Asparagine 706 and Glutamate 183 at the Catalytic Site of Sarcoplasmic Reticulum Ca2+-ATPase Play Critical but Distinct Roles in E2 States*

Mutants with alteration to Asn706 of the highly conserved 701TGDGVND707 motif in domain P of sarcoplasmic reticulum Ca2+-ATPase were analyzed for changes in transport cycle kinetics and binding of the inhibitors vanadate, BeF, AlF, and MgF. The fluorides likely mimic the phosphoryl group/Pi in the respective ground, transition, and product states of phosphoenzyme hydrolysis (Danko, S., Yamasaki, K., Daiho, T., and Suzuki, H. (2004) J. Biol. Chem. 279, 14991–14998). Binding of BeF, AlF, and MgF was also studied for mutant Glu183 → Ala, where the glutamate of the 181TGES184 motif in domain A is replaced. Mutations of Asn706 and Glu183 have in common that they dramatically impede the function of the enzyme in E2 states, but have little effect in E1. Contrary to the Glu183 mutant, in which E2P slowly accumulates (Clausen, J. D., Vilsen, B., McIntosh, D. B., Einholm, A. P., and Andersen, J. P. (2004) Proc. Natl. Acad. Sci. U. S. A. 101, 2776–2781), E2P formation was not detectable with the Asn706 mutants. Differential sensitivities of the mutants to inhibition by AlF, MgF, and BeF made it possible to distinguish different roles of Asn706 and Glu183. Hence, Asn706 is less important than Glu183 for gaining the transition state during E2P hydrolysis but plays critical roles in stabilization of E2P ground and E2·Pi product states and in the major conformational changes associated with the Ca2E1P → E2P and E2 → Ca2E1 transitions, which seem to be facilitated by interaction of Asn706 with domain A.

The sarco(endo)plasmic reticulum Ca 2ϩ -ATPase (1) is an energytransducing enzyme of the P-type that couples hydrolysis of ATP to translocation of Ca 2ϩ from the cytosol to the endoplasmic reticulum. In this control of cytosolic Ca 2ϩ concentration, the Ca 2ϩ -ATPase plays a vital role in cellular activation events, such as muscle contraction, hormone secretion, immune responses, cell migration, and protein synthesis. Ca 2ϩ transport is coupled to ATP hydrolysis by a reaction cycle (Scheme 1), in which the enzyme is transiently phosphorylated at a conserved aspartic acid residue and undergoes major conformational changes (2,3).
In recent years, several high resolution crystal structures of the Ca 2ϩ -ATPase, each thought to represent a particular intermediate state in the pump cycle, have been solved (4 -9). The Ca 2ϩ -ATPase consists of a membrane-spanning domain of ten helical segments and a large cytoplasmic head piece, comprising three distinct domains, named "N" (nucleotide binding), "P" (phosphorylation), and "A" (actuator). By combining crystallographic data with functional changes in site-specific mutants, an increasingly detailed picture of the mechanisms of energy interconversion and ion translocation in the Ca 2ϩ -ATPase is emerging. Thus, the catalytic function in E1 (autokinase activity) and E2 forms (autophosphatase activity), the movement of Ca 2ϩ ions across the membrane, as well as the major rate-limiting conformational changes of the cycle, i.e. E2 3 E1 and E1P 3 E2P, can all be understood on the basis of the sequential gathering and displacement of certain conserved amino acid motifs in domains N and A relative to the catalytic site in domain P and the coupling of these events to rearrangements of the transmembrane helices containing the high affinity Ca 2ϩ sites.
In the present study, we address the role of Asn 706 at the catalytic site of Ca 2ϩ -ATPase and revisit a previously examined mutant, E183A (10). Asn 706 and Glu 183 reside in the conserved 701 TGDGVND 707 (domain P) and 181 TGES 184 (domain A) motifs, respectively, and both residues are found in all known P-type ATPases (11). In fact, Asn 706 is highly conserved even in the superfamily of phosphohydrolases and phosphotransferases (the HAD superfamily), which, given the similarities in reaction mechanism, protein sequence, and structural architecture of the catalytic site, are believed to share a common evolutionary ancestor with the phosphorylation domain of the P-type ATPases (12)(13)(14). The side chains of Asn 706 and Glu 183 are both centrally located at the catalytic site in the E2 forms of Ca 2ϩ -ATPase, close to the phosphorylated aspartate, Asp 351 (6,7,9). In E1 conformations, Glu 183 has departed the phosphorylation site, whereas Asn 706 retains its close proximity to Asp 351 (4,5,7,8). In our previous study of Glu 183 (10), we demonstrated that substitution of the glutamate with alanine leads to a much reduced rate of both E2P hydrolysis and of the reverse phosphorylation of E2 with P i , suggesting that Glu 183 is critical for E2P transition state stabilization and catalysis. This proved to correlate well with the subsequently published E2⅐AlF 2 crystal structure, in which Glu 183 seems to coordinate and likely activates the attacking water molecule in the transition * This work was funded in part by grants from the Danish Medical Research Council, state complex (6). Here we show that Asn 706 likewise plays a crucial role in phosphorylation of E2 with P i , as well as in the major protein conformational changes involved in phosphoenzyme and dephosphoenzyme processing. We describe here, for the first time, measurements of the apparent binding affinities of mutants for the phosphoryl analogs AlF and MgF recently used for crystallization, as well as BeF, assumed to be an analog of phosphate in the E2P ground state (15). These measurements suggest different roles for Asn 706 and Glu 183 during catalysis of the E2P 7 E2 reaction.

