Interaction of Nucleotides with Asp351 and the Conserved Phosphorylation Loop of Sarcoplasmic Reticulum Ca2+-ATPase*

The nucleotide binding properties of mutants with alterations to Asp351 and four of the other residues in the conserved phosphorylation loop,351DKTGTLT357, of sarcoplasmic reticulum Ca2+-ATPase were investigated using an assay based on the 2′,3′-O-(2,4,6-trinitrophenyl)-8-azidoadenosine triphosphate (TNP-8N3-ATP) photolabeling of Lys492 and competition with ATP. In selected cases where the competition assay showed extremely high affinity, ATP binding was also measured by a direct filtration assay. At pH 8.5 in the absence of Ca2+, mutations removing the negative charge of Asp351 (D351N, D351A, and D351T) produced pumps that bound MgTNP-8N3-ATP and MgATP with affinities 20–156-fold higher than wild type (K D as low as 0.006 μm), whereas the affinity of mutant D351E was comparable with wild type. Mutations K352R, K352Q, T355A, and T357A lowered the affinity for MgATP and MgTNP-8N3-ATP 2–1000- and 1–6-fold, respectively, and mutation L356T completely prevented photolabeling of Lys492. In the absence of Ca2+, mutants D351N and D351A exhibited the highest nucleotide affinities in the presence of Mg2+ and at alkaline pH (E1 state). The affinity of mutant D351A for MgATP was extraordinarily high in the presence of Ca2+(K D = 0.001 μm), suggesting a transition state like configuration at the active site under these conditions. The mutants with reduced ATP affinity, as well as mutants D351N and D351A, exhibited reduced or zero CrATP-induced Ca2+ occlusion due to defective CrATP binding.

The Ca 2ϩ -ATPase of sarcoplasmic reticulum is a 10-trans-membrane helix Ca 2ϩ /H ϩ pump that hydrolyzes ATP through transient formation of an aspartyl phosphorylated intermediate (1)(2)(3)(4). The phosphorylated aspartate residue (Asp 351 ) and the binding site for MgATP are located in the large cytoplasmic domain of the pump protein, whereas the Ca 2ϩ transport sites are in the membrane domain (5). It is a well documented property of the pump that binding of Ca 2ϩ at the transport sites is required to activate phosphoryl transfer from ATP to Asp 351 . However, the long range intramolecular interaction between the Ca 2ϩ sites and the nucleotide binding site that triggers formation of a transition state for phosphoryl transfer and the nature of this transition state are not well understood. Electrostatic interactions in the vicinity of the phosphoryl groups of ATP and Asp 351 and other catalytic residues may be expected to dominate during ATP binding and in the transition state and possibly drive changes in the transport sites (6). The unphosphorylated Ca 2ϩ -ATPase exists in a Ca 2ϩ -and pH-dependent equilibrium (7) of several (E1/E2) conformational states (Scheme 1) that appear to interact differently with ATP. E2H 3 7 E2H 7 E1 7 E1Ca 7 E1Ca 2 SCHEME 1 Although in the presence of Mg 2ϩ all of the states indicated in Scheme 1 exhibit rather high affinities for ATP (K D in the range 0.5-20 M) (8,9), only the E1Ca 2 state is primed for transfer of the ␥-phosphoryl group to Asp 351 . As counterions to Ca 2ϩ (3), protons bind to the transport sites in place of Ca 2ϩ , stabilizing the E2 conformation. The fully protonated E2 form (E2H 3 in Scheme 1) is phosphorylated by P i at Asp 351 when the pump works in the reverse mode but cannot be phosphorylated by ATP. Since ATP accelerates P i binding (10) and dephosphorylation (11), modulates P i º HOH exchange (12), and binds fairly tightly to the vanadate complexed E2 form (13), E2H 3 must be able to bind ATP without preventing the access of P i to Asp 351 , suggesting that the ␥-phosphoryl group of the bound ATP is at some distance from Asp 351 in this conformation, in contrast to the E1Ca 2 state. A change in the interaction of bound nucleotide with Asp 351 related to enzyme activation by Ca 2ϩ is clearly demonstrated with TNP-8N 3 -ATP 1 that has been covalently attached to Lys 492 by light activation (14 -16). Tethering the nucleotide still permits Ca 2ϩ -dependent hydrolysis in the forward direction of catalysis, proving direct interaction with Asp 351 , and yet has little effect on P i -dependent phosphorylation in the absence of Ca 2ϩ , showing that the nucleotide, or at least a portion of it, shifts position with respect to the aspartyl residue upon Ca 2ϩ binding.
The phosphorylated aspartate and adjoining residues on the COOH-terminal side, segment 351 DKTGTLT 357 , termed the phosphorylation loop in this study, are highly conserved in P-type ATPases (17), and previous mutational analysis has documented their functional importance (18,19). Besides Asp 351 , also Lys 352 , Thr 355 , Leu 356 , and Thr 357 are critical to Ca 2ϩ transport as well as phosphorylation (19), and further clarification of the distinct roles of these residues in nucleotide binding, phosphoryl transfer, and long range interaction with the Ca 2ϩ sites may aid understanding of energy transduction in the pump. In this study, we assess the effects on nucleotide binding of mutations to the phosphorylation loop residues that previously were shown to result in severely impaired or inactive pumps (18,19). ATP binding is measured mainly through inhibition of TNP-8N 3 -ATP photolabeling of Lys 492 , which has recently been successfully applied to Ca 2ϩ -ATPase mutated in segment 487 FSRDRK 492 (16). In selected cases where the affinity is extremely high, binding is also measured by a direct filtration assay. In addition, the effects of the mutations on Ca 2ϩ occlusion induced with CrATP are determined. The results show that most of the mutations in the phosphorylation loop affect nucleotide binding and disrupt CrATP-induced Ca 2ϩ occlusion. Our analysis of mutations to Asp 351 reveals high intrinsic nucleotide binding energies when the negative charge is removed, particularly in the presence of Mg 2ϩ and Ca 2ϩ , i.e. in the E1Ca 2 state (K D ϭ 0.001 M for mutant D351A). These favorable interactions may be utilized to gain the transition state and to provoke conformational changes that communicate with the transport sites.

