ATP binding and hydrolysis by the multifunctional protein disulfide isomerase.

We previously reported the ability of protein disulfide isomerase (PDI) to undergo an ATP-dependent autophosphorylation. Our efforts to map the modification site have been hindered by the low abundance and instability of the labeling. Results are presented in this paper on the nature of phospho-PDI, which appears as an intermediate with a half-life of 2.5-8.8 min in an ATPase reaction. ATP binds to PDI with high affinity, K9.66 μM, and the kinetic parameters K and k of the ATPase reaction were measured by using a pyruvate kinase-lactate dehydrogenase-coupled assay under various conditions. Strikingly, the ATPase reaction is stimulated in the presence of denatured polypeptides, while the disulfide oxidization activity of PDI is not affected by ATP. However, PDI is known to participate in various unrelated functions in the endoplasmic reticulum, and ATP could be involved in the regulation of one of these. The results are discussed in light of recent findings on ATP-chaperone relationships.

We previously reported the ability of protein disulfide isomerase (PDI) to undergo an ATP-dependent autophosphorylation. Our efforts to map the modification site have been hindered by the low abundance and instability of the labeling. Results are presented in this paper on the nature of phospho-PDI, which appears as an intermediate with a half-life of 2.5-8.8 min in an ATPase reaction. ATP binds to PDI with high affinity, K d 9.66 M, and the kinetic parameters K m ATP and k cat of the ATPase reaction were measured by using a pyruvate kinase-lactate dehydrogenase-coupled assay under various conditions. Strikingly, the ATPase reaction is stimulated in the presence of denatured polypeptides, while the disulfide oxidization activity of PDI is not affected by ATP. However, PDI is known to participate in various unrelated functions in the endoplasmic reticulum, and ATP could be involved in the regulation of one of these. The results are discussed in light of recent findings on ATP-chaperone relationships.
Although there is no direct experimental evidence of the presence of ATP in the lumen of the endoplasmic reticulum (ER), 1 it has been previously established that it is required to support the correct folding and disulfide bond formation of proteins (1). Molecular chaperones of the ER, such as BiP, grp94, and calnexin, hydrolyze or at least bind ATP in the course of their activity in vitro (2), which also strongly suggests that an ATP pool should exist in the lumen in this compartment. Finally, the characterization of an ATP translocator in the ER membranes of mammalian (3) and yeast (4) cells has settled the question of how ATP accumulates in the ER. Our laboratory reported that protein disulfide isomerase (PDI), an abundant microsomal protein, can undergo an ATP-dependent autophosphorylation in vitro and likely represents another physiological target for ATP in the ER (5).
PDI was originally characterized as an enzyme (E.C. 5.3.4.1) involved in the catalysis of the protein-disulfide formation (6,7), but in recent years, it has been found to be more than just a thioredoxin-like oxidoreductase. Its role in protein folding and cell physiology are more complex than initially thought since PDI now appears as a multifunctional protein (8) and even a chaperone (9,10). These features could reflect the general ability of this protein to bind various peptides and proteins apparently regardless of their sequence (7,11). PDI associates with various proteins such as the ␣-subunit of prolyl-hydroxylase, N-glycosyl transferase, and triglyceride transfer complex and has been revealed to be identical to the T3-binding protein, P55 (see Ref. 8 for a review).
Interestingly, it has been found by using a genetic approach in yeast that the essential function of PDI does not reside in the thioredoxin-like domains (12). In this context, we previously suggested that PDI phosphorylation modulates its interaction with various partners (5). The precise relationship between ATP and PDI remains elusive. Preliminary attempts to find the effect of ATP on PDI catalysis (13) or ATP hydrolysis (14) were unsuccessful. Using affinity chromatography on various immobilized denatured proteins, Nigam et al. (15) reported the Ca 2ϩdependent association of several ER proteins, among them PDI, with these substrates and their elution by ATP. However, the experiments did not provide a direct association between PDI and ATP.
The following paper presents the data we obtained on ATP binding by PDI and the use of a sensitive spectrophotometric method to demonstrate the existence of an ATPase activity for this protein. The kinetic parameters of this activity were measured under various conditions. The results suggest a new role for ATP in the ER, which uses the multifunctional PDI as a target.

