Fast kinetic analysis of conformational changes in mutants of the Ca(2+)-ATPase of sarcoplasmic reticulum.

Rapid quench experiments at 25 degrees C were carried out on selected mutants of the sarco(endo)plasmic reticulum Ca(2+)-ATPase to assess the kinetics of the conformational changes of the dephosphoenzyme associated with ATP binding/phosphoryl transfer and the binding and dissociation of Ca(2+) at the cytoplasmically facing transport sites. The mutants Gly(233) --> Glu, Gly(233) --> Val, Pro(312) --> Ala, Leu(319) --> Arg, and Lys(684) --> Arg differed conspicuously with respect to the behavior of the dephosphoenzyme, although they were previously shown to display a common block of the transformation of the phosphoenzyme from an ADP-sensitive to an ADP-insensitive form. The maximum rate of the ATP binding/phosphoryl transfer reaction was reduced 3.6-fold in mutant Gly(233) --> Glu and more than 50-fold in mutant Lys(684) --> Arg, relative to wild type. In mutant Leu(319) --> Arg, the rate of the Ca(2+)-binding transition was reduced as much as 10-30-fold depending on the presence of ATP. In mutants Gly(233) --> Glu, Gly(233) --> Val, and Pro(312) --> Ala, the rate of the Ca(2+)-binding transition was increased at least 2-3-fold at acid pH but not significantly at neutral pH, suggesting a destabilization of the protonated form. The rate of Ca(2+) dissociation was reduced 12-fold in mutant Pro(312) --> Ala and 3.5-fold in Leu(319) --> Arg, and increased at least 4-fold in a mutant in which the putative Ca(2+) liganding residue Glu(309) was replaced by aspartate. The data support a model in which Pro(312) and Leu(319) are closely associated with the cation binding pocket, Gly(233) is part of a long-range signal transmission pathway between the ion-binding sites and the catalytic site, and Lys(684) is an essential catalytic residue that may function in the same way as its counterpart in the soluble hydrolases belonging to the haloacid dehalogenase superfamily.

The challenge of understanding active ion transport across biological membranes is well illustrated by the sarco(endo)plasmic reticulum Ca 2ϩ -ATPase. 1 This enzyme is made up of a single 110-kDa peptide chain that catalyzes uphill Ca 2ϩ trans-port at the expense of ATP being hydrolyzed (1)(2)(3)(4)(5). The coupling between energy utilization and ion movement depends on long-range functional linkages between cytoplasmic and transmembrane protein domains. ATP hydrolysis occurs through formation of an aspartyl-phosphorylated intermediate at Asp 351 located in the large cytoplasmic domain (Fig. 1). The transfer of the ␥-phosphoryl group of ATP to Asp 351 requires the binding of two calcium ions at sites formed by residues located predominantly in the transmembrane segments M4, M5, and M6 (3)(4)(5). Before phosphorylation, Ca 2ϩ has access to these sites from the cytoplasmic side of the membrane, but following the phosphorylation Ca 2ϩ can only be released on the luminal side, in connection with or following the conversion of the phosphoenzyme from an ADP-sensitive form ("E 1 -P") to an ADP-insensitive form ("E 2 -P"). Subsequently, the aspartyl phosphate bond is hydrolyzed with liberation of P i , and the dephosphoenzyme returns to a state that readily binds Ca 2ϩ from the cytoplasmic side. The latter events are presumably accompanied by countertransport of protons (see reaction cycle in Fig. 1), and the dissociation of the translocated protons and binding of Ca 2ϩ at the cytoplasmically facing sites of the dephosphoenzyme take place as a series of consecutive reaction steps involving enzyme conformational changes, which may collectively be denoted "the E 2 to E 1 transition" or "the Ca 2ϩbinding transition" (1,4,6,7).
As an approach toward a detailed characterization of the transport-associated conformational changes we have used site-directed mutagenesis analysis to map several critical residues. Beginning with the Gly 233 (8) and Pro 312 (9), about 20 residues have now been shown to be crucial to the E 1 -P to E 2 -P transformation of the phosphoenzyme (for reviews, see Refs. 4, 5, and 10). This step is the slowest in the cycle and can be studied at 0°C using a manual mixing technique (8,9). The residues critical to the E 1 -P to E 2 -P conversion are located in various regions of the protein, including the large cytoplasmic loop that makes up the catalytic site, the fourth stalk segment that links the catalytic site to transmembrane segment M4, the ␤-strand domain that interconnects the transmembrane segments M2 and M3, and, to a lesser extent, the transmembrane segments.