EXPERIMENTAL PROCEDURES
The mutant Ca 2ϩ -ATPase cDNAs used in this study were the same as those described previously (18,19) but were shuttled to vector pMT2 (20) to obtain higher expression levels in COS-1 cells (18,21). COS-1 cell microsomes containing expressed wild-type or mutated Ca 2ϩ -ATPase were isolated by differential centrifugation 48 -72 h after transfection (18). The exogenous Ca 2ϩ -ATPase content of the microsomal fraction was assayed with a specific sandwich enzyme-linked immunosorbent assay (21).
The synthesis of [␥-32 P]TNP-8N 3 -ATP, photolabeling of COS-1 cell microsomes, the inhibition by ATP, quantification of labeled bands by electronic autoradiography ("imaging") following SDS-polyacrylamide gel electrophoresis, curve fitting equations and calculations of the "true" K D(ATP) have been described previously (16,22). For fitting of the TNP-8N 3 -ATP labeling data, the Hill equation with or without a linear component was used, and the Hill coefficient was set to 1. The concentration of free Ca 2ϩ was set with 5 mM EGTA and variable amounts of total CaCl 2 as calculated according to Fabiato and Fabiato (23) taking the Mg 2ϩ concentration and pH into consideration. CrATP-dependent Ca 2ϩ occlusion was measured as before (16,24).
Equilibrium ATP binding to mutants D351N and D351A was also measured by filtration. COS-1 cell microsomes (1 l of stock microsomes in 1 ml; approximately 0.5 pmol of Ca 2ϩ -ATPase protein/ml) were incubated with [␥-32 P]ATP, 1 mM [ 3 H]sucrose, and other components as indicated in the Fig. 7 legend for 1 min at 25°C, and the sample was filtered on Millipore GS 0.22-m filters under mild vacuum. The radioactivity of the filter was measured by liquid scintillation counting. The wet volume of the filter was determined from the tritium radioactivity (range: 28 -42 l), allowing determination of the radioactivity of unbound nucleotide, which was subtracted from the total 32 P cpm to obtain the amount of ATP bound to the microsomes.
The formation of a slowly dissociating CrATP complex with the Ca 2ϩ -ATPase in the presence of Ca 2ϩ was followed through the inhibition of [␥-32 P]TNP-8N 3 -ATP photolabeling. Microsomes containing expressed wild-type or mutated Ca 2ϩ -ATPase were incubated at 37°C with CrATP for up to 1 h. Aliquots were taken at timed intervals and diluted 50-fold into irradiation medium with 0.5 M [␥-32 P]TNP-8N 3 -ATP. The samples were irradiated for 1 min and subjected to SDSpolyacrylamide gel electrophoresis, and the radioactivity was quantified by electronic autoradiography as described previously (16).

RESULTS
Twenty-four mutations have previously been introduced into the conserved phosphorylation loop of the Ca 2ϩ -ATPase between Ile 348 and Thr 357 (18,19). All the mutants with alteration to the aspartic acid residue Asp 351 receiving the phosphoryl group during catalysis are inactive, and so are the mutants with alterations to Lys 352 , even in the case of the most conservative replacement of Lys 352 with arginine. Activity is not affected by conservative replacements of Thr 355 or Thr 357 with serine, but replacement with alanine reduces the Ca 2ϩ transport activity as well as the level of phosphoenzyme. Replacement of Leu 356 with isoleucine is without effect on activity, but mutation to threonine inactivates the pump. Hence, these residues (nine mutants in all; see Table I) were selected for the present study of nucleotide binding properties.
The assay for nucleotide binding, which is based on specific [␥-32 P]TNP-8N 3 -ATP photolabeling of Lys 492 and nucleotide competition, has been validated previously (16). Results obtained under optimum labeling conditions at pH 8.5 demonstrated that this assay is able to produce highly accurate values for TNP-8N 3 -ATP and ATP binding affinities of Ca 2ϩ -ATPase expressed in COS-1 cell microsomes (16). In assessing the results to be described below, it is furthermore useful to know that TNP-8N 3 -ATP is a substrate of the Ca 2ϩ -ATPase, albeit a slow one, whether untethered or tethered to Lys 492 by photolabeling (15,16). This means that the position of the ␥-phosphoryl group of the bound nucleotide must be similar, although probably not identical, to that of bound ATP.