EXPERIMENTAL PROCEDURES
Enzyme Preparations-Recombinant human PDI (rh-PDI) was obtained from an Escherichia coli strain BL21(DE3) transformed with the plasmid pTM2-PDI (kindly provided by Prof. K. Kivirikko, University of Oulu, Finland), as described previously (16). DsbA was obtained from an E. coli K10 strain transformed with the plasmid pPB2190, which was built by introducing the BspHI-SspI dsba-containing cassette of pPB2212 (17) into the expression vector pTrc99A (Pharmacia Biotech Inc.). E. coli thioredoxin was purchased from Promega. All proteins used in this study were controlled and eventually repurified by HPLC using a TSK3000SW (TosoHaas) gel filtration column (flow rate, 1 ml/min; mobile phase, 20 mM sodium phosphate; 0.3 M NaCl, pH 7.4). They were more than 95% pure when analyzed by SDS-PAGE.
Measurement of the Phospho-PDI Half-life-100 g of rh-PDI was pulse labeled for 5 min at 37°C in 0.5 ml of 0.1 M Tris acetate, 10 mM magnesium chloride, pH 8, containing 1 mM GSH, 0.2 mM GSSG with 60 Ci of [␥-32 P]ATP (final concentration, 40 nM). The mixture was separated into two samples, which were complemented with either 1 mM unlabeled ATP or nothing. At various times, 20-l aliquots were removed, and the reaction was stopped by instant freezing in liquid nitrogen. The phospho-PDI in each sample was separated by SDS, 10% PAGE (1 g per well) and quantitated by densitometry of the autoradiogram.
Coupled ATPase Assay-Continuous spectrophotometric measurement of the ATP conversion into ADP was performed in a pyruvate kinase-lactate dehydrogenase-coupled assay. 100 g of rh-PDI (final monomer concentration, 1.8 M) was allowed to equilibrate at 30°C for 5 min in 1 ml of 0.1 M triethanolamine, pH 7.5, 0.16 M potassium chloride, 50 mM magnesium sulfate, 0.25 mM phosphoenolpyruvate, 0.15 mM NADH, 6 g/ml pyruvate kinase, 2 g/ml lactate dehydrogenase. When indicated, 1 mM GSH, 0.2 mM GSSG, or 100 g of denatured * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
RNase-A, either reduced or alkylated with N-ethylmaleimide, was added. The reaction was started by addition of ATP (1-100 M) and monitored at 340 nm for 15 min.
PDI-mediated RNase A Refolding-The assay was carried out essentially as already described (18) with minor modifications. Ribonuclease A (Sigma) was denatured at 10 mg/ml in 0.1 M Tris acetate, pH 8.0, 0.1 M DTT, 6 M guanidine chloride and then dialyzed overnight against 0.1 M acetic acid, pH 3.5, 3 M guanidine chloride to remove DTT. This solution can be kept as a stock solution at Ϫ20°C for weeks. For the assay, 115 g were put in 1 ml of reaction mixture containing 1 mM GSSG, 0.2 mM GSH, 1 M PDI in 0.1 M Tris-HCl, pH 8. The residual 30 mM guanidine chloride has no effect on the PDI activity. The concentration of active RNase C R at each time was calculated using the equation and constants reported in Ref. 18 but modified to take into account the inhibitory effect of ATP on RNase activity (Equation 1). The inhibition constant for ATP K iATP was determined at 0.8 mM.
ATP Binding by Equilibrium Dialysis-Experiments were performed with an equilibrium microvolume dialyzer (EMD 101, Hoefer Scientific Instruments), 24 chambers, mounted around 12-14-kDa cut-off membranes. The wells were filled with 60 g of rh-PDI on one side of the membrane, [␣-32 P]dATP on the other side, and then filled up to 60 l with 0.1 M Tris-HCl buffer, pH 7.5. The system was rotated overnight at room temperature, and the concentrations of bound and free ligand were estimated by liquid scintillation counting on 50-l samples from both sides of the chambers.

RESULTS
The ability of PDI to autophosphorylate in the presence of [␥-32 P]ATP has been shown previously (5). However, the phospho-PDI obtained in such conditions turned out to be very unstable, rendering it very difficult to handle for further experiments such as mapping of the phosphorylation site. The transient nature of the phosphorylation was quantitatively approached, and half-lives of 2.5 and 8.8 min were measured under ATP chase or without chase conditions, respectively (Fig. 1). The maximal quantity of 32 P radioactivity incorporated into PDI indicated that phospho-PDI never accumulates to more than 0.4%. The substrate specificity of the phosphate donor was investigated, and we observed that both [␥- 32  3Ϫ did not give rise to PDI labeling (data not shown). These preliminary data suggested that phospho-PDI forms by a reaction involving the ␥-phosphoryl of ATP or dATP and requires cleavage of the bond between the ␥and ␤-phosphates. The presence of Mg 2ϩ in the phosphorylation mixture is critical for the formation of phospho-PDI; therefore, the substrate of the reaction is certainly the Mg-ATP complex.