Much less is known about the positions of residues that participate in the conformational changes of the dephosphoenzyme. These conformational changes can be analyzed by fast kinetic studies of phosphorylation. In classic rapid quench experiments with the Ca 2ϩ -ATPase isolated from rabbit muscle, the time course of the Ca 2ϩ -binding transition of the dephosphoenzyme was monitored by following the phosphorylation upon addition of Ca 2ϩ and [␥-32 P]ATP to enzyme initially present in the Ca 2ϩ -deprived form (11)(12)(13)(14)(15). Likewise, the time course of Ca 2ϩ dissociation from the dephosphoenzyme has been examined in the muscle enzyme by following the loss of phosphorylation activity upon removal of Ca 2ϩ from the me-dium, taking advantage of the dependence of the phosphoryl transfer reaction on the occupancy of the Ca 2ϩ sites (11,12,16,17). In the present work, we have for the first time been able to study phosphorylation of expressed mutant Ca 2ϩ -ATPases on a millisecond time scale. The aim was to determine the consequences of selected mutations for the kinetics of the conformational changes associated with the ATP binding/phosphoryl transfer reactions and the binding and dissociation of Ca 2ϩ at the cytoplasmically facing sites. One of the questions we have asked is to what extents these conformational changes of the dephosphoenzyme are affected by mutations previously shown to block the E 1 -P to E 2 -P transformation of the phosphoenzyme. Thus, we have selected a series of mutants with a very prominent block of the E 1 -P to E 2 -P transition: Gly 233 3 Glu, Gly 233 3 Val, Pro 312 3 Ala, Leu 319 3 Arg, and Lys 684 3 Arg (Fig. 1, cf. Refs. 5, 8, 9, 18, and 19). The phosphorylation kinetics of the Lys 684 3 Arg mutant is particularly interesting because Lys 684 is highly conserved within the whole haloacid dehalogenase superfamily of hydrolases, to which the Ca 2ϩ -ATPase and other P-type ATPases recently were assigned (20), suggesting a critical role of the lysine side chain in catalysis. In addition, we have included in some of the present studies mutant Glu 309 3 Asp, in which the ability to form a stable Ca 2ϩ -occluded form in the presence of the ␤,␥-bidentate chromium(III) complex of ATP was previously found to be lost even though the Ca 2ϩactivated phosphorylation by ATP is retained (21). These mutants display a surprising variability with respect to the characteristics of the reaction steps associated with the dephosphoenzyme.

Expression
The previously described mutant cDNAs (8,9,18,19,21), derived from the cDNA encoding the rabbit fast-twitch muscle Ca 2ϩ -ATPase (SERCA1a isoform), were transfected into COS-1 cells by the calcium phosphate procedure (22) and microsomal membranes were harvested by differential centrifugation as before (8,9). The amount of expressed enzyme was quantified by a specific sandwich enzyme-linked immunosorbent assay (18) using as standard expressed wild type for which the concentration was determined by measurement of the capacity for phosphorylation by inorganic phosphate in the presence of 30% dimethyl sulfoxide (23). The enzyme-linked immunosorbent assay was usually supplemented with a determination of the capacity for phosphorylation from ATP ("active site concentration"). The expression levels varied between 200 and 400 pmol of Ca 2ϩ -ATPase/mg of total microsomal protein. This is several hundredfold higher than the level of endogenous Ca 2ϩ -ATPase in the COS-1 cells and the contribution of the endogenous enzyme to the phosphorylation signals is, thus, negligible.

Quench-flow Instrumentation
All fast kinetic experiments were performed using a Bio-Logic quench-flow module QFM-5 (Bio-Logic Science Instruments, Claix, France). The complete experimental setup also included a control unit/ power supply (MPS-5) and a PC equipped with an RS-232 serial interface. The apparatus is designed to work with reaction volumes in the microliter range and can perform both single and double mixing experiments using three or four reaction syringes, respectively. In addition, a fifth waste syringe allows a design of the instrument that makes no use of valves to divert the mixture to be wasted or collected. All five syringe plungers are driven by independent microprocessor-controlled stepping motors. The reaction flow-line is constructed to drastically reduce dead volumes and contains 3 Berger ball mixers (24) connected by interchangeable reaction loops. The apparatus is enclosed in a water jacket to allow temperature regulation of reactant chambers and the reaction flow-line.
Two modes of operation were used for single mixing experiments, with only 2 mixers and one reaction loop in use. In the continuous mode, the flow rate and reaction loop volume were varied (0.2 to 4 l/ms and 7 to 200 l, respectively) to obtain reaction times between 10 and 800 ms. In the interrupted mode, the flow through the reaction loop (200 l) was stopped for serial intervals to obtain reaction times between 100 ms and 10 s. In double mixing experiments, all 3 mixers and both reaction loops were in use. The volume of the first reaction loop was varied (7 to 200 l), and both the continuous and the interrupted mode were used to vary the reaction times of the first reaction in the double mixing procedure between 19 ms and 10 s. To ensure a constant reaction time for the second reaction in the double mixing procedure, the flow rate remained unchanged as did the volume of the second reaction loop.

Mixing Protocols for Rapid Kinetic Studies
The quench-flow module was used to study the time course of phosphorylation from [␥-32 P]ATP of the expressed wild-type and mutant Ca 2ϩ -ATPases according to the protocols described below. All experiments were carried out at 25°C. In all four protocols, the acid-precipitated microsomal protein was washed by centrifugation and subjected to SDS-polyacrylamide gel electrophoresis at pH 6.0 as described previously (8,9) and the radioactivity associated with the separated Ca 2ϩ -ATPase band was quantified by "imaging" using a Packard Cyclone TM Storage Phosphor System. Background phosphorylation levels, determined by reversing the order of addition of [␥-32 P]ATP and quench solution or by substituting 2 mM EGTA for CaCl 2 , was subtracted from all data points.