The concentration dependence of TNP-8N 3 -ATP photolabeling of wild-type and mutant Ca 2ϩ -ATPases at pH 8.5 in the presence of Mg 2ϩ and absence of Ca 2ϩ (presence of EGTA) is shown in Fig. 1A. The data could be fitted satisfactorily to the sum of a simple hyperbolic binding function and a linear component, the latter representing nonspecific labeling as previpressure liquid chromatography; TES, 2-{[2-hydroxy-1,1-bis(hydroxymethyl)ethyl]amino}ethanesulfonic acid; TMAH, tetramethyl ammonium hydroxide.   Ͼ50 or no specific labeling Ϫ T357A 6.4 ϳ20 Ϫ a The "true" K D calculated under the assumption of competitive inhibition as described in Ref. 16; the concentration of TNP-8N 3 -ATP was 3 ϫ K 0.5 except for mutants K352Q and T357A where it was equal to the K 0.5 . b Ϫ, no occlusion; ϩ, partial occlusion; ϩϩ, full occlusion; ND, not determined. Data are shown in Fig. 2. ously explained (16). For most of the mutants, the linear component was small and insignificant, but as seen in Fig. 1 the linear component was rather prominent for mutant K352Q, for unknown reasons. The derived K 0.5 values corresponding to the hyperbolic component are listed in Table I. It can been seen that removal of the negative charge on Asp 351 , as shown by mutants D351N, D351A, and D351T, led to a pronounced increase in TNP-8N 3 -ATP affinity, with D351N exhibiting the largest increase of 156-fold. By contrast, mutation D351E, which conserves the negative charge, did not significantly affect the TNP-8N 3 -ATP binding affinity. The concentration of Ca 2ϩ -ATPase in the irradiation assay was approximately 0.4 nM for the tightly binding mutants and approximately 2 nM for the rest to ensure a reasonably high ratio of free to bound nucleotide, thereby allowing the total concentration to be equated with the free concentration.
Mutation K352Q, which removes the positive charge of Lys 352 , lowered the affinity for TNP-8N 3 -ATP at least 5-fold, whereas the more conservative replacement with arginine, K352R, was without significant effect. Mutation T355A was likewise silent, but T357A lowered the affinity for the TNP nucleotide about 6-fold. Mutant L356T exhibited a low level of labeling that was linear with increasing concentrations of TNP-8N 3 -ATP up to 30 M. This indicates that either Lys 492 was not being labeled or the affinity was extremely poor (K 0.5 estimated to be Ͼ50 M).
The inhibition of photolabeling by ATP under the same buffer conditions is shown in Fig. 1B, and the derived true K D values assuming competitive inhibition are listed in Table I. Usually, the concentration of TNP-8N 3 -ATP was fixed at 3 ϫ K 0.5 to ensure that the binding site is close to saturation; however, in the case of mutants K352Q and T357A, inhibition was so poor at these concentrations that the TNP-8N 3 -ATP concentrations were lowered to the K 0.5 in each case. It is apparent that removal of the negative charge on Asp 351 caused a large increase in affinity for ATP (83-fold, 45-fold, and 22-fold for D351N, D351T, and D351A, respectively) similar to that seen for TNP-8N 3 -ATP. Mutations D351E and T355A appeared to slightly decrease the affinity for ATP (1.5-2-fold compared with wild type). While mutation K352R resulted in a 13-fold reduction of ATP affinity, K352Q led to a spectacular effect, reducing the ATP affinity close to 1000-fold. Mutation T357A decreased the affinity for ATP at least 40-fold. Because the affinity of the latter two mutants was too low for complete . The samples were subjected to SDS-polyacrylamide gel electrophoresis, and the relevant radioactive bands were quantified by imaging. The data were fitted to the sum of a hyperbolic function and a linear function, the latter representing nonspecific labeling, and the derived K 0.5 values for the hyperbolic function are shown in Table I. B, experiments were performed as above at a concentration of TNP-8N 3 -ATP of 3 ϫ K 0.5 for each mutant Ca 2ϩ -ATPase (see Table I), except for T357A and K352Q, where it was equal to the K 0.5 , and ATP was included at the concentrations shown. In two cases, namely mutations K352Q and T357A, additional Mg 2ϩ was included at 1 and 3 mM ATP to a total concentration of 2 and 4 mM, respectively. The data were fitted to a simple binding function with an offset representing nonspecific labeling. The K D(ATP) values calculated from the derived K 0.5 values as described (16) are listed in Table I inhibition to be reached, the choice of the offset of the binding curve was somewhat uncertain in these cases, resulting in corresponding uncertainties with respect to the exact K D(ATP) values.
Hence, the results shown in Fig. 1 indicate that the most disruptive mutation in terms of nucleotide binding is L356T, followed by K352Q, T357A, and K352R. The latter three mutations affected ATP binding much more than the binding of TNP-nucleotide, similar to the situation with mutations close to Lys 492 (16). All of the mutants in which the negative charge on Asp 351 was removed exhibited a large increase in affinity for both nucleotides, with D351N being the most dramatic. On the other hand, mutation D351E had little effect on the binding of either nucleotide.