Mutagenesis, Expression, and Assays of the Overall Reaction-Oligo-
nucleotide-directed mutagenesis of cDNA encoding the rabbit fasttwitch muscle Ca 2ϩ -ATPase (SERCA1a isoform) was carried out as described previously (16). The cDNA encoding the Ca 2ϩ -ATPase mutant E183A was the same as that applied in two previous studies (10,17). For expression, the wild-type or mutant cDNA, inserted in the pMT2 vector (18), was introduced into COS-1 cells (19) by transfection using the calcium phosphate precipitation method (20). The microsomal fraction containing expressed wild-type or mutant Ca 2ϩ -ATPase was isolated by differential centrifugation (21). The concentration of expressed Ca 2ϩ -ATPase was quantified by a specific enzyme-linked immunosorbent assay (22,23). Transport of 45 Ca 2ϩ into the microsomal vesicles was measured by filtration, and the ATPase activity was determined by measuring the amount of P i liberated in the presence of ionophore A23187 to prevent inhibition caused by rebinding of Ca 2ϩ to the luminally facing Ca 2ϩ sites (24,25).
Phosphorylation from [␥-32 P]ATP and 32 P i -Manual mixing experiments at various buffer and temperature conditions (detailed in the figure legends) were carried out according to the principles described previously (16,23,24,26). Transient kinetic experiments at 25°C were performed using a Bio-Logic quench-flow module QFM-5, as described (27). Acid quenching of phosphorylated enzyme was performed with 0.5-2 volumes of 25% (w/v) trichloroacetic acid containing 100 mM H 3 PO 4 . The acid-precipitated protein was washed by centrifugation and subjected to SDS-polyacrylamide gel electrophoresis in a 7% polyacrylamide gel at pH 6.0 (26,28), 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 were subtracted from all data points. The background was determined in parallel experiments, either with control microsomes isolated from mock-transfected COS-1 cells or by adding an excess of EGTA (in ATP-phosphorylation experiments) before initiating the phosphorylation. In some experiments, the constant phosphorylation level reached after an exponential decay was taken as background (usually ϳ5% of the initial phosphorylation).
Assays for Binding of Vanadate and Fluorides-We used the previously described assay for vanadate binding (29), and this assay, which is based on the inability of the inhibitor-bound Ca 2ϩ -ATPase to be phosphorylated by ATP, was modified to study also the binding of AlF, BeF, MgF, and ADP⅐AlF to wild-type and mutant Ca 2ϩpumps. Microsomes were pre-equilibrated at 25°C and pH 7, either in the absence of Ca 2ϩ to allow accumulation of the E2 state or in the presence of Ca 2ϩ for accumulation of Ca 2 E1. The reaction with inhibitor was then initiated by addition of varying concentrations of AlCl 3 , BeSO 4 , or MgCl 2 at a fixed concentration of NaF with or without ADP. In experiments involving AlF, BeF, or ADP⅐AlF, the concentration of Mg 2ϩ (a cofactor of the reaction with the inhibitor (30 -32)) was kept as low as 200 M during the inhibition step to avoid formation of MgF. Following incubation with the inhibitor for 30 min at 25°C and subsequent cooling for 10 min at 0°C, the amount of inhibitor-free enzyme was determined by phosphorylation for 10 s at 0°C with 5 M [␥-32 P]ATP. In the experiments where inhibition was carried out with enzyme pre-equilibrated in the absence of Ca 2ϩ , excess Ca 2ϩ was added prior to the phosphorylation step (ϳ5 s before the addition of ATP). The data presented in the supplemental material (Figs. SI and SII) validate the above described method.  Fig. SII that dissociation of the complexes of Ca 2ϩ -ATPase with vanadate, AlF, BeF, and MgF is very slow at 0°C, demonstrating that the 5-s incubation with Ca 2ϩ prior to phosphorylation does not cause significant dissociation of en-zyme⅐inhibitor complex. However, because the enzyme⅐BeF complex was more Ca 2ϩ -sensitive than the complexes with AlF and MgF, we used a free Ca 2ϩ concentration of only 100 M for activation of phosphorylation in the experiments with BeF, whereas the Ca 2ϩ concentration was 500 M in the experiments with AlF and MgF.
Assays for Nucleotide Binding-The synthesis of [␥-32 P]TNP-8N 3 -ATP, the photolabeling of COS-1 cell microsomes containing wild-type or mutant Ca 2ϩ -ATPase, the inhibition of photolabeling by ATP, and the quantification of labeled bands by radioimaging following SDSpolyacrylamide gel electrophoresis were carried out as described previously (33)(34)(35).
Calculations and Data Analysis-Experiments were conducted at least twice, and the complete set of data was analyzed by nonlinear regression using the SigmaPlot program (SPSS, Inc.) or by computation using the SimZyme program (27,36). The analysis of ligand concentration dependences was based on the Hill equation, EP ϭ EP max ⅐[L] n /(K 0.5 n ϩ [L] n ), or, for the concentration dependences of inhibitory ligands, on the Hill equation . The "true" dissociation constant for ATP binding in [␥-32 P]TNP-8N 3 -ATP labeling experiments was calculated using the validated equation for competitive inhibition (33).

RESULTS
Expression, Overall Activity, and Phosphorylation of Ca 2ϩ Pumps with Alterations to Asn 706 -Asn 706 of the Ca 2ϩ -ATPase was replaced by either alanine, cysteine, or serine, these mutations being chosen to either completely remove the side-chain carboxamide function (alanine) or remove the nitrogen, while retaining an oxygen in the side chain (serine), or introduce minimal negative charge (cysteine, the ionized fraction depending on the pK and pH), in all cases with only minor changes of the volume of the side chain. The mutant proteins were expressed in COS-1 cells, and the expression level of the mutants was similar to that obtained with wild-type Ca 2ϩ -ATPase, as evaluated by their immunoreactivity in a specific enzyme-linked immunosorbent assay (data not shown). To obtain an initial overview of the functional consequences of the mutations, we measured the overall rates of 45 Ca 2ϩ transport into the microsomal vesicles and ATP hydrolysis at 37°C and saturating substrate conditions. In either assay, the three mutants displayed no significant activity above the background level obtained with control microsomes harvested from mock-transfected COS-1 cells. For N706A, the lack of ATPase activity confirms a previously published result (37).