The amount of expressed Ca 2ϩ -ATPase protein used per experimental point was 0.05-0.2 pmol. Because of the small reaction volumes, the amount of radioactivity used per experimental point could be restricted to less than 5 ϫ 10 4 Bq.

Curve Fitting and Simulation
The phosphorylation time courses were fitted to a monoexponential time function corresponding to first-order kinetics, using the SigmaPlot program (SPSS, Inc.), and when good fits could be obtained, the extracted rate constants were used to characterize the data. In many cases this procedure was feasible, and for the sake of simplicity we refrained from using more complex descriptions with several parameters.
In cases where there was an obvious deviation from monophasic kinetics, the data were fitted by computer simulation of the simplified reaction cycle in Scheme 1 using a kinetic simulation software, SimZyme, developed in this laboratory. SimZyme is based on the prin-SCHEME 1.
Fast Kinetics of Ca 2ϩ -ATPase Mutants ciples described in Ref. 25. For any choice of reaction cycle and rate constants, the program solves the relevant differential equations, using the 4th order Runge-Kutta numerical method, and provides a graphical representation of the time dependence of the concentrations of the reaction intermediates which can be compared with experimental data points. A detailed description of a program with similar characteristics has previously been published (26). Initial rates per ATPase molecule in units of s Ϫ1 were obtained from the slopes of the fitted curves at time 0 after normalizing the curves to the enzyme concentration.

RESULTS
Phosphorylation of Enzyme Preincubated with Ca 2ϩ -The first step in our fast kinetic characterization of the mutants was to study the phosphorylation of enzyme pre-equilibrated with Ca 2ϩ ("Ca 2 E 1 " in Fig. 1 or "Ca 2 E" in Scheme 1). Microsomes containing expressed wild type or mutant Ca 2ϩ -ATPase were preincubated in the presence of 100 M Ca 2ϩ to saturate the Ca 2ϩ -transport sites, and the time course of phosphorylation was monitored following addition of 5 M [␥-32 P]ATP (Fig.  2). Under these conditions an initial overshoot was observed for the wild type, corresponding to a phosphorylation level 5-10% higher than the steady-state level. The initial overshoot and the phosphorylation rate of the expressed wild-type Ca 2ϩ -ATPase are in good agreement with the behavior under comparable conditions of Ca 2ϩ -ATPase purified from rabbit skeletal muscle (14,27). The data for the wild type, including the overshoot, could be reproduced satisfactorily by computer simulation of the simplified reaction cycle in Scheme 1, using the SimZyme program described under "Experimental Procedures," and there was not much latitude in the choice of the rate coefficients. As shown by the line, the following rate coefficients gave a good fit: k A ϭ 35 s Ϫ1 , k B ϭ 5 s Ϫ1 , and k C ϭ 25 s Ϫ1 . The reverse reactions can be neglected due to the absence of added ADP and the 1000-fold lower Ca 2ϩ affinity of the phosphoenzyme relative to the dephosphoenzyme (1, 2).
The overshoot was lacking in the mutants, as expected on the basis of the previously observed block of dephosphorylation.
The rate of dephosphorylation is reduced 20 -100-fold or more, relative to the wild type in the mutants Gly 233 3 Glu, Gly 233 3 Val, Pro 312 3 Ala, Leu 319 3 Arg, and Lys 684 3 Arg, as a consequence of the block of the E 1 -P to E 2 -P conversion (8,9,18,19). In mutant Glu 309 3 Asp, the dephosphorylation rate is also very low, due to a block of the hydrolysis of the E 2 -P intermediate (21). For each mutant, the appearance of phosphoenzyme following addition of ATP to the Ca 2ϩ -saturated enzyme could be fitted to a monoexponential function, as shown by the lines in Fig. 2 for mutants Gly 233 3 Glu, Gly 233 3 Val, Pro 312 3 Ala, Leu 319 3 Arg, and Lys 684 3 Arg. The rate coefficient determined in this way corresponds to that termed k A for the wild type, since k B (the rate coefficient corresponding to dephosphorylation) is insignificant for the mutants. The apparent monophasic time course may suggest that under the present conditions a single rate-limiting step exists in the reaction path consisting of ATP binding and phosphoryl transfer to the Ca 2ϩ -saturated enzyme. However, it is impossible to completely rule out the existence of a small lag phase indicative of two or more consecutive slow reactions (17). For the mutants Gly 233 3 Val, Pro 312 3 Ala, and Leu 319 3 Arg, the fits of the data to monoexponential functions gave rate coefficients of 40 s Ϫ1 , 38 s Ϫ1 , and 34 s Ϫ1 , respectively, i.e. indistinguishable from the value k A ϭ 35 s Ϫ1 determined for the wild type. A lower rate coefficient of 25 s Ϫ1 was found for the Gly 233 3 Glu mutant. The significance of this difference was attested by its consistency over a range of ATP concentrations (not shown, but see, Fig. 3).