Nucleotide binding and the coupling between the catalytic and transport sites can be assessed by measuring CrATPinduced Ca 2ϩ occlusion. CrATP slowly forms a complex with the Ca 2ϩ -ATPase at the catalytic site without phosphorylating Asp 351 and causes simultaneous occlusion of Ca 2ϩ at the transport sites (24 -27). Because this complex is very stable, requiring hours to dissociate, 45 Ca 2ϩ occlusion with CrATP can be measured in Ca 2ϩ -ATPase expressed in COS-1 cell microsomes by size exclusion HPLC following detergent solubilization of the microsomes (24). The results of such measurements on the wild type and selected mutants are shown in Fig. 2 and summarized in Table I. The microsomes were incubated 1 h at 37°C with 1 mM CrATP and 40 M 45 Ca 2ϩ , i.e. just about enough to ensure saturation of the occlusion reaction in the wild-type enzyme, before solubilization and chromatography. Equal amounts of expressed Ca 2ϩ -ATPase, according to en-zyme-linked immunosorbent assay measurements, were chromatographed, so the elution profiles are comparable. The control represents microsomes without expressed Ca 2ϩ -ATPase, i.e. harvested from COS-1 cells mock-transfected with the expression vector without insert. The amount of control microsomes in mg of total membrane protein corresponds to that chromatographed in the case of the wild type. The distinct peak of radioactivity eluting between 14 and 15 min in the experiment with the wild type and in some of the experiments with mutants represents Ca 2ϩ occluded in the detergent-solubilized monomeric enzyme (24). As can be seen, there was no such peak for mutants D351N and D351A, and hence these mutants did not occlude Ca 2ϩ . Mutants K352R and T355A showed partial occlusion, and mutants K352Q, L356T, and T357A also failed to occlude Ca 2ϩ .
Thus, for some mutations, notably those that slightly or grossly lower the affinity for ATP, the effect on CrATP-induced Ca 2ϩ occlusion seems to be correlated with the change in ATP binding affinity. However, mutations to Asp 351 , which caused a huge increase in affinity for ATP, also appeared to prevent CrATP-induced Ca 2ϩ occlusion. In order to elucidate whether this was due to defective CrATP complexation at the catalytic site or an uncoupling of CrATP complexation and Ca 2ϩ occlusion, we devised an assay wherein the microsomes were incubated with CrATP for up to 1 h at 37°C and then diluted substantially prior to photolabeling with TNP-8N 3 -ATP. The affinity of the Ca 2ϩ -ATPase for CrATP is not high, and a concentration of CrATP in the millimolar range is required to obtain saturation so that occlusion occurs in a reasonable period of time. To minimize the competitive binding of contaminant ATP that might be present in the CrATP preparation, the samples were irradiated under conditions where the affinity for TNP-8N 3 -ATP is reasonably high and that for ATP is fairly low (pH 8.5 in the presence of EDTA). Also with this in mind, a concentration of TNP-8N 3 -ATP of approximately 10ϫ K 0.5 was The graphs show the size exclusion HPLC elution profiles of 45 Ca associated with solubilized protein originating from microsomes harvested from cells transfected with wild-type or mutant Ca 2ϩ -ATPase cDNA and from control microsomes from cells transfected with the expression vector without insert. The difference between the test and control curves corresponding to the elution time of the monomeric Ca 2ϩ -ATPase at 14 -15 min provides a measure of the occluded Ca 2ϩ . Equivalent amounts of wild-type and mutant Ca 2ϩ -ATPase protein (as determined by enzyme-linked immunosorbent assay) were applied to the column in the experiments shown. A summary of the results is shown in Table I. chosen. As seen in Fig. 3, photolabeling of the wild-type Ca 2ϩ -ATPase was inhibited by CrATP in a time-dependent manner, indicative of a gradual and effectively irreversible complexation of CrATP at the catalytic site. In mutants D351N and D351A, CrATP failed to inhibit photolabeling, showing that the irreversible binding of CrATP had not occurred. This explains the lack of CrATP-induced Ca 2ϩ occlusion in these mutants.
The buffer conditions in the above binding experiments with TNP-8N 3 -ATP and ATP (pH 8.5, in the presence of Mg 2ϩ and absence of Ca 2ϩ ) largely favor accumulation of the E1 state of the wild-type Ca 2ϩ -ATPase, whereas the E2H and E2H 3 forms (cf. Scheme 1) would predominate at neutral and acid pH in the absence of Ca 2ϩ (7). To better understand the huge increase in nucleotide affinity induced by mutations D351N and D351A and the implications for the catalytic mechanism, we investigated the influence of pH, Mg 2ϩ , and Ca 2ϩ , as well as thapsigargin, a tightly binding inhibitor that appears to lock the enzyme into E2 or an "E2-like" state (28). Results of these studies are presented in Figs. 4 -6 and summarized in Table II in terms of derived dissociation constants.
As previously noted (16), the efficiency of the specific incorporation of TNP-8N 3 -ATP photolabel into the Ca 2ϩ -ATPase expressed in COS-1 cell membrane is almost as high at neutral pH compared with alkaline pH, contrary to the situation for Ca 2ϩ -ATPase in sarcoplasmic reticulum. This made it feasible to extend the photolabeling assay to pH 7.0 for the expressed wild-type and mutant Ca 2ϩ -ATPases. The TNP-8N 3 -ATP concentration dependence of photolabeling of the wild-type Ca 2ϩ -ATPase at pH 7.0 in the presence of Mg 2ϩ is shown in Fig. 4A. The concentration dependence is characterized by hyperbolic and linear phases, with the specific hyperbolic component dominating in the low concentration range. This is similar to the results obtained at pH 8.5 (see Ref. 16 for a detailed description of the wild-type data at pH 8.5). The inset shows autoradiographs of the same experiments, and it can be seen that even at pH 7.0 the ratio of Ca 2ϩ -ATPase labeling to background labeling is reasonably high. The inhibition by ATP of the photolabeling of the wild type at pH 7.0 in the presence of a concentration of TNP-8N 3 -ATP of 3 ϫ K 0.5 is shown in Fig. 4B. Similar experiments at pH 7.0 were performed in the absence of Mg 2ϩ (in the presence of 2 mM EDTA), the amount of extra labeling in the 10 -30 M range being much less in this condition (data not shown). In addition, experiments in the absence of Mg 2ϩ were carried out at pH 8.5 for comparison. As seen in Table II, the decrease in pH from 8.5 to 7.0 enhanced the affinity of the wild-type Ca 2ϩ -ATPase for TNP-8N 3 -ATP 2-fold in the pres-ence of Mg 2ϩ , whereas the affinity for ATP was reduced 16-fold under these conditions. The effects of Mg 2ϩ were also quite different for the two nucleotides, the divalent cation strongly enhancing ATP binding (33-fold at pH 8.5 and 9-fold at pH 7.0) but diminishing the affinity for TNP-8N 3 -ATP, especially at pH 8.5. A trend to higher affinity for ATP in the presence of Mg 2ϩ and at alkaline pH has previously been reported for the Ca 2ϩ -ATPase in the native sarcoplasmic reticulum membrane (9,29,30), although the effects appear to be somewhat larger in the present study, possibly because of the presence of glycerol (31), which is added as a nitrene scavenger.
