Ca Occlusion and Gating Function of Glu 309 in the ADP-Fluoroaluminate Analog of the Ca 2 (cid:1) -ATPase Phosphoenzyme Intermediate*

In the absence of ATP the sarcoplasmic reticulum ATPase (SERCA) binds two Ca 2 (cid:1) with high affinity. The two bound Ca 2 (cid:1) rapidly undergo reverse dissociation upon addition of EGTA, but can be distinguished by isotopic exchange indicating fast exchange at a superfi-cial site (site II), and retardation of exchange at a deeper site (site I) by occupancy of site II. Site II mutations that allow high affinity binding to site I, but only low affinity binding to site II, show that retardation of isotopic exchange requires higher Ca 2 (cid:1) concentrations with the N796A mutant, and is not observed with the E309Q mutant even at millimolar Ca 2 (cid:1) . Fluoroaluminate forms a complex at the catalytic site yielding stable analogs of the phosphoenzyme intermediate, with properties similar to E 2-P or E 1-P (cid:1) Ca 2 . Mutational analysis indicates that Asp 351 , Lys 352 , Thr 353 , Asp 703 , Asn 706 , Asp 707 , Thr 625 , and Lys 684 participate in stabilization of fluoroaluminate and Mg 2 (cid:1) at the phosphorylation site. In the presence of fluoroaluminate and Ca 2 (cid:1) , ADP (or AMP-PCP) favors formation of a stable ADP (cid:1) E 1-P (cid:1) Ca 2 analog. SERCA concentrations as determined by Western blotting. Contaminant Ca 2 (cid:1) in various reaction mixtures was estimated by ti-tration with EGTA in the presence of metallochromic indicators in a double wavelength spectrophotometer. Ca 2 (cid:1) dependent hydrolytic activity and Ca 2 (cid:1) transport were measured as previously described (13). Ca 2 (cid:1) binding at equilibrium was measured by incubating SR vesicles (40 (cid:1) g/ml) or COS-1 cell microsomes (60 (cid:1) g/ml) in reaction mixtures containing 20 m M MOPS, pH 7.0, 80 m M KCl, 5 m M MgCl 2 , 0.2 m M EGTA, [ 45 Ca]CaCl 2 to yield various concentrations of free Ca 2 (cid:1) (14), and 5 (cid:1) M A23187 calcium ionophore. Thapsigargin (TG) was added to half of the samples (1 (cid:1) M ) to provide controls exhibiting no specific Ca 2 (cid:1) binding. Following 15 min incubation at 25 °C, 1-ml samples were filtered by suction on 0.45- (cid:1) m Millipore filters, and the filters were collected, blotted, and processed for scintillation counting. at 25 °C, and the samples were then used directly for ATPase measurements by adding 0.95 m M Ca 2 (cid:1) (when Ca 2 (cid:1) was not present), and ATP (1 m M ). Reaction with fluoroaluminate to determine Ca 2 (cid:1) binding or dissociation of bound Ca 2 (cid:1) from the ATPase was obtained by incubating SR vesicles with [ 45 Ca]Ca 2 (cid:1) as explained above, but with the addition of 2 m M potassium fluoride and 5 or 50 (cid:1) M AlCl 3 , in the absence or presence of various concentrations of ADP or AMP-PCP. Following 30 min incubation at 25 °C, 1-ml samples were placed on a 0.65- (cid:1) m Millipore filter and the medium was removed by suction. The filters were then processed for determination of bound calcium. Alternatively, the vesicles were perfused with the same medium as for preincubation, but containing 10 m M EGTA for dissociation of bound [ 45 Ca]Ca 2 (cid:1) . The experiments were performed in a BioLogic rapid filtration apparatus, and increasing perfusion times were applied to individual samples. The filters were then collected, blotted, and processed for scintillation counting. Controls with TG were obtained for each perfusion time. For the experiments on limited proteolytic digestion, microsomes containing recombinant (chicken or rabbit) ATPase were preincubated for 40 min at 25 °C in media containing 50 m M MOPS, pH 7.0, 50 m M NaCl, 1.2 mg of microsomal protein/ml, and 2.0 m M EGTA, in the presence or absence of 2.1 m M CaCl 2 , 5 m M MgCl 2 , 2 m M NaF, 5 or 50 (cid:1) M AlCl 3 , and 0.1 m M ADP. After 60 min, limited proteolytic digestion was started by the addition of either 0.04 mg of proteinase K or 0.012 mg of trypsin/ml. The reaction was quenched after 50 min of digestion by the addition of trichloroacetic acid to reach a 2.5% final concentration. The quenched protein was then solubilized by adding sodium dodecyl sulfate (1%), Tris (0.312 M ), pH 6.8, sucrose (3.75%), (cid:2) -mercap-toethanol (1.25 m M ), and bromphenol blue (0.025%). The samples were then subjected to electrophoretic analysis (15) on 12% gels followed by staining with Coomassie Blue. Western blots were obtained with mono- clonal antibody CaF3–5C3 for chicken SERCA-1 or MA911 for rabbit SERCA1, followed by goat anti-mouse IgG horseradish peroxidase- conjugated secondary antibodies and visualization with an enhanced chemiluminescence-linked detection system (Amersham Biosciences).