As further seen in Fig. 2, the Lys 684 3 Arg mutant phosphorylated with a rate coefficient of only 1.2 s Ϫ1 , i.e. an almost 30-fold reduction compared with the wild type. This feature of the Lys 684 3 Arg mutant indicates that Lys 684 plays an important role in ATP binding and/or catalysis. The Glu 309 3 Asp mutant was found to phosphorylate with a rate coefficient about 8-fold lower than that of the wild type (4.3 s Ϫ1 , data not shown), in good agreement with previous studies at 0°C (21).
ATP Dependence of Wild-type and Mutants Gly 233 3 Glu and Lys 684 3 Arg-To further analyze the mechanism behind the reduction of the rate constant for phosphorylation in the mutants Gly 233 3 Glu and Lys 684 3 Arg, we investigated the ATP concentration dependence. At each ATP concentration tested, the phosphorylation rates of the Gly 233 3 Glu and Lys 684 3 Arg mutants were lower than that of the wild type. Fig. 3 shows double reciprocal plots of the initial phosphorylation rate versus ATP concentration (1, 2, 5, and 10 M). The initial rates are in units of s Ϫ1 , because the absolute rates have been divided by the enzyme concentration to obtain the rate per ATPase molecule ("molecular rate"). For the wild type as well as the mutants, the data could be fitted satisfactorily by a straight line in accordance with the results obtained with the purified muscle enzyme (26). The extrapolated maximum rate of 225 s Ϫ1 corresponding to infinite ATP concentration and the K m value of 13 M determined for the wild type are fairly consistent with the literature on the purified muscle enzyme (17,26,27). For mutant Lys 684 3 Arg, the maximum rate was reduced more than 50-fold relative to wild type, suggesting that Lys 684 is highly important for the phosphoryl transfer reaction. For Gly 233 3 Glu, the maximum rate was reduced 3.6-fold. The K m value was found to be 7 M for each mutant and, thus, smaller than that of the wild type. It should be noticed that because K m is a function of several rate constants, including those that determine the maximum rate, the reduction of K m does not imply that the affinity for ATP is increased in the mutants.
The Ca 2ϩ -binding Transition at pH 7.0 -Having determined the rate of phosphorylation of Ca 2ϩ -saturated enzyme, we turned to the partial reaction steps associated with Ca 2ϩ binding. The two calcium ions required for phosphorylation bind consecutively and at least one relatively slow conformational change seems to be involved (1,2,6,11,14,15,26,28). When the enzyme is preincubated in the absence of Ca 2ϩ (presence of EGTA) followed by simultaneous addition of [␥-32 P]ATP and excess Ca 2ϩ , the rate of appearance of phosphoenzyme is limited not only by the steps associated with ATP binding and phosphoryl transfer, as in the above described experiments, but also by any slow step in the Ca 2ϩ -binding reaction sequence ("the Ca 2ϩ -binding transition"). Fig. 4 shows the results of experiments carried out according to this protocol at pH 7.0, with microsomes containing wild type or the Gly 233 3 Glu, Gly 233 3 Val, Pro 312 3 Ala, Leu 319 3 Arg, and Lys 684 3 Arg mutant ATPases. The initial phosphorylation overshoot seen for the wild type after preincubation in the presence of Ca 2ϩ disappeared when Ca 2ϩ was removed from the preincubation medium, as has previously been described for the purified muscle enzyme (12)(13)(14). This is explained by the rate limitation imposed by the Ca 2ϩ -binding transition, preventing the transient accumulation of phosphoenzyme in excess of the steadystate level (12)(13)(14). As illustrated by the lines in Fig. 4, good fits of the data to a monoexponential function could be obtained for the wild type as well as the mutants. The wild-type enzyme phosphorylated with an apparent rate constant of 21 s Ϫ1 corresponding to a 1.7-fold reduction relative to the value found when phosphorylation was initiated after preincubation with Ca 2ϩ . This shows that a relatively slow step had indeed been introduced by omitting the pre-equilibration with Ca 2ϩ . For mutant Leu 319 3 Arg, the rate coefficient observed after preincubation without Ca 2ϩ was only 0.70 s Ϫ1 , an almost 50-fold reduction compared with the rate coefficient determined after preincubation in Ca 2ϩ -containing medium and 30-fold lower than that of the wild type. This is a very significant effect showing that the introduction of an arginine as substitute for Leu 319 severely impairs the ability of the enzyme to undergo one or more steps in the Ca 2ϩ -binding transition. The mutants Gly 233 3 Glu, Gly 233 3 Val, and Pro 312 3 Ala displayed rate coefficients of 28 s Ϫ1 , 24 s Ϫ1 , and 27 s Ϫ1 , respectively. Although these values appear slightly higher than that of 21 s Ϫ1 determined for the wild type, the increase cannot be considered significant given the scatter of the experimental points. The rate coefficient of 28 s Ϫ1 for the Gly 233 3 Glu mutant is indistinguishable from the value observed for the same mutant following pre-equilibration with Ca 2ϩ (25 s Ϫ1 , see above). Thus, in the Gly 233 3 Glu mutant the Ca 2ϩ -binding transition does not seem to be rate-limiting for the reaction sequence leading to phosphorylation. For the Lys 684 3 Arg mutant, the rate coefficient observed after preincubation in the absence of Ca 2ϩ was 0.74 s Ϫ1 , i.e. even lower than the 1.2 s Ϫ1 described above for preincubation with Ca 2ϩ .