We then proceeded to test the phosphorylation of the mutants, from [␥-32 P]ATP in the presence of Ca 2ϩ , as well as from 32 P i without Ca 2ϩ in the backward direction of the normal reaction cycle (Scheme 1). The steady-state level of phosphoenzyme formed in the presence of 5 M [␥-32 P]ATP at 0°C was wild-type-like for all three mutants (Fig. 1, upper  gel). However, no significant phosphoenzyme was formed in the mutants in the presence of 500 M 32 P i (Fig. 1, lower gel), under conditions where E2P accumulates for wild type (25°C, presence of Me 2 SO, absence of Ca 2ϩ , and incubation for 10 min). Under these conditions, the concentration of P i giving half-maximal phosphorylation of wildtype enzyme is ϳ10 M, and with 500 M P i the reaction reaches steadystate within ϳ10 s (10). Thus, it is clear from Fig. 1 that the replacement of Asn 706 affects the ability to form the E2P state dramatically. It should be noted that this finding is at variance with a recent study (38) reporting that mutant N706A phosphorylates with P i to ϳ50% of the extent seen for wild-type Ca 2ϩ -ATPase. The latter study (38) also reported 19% phosphorylation with P i of another mutant (T353A), which we previously found unable to phosphorylate from P i (39). We have no explanation of this apparent discrepancy. Our results were reproducible in several independent experiments carried out with different microsome preparations, and our background phosphorylation level was always rather low for mutants as well as wild-type (Ͻ5% of the maximum phosphorylation level obtained with ATP).
Ca 2ϩ and MgATP Dependence of Phosphorylation from ATP-The ability of the mutants to form a phosphoenzyme from ATP allowed us to further study the partial reactions of the pump cycle in phosphorylation experiments. First, we tested the Ca 2ϩ concentration dependence of steady-state phosphorylation from [␥-32 P]ATP ( Fig. 2A). Small, but significant, deviations from the wild-type enzyme were found with the mutants, with the activation curve of N706S being slightly left-shifted relative to wild type and that of N706C being right-shifted. Similar effects were seen for the MgATP dependence of phosphorylation (Fig.  2B), corresponding to a 3-fold increase of affinity for N706S and a 4-fold reduction for N706C, relative to wild type. The observed changes of apparent affinity for Ca 2ϩ as well as MgATP seem to be kinetic effects, resulting from a reduced rate of phosphoenzyme turnover in combination with various degrees of slowing of the phosphorylation rate (see below and "Discussion").
Nucleotide Binding-In a recent mutagenesis study (35) we investigated the importance of several other conserved domain P residues for the nucleotide binding properties of the Ca 2ϩ -ATPase by studying the nucleotide concentration dependence of TNP-8N 3 -[␥-32 P]ATP photolabeling and ATP/MgATP competitive inhibition thereof. Asp 703 and Asp 707 , which are close to Asn 706 both in the primary and the threedimensional structure, proved of importance for ATP/MgATP binding. Thus, charge-removal from the side chains of these aspartates enhanced ATP binding up to 14-fold in the absence of Mg 2ϩ , and inhibited ATP binding up to 8-fold in the presence of Mg 2ϩ . Table 1 shows the results of similar experiments carried out with the Asn 706 mutants. It is clear that Asn 706 is not a critical residue for TNP-8N 3 -ATP or ATP binding, neither in the absence nor presence of Mg 2ϩ , as the reduction of affinity was maximally 2-fold relative to wild type. It is noteworthy, however, that the largest reduction of affinity was seen in the presence of Mg 2ϩ and for N706C, which also showed a significant reduction of apparent MgATP affinity in the titration of steady-state phosphorylation described above.
Time Course of ATP Phosphorylation of Ca 2ϩ -saturated Enzyme- Fig.  3A shows the results of rapid kinetic measurements at 25°C of the time course of phosphorylation from 5 M MgATP of enzyme pre-equilibrated with Ca 2ϩ . Compared with wild type, the phosphorylation rate was significantly reduced for N706A (1.8-fold) and N706C (4.4-fold) but unaltered for N706S (Fig. 3A). Because a subsaturating ATP concentration was used, the measured phosphorylation rate depends not only on the rate constant for transfer of the ␥-phosphate of ATP to Asp 351 at the catalytic site but also on the affinity of the catalytic site for ATP. Hence, at least for N706C, the slight reduction in MgATP affinity described above could contribute to determine the observed reduction of phosphorylation rate.  Fig. 3B shows the time course of phosphorylation determined under conditions similar to those corresponding to Fig. 3A, except that the enzyme was pre-equilibrated in the absence of Ca 2ϩ (presence of EGTA), and phosphorylation was initiated by mixing with a buffer containing [␥-32 P]ATP and excess Ca 2ϩ . For wild type, the rate obtained under these conditions is 1.8-fold lower than that obtained following pre-equilibration with Ca 2ϩ , reflecting the ratelimiting nature of the Ca 2ϩ binding transition (i.e. E2 3 E1 conformational change and accompanying Ca 2ϩ binding) preceding the phosphorylation reaction (27). The three mutants displayed 4-to 6-fold lower rates of phosphorylation when starting from Ca 2ϩ -deprived enzyme as compared with Ca 2ϩ -pre-equilibrated enzyme (Fig. 3, compare A and B). Relative to wild type, the Ca 2ϩ binding transition was 4.4-, 7.4-, and 3.2-fold slowed in N706A, N706C, and N706S, respectively (Fig. 3B).