The sarcoendoplasmic reticulum Ca 2ϩ -ATPase (SERCA) 1 is a ubiquitous Ca 2ϩ pump, required for intracellular Ca 2ϩ storing and Ca 2ϩ signaling mechanisms. The SERCA1 isoform is a 994-amino acid protein that includes 10 transmembrane segments and a cytosolic headpiece comprising three (N, P, and A) distinct domains (1,2). The reaction sequence (Scheme I) of the catalytic and transport cycle (3) includes high affinity binding of 2 Ca 2ϩ , ATP binding, formation of a phosphorylated intermediate, release of ADP, isomerization of the phosphorylated intermediate from a state of high affinity (E1-P⅐Ca 2 ) to a state of low affinity for Ca 2ϩ (E2-P⅐Ca 2 ), vectorial dissociation of bound Ca 2ϩ , and finally, hydrolytic cleavage of phosphate.
The initial binding of two Ca 2ϩ (in exchange for H ϩ ) occurs sequentially (4 -6) and involves two neighboring sites (I and II) within the ATPase membrane region where stabilization is provided by amino acid residues deriving from the M4, M5, M6, and M8 transmembrane helices (2,7). On the other hand, binding and catalytic utilization of ATP occurs within the cytosolic headpiece of the enzyme with direct or indirect intervention of several residues of the N, P, and A domains. In the absence of ATP, the two bound Ca 2ϩ undergo rapid reverse dissociation when the Ca 2ϩ concentration in the medium is lowered by the addition of EGTA (6). However, upon addition of ATP, occlusion (i.e. lack of reverse dissociation) of the two bound Ca 2ϩ is observed upon enzyme phosphorylation (8,9).
It was previously reported (10) that incubation of ATPase with fluoroaluminate yields an ATPase complex with characteristics attributed to E2-P or E1-P⅐Ca 2 enzyme intermediates. We have therefore studied the reverse dissociation and exchange of bound Ca 2ϩ in the absence of ATP, and the Ca 2ϩ dissociation following formation of a phosphoenzyme analog with fluoroaluminate. In addition, we performed a mutational analysis of the Ca 2ϩ binding site and the catalytic domain, to clarify the role of various residues in stabilization of fluoroaluminate and in the mechanism of Ca 2ϩ occlusion.

MATERIALS AND METHODS
Native sarcoplasmic reticulum (SR) vesicles were obtained by homogenization and differential centrifugation of rabbit white muscle (11). Recombinant ATPase protein was obtained with the microsomal fraction of COS-1 cells infected with adenovirus vectors carrying WT or mutated (chicken or rabbit) SERCA1 cDNA. The methods for construction of vectors, cultures, and preparation of microsomes were previously described in detail (12). The total microsomal protein was determined using bicinchoninic acid with the biuret reaction (Pierce). In all experiments with recombinant ATPase, the total protein concentrations in reaction mixtures, or the experimental results, were adjusted to reflect comparable SERCA concentrations as determined by Western blotting. Contaminant Ca 2ϩ in various reaction mixtures was estimated by titration with EGTA in the presence of metallochromic indicators in a double wavelength spectrophotometer. Ca 2ϩ dependent hydrolytic activity and Ca 2ϩ transport were measured as previously described (13).
Ca 2ϩ binding at equilibrium was measured by incubating SR vesicles (40 g/ml) or COS-1 cell microsomes (60 g/ml) in reaction mixtures containing 20 mM MOPS, pH 7.0, 80 mM KCl, 5 mM MgCl 2 , 0.2 mM EGTA, [ 45 Ca]CaCl 2 to yield various concentrations of free Ca 2ϩ (14), and 5 M A23187 calcium ionophore. Thapsigargin (TG) was added to half of the samples (1 M) to provide controls exhibiting no specific Ca 2ϩ binding. Following 15 min incubation at 25°C, 1-ml samples were filtered by suction on 0.45-m Millipore filters, and the filters were collected, blotted, and processed for scintillation counting.