The Ca 2ϩ Binding Transition in the Leu 319 3 Arg Mutant in the Absence of ATP-Because even micromolar concentrations of ATP accelerate a slow step associated with Ca 2ϩ binding in the wild-type enzyme (15), we inquired whether the low rate of the Ca 2ϩ -binding transition in the Leu 319 3 Arg mutant might be a consequence of a defect in this ATP modulation. To examine the rate of the Ca 2ϩ -binding transition in the absence of ATP, we performed a series of measurements using the double mixing technique described by Guillian et al. (15). This approach is feasible when the phosphorylation rates measured with and without pre-equilibration with Ca 2ϩ differ. The en-zyme is preincubated in the absence of Ca 2ϩ as in the above described experiments, but instead of adding [␥-32 P]ATP and Ca 2ϩ simultaneously, Ca 2ϩ binding is initiated by addition of Ca 2ϩ without ATP. After a variable time interval, t, [␥-32 P]ATP is included to assess the amount of enzyme that during t has acquired the ability to react with ATP and become phosphorylated (i.e. which has bound two calcium ions), and acid quenching is performed 34 ms later. As discussed by Guillian et al. (15), the amount of phosphoenzyme measured at t ϩ 34 ms follows the same function of t as the appearance of ATP-reactive enzyme, and, hence, this protocol allows determination of the basic rate of the Ca 2ϩ -binding transition in the absence of ATP modulation. The results are shown in Fig. 5. The data could be fitted to a monoexponential time function, but in the case of the wild type with an offset indicating that part of the enzyme was initially able to react with Ca 2ϩ and become phosphorylated during the 34-ms incubation with ATP. The data are consistent with a rate coefficient of 4.3 s Ϫ1 for the Ca 2ϩbinding transition in the wild type, i.e. 5-fold lower than the ATP-modulated transition rate of 21 s Ϫ1 . For mutant Leu 319 3 Arg, the data are consistent with a rate coefficient of 0.43 s Ϫ1 , i.e. 10-fold lower than that of the wild type and 1.6-fold lower than the ATP-modulated rate of 0.70 s Ϫ1 in the mutant. Thus, both the basic rate of the Ca 2ϩ -binding transition measured in the absence of ATP and the modulatory effect of ATP are much reduced in the mutant compared with the wild type.
The Ca 2ϩ Binding Transition at pH 6.0 -It has previously been demonstrated that the rate of the Ca 2ϩ -binding transition in the purified muscle Ca 2ϩ -ATPase depends strongly on the pH (7,14,29,30). The reduction in the rate observed when the pH is lowered is probably a consequence of accumulation of an enzyme form ("EH 3 " or "E 2 ") with protons bound at the transport sites, perhaps in an occluded state (cf. Refs. 2, 4, 6, 7, and "Discussion"). Fig. 6 shows the results of experiments in which the Ca 2ϩ -binding transition of the expressed wild type and mutants Gly 233 3 Glu, Gly 233 3 Val, and Pro 312 3 Ala was examined at pH 6.0, using an experimental protocol otherwise similar to that described for Fig. 4 (preincubation in the absence of Ca 2ϩ followed by simultaneous addition of Ca 2ϩ and [␥-32 P]ATP). As illustrated by the line, the apparent rate constant, k obs , determined for the wild type by fitting the data in Fig. 6 to a monoexponential time function, is 3.8 s Ϫ1 , i.e. 5-6fold lower than that determined at pH 7.0 (cf. Fig. 4). For the mutants Gly 233 3 Glu, Gly 233 3 Val, and Pro 312 3 Ala, the fits to monoexponential functions illustrated in Fig. 6 gave apparent rate constants of 7.8 s Ϫ1 , 9.9 s Ϫ1 , and 8.1 s Ϫ1 , respectively, i.e. 2-3-fold higher values compared with wild type. When experiments at pH 6.0 were carried out according to the same protocol as in Fig. 2 (i.e. preincubation in the presence of Ca 2ϩ ), the rate constants for phosphorylation did not differ significantly from those determined at pH 7.0 (k A ϭ 35 s Ϫ1 , 30 s Ϫ1 , 45 s Ϫ1 , and 37 s Ϫ1 , for the wild type and mutants Gly 233 3 Glu, Gly 233 3 Val, and Pro 312 3 Ala, respectively, data not shown). Therefore, the difference between the wild type and the mutants seen in Fig. 6 can be ascribed to mutational effects on the Ca 2ϩ -binding transition. In fact, the observed 2-3-fold enhancement of the apparent rate constant k obs in Fig. 6 represents a minimum estimate of the mutational effects on the "true" rate coefficient k C for the Ca 2ϩ -binding transition. This is because k obs , describing the approach to steady state, approximately equals the sum of k C and the rate coefficient of dephosphorylation, k B . Thus, k obs is higher than k C in the wild type where the dephosphorylation is significant (the rate coefficients determined by computer simulation of the data for the wild type at pH 6.0 are k A ϭ 35 s Ϫ1 , k B ϭ 2 s Ϫ1 , and k C ϭ 2 s Ϫ1 , not shown), but in the mutants, where the dephosphorylation is negligible, k obs is very close to k C . Therefore, the enhancement of k C induced by the mutations may actually amount to as much as 4 -5-fold.