Rate of the Ca 2ϩ -binding Transition in the Presence or Absence of ATP-
In the experiment corresponding to Fig. 3B, 5 M MgATP was present during the course of the Ca 2ϩ binding transition. Because even micromolar concentrations of ATP accelerate E2 3 E1 (40), we speculated whether the inhibition of the Ca 2ϩ binding transition in the Asn 706 mutants could result from defective ATP modulation of E2 3 E1. We therefore examined the rate of the Ca 2ϩ binding transition using an assay in which MgATP is absent during Ca 2ϩ binding (Fig. 4A). Enzyme pre-equilibrated with EGTA was mixed with an excess amount of Ca 2ϩ and incubated for varying time intervals (t in the mixing protocol at the top of Fig. 4A). The amount of phosphorylatable Ca 2 E1 accumulated during the Ca 2ϩ incubation step was then determined for each time interval by a further 34-ms incubation with 5 M [␥-32 P]ATP and 5 mM Mg 2ϩ , followed by acid quenching. Because the temperature and buffer conditions during the Ca 2ϩ binding transition in this assay were identical to those applied in the experiment corresponding to Fig. 3B, except for the absence of MgATP, the rate constants obtained in the two assays can be compared directly, and their ratio reflects the modulatory influence of MgATP on the Ca 2ϩ binding transition. For both wild type and mutants, the Ca 2ϩ binding transition was ϳ2-fold slower in the absence of MgATP as compared with its presence, implying that MgATP modulation of the E2 3 E1 transition is unaffected by the Asn 706 mutations. Hence, relative to wild type the Ca 2ϩ binding transition in the absence of MgATP was 6.6-, 8.6-, and 4.5-fold slowed in N706A, N706C, and N706S, respectively (Fig. 4A).
Ca 2ϩ Dissociation from the High Affinity Binding Sites in Ca 2 E1-The function of the two high affinity Ca 2ϩ sites in E1 was studied by measurement of the rate of Ca 2ϩ dissociation from Ca 2 E1, taking advantage of the fact that only the E1 form with two bound Ca 2ϩ ions can be phosphorylated by ATP (41). As illustrated by the mixing protocol at the top of Fig. 4B, enzyme pre-equilibrated with Ca 2ϩ was mixed with excess EGTA and incubated for varying time intervals. The amount of phosphorylatable Ca 2 E1 remaining after the EGTA incubation step was then determined for each time interval by a further 34-ms incubation with 5 M [␥-32 P]MgATP, followed by acid  quenching. Because only the Ca 2ϩ -bound enzyme fraction phosphorylates, the rate of disappearance of ability to phosphorylate reflects the rate of Ca 2ϩ dissociation. As seen in Fig. 4B, there were only minor differences between the Ca 2ϩ dissociation rates of the wildtype and the Asn 706 mutants, apparently excluding a role for Asn 706 in Ca 2ϩ binding or in the conformational changes involved in Ca 2ϩ dissociation from the high affinity sites. The Ca 2 E1P 3 E2P Conformational Transition of the Phosphoenzyme-To investigate the processing of the phosphoenzyme, the enzyme was phosphorylated with [␥-32 P]ATP under conditions where Ca 2 E1P accumulates as the major steady-state intermediate in the wild-type enzyme (0°C, presence of K ϩ , neutral pH). Phosphoenzyme decay in the forward direction of the pump cycle was examined by addition of excess EGTA to terminate phosphorylation by removing Ca 2ϩ , followed by acid quenching at varying time intervals. As shown in Fig. 5 (open symbols), all three mutants displayed reduced rates of phosphoenzyme turnover relative to wild type (2.4-, 8.8-, and 2.9-fold for N706A, N706C, and N706S, respectively). For wild type, the conformational change of the phosphoenzyme, the Ca 2 E1P 3 E2P transition, is rate-limiting for the overall ATPase reaction, whereas the ensuing dephosphorylation of E2P (cf. Scheme 1) is much faster in the presence of K ϩ at neutral pH. The Ca 2 E1P state is characterized by being able to donate its phosphoryl group back to ADP, forming ATP, whereas E2P is ADP insensitive and dephosphorylates only by hydrolysis of the acyl phosphate. To determine whether the Ca 2 E1P 3 E2P step or E2P 3 E2 is rate-limiting in the mutants, we measured the ADP sensitivity of the accumulated phosphoenzyme. The dephosphorylation solution was supplemented with 1 mM ADP, and phosphoenzyme decay was again followed (Fig. 5, solid symbols). For the three mutants, as well as the wild type, all phosphoenzyme disappeared completely within 5 s of the addition of ADP, demonstrating that the accumulated phosphoenzyme was entirely ADP-sensitive Ca 2 E1P. Similar experiments were carried out under buffer conditions where a high level of E2P accumulates at steady state in wild type (pH 8 and K ϩ The lines show the best fit of a monoexponential decay function, giving the rate constants indicated in parentheses: circles, wild type (3.2 s Ϫ1 ); squares, N706A (3.8 s Ϫ1 ); triangles pointing upward, N706C (2.0 s Ϫ1 ); triangles pointing downward, N706S (2.9 s Ϫ1 ). In each case, the phosphorylation level corresponding to zero time was taken as 100%. replaced by Li ϩ , cf. Ref. 10), and again no significant level of E2P was detected in the Asn 706 mutants (data not shown). The fact that no ADPinsensitive E2P had accumulated shows that the low rate of phosphoenzyme processing in the mutants, corresponding to the open symbols in Fig. 5, results from a block of the Ca 2 E1P 3 E2P transition, and that E2P hydrolysis is not grossly slowed in the mutants.
