closed state (with respect to the position of the subdomain helix 1-4-5) in the apo-form, similar to the structure observed in the presence of inhibitors (34). In contrast, in the presence of ATP, the open state is observed (32).
The kinetic dissection of the ATPase cycle presented here suggests that during the ATPase cycle Grp94 is predominantly in the open state (97%) concerning ATP binding. This is in marked contrast to the situation reported for yeast Hsp90 where the open state is only populated to 20% (13), and accordingly a closed structure is observed in the presence of nucleotides (14,50). Interestingly, a combined EM crystallography analysis of HtpG, the Hsp90 member from E. coli, also shows closure of the enzyme upon nucleotide binding (15). We assume that, despite differences in the kinetics/thermodynamics, the corresponding conformational rearrangements of the ATPase cycle are structurally similar for all Hsp90 proteins. However, the energetic balance between individual steps may be different, possibly also reflecting different types of regulation by cochaperones or substrates.