Dissociation of Ca 2ϩ -Finally, we focused on the rate of Ca 2ϩ dissociation at the cytoplasmically facing transport sites of the dephosphoenzyme. When a Ca 2ϩ chelator such as EGTA is added to Ca 2ϩ -saturated enzyme, the ability to phosphorylate will disappear at a rate corresponding to the dissociation of the calcium ion that is first to leave in the sequential mechanism, because both Ca 2ϩ sites must be occupied to allow phosphoryl transfer from ATP (17). It is therefore possible to determine the rate of this Ca 2ϩ dissociation step by rapid-quench phosphorylation measurements (11,12,17). We have used an experimental setup where the Ca 2ϩ -saturated enzyme is mixed simultaneously with [␥-32 P]ATP and excess EGTA at pH 7.0. When free Ca 2ϩ in the medium is removed with EGTA, the enzyme partitions between phosphorylation and dissociation of Ca 2ϩ , as illustrated in Scheme 2, and, hence, the amount of phosphoenzyme formed will depend on the rate constants of both processes. The results of such experiments with the wild type are illustrated in the upper panel of Fig. 7, where the phosphoenzyme level at various time intervals after addition of [␥-32 P]ATP together with EGTA is shown as percentage of the maximum level reached in the presence of Ca 2ϩ in the medium (the 100% value in Fig. 2). As a consequence of dissociation of part of the bound Ca 2ϩ during the incubation with [␥-32 P]ATP, the phosphorylation reaches a lower peak value relative to that seen in the presence of Ca 2ϩ in the medium (compare Fig. 7 with Fig. 2). Furthermore, the phosphorylation level tends toward zero after a few hundred milliseconds, because the removal of Ca 2ϩ prevents rephosphorylation after the dephosphorylation has occurred. As shown by the line, the data in the upper panel of Fig. 7 could be reproduced satisfactorily by computer simulation of Scheme 2, using the rate constants k A ϭ 35 s Ϫ1 and k B ϭ 5 s Ϫ1 , determined in Fig. 2, together with a value of k ϪCa ϭ 27 s Ϫ1 for the rate constant corresponding to Ca 2ϩ dissociation, thus demonstrating that Scheme 2 provides a sound description of the phosphorylation under these conditions. The value of k ϪCa ϭ 27 s Ϫ1 giving the best fit to the data lies within the range of 15-55 s Ϫ1 reported in the literature for the purified muscle enzyme under conditions that differ only slightly from ours (12,17,31).
On the basis of the above described principles, we designed a simpler approach in which k ϪCa is determined from the ratio EP ATPϩEGTA /EP ATP obtained corresponding to a single time point of 34 ms. EP ATPϩEGTA is the amount of phosphoenzyme measured 34 ms after simultaneous addition of [␥-32 P]ATP and EGTA, whereas EP ATP is the amount of phosphoenzyme measured after 34 ms incubation of the Ca 2ϩ -saturated enzyme with [␥-32 P]ATP in the absence of EGTA. The time point of 34 ms was selected, because it is convenient that the dephosphorylation is negligible, so that the ratio EP ATPϩEGTA /EP ATP can be translated to a value for k ϪCa by numerically solving Equation 1, where k A is known from the data in Fig. 2. This equation has been derived from Scheme 2 under the assumption that k B [Ca 2 EP] Х 0. To examine whether the dephosphorylation can actually be considered insignificant under the present conditions, we have in the lower panel of Fig. 7 for t ϭ 34 ms and various values of k ϪCa compared the expected ratios EP ATPϩEGTA /EP ATP calculated by using Equation 1 and k A ϭ 35 s Ϫ1 (line indicated "calculated") with the corresponding ratios generated by computer simulation under the assumption that k A ϭ 35 s Ϫ1 and k B ϭ 5 s Ϫ1 (line indicated "simulated"). It can be seen that the two lines are indistinguishable, thus indicating that it is legitimate to neglect the dephosphorylation Ala mutants was carried out as described for Fig. 4, except that pH was 6.0 (see "Experimental Procedures"). The data for the wild type and each mutant were normalized separately taking the maximum level of phosphorylation as 100%. The lines show the best fits to a monoexponential function, giving the following rate constants: wild type, 3.8 s Ϫ1 ; Gly 233 3 Val, 9.9 s Ϫ1 ; Gly 233 3 Glu, 7.8 s Ϫ1 ; Pro 312 3 Ala, 8.1 s Ϫ1 . SCHEME 2.