As seen in Figs. 5 and 6 and Table II, the observation of higher nucleotide affinities of mutants D351N and D351A relative to wild type could be confirmed at pH 7.0 as well as in the absence of Mg 2ϩ , although the pH and Mg 2ϩ influenced to some extent the magnitude of the affinity increase induced by the mutations. At pH 8.5, the mutations seem to negate the anomalous inhibitory effect of Mg 2ϩ on TNP-8N 3 -ATP binding to wild-type Ca 2ϩ -ATPase, the divalent cation actually enhancing the affinity 2-3-fold in the case of D351N and having no significant effect for mutation D351A. The Mg 2ϩ -induced enhancement of ATP affinity was even more pronounced in mutant D351N (200-fold) than in the wild type at pH 8.5 but was much less (only 2-fold) at pH 7.0.
It was previously demonstrated that thapsigargin is without effect on TNP-8N 3 -ATP photolabeling of the wild-type Ca 2ϩ -ATPase but lowers the affinity for ATP more than 100-fold (16). In the presence of thapsigargin, both of the mutants D351N and D351A displayed much lower affinity for either nucleotide than in the absence of the inhibitor (Table II), but the effect of the inhibitor was most pronounced for ATP (more than 700-fold decrease in ATP affinity of both mutants). Importantly, even in the thapsigargin-bound state, the mutations induced a significant increase in affinity for both nucleotides. Since thapsigargin is believed to stabilize E2 or an "E2-like state" (28), this finding clearly demonstrates that the affinity increase resulting from removal of the negative charge of Asp 351 is not an exclusive property of E1 forms. On the other hand, it is noteworthy that the magnitudes of the effects induced by the mutations were somewhat smaller in the presence of thapsigargin than in its absence (cf. also Table III and below).
Because Ca 2ϩ binding to the transport sites on the wild-type Ca 2ϩ -ATPase activates rapid transfer of the ␥-phosphoryl group of bound ATP to the enzyme, the affinity of the wild-type enzyme for ATP in the presence of Ca 2ϩ cannot be determined under equilibrium conditions, but a value of 3 M for the K D(ATP) of the wild-type enzyme in the presence of Ca 2ϩ has been estimated on the basis of rapid kinetic measurements of association and dissociation rate constants (8). The inability of mutants D351N and D351A to phosphorylate offered a unique possibility for studies of the effect of Ca 2ϩ on equilibrium ATP binding to these mutants. As seen in Figs. 5 and 6 and Table II, we determined the affinities for TNP-8N 3 -ATP as well as ATP in the presence of Mg 2ϩ and Ca 2ϩ at both pH 8.5 and pH 7.0. For mutant D351N, Ca 2ϩ binding caused a significant 4-fold reduction of the affinity for ATP at pH 8.5, whereas, surprisingly, at pH 7.0 the affinities for both nucleotides were increased (10-fold for TNP-8N 3 -ATP and 20-fold for ATP). For mutant D351A, Ca 2ϩ binding caused a tremendous increase in ATP affinity, resulting in a K D of about 0.001 M at both pH values (300-fold increase in affinity at pH 7.0 and 23-fold at pH 8.5). Also, the affinity for the TNP-nucleotide increased very much in the presence of Ca 2ϩ in this mutant (10-fold at pH 8.5 and 40-fold at pH 7.0).
The important implications of these Ca 2ϩ -dependent changes in ATP binding and the extremely high affinities for ATP prompted us to try to verify the photolabeling results by measuring ATP binding directly by filtration. This method is not generally applicable to expressed wild-type or mutant pro-teins, since each data point requires several g of protein when the K D is in the micromolar range or above, and also there could be a problem with the specificity under these conditions considering the large number of proteins in the COS cell microsomes. Fig. 7 shows the results obtained at pH 7.0, and as can be seen extremely tight [␥-32 P]ATP binding to mutants D351A and D351N could indeed be measured in the presence of Ca 2ϩ . The K D values (0.0012 and 0.030 M, respectively) are the same or very similar to those found by photolabeling (0.0014 and 0.028 M, respectively). The binding in the absence of Ca 2ϩ was much weaker, and the data points are compatible with the K D values obtained by photolabeling (0.42 and 0.57 M, respectively, Table II). It could be noted that the filtration experiments were performed in the absence of glycerol, and the similarity of the photolabeling and filtration measurements indicates that the cosolvent is not affecting the K D values significantly under these conditions. Using the filtration assay, we also obtained a few data points at pH 8.5, confirming that the addition of Ca 2ϩ increases the ATP affinity of mutant D351A and decreases the ATP affinity of mutant D351N (results not shown).