The time course of [ 45 Ca]Ca 2ϩ dissociation from the ATPase, was determined by first incubating SR vesicles (40 g/ml) or COS-1 cell microsomes (100 g/ml) in a medium containing 50 mM MOPS, pH 7.0, 80 mM KCl, 5 mM MgCl 2 , 5 M A23187 calcium ionophore, and 20 M [ 45 Ca]Ca 2ϩ (including contaminant Ca 2ϩ ). TG was added to half of the samples (1 M) to provide controls exhibiting no specific Ca 2ϩ binding. Following 10 min incubation at 25°C, 1-ml samples were placed on a 0.65-m Millipore filter and the medium was removed by suction. The vesicles were then perfused with the same medium, but containing 10 mM EGTA for dissociation of bound [ 45 Ca]Ca 2ϩ , or various concentrations of [ 40 Ca]Ca 2ϩ for exchange with bound [ 45 Ca]Ca 2ϩ . The experiments were performed in a BioLogic rapid filtration apparatus, and increasing perfusion times were applied to individual samples. The filters were then collected, blotted, and processed for scintillation counting. Controls with TG were obtained for each perfusion time.
Reaction with fluoroaluminate to test ATPase inhibition was obtained by incubating SR vesicles (0.06 mg/ml) in media containing 40 mM MOPS, pH 7.0, 80 mM KCl, 2 mM MgCl 2 , 2 M A23187 Ca 2ϩ ionophore, 1 mM EGTA, 2 mM potassium fluoride, and various concentrations (0 -50 M) of AlCl 3 , in the presence or absence of 0.95 mM CaCl 2 . The incubation was carried out for 30 min at 25°C, and the samples were then used directly for ATPase measurements by adding 0.95 mM Ca 2ϩ (when Ca 2ϩ was not present), and ATP (1 mM).
Reaction with fluoroaluminate to determine Ca 2ϩ binding or dissociation of bound Ca 2ϩ from the ATPase was obtained by incubating SR vesicles with [ 45 Ca]Ca 2ϩ as explained above, but with the addition of 2 mM potassium fluoride and 5 or 50 M AlCl 3 , in the absence or presence of various concentrations of ADP or AMP-PCP. Following 30 min incubation at 25°C, 1-ml samples were placed on a 0.65-m Millipore filter and the medium was removed by suction. The filters were then processed for determination of bound calcium. Alternatively, the vesicles were perfused with the same medium as for preincubation, but containing 10 mM EGTA for dissociation of bound [ 45 Ca]Ca 2ϩ . The experiments were performed in a BioLogic rapid filtration apparatus, and increasing perfusion times were applied to individual samples. The filters were then collected, blotted, and processed for scintillation counting. Controls with TG were obtained for each perfusion time.
For the experiments on limited proteolytic digestion, microsomes containing recombinant (chicken or rabbit) ATPase were preincubated for 40 min at 25°C in media containing 50 mM MOPS, pH 7.0, 50 mM NaCl, 1.2 mg of microsomal protein/ml, and 2.0 mM EGTA, in the presence or absence of 2.1 mM CaCl 2 , 5 mM MgCl 2 , 2 mM NaF, 5 or 50 M AlCl 3 , and 0.1 mM ADP. After 60 min, limited proteolytic digestion was started by the addition of either 0.04 mg of proteinase K or 0.012 mg of trypsin/ml. The reaction was quenched after 50 min of digestion by the addition of trichloroacetic acid to reach a 2.5% final concentration. The quenched protein was then solubilized by adding sodium dodecyl sulfate (1%), Tris (0.312 M), pH 6.8, sucrose (3.75%), ␤-mercaptoethanol (1.25 mM), and bromphenol blue (0.025%). The samples were then subjected to electrophoretic analysis (15) on 12% gels followed by staining with Coomassie Blue. Western blots were obtained with monoclonal antibody CaF3-5C3 for chicken SERCA-1 or MA911 for rabbit SERCA1, followed by goat anti-mouse IgG horseradish peroxidaseconjugated secondary antibodies and visualization with an enhanced chemiluminescence-linked detection system (Amersham Biosciences).

Ca 2ϩ
Binding and Dissociation in the Absence of ATP-Equilibrium binding of Ca 2ϩ to the native ATPase of SR vesicles has been previously observed within the micromolar concentration range (4). However, measurements of Ca 2ϩ binding to recombinant ATPase are more difficult because of low concentrations of the enzyme in COS-1 cell microsomes, and restrictive conditions required for subsequent kinetic experiments (i.e. filtration and low Ca 2ϩ /EGTA concentrations, as opposed to column chromatography and high Ca 2ϩ /EGTA buffer). Under our present conditions, we obtained equilibrium binding isotherms that yield a maximal stoichiometry of two Ca 2ϩ per ATPase, and are best fitted with a cooperative equation assuming K eq values of 7 ϫ 10 5 and 2 ϫ 10 6 M Ϫ1 for two interacting sites (16). Identical values, adjusted for the ATPase molar concentration, are obtained with native and recombinant WT enzyme.