Fast Kinetics of Ca 2ϩ -ATPase Mutants
for times Յ34 ms, when k B is 5 s Ϫ1 (corresponding to wild type) or lower (corresponding to mutants). Fig. 8 shows experimental data obtained for the wild type and all the mutants at t ϭ 34 ms and the corresponding k ϪCa values derived using Equation 1. The k ϪCa value of 27 s Ϫ1 for the wild type is identical to that obtained independently in Fig.  7, hence validating the simpler approach based on a single time point at 34 ms. It is also seen in Fig. 8 that k ϪCa is reduced as much as 12-fold in mutant Pro 312 3 Ala relative to wild type. A significant reduction (3.5-fold) was also found for mutant Leu 319 3 Arg, whereas the mutants with alterations to Gly 233 or Lys 684 displayed k ϪCa values indistinguishable from that of the wild type. For comparison, we also included the Glu 309 3 Asp mutant in these experiments. This mutant has previously been shown to be defective with respect to Ca 2ϩ occlusion (21) and in line with this finding the rate of Ca 2ϩ dissociation determined in Fig. 8 was conspicuously increased (at least 4-fold) relative to wild type. DISCUSSION In the present work, we have used the quench flow technique to analyze mutational effects on reactions associated with the dephosphoenzyme in mutants showing a common block of the E 1 -P to E 2 -P transition of the phosphoenzyme. As summarized in Table I, our analysis reveals a surprising variability among these mutants with respect to the behavior of their dephos-phoenzymes. We have focused on three reaction sequences: "the ATP-binding/phosphoryl transfer reactions" (studied in Figs. 2 and 3), "the Ca 2ϩ -binding transition" (studied in Figs. 4, 5, and 6), and "Ca 2ϩ dissociation" (studied in Figs. 7 and 8), each of which consists of ligand addition or dissociation step(s) as well as one or more conformational changes, and in Table I the mutational effects are expressed in terms of changes to the "overall" rate constants for these reaction sequences, k A , k C , and k ϪCa .
The Ca 2ϩ -binding transition was conspicuously slowed down in mutant Leu 319 3 Arg relative to wild type and accelerated in mutants Gly 233 3 Glu, Gly 233 3 Val, and Pro 312 3 Ala at pH 6.0, but not significantly at pH 7.0. It has been proposed that before Ca 2ϩ binding the wild-type enzyme exists in a pH-dependent pre-equilibrium of two conformational states, E 1 and E 2 , of which only E 1 is able to bind Ca 2ϩ with high affinity and become phosphorylated by ATP (1,7,14,29). According to the detailed analysis by Forge et al. (6,30), there are as many as three Ca 2ϩ free species with different degrees of protonation, EH 3 (equivalent to E 2 ), EH, and E (together constituting E 1 ). EH can bind one calcium ion, but only the fully deprotonated form, E, is able to bind both calcium ions as required for phosphorylation by ATP (6,30). The deprotonation steps are relatively slow (30) and may represent conformational changes associated with "deocclusion" of H ϩ countertransported in a way analogous to K ϩ transport by the closely related Na ϩ ,K ϩ -ATPase (2)(3)(4)(5). At pH 6.0 in the absence of Ca 2ϩ , almost all the wild-type enzyme resides in the EH 3 form, whereas EH prevails at neutral pH and E at alkaline pH, explaining the pH dependence of Ca 2ϩ binding in the wild type (6,30). The increased rate of the Ca 2ϩ -binding transition observed for mutants Gly 233 3 Glu, Gly 233 3 Val, and Pro 312 3 Ala at pH 6.0 may, therefore, be the result of a destabilizing effect of the mutations on the fully protonated EH 3 species, leading to accumulation of the Ca 2ϩ -reactive EH and/or E forms. In the E 1 -E 2 notation, this would correspond to a displacement of the E 1 -E 2 pre-equilibrium in favor of E 1 . This hypothesis predicts that the accelerating effect of the mutations on the Ca 2ϩbinding transition should be much less pronounced at pH 7.0 compared with pH 6.0, because at pH 7.0 the E 1 forms accumulate in the wild type as well. The actual finding is in accordance with the prediction. The severely reduced rate of the Ca 2ϩ -binding transition observed for mutant Leu 319 3 Arg could be due to a stabilization of the protonated EH 3 species (E 2 ). It is possible that the ATP-induced acceleration of the deprotonation (32) is defective in this mutant, since the rate coefficients determined in the presence and absence of ATP differed much less in the mutant than in the wild type (compare Figs. 4 and 5). An alternative interpretation would be that another partial reaction step in the Ca 2ϩ -binding transition, not modulated by ATP, is slowed down by the mutation so that it contributes more significantly to rate limitation in the mutant than in the wild type. In this connection it is interesting to note that mutant Leu 319 3 Arg in addition to the reduced rate of the Ca 2ϩ -binding transition displayed a 3.5-fold reduced rate of Ca 2ϩ dissociation (Fig. 8).
Either observation would be explained if the bulky and positively charged side chain of arginine imposes steric or electrostatic hindrance to Ca 2ϩ diffusion to and from the binding sites or to the local conformational changes associated with opening and closure of the binding pocket in the E 1 form (33). This would imply that Leu 319 is rather close to the cytoplasmic inlet to the ion binding pocket, which is not unrealistic given the structural model in Fig. 1 where Leu 319 is located right at the boundary between the fourth transmembrane segment and its cytoplasmic extension.