The increase in the affinity for TNP-8N 3 -ATP upon Ca 2ϩ binding to the transport sites of mutants D351N and D351A at pH 7.0 (10-and 40-fold, respectively) permitted investigation of the Ca 2ϩ concentration dependence of this phenomenon (Fig.  8). While the Hill coefficients of 1.3 are similar to those observed for Ca 2ϩ binding to the wild-type Ca 2ϩ -ATPase in sarcoplasmic reticulum, the K 0.5 values of 0.10 and 0.05 M for mutants D351N and D351A, respectively, are more than 10fold lower than those determined for sarcoplasmic reticulum Ca 2ϩ -ATPase under comparable buffer conditions by equilibrium binding studies in the absence of nucleotide or in the presence of nonhydrolyzable nucleotides such as AMPPCP (32,33).
For nondisruptive mutations, the interaction energies (⌬⌬G Interaction ) between a side chain and ligand can be obtained from the ratio of the dissociation constants of wild type and mutant proteins (34). Table III shows the results of such calculations of nucleotide-Asp 351 interaction energies for mutants D351N and D351A based on the data in Table II for wild type and mutants in the absence of Ca 2ϩ and some additional results of similar studies with ADP and AMPPCP in place of ATP (see below). The interaction energies for ATP in the presence of Ca 2ϩ and Mg 2ϩ were also calculated for pH 7.0 (last line in Table III) using the data in Table II for the mutants and the K D(ATP) of 3 M for the wild type determined by rapid kinetic measurements (8). In the absence of Ca 2ϩ , the interaction energies calculated for mutation D351N tend to be higher than those calculated for mutation D351A. The TNP-8N 3 -ATP data show no pH dependence in the absence of Mg 2ϩ but a strong dependence in its presence. Thus, for both mutants there is a pronounced Mg 2ϩ -induced increase in the interaction energy at pH 8.5, but not at pH 7.0. Thapsigargin binding at pH 8.5 has a similar effect to lowering the pH or removing the divalent cation. As for the ATP data, the interaction energy calculated for mutant D351N appears to follow the general trend, being enhanced from 1.6 to 2.6 kcal/mol by Mg 2ϩ at pH 8.5 and by the pH increase in the presence of Mg 2ϩ , whereas the interaction energy calculated for mutant D351A remains close to the lower of these two values, unaffected by Mg 2ϩ and pH. Ca 2ϩ , on the other hand, seems to increase the interaction energy most strongly for mutant D351A, to as much as 4.5 kcal/mol.
The K D values determined for the binding of ADP and AMP-PCP at pH 8.5 in the presence of Mg 2ϩ were for the wild type 7 and 63 M, respectively, and for mutant D351N 0.23 and 0.8 M, respectively (data not shown, but see the derived interaction energies in Table III). Thus, the interaction energy with MgADP is lower than with MgATP but still significant. The energy with MgAMPPCP is similar to MgATP. DISCUSSION Our results document the close interaction of bound nucleotides with Asp 351 and other residues in the conserved phosphorylation loop of sarcoplasmic reticulum Ca 2ϩ -ATPase. Elimination of the charge on Asp 351 , as in mutations D351N, D351A, and D351T, enhanced the affinity for ATP and TNP-8N 3 -ATP up to 156-fold in the absence of Ca 2ϩ and even more in the presence of Ca 2ϩ , suggesting strong electrostatic effects between the ␥-phosphate and the aspartate in the wild-type pump. In contrast, fairly conservative changes to Lys 352 , Thr 355 , Leu 356 , and Thr 357 impaired nucleotide binding to various extents, suggesting that these residues either directly ligate nucleotide or influence those that do. The removal of the positive charge of Lys 352 in mutant K352Q gave rise to a very pronounced decrease in ATP affinity of at least 1000-fold, again revealing the importance of electrostatic interactions at the active site. The disruptive mutations, as well as D351N and D351A, reduced or eliminated CrATP-induced Ca 2ϩ occlusion.
Recently, bacterial haloacid dehalogenases have been found to be homologous with P-type ATPases (35), and the crystal structures of the dehalogenases provide insights into possible Ca 2ϩ -ATPase structure. A catalytic aspartate of the dehalogenases is homologous to Asp 351 of Ca 2ϩ -ATPase and is positioned at the end of a ␤-strand, which is centrally located in a ␤-sheet (36). The critical aspartate leads into a spiral loop and hence to a secondary domain, which acts like a cap over the aspartate and active site. In the Ca 2ϩ -ATPase, the nucleotide substrate must be positioned between the two domains, but it is not yet clear how it would be oriented with respect to the aspartate and the phosphorylation loop. Crude modeling suggests that Lys 352 may be involved in direct interaction with the nucleotide, and the disruptive influence of even K352R on ATP binding is compatible with a role for this lysine in direct ligation. Leu 356 and Thr 357 do not appear to be able to interact directly with the nucleotide but rather appear to brace Asp 351 and Lys 352 and lead into the hinge segment between the main phosphorylation domain and the secondary cap domain. Mutation L356T was severely disruptive and seemed to prevent photolabeling of Lys 492 (expected to be part of the cap domain and some distance away from Leu 356 ). If this is occurring by shifting the position of Lys 492 , it would indicate interdomain communication between critical regions of the protein and point to Leu 356 being in a pivotal position. Alternatively, the nucleotide affinity may have been reduced as a result of displacement of Asp 351 and Lys 352 . Thr 355 appears to play a minor role in ATP binding, and perhaps the ATP molecule is oriented away from this residue, and its apparently crucial role in phosphorylation (19) may be catalytic, a conclusion supported by the strong conservation of this residue in both P-type pumps and dehalogenases (35).