Dissociation of bound Ca 2ϩ was studied by preincubating microsomes with 20 M [ 45 Ca]Ca 2ϩ for 15 min, which results in saturation the high affinity binding sites (2 Ca 2ϩ /ATPase). We placed preincubated SR vesicles on filters for removal of the medium by suction, and subjected them to perfusion with 10 mM EGTA and no added calcium (2 ϫ 10 Ϫ10 M free Ca 2ϩ ) for various time intervals in the rapid filtration apparatus. We found that, following addition of EGTA, most of the bound calcium dissociates with a 21-26 s Ϫ1 rate constant ( Fig. 1), whereas ϳ10% of the total signal dissociates slightly slower (see also Ref. 6). It is apparent that under these conditions the bound Ca 2ϩ dissociates rapidly and the two sites cannot be distinguished.
Contrary  Effects of Site-directed Mutations-Consistent with the cooperative mechanism, single mutations in site I interfere with Ca 2ϩ binding on both sites I and II, even when the residue undergoing mutation does not participate directly in Ca 2ϩ binding at site II (17). On the other hand, high affinity binding (K eq ϭ 10 6 M Ϫ1 ) of one Ca 2ϩ on site I is still retained following single mutations on site II (i.e. E309Q or N796A), whereas site II looses high affinity binding (16). Consequently, no binding to site II is observed at micromolar Ca 2ϩ concentrations.
We found that, following addition of EGTA, Ca 2ϩ dissociation from these mutants (i.e. from site I) occurs with a single rate constant of 2-4 s Ϫ1 (Fig 2, A and B). This is significantly slower than observed with the WT enzyme, and may reflect additional stabilization of Ca 2ϩ in site I by oxygen functions of site II. For instance, the Asp 800 side chain, which contributes one oxygen function to site I and one oxygen function to site II in the WT enzyme, is likely to contribute both its oxygen functions to site I in the site II mutants.
We measured calcium isotope exchange in the two mutants N796A and E309Q. As explained above, Asn 796 and Glu 309 residues contribute side chain oxygen atoms for Ca 2ϩ complexation on site II. Therefore, while retaining high affinity Ca 2ϩ binding to site I, these mutants loose high affinity Ca 2ϩ binding to site II. On the other hand, low affinity binding is still retained by the altered site II, as suggested by inhibition of enzyme phosphorylation with P i (18). We then tested if addition of [ 40 Ca]Ca 2ϩ would result in isotopic exchange with [ 45 Ca]Ca 2ϩ bound to site I. We found that the two mutants exhibit a different behavior, as dissociation (i.e. isotopic exchange) of bound [ 45 Ca]Ca 2ϩ is increasingly retarded by [ 40 Ca]Ca 2ϩ within the 0.1-1.0 mM range in the N796A mutant ( Fig. 2A), whereas the E309Q mutant exhibits dissociation as rapid as that observed in the presence of EGTA even following addition of 1 mM [ 40 Ca]Ca 2ϩ (Fig. 2B). This indicates that Glu 309 makes a greater contribution than Asn 796 to the affinity of site II for Ca 2ϩ . Therefore, the Glu 309 side chain plays a very important role in retardation of Ca 2ϩ dissociation from site I and in capping the calcium binding cavity.
Formation of Fluoroaluminate-ATPase Complex-Intrinsic fluorescence and Ca 2ϩ binding measurements were previously reported (10) indicating that incubation of ATPase with fluoroaluminate yields a fluoroaluminate-ATPase complex with characteristics attributed to the E2-P or E1-P⅐Ca 2 enzyme intermediate. In our experiments, we tested the formation of stable intermediate analogs by incubating the ATPase with various concentrations of fluoroaluminate in the absence or presence of Ca 2ϩ , and then measured the residual Ca 2ϩ -dependent ATPase activity. It is shown in Fig. 3 that ATPase inactivation occurs in either case, but higher concentrations of fluoroaluminate are required when the incubation is carried out in the presence of Ca 2ϩ . This demonstrates that a stable complex is formed in either case, most likely corresponding to E2-P in the absence of Ca 2ϩ , and to E1-P⅐Ca 2 in the presence of Ca 2ϩ . Formation of the latter analog requires higher fluoroaluminate concentrations.