Affinity of E2 for Vanadate-As demonstrated above (Fig. 1), none of the Asn 706 mutants showed any phosphorylation from inorganic phosphate, even after incubation for a long time at a concentration of P i 50-fold higher than the concentration required for half-maximal phosphorylation of wild type. Because the mutants could easily undergo phosphorylation from ATP, the binding of the phosphate analog vanadate could be examined by taking advantage of the competition between ATP and vanadate, as previously described (29). Vanadate, which often is considered to mimic a penta-coordinated transition state of the phosphoryl group, binds to E2 in a slow reaction that requires Mg 2ϩ (42,43). Ca 2ϩ -deprived enzyme was equilibrated at 25°C with varying concentrations of vanadate in the presence of Mg 2ϩ , and the level of vanadatefree enzyme was determined for each vanadate concentration after cooling to 0°C (to slow vanadate dissociation as much as possible), by measuring the phosphorylation level obtained upon addition of excess Ca 2ϩ and [␥-32 P]ATP. As seen in Fig. 6, the wild-type enzyme displayed half-maximal inhibition of phosphorylation at 0.16 M vanadate. In contrast, the three mutants with alterations to Asn 706 displayed K 0.5 values at least three orders of magnitude higher than that of wild type ( Fig. 6 and Table 2). A similar insensitivity to inhibition by vanadate was previously described for a mutant in which Glu 183 in domain A had been replaced by alanine (Ref. 10; for comparison the previously reported data for E183A is included in Fig. 6 and Table 2). However, the E183A mutant differed from the mutants with alterations to Asn 706 by being able to slowly form E2P from P i , i.e. the rate of catalysis was more strongly affected than the stability of E2P in this case. For further comparison, the E183A mutant was included in the experiments described below.
Affinity of E2 for the Phosphate Analogs AlF, BeF, and MgF-Like vanadate, the complexes of Al 3ϩ , Be 2ϩ , and Mg 2ϩ with fluoride are considered phosphate analogs that bind to the Ca 2ϩ -deprived E2 state of Ca 2ϩ -ATPase (15, 30 -32 Table 2). In light of the arrangements of AlF and MgF in crystal structures of other phosphotransferases, the E2⅐AlF form was suggested to represent the transition state occurring during E2P hydrolysis, either as AlF 4 Ϫ or as AlF 3 (in both cases with planar geometry), with two oxygen atoms coordinating to the aluminum at apical positions to produce a state superposable (AlF 3 ) or analogous (AlF 4 Ϫ ) to the trigonal bipyramidal structure of the penta-coordinated phosphorous in the transition state of an in-line associative acylphosphate hydrolysis mechanism. The E2⅐MgF form was, on the other hand, suggested to mimic the product state of E2P hydrolysis (E2⅐P i ) with the non-covalently bound P i being represented by MgF 4 2Ϫ of tetrahedral geometry (15), see Table 2. The recently published crystal structures of Ca 2ϩ -   ATPase in E2 with bound AlF or MgF (6, 7) seem to confirm the proposals by Danko et al. (15) with respect to the structure of the bound AlF and MgF complexes, whereas no crystal structure of Ca 2ϩ -ATPase with bound BeF has yet been solved. Thus, the structure analogous to genuine E2P is still missing. The distinctions are important for the pumping mechanism, because the AlF and MgF protein complexes do not have an obvious passage for Ca 2ϩ access from the lumen, whereas such a passage may exist in E2⅐BeF and E2P (15).
To obtain more detailed information about the roles of Asn 706 and Glu 183 in interaction with the phosphoryl group in various E2 states, and to learn more about the differential characteristics of the fluoride complexes and vanadate, we studied the binding of AlF (Fig. 7A), BeF (Fig. 7B), and MgF (Fig. 7, C and D) to wild type and mutants, using the same method as described above for vanadate, which takes advantage of the competition at the catalytic site. This was feasible, because these phosphate analogs, like vanadate, all dissociate slowly from the Ca 2ϩ -ATPase (BeF was found to dissociate at approximately the same rate as vanadate under the conditions applied, and AlF and MgF dissociated even slower, see Fig. SII in supplemental materials). To study AlF binding, Ca 2ϩ -deprived enzyme was incubated for 30 min at 25°C with 2 mM NaF, 200 M Mg 2ϩ , and varying concentrations of AlCl 3 , followed by cooling on ice. Mg 2ϩ at low concentration is required for the formation of the enzyme complex with AlF, but only 200 M was added during the AlF binding step to avoid formation of MgF, cf. Fig. 7 (C and D). The uncomplexed fraction of the enzyme was then determined by measuring the phosphorylation occurring during a 10-s incubation with 5 M [␥-32 P]ATP at 0°C after supplementing the medium with excess Ca 2ϩ and 5 mM Mg 2ϩ . Wild-type Ca 2ϩ -ATPase displayed halfmaximal inhibition of phosphorylation at 8.8 M AlCl 3 (Fig. 7A). In control experiments where no NaF had been added, Ͼ200 M AlCl 3 (probably binding at Ca 2ϩ sites) was required to obtain half-maximal inhibition (broken line in Fig. 7A; data points shown only for wild type, but identical data were obtained for the mutants). The mutants with alterations to Asn 706 were also very sensitive to inhibition by AlF, contrasting their lack of sensitivity to vanadate (cf. Fig. 6). Thus, the AlF inhibition profile of N706S was indistinguishable from that of the wild type, and N706A and N706C displayed ϳ3-fold reduced apparent affinities for AlF, relative to wild type ( Fig.  7A and Table 2). E183A, on the other hand, displayed very low AlF affinity (16-fold reduced relative to wild type). In fact, there was little difference between the result of the AlF titration experiment with E183A and the control experiment, where Al 3ϩ was added in the absence of NaF.