An even more pronounced reduction (12-fold) of the rate of Ca 2ϩ dissociation from the dephosphoenzyme was found for the Pro 312 3 Ala mutant (Fig. 8). Such a low rate of Ca 2ϩ dissociation under conditions where the rate of Ca 2ϩ binding appears to be normal (Fig. 4) suggests that the stability of the Ca 2ϩ bound state, perhaps the occluded form, is increased in the mutant relative to the wild type. The effect of the Pro 312 3 Ala mutation on Ca 2ϩ dissociation is most likely associated with the position of Pro 312 in the M4 helix right above one of the putative Ca 2ϩ liganding residues, Glu 309 (Fig. 1). Presumably, the alanine substituent stabilizes the Ca 2ϩ bound state by removing the kink of the M4 helix introduced by the proline.
In previous work, we have demonstrated that Glu 309 is involved in the binding of the most superficially located calcium ion (the one that binds last in the sequential binding process) (4), and measurements of Ca 2ϩ occlusion following incubation with the ␤,␥-bidentate chromium(III) complex of ATP and removal of non-occluded Ca 2ϩ by gel filtration in detergent solution suggested that the bound Ca 2ϩ does not become properly occluded in the Glu 309 3 Asp mutant (21). This hypothesis seems to be further substantiated by the present finding of a conspicuous enhancement by the Glu 309 3 Asp mutation of the rate of Ca 2ϩ dissociation from the dephosphoenzyme in the presence of the natural substrate MgATP and in the absence of the perturbing influence of detergent. The side chain of Glu 309 , which is shortened by one methylene group in this mutant, may possibly constitute an essential element in a molecular gate at the entrance to the Ca 2ϩ binding pocket.
Contrary to the Glu 309 , Pro 312 , and Leu 319 mutants, the mutants with alteration to Gly 233 or Lys 684 showed no significant change in the rate of Ca 2ϩ dissociation relative to wild type. This is consistent with the proposed location of the latter two residues outside the membrane far from the ion-binding sites (Fig. 1). The above discussed effect of the Gly 233 mutations on the E 1 -E 2 conformational equilibrium suggests a role of the glycine in long-range conformational changes. It is noteworthy in this connection that Gly 233 is highly conserved among P-type ATPases with various cation specificities, and studies of proteolytically cleaved Na ϩ ,K ϩ -ATPase have suggested that intactness of the peptide segment right on the C-terminal side of the glycine is important for the coordination between structural changes of the catalytic site and the ion-binding sites (34,35). Interestingly, the present results demonstrate a specific effect of the Gly 233 3 Glu mutation on ATP binding/phosphoryl transfer. In the Gly 233 3 Glu mutant, the phosphorylation rate measured after pre-equilibration with Ca 2ϩ was significantly lower than that of the wild type over a range of ATP concentrations, and our analysis indicates that the maximal rate of phosphorylation is lower than one-third that of the wild type. This effect, which may be due to the negative charge of the glutamate as it was not observed with valine as substituent, is in line with the hypothesis that the glycine is part of a longrange signal transmission pathway between the ion-binding sites and the catalytic site.
Finally, our fast kinetic studies reveal the importance of Lys 684 for the ATP-binding/phosphoryl transfer reactions. This lysine is also highly conserved among the P-type ATPases and on the basis of sequence and structural alignment studies it was recently suggested to be equivalent to a lysine that plays a central role in the catalytic mechanism of soluble hydrolases belonging to the haloacid dehalogenase superfamily (20,36). We previously found that the ability of the Ca 2ϩ -ATPase to undergo phosphorylation by ATP is retained following substitution of Lys 684 with arginine, and the most noticeable characteristic of the Lys 684 3 Arg mutant observed previously was its inability to form the E 2 -P phosphoenzyme intermediate, both in the forward direction of the reaction cycle, from E 1 -P, as well as in the backward reaction with P i (18). The present investigation has, however, revealed that the maximal rate of phosphorylation is reduced more than 50-fold relative to wild type. This finding is in accordance with a critical role for Lys 684 in stabilization of the excess of negative charge developed in the transition state complex between the active-site aspartate and the ␥-phosphoryl group of ATP, as predicted on the basis of the catalytic mechanism of the haloacid dehalogenases (20). Hence, a structural and functional homology between the P-type AT- a The number in parentheses indicates the maximal molecular rate (cf. Fig. 3). b The number in parentheses indicates the rate determined in the absence of ATP (cf. Fig. 5). c For the wild type the k obs value is higher than k C , see text. d ND, not determined.
Pases and the soluble hydrolases belonging to the haloacid dehalogenase superfamily seems to be corroborated by the present data and it is likely that the side chain of Lys 684 is close to that of Asp 351 in the three-dimensional structure. In the light of this information, the inability of the Lys 684 3 Arg mutant to form E 2 -P may as well be explained by a requirement for the lysine side chain to participate in the interaction with the phosphoryl group.