We also explored whether formation of the fluoroaluminate complex would produce the ATPase conformational changes as expected of phosphoenzyme intermediate formation. For this  purpose we conducted experiments using partial digestion with proteinase K or trypsin, which may reveal protection of specific sites on the peptide loops intervening between the A domain and the M3 transmembrane segment. This protection is observed following reaction of the ATPase with phosphate analogs, suggesting large conformational changes that include various degrees of A domain rotation and gathering of the cytosolic domains (19). Accordingly we found that, although the pattern of digestion by proteinase K is different in the absence or presence of Ca 2ϩ , protection is obtained following reaction with 50 M fluoroaluminate in either case (Fig. 4). In fact even Mg 2ϩ makes little difference, indicating that while Mg 2ϩ is required for electrophilic assistance in the catalytic mechanism, its contribution to steric stabilization of fluoroaluminate is not apparent in our experiments with WT enzyme. On the other hand, following mutation of residues (i.e. Asp 601 , Lys 352 , Asn 706 , and Lys 684 ) that contribute to fluoroaluminate stabilization in the WT enzyme (see below and Table II), the Mg 2ϩ contribution to stabilization of the complex becomes apparent.
A conformational effect of Ca 2ϩ on the phosphoenzyme analog is demonstrated by the experiment shown in Fig. 5. In fact, protection from trypsin (as opposed to proteinase K) is obtained when the reaction with fluoroaluminate is performed in the absence of Ca 2ϩ , but not in the presence of Ca 2ϩ (Fig. 5, A and B). These results are consistent with conformational changes because of the fluoroaluminate reaction, and a significantly different A domain rotation in the E2-P and E1-P⅐Ca 2 analogs allowing protection of the Arg 198 trypsin site only in the former analog. Furthermore, it is shown in Fig. 5C that, in the presence of Ca 2ϩ , low fluoroaluminate concentrations (5 M) afford only partial protection from proteinase K. In this case, addition of ADP produces total protection, as acquisition of the ADP⅐E1-P⅐Ca 2 conformation is highly favored. It is noteworthy that formation of the E2-P analog is not influenced by ADP, as complete protection is already observed with 5 M fluoroaluminate.
We found that protection is not observed at all when the fluoroaluminate reaction is carried out with the D351N or D351A mutants (Table II). Considering that Asp 351 is the residue involved directly in the phosphorylation reaction, the lack of reactivity of Asp 351 mutants demonstrates unambiguously that the fluoroaluminate complex obtained with the WT enzyme is in fact a phosphoenzyme analog. It is remarkable that the same Asp 351 mutations favor protection by nucleotide binding, while interfering with protection by fluoroaluminate (see "Discussion").
We then performed a mutational analysis of several residues whose involvement in substrate binding and/or utilization was indicated by previous studies with SERCA (2,13) or with the highly analogous phosphoserine phosphatase (PSPase) enzyme (20). All of these mutations produce catalytic inhibition when either ATP or acetylphosphate are used as the substrate (Table  II). A special case is the R560A mutant in which a greater inhibition of ATP utilization is observed, evidently because of a prevalent role of Arg 560 in ATP (13), as compared with acetylphosphate binding. It is noteworthy that derivatization of Lys 515 (N domain) with fluoroisothiocyanate produces total inhibition of ATP utilization, while allowing partial utilization of acetylphosphate (not shown).
Various mutations within the P domain interfere with nucleotide or fluoroaluminate protection from proteinase K (Table  II). This is the case of K352A, T353A, T625A, K684A, D703A, N706A, and D707A. It is noteworthy that the Thr 353 , Asp 707 , and Thr 625 mutations have approximately the same degree of interference with fluoroaluminate protection in the absence or presence of Ca 2ϩ and/or Mg 2ϩ . On the other hand, interference by the Lys 352 , Asp 703 , Asn 706 , and Lys 684 mutations is relieved to some extent by the presence of Mg 2ϩ . Furthermore, in the K684A, N706A, and D703A mutants the effect of Mg 2ϩ is much more prominent when Ca 2ϩ is also present, suggesting a different contribution of fluoroaluminate stabilization by these residues in the E1-P⅐Ca 2 as compared with the E2-P conformation. Mutation of Gly 626 has very little effect on protection by fluoroaluminate, although it interferes with protection by nucleotide and produces catalytic inhibition.
A case of special interest is the N706A mutation that, in analogy to the Asp 351 mutation, favors protection by nucleotide (see "Discussion") while interfering with protection by fluoroaluminate only in the presence of Ca 2ϩ . This interference is totally relieved by Mg 2ϩ (Table II). The Asn 706 mutation produces complete inhibition of catalytic activity (see "Discussion").