We then proceeded to study the binding of BeF to E2 (Fig. 7B and Table 2). The assay was carried out as described above for AlF, except that AlCl 3 was replaced by BeSO 4 . Wild-type Ca 2ϩ -ATPase displayed half-maximal inactivation at 0.8 M BeSO 4 , i.e. 11-fold higher apparent affinity than with AlCl 3 . The effects of the Asn 706 mutations on BeF binding were markedly stronger than the effects on AlF binding. Thus, N706C displayed a 32-fold reduced affinity for BeF relative to wild type (compare with the 3.5-fold reduced affinity of N706C for AlF). A similar pattern was seen for N706A and N706S (15-and 6.1-fold reduced affinities for BeF, respectively, compare with the respective 2.9-fold reduced and wild type-like affinities for AlF). E183A was rather similar to the Asn 706 mutants with respect to BeF binding, displaying 9.3-fold reduced affinity for BeF relative to wild type. As was the case with Al 3ϩ , Be 2ϩ in the absence of NaF also inhibited the enzyme at high concentrations (broken line in Fig. 7B; data points only shown for wild type, but identical data were obtained for the mutants).
The binding of MgF to E2 was likewise examined (Fig. 7, C and D, and Table 2). The results shown in Fig. 7C were obtained as described above for AlF, except that AlCl 3 was replaced by MgCl 2 (and Mg 2ϩ thus excluded from the standard reaction buffer), and that 5 mM NaF was used instead of 2 mM to increase the amount of MgF formed as much as possible (the formation constant for MgF is considerably lower that those pertaining to AlF and BeF (44), with resulting lower expected apparent affinity for MgF as compared with the other fluoride complexes). In Fig. 7D, the roles of NaF and MgCl 2 were reversed: the added MgCl 2 concentration being kept constant at 5 mM while the added NaF concentration was varied (note the broken line, demonstrating that fluoride alone at concentrations up to 10 mM is without effect, this also serves as fluoride control in the other experiments). Irrespective of whether MgCl 2 or NaF was varied, the results looked the same, consistent with the idea that the inhibitory compound is a complex between Mg 2ϩ and F Ϫ and not any of these species in their uncomplexed state. All four mutants displayed significant affinity shifts relative to wild type, with N706A and N706C showing the most marked effects (7.4-to 16-fold and 3.9-to 8.3-fold reduction of apparent affinity, respectively), N706S (1.5-to 1.9-fold reduction of apparent affinity) being only slightly less sensitive to MgF inhibition than wild type, and E183A displaying an intermediate behavior (2.7-to 4.1-fold reduction of apparent affinity).
Affinity of Ca 2 E1 for ADP⅐AlF-ADP and AlF bind together with high affinity in a complex with the Ca 2 E1 state of wild-type Ca 2ϩ -ATPase, with concomitant tight occlusion of the two Ca 2ϩ ions at the transport sites (31). Because the Ca 2 E1 enzyme complex with ADP⅐AlF is believed to mimic the transition state in the transfer of the ␥-phosphoryl group from ATP (5, 7), we also studied the effects of the Asn 706 and Glu 183 mutations on formation of this complex, using the same assay as described above, but including 100 M Ca 2ϩ and 1 M ADP during complex formation (Fig. 8). The presence of Ca 2ϩ ensures that the enzyme is in Ca 2 E1. Hence, the E 2 conformation reacting with AlF in the absence of ADP (cf. Fig. 7A) is depleted (as illustrated by the broken line in Fig. 8, inhibition by AlF is rather weak in the absence of ADP when Ca 2ϩ is present). The ADP concentration of 1 M was chosen as a compromise, being sufficient to cause marked inhibition of wild-type Ca 2ϩ -ATPase in the presence of AlCl 3 and NaF, without interfering significantly with the phosphorylation of uncomplexed enzyme from MgATP (data not shown). Fig. 8 shows that the wild type displayed a K 0.5 of 3.7 M for inhibition under these conditions. N706A and N706C displayed markedly increased K 0.5 values (reduced apparent affinities) relative to wild type (4.1-and 15-fold, respectively, see Table 2), whereas N706S differed only slightly from wild type. E183A likewise displayed only 1.8-fold reduced apparent affinity for ADP⅐AlF relative to wild type ( Fig. 8 and Table 2).

DISCUSSION
In this study, we have explored the functional consequences of mutations of Asn 706 in domain P of Ca 2ϩ -ATPase. It is informative to analyze the results obtained with the Asn 706 mutants in relation to the crystal structures of the Ca 2ϩ -ATPase (Fig. 9), and furthermore to compare with the results obtained previously (10), and in the present study, with mutant E183A, where the glutamate of the conserved TGES motif of domain A is replaced. Mutations of Asn 706 and Glu 183 have in common that they dramatically impede the function of the enzyme in E2 forms, but have less striking effects on E1. In our previous study (10), we showed that replacement of Glu 183 with alanine leads to a reduced rate of both E2P dephosphorylation and the reverse phosphorylation of E2 with P i , suggesting that Glu 183 is very critical for catalysis and, thus, for transition state stabilization. The functional analysis turned out to correlate well with the location of Glu 183 in the subsequently published E2⅐AlF crystal structure (Fig. 9). In this structure, the AlF complex  (modeled as AlF 4 Ϫ ) is planar and is positioned linearly between one of the carboxylate oxygens of Asp 351 and a water molecule, held and likely activated by Glu 183 for nucleophilic attack on the phosphorous atom (6). It is clear from the present results that Asn 706 is also a critical residue for E2P formation. However, whereas for mutant E183A E2P phosphoenzyme formed from P i did, in fact, accumulate (though slowly relative to wild type), none of the mutants with alterations to Asn 706 showed any accumulation of E2P, even after incubation for a long time at a concentration of P i 50-fold higher than the concentration required for halfmaximal phosphorylation of wild type (Fig. 1). The lack of ability to phosphorylate from P i prevented us from studying the rate of E2P hydrolysis directly in the Asn 706 mutants, but some of the data obtained with the phosphoenzyme formed from ATP suggest that the E2P hydrolysis step is not as inhibited in these mutants as it is in E183A. Asn 706 therefore seems to play a less critical role than Glu 183 in stabilization of the transition state