Mutations D351N, D351A, and D351T generally caused large increases in affinity for TNP-8N 3 -ATP and ATP, suggesting the existence of a strong electrostatic repulsion between the aspartate and the ␥-phosphoryl group of ATP, which obscures a  In many of the conditions tested, we found larger affinity increases for the asparagine substitution than for the alanine substitution, in contrast to what was found for the Na ϩ ,K ϩ -ATPase, where the alanine substitution appears to  be the more effective. It may furthermore be noted that the Na ϩ ,K ϩ -ATPase D369N mutant showed no significant change (or possibly a slight decrease) in affinity for ADP relative to the wild type, whereas a 30-fold increase in affinity for MgADP was demonstrated for the corresponding Ca 2ϩ -ATPase mutant. This difference may possibly be ascribed to the presence of Mg 2ϩ (see below).
The results obtained with mutant D351N in the absence of Ca 2ϩ reveal that the increase in nucleotide affinity induced by the mutation (nucleotide-Asp 351 interaction energies, ⌬⌬G Interaction, Table III) is largest at alkaline pH in the presence of Mg 2ϩ . Mg 2ϩ increases the interaction energy (or electrostatic repulsion) between the ␥-phosphoryl group and Asp 351 at pH 8.5 but has little effect at pH 7.0. Since the E1 form of the pump is expected to prevail at pH 8.5, whereas the E2H and E2H 3 forms accumulate at pH 7.0 when Ca 2ϩ is absent (7), our data suggest that the Mg 2ϩ effect depends on the enzyme being in the E1 form.
ATP has two negative charges on the ␥-phosphoryl group at pH 8.5 and is partially protonated at pH 7.0 (pK a ϭ 6.63), whereas MgATP can be expected to have a single negative charge at both pH values (pK a ϭ 4.72) (37). Mg 2ϩ might have been expected to decrease the electrostatic interaction at both pH values. Also, Mg 2ϩ is expected to polarize the P-O bond, withdrawing electronegativity from the phosphorus atom, which should further reduce the repulsion. But these ameliorating effects of Mg 2ϩ appear to be counteracted by an increased binding interaction and a rise in repulsion as reflected in the change in interaction energy from 1.6 kcal/mol to 2.6 kcal/mol, resulting from Mg 2ϩ binding at pH 8.5 in the case of mutant D351N. This is probably the result of Mg 2ϩ assisting a close approach of the ␥-phosphoryl group and the carboxylate anion in the E1 state.
In the presence of Mg 2ϩ , the decrease in nucleotide-Asp 351 interaction energy observed for the D351N mutant when the pH is reduced from 8.5 to 7.0 is roughly equivalent to the decrease seen upon thapsigargin binding or replacement of MgATP with MgADP. This is consistent with the hypothesis that, in the first two instances, the ␥-phosphoryl group is withdrawn 3-4 Å from Asp 351 , or approximately the length of a phosphate group. Since E2 states are expected to prevail both at the lower pH and in the presence of thapsigargin, the displacement of nucleotide from the phosphorylation loop under these conditions is compatible with other findings that ATP and P i can bind simultaneously at the active site in E2 states (10 -14).
The ATP-Asp 351 interaction energy calculated for mutant D351A at alkaline pH in the presence of Mg 2ϩ is similar to the interaction energy calculated for D351N at pH 7.0 in the presence of Mg 2ϩ . For mutant D351A, there is furthermore little effect of reducing the Mg 2ϩ concentration or pH. The reason could be that mutation D351A to some extent counteracts the influence of alkaline pH on the E1-E2 conformational equilibrium in the absence of Ca 2ϩ . This would be equivalent to the displacement of the equilibrium in favor of E2 reported for the D369A mutation in the Na ϩ ,K ϩ -ATPase (6).
A central observation in the present study is that Ca 2ϩ binding to the transport sites causes a substantial increase in the intrinsic nucleotide binding energy of either of the mutants D351N and D351A at pH 7.0 and of D351A at pH 8.5. The same results were obtained both with the photolabeling assay and a direct filtration assay. In the case of mutant D351A at pH 7.0, the affinity for MgATP is increased 300-fold upon Ca 2ϩ binding. The affinity of the wild-type Ca 2ϩ -ATPase for MgATP in the presence of Ca 2ϩ cannot be measured directly in equilibrium binding experiments due to the activation of phosphoryl transfer by Ca 2ϩ , but it has been estimated from rapid kinetic determinations of association and dissociation rate constants to be very similar to the affinity in the absence of Ca 2ϩ at pH 7.0 (8). The very large increase in MgATP affinity induced by Ca 2ϩ binding in mutant D351A may possibly be ascribed to a close resemblance of the E1Ca 2 complex of mutant D351A with MgATP to the transition state for phosphoryl transfer (a "pseudotransition state"), the only difference from the transition state in the wild type being the absence of strong electrostatic effects corresponding to an interaction energy of as much as 4.5 kcal/mol (Table III). The reason why the corresponding complex of mutant D351N is less tight may be that the amino hydrogens of the asparagine impose steric hindrance to the closest possible approach of the ␥-phosphoryl group of ATP, whereas the shorter alanine side chain with attendant void can be accommodated. This would be especially critical in an associative mechanism of phosphoryl transfer (where there is substantial bond formation between the oxyanion of the carboxylate and the phosphorus atom while the bond between the ␥-phosphorus and the bridging oxygen is being broken) compared with a dissociative mechanism (where a distinct metaphosphate intermediate is generated) (see Scheme 2).