Mutations of Arg 560 , Glu 439 , Arg 489 (N domain), Arg 174 , and Glu 183 (A domain) do not interfere significantly with fluoroaluminate protection from proteinase K, indicating that these residues do not contribute to stabilization of fluoraluminate. On the other hand, interference of nucleotide protection is produced by the Arg 560 and Glu 439 (but not Arg 489 ) mutations, suggesting that these residues are involved in nucleotide binding (13).
Ca 2ϩ Binding and Dissociation in the Fluoroaluminate-ATPase Complex-To study Ca 2ϩ dissociation from the fluoroaluminate complex, we first obtained a preliminary assessment of Ca 2ϩ binding when the ATPase is incubated for 30 min in the presence of 20 M Ca 2ϩ with fluoroaluminate at concentrations yielding enzyme inactivation (Fig. 3). We found that when the ATPase is incubated in the presence of 20 M [ 45 Ca]Ca 2ϩ in the absence or presence of fluoroaluminate, the maximal level of high affinity Ca 2ϩ binding (2 Ca 2ϩ -ATPase) is reduced to 85% if 5 M fluoroaluminate is present, and to 60% if 50 M fluoroaluminate is present (Table III). This reduction is evidently related to formation of the E2-P analog that does not retain high affinity binding, and accounts for 15 and 40% following incubation with 5 or 50 M fluoroaluminate, respectively (Table III).
When we studied the time course of Ca 2ϩ dissociation from the fluoroaluminate-ATPase upon addition of EGTA, we found a fast and a slow kinetic component, occurring with 20 -30 s Ϫ1 and 0.3-0.4 s Ϫ1 rate constants (Fig. 6). It is apparent that the fast dissociation is quite similar to that observed in the absence of fluoroaluminate (E1-Ca 2 ), whereas the slow component is 85 or 60% by 5 or 50 M fluoroaluminate, respectively, we conclude that incubation with 5 M fluoroaluminate yields 15% E2-P (no Ca 2ϩ bound), 52% E1-Ca 2ϩ (rapid dissociation), and 33% E1-P⅐Ca 2 (slow dissociation). On the other hand, incubation with 50 M fluoroaluminate yields 40% E2-P (no bound Ca 2ϩ ), 13% E1-Ca 2 (rapid dissociation), and 47% E1-P⅐Ca 2 (slow dissociation). These levels are summarized in Table III.
Because ADP favors formation of the E1-P⅐Ca 2 analog (Fig.   5C), we tested whether dissociation of bound Ca 2ϩ would be also affected. We found that addition of ADP to the incubation mixture with a fluoroaluminate concentration as low as 5 M allowed 100% Ca 2ϩ binding. In addition, it produced total occlusion of bound Ca 2ϩ with no significant dissociation through-  Incubation with fluoroaluminate, in the presence of Ca 2ϩ and/or ADP, was carried out as described under "Materials and Methods." A, full protection from proteinase K is produced by preincubation with 50 M fluoroaluminate both in the presence or absence of Ca 2ϩ . B, full protection from trypsin (A1 fragment) is produced by preincubation with 50 M fluoroaluminate in the absence of Ca 2ϩ , but only partial protection in the presence of Ca 2ϩ . C, full protection from proteinase K is produced by preincubation with fluoroaluminate at a concentration as low as 5 M, in the absence of Ca 2ϩ ; ADP makes no difference. In the presence of Ca 2ϩ , however, 5 M fluoroaluminate protects only partially in the absence of Ca 2ϩ , but full protection is obtained by the addition of ADP.  out a 2-s perfusion with EGTA (Fig. 7A). Interestingly, the effective ADP concentration was in the micromolar concentration range. A similar high affinity (Fig. 7B) for the fluoroaluminate complex was exhibited by AMP-PCP, an ATP analog that is not utilized for enzyme phosphorylation. It is noteworthy that in the absence of fluoroaluminate, ADP and AMP-PCP produced no retardation of Ca 2ϩ dissociation even at millimoloar concentrations. The levels of Ca 2ϩ bound, the rates of dissociation, and the distribution of enzyme states under these conditions are summarized in Table III.
Finally, we tested whether occlusion of bound Ca 2ϩ was produced by fluoroaluminate and ADP in the E309Q and N796A mutants. It is shown in Fig. 8 that dissociation of bound Ca 2ϩ occurs rapidly following perfusion of the E309Q mutant with EGTA, demonstrating that the gating role of Glu 309 plays an important role for Ca 2ϩ occlusion in the ADP⅐E1-P⅐Ca 2 analog. Interestingly, the N796A mutant does not display tight occlusion as the WT enzyme does (Fig. 7). Rather, a slow dissociation of Ca 2ϩ is noted (Fig. 8) upon addition of EGTA. It is then apparent that even though Glu 309 plays a direct role in gating, engagement of all site II residues is required to obtain tight occlusion of bound Ca 2ϩ in the ADP⅐E1-P⅐Ca 2 analog.