in E2P hydrolysis. Thus, virtually all phosphoenzyme accumulated at steady state following reaction of the three Asn 706 mutants with ATP was ADP-sensitive Ca 2 E1P (Fig. 5). In comparison, 43% of the phosphoenzyme accumulated under identical conditions with E183A was ADP-insensitive E2P, despite a reduced rate of the Ca 2 E1P 3 E2P transition in E183A (10), thus reflecting the reduced rate of E2P hydrolysis in the latter mutant. Even in phosphorylation experiments with ATP carried out under buffer conditions where a high level of E2P accumulates at steady state in wild type (i.e. pH 8 and K ϩ replaced by Li ϩ ), no significant level of E2P accumulated in the mutants with alterations to Asn 706 . Thus, unlike E183A, the mutations of Asn 706 do not seem to impair E2P hydrolysis markedly; in fact it may even be enhanced, due to destabilization of E2P. Danko et al. (15) proposed a distinction between E2⅐BeF, E2⅐AlF, and E2⅐MgF as analogs of the ground state, transition state, and E2⅐P i product state, respectively, in E2P hydrolysis. In contrast to the planar tetragonal AlF complex of the E2⅐AlF crystal structure, that seems to mimic the transition state of E2P hydrolysis, the MgF complex (i.e. MgF 4 2Ϫ ) is in a tetrahedral arrangement in the E2⅐MgF crystal structure (see Fig. 9 and Table 2), implying that E2⅐MgF represents the E2⅐P i product state of E2P hydrolysis. The E2⅐BeF complex (of which no crystal structure has yet been published) shares several features with the E2P ground state, including an increased hydrophobicity of the nucleotide site relative to E2, E2⅐AlF, E2⅐MgF, and E2⅐vanadate. Furthermore, E2⅐BeF displays a high sensitivity to luminal Ca 2ϩ (a feature not seen with E2⅐AlF and E2⅐MgF), as one would expect from a true E2P ground state with luminally exposed Ca 2ϩ sites (15). With respect to BeF binding (Fig. 7B), all four mutants displayed significant shifts toward lower apparent affinity relative to wild type (6-to 32-fold), suggesting that both Asn 706 and Glu 183 contribute significantly to stabilization of the E2P ground state. The affinities for AlF and MgF differed more significantly among the four mutants. The affinity for AlF (Fig. 7A) was markedly reduced in E183A relative to wild type, consistent with Glu 183 being involved in stabilization of the E2P transition state. The affinity for AlF was much less affected in the mutants with alterations to Asn 706 than in E183A, N706S, in particular, being completely wild type-like (Fig.  7A), thus suggesting that the side chain of Asn 706 contributes marginally or not at all to the stability of the E2P transition state, in line with the results of studies of the phosphoenzyme discussed above. This is somewhat surprising considering that the side-chain nitrogen atom of Asn 706 is located within 2.8 Å of AlF in the E2⅐AlF crystal structure (Fig. 9) and that the side-chain oxygen of Asn 706 is within bonding distance of a water molecule that coordinates the catalytic Mg 2ϩ ion (indicated by orange broken lines in Fig. 9), although the link to Mg 2ϩ might persist in mutant N706S, due to the oxygen in the serine side chain.
In the E2⅐MgF crystal structure, thought to mimic the E2⅐P i product state, the disposition of the Asn 706 side chain in relation to the fluoride and the water molecule coordinating the catalytic Mg 2ϩ is very similar to that seen in the E2⅐AlF crystal structure (Fig. 9). Nevertheless, the inhibition data obtained with MgF (Fig. 7, C and D) showed a picture quite different from that seen for AlF. Hence, even mutation N706S lowered the affinity for MgF significantly, and the order of E183A and N706A/N706C was reversed, such that the latter two mutants displayed the most severely reduced binding affinities of the four mutants. The MgF inhibition data suggest that the Asn 706 side chain contributes significantly to stabilization of the E2⅐P i , product state, consistent with the E2⅐MgF crystal structure.
To understand the differential effects of Asn 706 mutation on the binding of AlF and MgF, it could be important that there is increased negative charge density around the phosphoryl group in an associative transition state (see schematic diagram in Ref. 34). Perhaps for this reason the role of the neutral asparagine side chain is diminished during the transition and in the E2⅐AlF complex. In the E2⅐AlF crystal structure, the distance from A1 to the coordinating oxygen of Asp 351 or the water molecule positioned by Glu 183 is only 2.1 Å, consistent with covalent bonding (6), which means that any interaction with Asn 706 is relatively less important for stabilization of the E2⅐AlF structure compared with E2⅐MgF, where the interaction between the protein and the fluoride does not have the character of a covalent bond.
Turning now to the mutational effects on the E1 states, it is clear from our nucleotide binding data ( Table 1) that the Asn 706 side chain is not critical for high affinity nucleotide binding in E1. Thus, the mutations of Asn 706 had no effect on ATP binding in the absence of Mg 2ϩ , and in the presence of Mg 2ϩ we observed a less than 2-fold reduction of MgATP affinity, relative to wild type, for N706A and N706C, and wild type-like behavior for N706S. Likewise, the rate of phosphorylation of Ca 2ϩsaturated enzyme from [␥-32 P]ATP (Fig. 3A) as well as the binding affinity for ADP⅐AlF (Fig. 8) were wild type-like in N706S. The effects of the other two Asn 706 mutations were larger, in particular N706C showed 4-to 5-fold reduced phosphorylation rate with [␥-32 P]ATP (Fig.  3A) and 15-fold reduced affinity for ADP⅐AlF (Fig. 8). In either of the two published crystal structures of the Ca 2 E1⅐AlF⅐ADP state of Ca 2ϩ -ATPase, supposed to mimic the transition state in the phosphoryl transfer from ATP (5, 7), the side chain of Asn 706 is in a very similar position to that seen in the E2⅐MgF and E2⅐AlF crystal structures, i.e. within bonding distance of the fluoride and linked to the catalytic Mg 2ϩ via a water molecule (cf. Fig. 9). The small effect of N706A and lack of effect of N706S on ADP⅐AlF binding suggest that Asn 706 plays a relatively minor role in E1P transition state stabilization, exactly as found for the E2P transition state. The larger effects of mutation N706C may be due to the cysteine side chain being ionized, thereby leading to electrostatic repulsion of the ␥-phosphate/AlF.