In light of the Ca 2ϩ effects on nucleotide affinities described here, it is noteworthy that the ATP affinity of the Na ϩ ,K ϩ -ATPase mutant D369A appeared to be unaffected by the presence of Na ϩ (6), although this cation plays a role as activator of the phosphorylation reaction in Na ϩ ,K ϩ -ATPase equivalent to that of Ca 2ϩ in the Ca 2ϩ -ATPase. A likely explanation is that the studies on Na ϩ ,K ϩ -ATPase mutants were conducted in the absence of Mg 2ϩ , which is also required at the catalytic site for transition state formation and, thus, for attainment of the extremely high nucleotide affinity.
The Ca 2ϩ activation of phosphorylation in the wild-type Ca 2ϩ -ATPase and of nucleotide binding with very high affinity in the D351A and D351N mutants implies that signals originating in the Ca 2ϩ transport sites in the membrane are conveyed by protein conformational changes over long distance to the catalytic site in the cytoplasmic protein domain. Signals originating at the catalytic site are likewise transmitted in the reverse direction from the cytoplasmic portion down the stalk to the Ca 2ϩ sites (5,38). Hence, in the wild-type enzyme the phosphorylation of Asp 351 (or perhaps already the formation of the transition state) leads to occlusion of Ca 2ϩ . The occluded Ca 2ϩ is released from the phosphoenzyme only following conformational changes that open the binding pocket toward the luminal side of the membrane (38,39). An intriguing question was, therefore, whether mutations D351N and D351A elicited events at the catalytic site that were transmitted to the membrane domain, affecting the Ca 2ϩ binding properties of the transport sites. We tested the affinity for Ca 2ϩ in mutants SCHEME 2 D351N and D351A by making use of the Ca 2ϩ -induced increase in affinity for TNP-8N 3 -ATP at pH 7.0. The results show that the apparent Ca 2ϩ affinities of the mutants were significantly higher than the literature values for the wild-type Ca 2ϩ -ATPase in sarcoplasmic reticulum determined under comparable conditions in the presence or absence of nonhydrolyzable nucleotide (32,33) and highest in the case of mutation D351A. In our assay for Ca 2ϩ binding (Fig. 8), the concentration of TNP nucleotide present was by necessity only partially saturating for the most part, and therefore the K D values for Ca 2ϩ binding to mutant proteins with bound nucleotide should be even lower than the 0.1 and 0.05 M indicated by the data in the figure. Also, the "principle of linked functions" (40) resulting from basic thermodynamics predicts that when Ca 2ϩ binding increases the affinity for nucleotide, then nucleotide binding should correspondingly increase the affinity for Ca 2ϩ . Hence, the results support the hypothesis that in the pseudotransition state induced by the mutations in the presence of nucleotide, Ca 2ϩ is bound with increased affinity relative to the normal E1 state of the wild-type enzyme.
Incubation of the Ca 2ϩ -ATPase with CrATP as substitute for ATP leads to gradual formation of a very slowly dissociating complex with the active site without phosphorylating Asp 351 , and simultaneously Ca 2ϩ bound at the transport sites becomes stably occluded (24 -27). The complex with CrATP may mimic to some extent a pretransition state, the transition state, or the phosphorylated state (38). Our findings with the D351N and D351A mutants seem to suggest that Asp 351 is critical for CrATP complexation and, thus, for the accompanying Ca 2ϩ occlusion. The mechanism behind the development of the slowly dissociating complex over time is unknown, but it might involve replacement of the Cr 3ϩ -coordinated waters with protein ligands, and if this is the case then the carboxylate of Asp 351 appears to be one of these. Caution is needed in interpreting these results, because even traces of ATP present as contaminant in the CrATP preparation might be able to compete efficiently with CrATP for binding to the mutants due to their very high ATP affinity. However, we have not been able to detect any significant level of ATP contamination in the purified CrATP preparation that was used.
In conclusion, the present results seem to support the following ideas concerning the mechanism of catalysis and energy transduction in the initial part of the Ca 2ϩ -ATPase reaction cycle. At pH 7.0 and before the binding of Ca 2ϩ , MgATP binds fairly tightly such that the nucleotide and Asp 351 may be poised electrostatically repelling each other to the extent of approximately 1.6 kcal/mol. Ca 2ϩ binding at the transport sites causes a significant increase in the favorable nucleotide-active site interactions, to the extent of approximately 2.9 (4.5 Ϫ 1.6) kcal/mol, which in the wild type may be used to balance an increase in the electrostatic repulsion between the ␥-phosphate and Asp 351 , thereby facilitating the phosphoryl transfer. The Ca 2ϩ -induced conformational change could involve the repositioning of a positively charged residue, such as a lysine, in proximity to the ␥-phosphate and Asp 351 to create a salt linkage with the phosphate. This mechanism finds support in the Ca 2ϩ -dependent cross-linking of a lysyl residue to Asp 351 (41) and the involvement of a lysyl residue in the catalytic mechanism of haloacid dehalogenases (35). In the Ca 2ϩ -ATPase, the events at the catalytic site elicit further conformational changes in the Ca 2ϩ binding domain that increase the Ca 2ϩ affinity and bring about occlusion of the Ca 2ϩ sites.