DISCUSSION
The Ca 2ϩ Binding Mechanism-The stoichiometry (2 Ca 2ϩ per ATPase) and the cooperative character of high affinity Ca 2ϩ binding to SERCA was established by early equilibrium measurements in the absence of ATP (4). Mutational (7) and crystallographic studies (2) then showed that the two bound Ca 2ϩ reside within the transmembrane region of the ATPase (2), separated by a 5.7-Å distance, and surrounded by the M4, M5, M6, and M8 helices (Fig. 9). The side chains of Glu 771 (M5), Asn 768 (M5), Thr 799 (M6), Asp 800 (M6), and Glu 908 (M8), as well as two water molecules, contribute oxygen atoms for stabilization of one Ca 2ϩ (site I). Stabilization of the other Ca 2ϩ (site II) is obtained with side chain oxygen atoms of Glu 309 (M4), Asn 796 (M6), and Asp 800 (M6), and main chain carbonyl oxygen atoms of Val 304 , Ala 305 , and Ile 307 . Asp 800 contributes its two acidic side chain oxygen atoms to coordinate both calcium ions (sites I and II). Participation of all these residues is rendered possible by unwinding of the M4 and M6 helices. Our present experiments were designed to characterize the mechanism of reverse dissociation of the bound Ca 2ϩ and its dependence on local effects, as well as its occlusion through long range effects of phosphorylation site occupancy in a stable analog of the phosphoenzyme intermediate.
It is clear that, as a consequence of Ca 2ϩ binding, the side chains of residues involved in binding undergo a change in orientation, most prominently that of Glu 309 (Fig. 9). These changes are permitted by responsive displacement and reorganization of pertinent transmembrane helices, such as M4, M5, and M6, and the resulting coordination geometry is different for the two sites. The denomination I and II assumes that the two sites are occupied sequentially, as predicted by the cooperative mechanism. In fact, single mutations of site I residues prevent binding on both sites I and II. On the other hand, single mutations of site II residues interfere with binding on site II, but not on site I (12).
Ca 2ϩ Dissociation and Exchange-When the Ca 2ϩ concentration in the medium is lowered below 10 Ϫ8 M by addition of EGTA, reverse dissociation of bound Ca 2ϩ occurs rapidly, and a distinct behavior of the two bound Ca 2ϩ cannot be demonstrated. However, a clear distinction is obtained by the exchange experiments. It is shown in Fig. 1 (24).
An important finding is that retardation of exchange of Ca 2ϩ on site I is still obtained with the N796A mutant when 0.1-1.0 mM [ 40 Ca]Ca 2ϩ is added, but is not observed with the E309Q mutant even in the presence of 1 mM [ 40 Ca]Ca 2ϩ (Fig. 2). This indicates that the binding affinity of site II is lower in the Glu 309 than in the Asn 796 mutant, and the Glu 309 side chain has a very important role in Ca 2ϩ binding and capping the binding cavity. Structural studies (21) have demonstrated clearly that the Glu 309 side chain undergoes a pronounced change in orientation upon Ca 2ϩ binding, thereby sealing site II (Fig. 9).
Fluroaluminate Analog of the Phosphorylated Enzyme Intermediate-An early and important step of the catalytic cycle is the rapid occlusion of bound Ca 2ϩ that occurs in parallel with phosphoenzyme formation by utilization of nucleotide substrate (8,9). Ca 2ϩ occlusion is also produced by reaction of ATPase monomers with Cr-ATP (25). In the experiments reported here we have used fluoroaluminate to obtain a stable analog of the phosphorylated enzyme intermediate (10), and to define its effects on the ATPase conformational state, as well as its ability to bind and occlude Ca 2ϩ . No effects were observed with Asp 351 mutants, demonstrating that interaction of fluoroaluminate with this residue is required and the fluoroaluminate-ATPase complex is in fact an analog of the phosphoenzyme intermediate. It is of interest that mutation of Asp 351 appears to relieve a repulsive effect of its acidic side chain on the nucleotide ␥-phosphate (Ref. 13 and Table II) although it interferes strongly with fluoroaluminate stabilization. It is then clear that stabilization occurs only in concomitance with the covalent phosphoryl transfer reaction, or the transition state thereof. It is interesting that mutation of Asn 706 has an effect similar to the Asp 351 mutation, suggesting that in the WT enzyme Asn 706 may contribute to the relative positioning of ␥-phosphate and the Asp 351 side chain.