The conspicuous effects of the Asn 706 mutations on vanadate binding to E2 were very similar to the effects on P i reactivity. Both vanadate and AlF have been considered phosphoryl transition state analogs, but when one compares the results in Figs. 6 and 7A, it is clear that vanadate binding was much more strongly affected than AlF binding by all four mutations studied here. In particular, N706S was almost completely insensitive to vanadate and, at the same time, wild type-like with respect to inhibition by AlF. Hence, vanadate must bind and exert its inhibitory effect differently from AlF. As judged from the mutational effects, the E2⅐vanadate state seems to be a closer mimic of the enzyme with bound phosphoryl group than the enzyme complexes with fluoride com-pounds, which may have to do with the fact that vanadate adopts a trigonal bipyramidal structure and has oxygens like phosphate, and not the strongly electronegative fluorines. The fluoride compound that resembles vanadate the most in our functional assays is BeF, with its more than 6-fold reduced apparent affinity for all four mutants relative to wild type (cf. Fig. 7B).
Besides their strong effects on P i and vanadate reactivity, Asn 706 mutations, including N706S, also caused significant reductions of the rates of the major protein conformational changes involved in Ca 2 E1P 3 E2P and E2 3 Ca 2 E1 partial reactions. It is notable that, in the E2⅐AlF and E2⅐MgF crystal structures, the side chain of Asn 706 lines up with the TGES backbone of domain A, such that the Thr 181 main-chain carbonyl comes close to the Asn 706 side-chain nitrogen and the Gly 182 main-chain amine comes close to the Asn 706 side-chain carbonyl (interaction distances between 3.4 and 3.8 Å in the two structures, indicated by blue broken lines in Fig. 9). If Asn 706 mediates similar van der Waals interactions between domains A and P in the E2P ground state of the native enzyme (or the residues are actually even closer and participate in hydrogen bonding), these interactions might be essential for stabilizing the docking of domain A into the catalytic site, thus explaining the slowing of the Ca 2 E1P 3 E2P transition and destabilization of E2P (as well as E2⅐vanadate and E2⅐P i ) in the Asn 706 mutants. In the functioning enzyme, most of the motions that subsequently disengage domain A from its docking site with domain P take place during the Ca 2ϩ binding transition of the dephosphoenzyme (E2 3 Ca 2 E1), although the departure of the TGES motif from the catalytic site has apparently already been initiated in E2 (7,9). We observed a considerable slowing (5-to 9-fold in the absence of ATP) of the Ca 2ϩ binding transition in the Asn 706 mutants (Figs. 3B and 4A). Because in E2 the side chain of Asn 706 is still within bonding distance (2.6 Å) of the main-chain carbonyl of Leu 180 , it is not straightforward to understand why mutations of Asn 706 , that undoubtedly disrupt this interaction, also seem to inhibit the E2 3 Ca 2 E1 transition, rather than enhance it. Interestingly, a similar inhibition of E2 3 Ca 2 E1 was found in E183A (10), as well as in mutants with alterations to certain other residues at the domain A-P interface (45). It appears that disruption of the normal interactions between domains A and P captures the ATPase in an E2-like conformation, from which departure is slow relative to the normal E2 state. Thus, these particular interactions seem to facilitate the normal fast cycling through E2.
In the light of the effects of the Asn 706 mutations on the Ca 2 E1P 3 E2P transition it becomes feasible to understand the changes of apparent affinity for Ca 2ϩ and MgATP, which showed strikingly similar patterns (Fig. 2, A and B). As previously demonstrated by computation (36), a shift in the K 0.5 for Ca 2ϩ activation of phosphorylation does not necessarily reflect a "true" change of the Ca 2ϩ -binding properties. In fact, a mutant for which phosphoenzyme turnover is blocked, either by inhibition of Ca 2 E1P 3 E2P or by inhibition of E2P hydrolysis, should show an increased apparent Ca 2ϩ affinity, because lower Ca 2ϩ concentrations are required for phosphoenzyme to accumulate. A similar argument holds with respect to the apparent affinity for MgATP activation of phosphorylation, and this mechanism explains the left shift of the activation curves seen for N706S in Fig. 2. For N706A and N706C, the reduced rate of phosphorylation from ATP in addition contributes to determine the apparent affinity for Ca 2ϩ and MgATP, tending to right shift the curves. For N706A, the result of the two opposing effects is a K 0.5 rather similar to that of the wild type, whereas for N706C the phosphorylation rate is sufficiently low for the final result to be a right shift of the curves with increased K 0.5 relative to wild type.
In conclusion, by studying the phosphoenzyme and the binding of the three phosphate analogs, AlF, MgF, and BeF, believed to represent different states of the phosphoryl group during E2P dephosphorylation, we have been able to distinguish different roles of Asn 706 and Glu 183 during catalysis of the E2P 7 E2 reaction. The extremely strong effects of the Asn 706 mutations on the reaction of the enzyme with vanadate and P i cannot be fully accounted for by reference to the effects on binding of the fluorides, possibly due to their strong electronegativity. However, it appears that Asn 706 is critical for stabilizing E2P and, possibly therefore, also for the Ca 2 E1P 3 E2P transition, and that it likely loses its importance in the transition state, where charge effects are greatest, but regains consequence in E2⅐P i and E2. The latter conformation is only slowly transformed into Ca 2 E1 in the Asn 706 mutants. Interaction of Asn 706 with domain A may be an important aspect of these effects.