The properties of this ATPase activity of PDI were assayed in a pyruvate kinase-lactate dehydrogenase-coupled reaction. A significant deviation of the absorbance at 340 nm was observed in presence of PDI, which implies a significant consumption of ATP (Fig. 2). Prokaryotic oxidoreductases related to PDI, i.e. thioredoxin and DsbA, are devoid of this activity. This is in good agreement with our preliminary report that the site of phosphorylation, and thus probably the ATPase active site, lies somewhere within the central domain of the protein (5), which is specific for the mammalian enzyme. This site is far away from the redox active sites in the sequence. Furthermore, the measurements of the rates of PDI-catalyzed refolding of denatured-reduced RNase A in the absence (0.70 Ϯ 0.02 mmol RNaseA⅐min Ϫ1 ⅐mmol PDI Ϫ1 ) or in the presence of ATP (0.70 Ϯ 0.02 mmol RNaseA⅐min Ϫ1 ⅐mmol PDI Ϫ1 ) show that ATP does not, or only very slightly, affects this PDI activity (Fig. 3). We are thus in agreement with a previous report describing the lack of direct effect of ATP on the disulfide formation catalysis (13).
However, when we measured the kinetic parameters of this ATPase reaction under various conditions, we saw an unexpected dependence of this parameter on the assay conditions. While the rate of hydrolysis is 0.057 M⅐min Ϫ1 ⅐M monomer Ϫ1 for PDI alone, it increases 6.7-fold when the hydrolysis is measured under the conditions of the RNase refolding assay, i.e. in the presence of GSH, GSSG, and RNase A (Fig. 4). A comparison of various effectors was carried out to clarify how the PDI-ATPase depends on assay conditions (Table I). The redox state of the medium does not seem to be the basis of this effect since GSH, GSSG, or DTT does not allow the hydrolysis to reach the optimal value. Nevertheless, both GSH and GSSG lead to a significant 2-fold increase. The presence of denatured RNase A, either alone or associated with the redox partners GSH and GSSG, enhances the ATPase 5-7-fold without significant alteration of the K m . The alkylation of the eight RNase thiol groups with N-ethylmaleimide did not alter its stimulatory effect on the hydrolysis. Furthermore, a control experiment performed with lysozyme displayed similar activity. The redox potential of the medium, and as a result the redox state of the PDI active site, is not responsible for the stimulatory effect on the ATPase activity. The complete chemical modification of the PDI active site with N-ethylmaleimide (verified by using the Ellman reagent) does not completely abolish the ATPase activity but decreases it by a factor 4 (data not shown). To sum up, the ATPase promotion likely ensues from the association with a polypeptidic effector in the assay mixture. In this respect, the difference observed between DTT and GSH could rely on the peptidic nature of glutathione. The k cat determined under optimal conditions, 0.37 min Ϫ1 (Table I), is in the same range as the rate measured for the PDI-catalyzed formation of native RNase A from either its reduced-denatured form (see above and Ref. 18) or its scrambled form (19).
However, ATPase and redox sites certainly function independently since various simultaneous measurements of these activities gave stoichiometric relationships between S-S bridge formation and ATP consumption ranging from 5 to 10, depending on the amount of PDI present in the assay (data not shown). When analyzed as a reciprocal plot 1/v ϭ f(1/S), the data contained in Fig. 4A depicted an upward curvature at very low ATP concentrations, which is indicative of a possible cooperativity in the ATP hydrolysis reaction (data not shown). This departure from the linearity neither results from the auxiliary system, since the proportionality curve was linear and intercepted the y axis at zero, nor from the enzyme preparation, since similar results were obtained for two preparations of rh-PDI and for bovine liver PDI. A Hill plot (Fig. 4B) gives a cooperativity number of 1.49 for the PDI dimer, which can be assumed to derive from the vicinity of two monomeric ATP binding sites within the dimer. We measured the ATP binding properties of PDI by using [␣-32 P]dATP as a ligand and in the absence of Mg 2ϩ to prevent hydrolysis during the time necessary to reach equilibrium. The intercept with the x axis of the Scatchard plot (Fig. 5), from which we calculated 0.7 ATP binding site per PDI monomer, confirms the presumption anticipated before from kinetics data. The dissociation constant for ATP is K D ϭ 9.66 Ϯ 0.48 M. Under comparable conditions, except for the presence of Mg 2ϩ , we measured a K m for Mg-ATP of 7.1 M (Table I), which suggests that ATP complexed with magnesium would have a better affinity for PDI. Otherwise, substrate binding will become a very limiting step in the reaction mechanism.
An additional experiment showing the coelution of PDI and ATPase activity on an HPLC column is presented to strengthen previous experiments. When pure rh-PDI is submitted to gel filtration (Fig. 6), three major peaks are detectable that correspond to various oligomeric states of the protein as described previously (20), namely high molecular weight form (A), tetramer (B), and dimer (C). The ATPase activity profile superimposed nicely with absorbance peaks, indicating that each oligomer certainly has a similar specific activity.

DISCUSSION
These results describe a new property for the multifunctional protein PDI. Both ATP binding and hydrolysis properties are clearly exhibited and were extensively characterized. The data presented here corroborate the preliminary evidence published in our first paper on bovine liver PDI (5). Several soluble and membrane-bound ATPases have been found in microsomal extracts that could contaminate our enzyme preparations. We ruled out this possibility for the following reasons: similar V m and K m were measured for both human recombinant PDI expressed in E. coli and for the bovine PDI extracted from ox liver microsomes (Fig. 4, Table I), the calculation of the concentra-  tion of ATP binding sites from the Scatchard plot gives a concentration very close to that of PDI and which cannot be that of a contaminant, and ATPase copurifies with PDI in HPLC. The rate of ATP hydrolysis could appear low compared with other kinases, even under the optimal conditions. However, it is in the same range as those measured for chaperones, i.e. less than min Ϫ1 , and its efficiency is quite high when expressed in term of k cat /K m , i.e. 85 ϫ 10 3 min Ϫ1 ⅐M Ϫ1 . Phospho-PDI appears as a covalent intermediate in the hydrolysis reaction. Its short half-life, low abundance, and instability partly explain why we did not succeed in purifying sufficient amounts of material to identify, at the sequence level, the site of phosphorylation. The PDI sequence was searched extensively for homologies with kinase, ATPase, and chaperone subsets from the EMBL data base as well as for motifs from the Prosite base without any significant hit. So, the mechanism of ATP hydrol-ysis might be original. The presence of peptide substrate in the assay increases the rate of hydrolysis while ATP by itself does not promote the apparent activity of PDI on RNase refolding. In some respects, the situation could be close to that of Hsc70-ATPase, which is stimulated 2-3-fold upon addition of an unfolded polypeptide (21). There is no apparent stoichiometry between ATP hydrolysis and S-S bridge formation. From a thermodynamic point of view, disulfide bond oxidization is exergonic and thus, a priori, does not require the energy input provided by the ATP hydrolysis. In the case of DnaK, it was clearly demonstrated that the ATPase is not stoichiometrically coupled to a peptide binding/release cycle (22,23). The requirement for ATP hydrolysis is not even absolute in the chaperone function of either GroEL (24) or DnaK (25). Both ATP and ADP are able to change the conformation of GroEL (26,27), whose protein binding properties are also regulated by heat shockinduced phosphorylation (28). The relationship between PDI and ATP is puzzling and might reveal another possible likeness with molecular chaperones. However, we regard this with extreme reservation and we are at work on it to validate this presumption. The function of PDI, which is possibly under ATP control, and the role of hydrolysis remain to be determined as well as whether the conclusions drawn in vitro are also relevant in the cellular context. FIG. 6. HPLC separation of PDI-associated ATPase. Five runs with 50 g of pure rh-PDI were performed on a gel filtration column (a typical chromatogram is shown). 25 0.5-ml fractions were collected across the separation and then pooled and concentrated to 50 l using Centricon 10 devices. ATPase activity was assayed on 25 l of each fraction in the presence of GSH, GSSG, and RNase. As depicted on the 10% SDS-PAGE, all peaks correspond to the PDI monomer (57 kDa) in various oligomeric states. Lane H represents rh-PDI before injection.