Experiments on limited proteolytic digestion are a useful device to study conformational changes associated with formation of phosphorylated enzyme intermediates, because protection of proteolytic sites within the ATPase loop connecting the A domain and the M3 transmembrane segment are indicative of cytosolic domain rotation and gathering (19). Using this technique, we found that resistance to proteinase K was obtained by preincubating the ATPase with high concentrations (50 M) of fluoroaluminate in the absence or presence of Ca 2ϩ (Fig. 4A). However, much lower protection to trypsin (T2 site) was observed when the fluoroaluminate reaction was conducted in the presence of Ca 2ϩ , as compared with the absence of Ca 2ϩ (Fig. 4B). This difference is attributed to a specific conformational difference because of rotation of the A domain, resulting in protection of the T2 site (Arg 198 ) in E2-P but not in E1-P⅐Ca 2 (19). This demonstrates that conformational analogs of E2-P or E1-P⅐Ca 2 are prevalently obtained when Ca 2ϩ is absent or present during the incubation with fluoroaluminate. In the presence of low (5 M) concentrations of fluoroaluminate (Fig. 5C), total protection from proteinase K is observed in the absence of Ca 2ϩ , but only partial protection in the presence of Ca 2ϩ because of low production of the E-P⅐Ca 2 analog. On the other hand, full protection is obtained by the addition of ADP, demonstrating unambiguously that under these conditions formation of the ADP⅐E-P⅐Ca 2 analog is favored and its conformation is highly resistant to digestion.
Mutational analysis of the fluoroaluminate protection of SERCA from proteinase K can be interpreted with reference to the high resolution structure of the fluoroaluminate intermediate of the analogous PSPase (20). For this purpose, the SERCA residues corresponding to the PSPase residues are shown in parentheses in Fig. 10. Comparison of our mutational analysis (Table II) with the PSPase intermediate structure (Fig. 10) shows that the aluminum atom is stabilized by the three fluorine atoms, and by oxygen atoms from the Asp 351 side chain and a water molecule. In turn, the fluorine atoms are stabilized by hydrogen bonding with backbone amino groups of Gly 626 , Lys 352 , and Thr 353 , as well as with Lys 684 , Asn 706 , and Thr 625 side chains. Furthermore, SERCA Thr 353 can still provide side chain oxygen for hydrogen bonding with water, even though Thr 353 corresponds to Asp 13 of the PSPase. It should be pointed out that although the residue corresponding to Asp 707 is not included in the interactions shown in the structure of the PSPase intermediate (Fig. 10), mutation of this residue interferes strongly with binding of ATP and fluoroaluminate in the SERCA enzyme (Table II). Asp 707 resides in close proximity of FIG. 9. The coordination geometry of the two Ca 2؉ sites. Site I includes six side chain oxygen atoms and two water molecules. Site II includes four side chain and three main chain oxygen atoms. The side chains are shown with magenta color for E1⅐Ca 2 , and yellow color for E2-TG (2,21). Note the wide swing of the Glu 309 side chain upon Ca 2ϩ binding. The figures were derived from Protein Data Bank accession numbers 1EUL for E1⅐Ca 2 and 1IWO for E2⅐TG. Water and main chain oxygen atoms are left out to show clearly the movements of side chains upon Ca 2ϩ binding or dissociation. A complete representation is given by Toyoshima and Inesi (22).
Asp 351 , and its mutation produces complete catalytic inactivation of SERCA (13). Mutation of Asp 601 also interferes with nucleotide and fluoroaluminate protection (Table II). In fact this residue is likely to provide stabilization to Lys 352 that, in turn, is involved in stabilization of one of the fluorine atoms.
The very small effect of the E183Q mutation on fluoroaluminate protection from proteinase K (Table II) indicates that Glu 183 is not directly involved in fluoroaluminate stabilization. It is apparent that the catalytic inactivation produced by Glu 183 mutation is related to the role of this residue in coordination of water (Fig. 10) for the final hydrolytic reaction (26).
It should be pointed out that structural studies indicate (Fig.  10) that Mg 2ϩ is coordinated with side chain oxygen atoms of Asp 703 and Asp 351 , the backbone oxygen of Thr 353 , and a fluorine atom. However, our proteinase K protection experiments with WT enzyme (Table II) do not reveal an important role for Mg 2ϩ in stabilization of fluoroaluminate, despite the known Mg 2ϩ requirement for electrophilic assistance in catalysis. On the other hand, an effect of Mg 2ϩ is observed in the protection experiments with the Lys 352 , Asp 706 , Lys 684 , and Asp 601 mutants (Table II). This suggests that Mg 2ϩ coordination can compensate for interference by mutations of residues that are normally involved (directly or indirectly) in stabilization of fluoroaluminate, and its importance in conformational stabilization becomes evident in experiments with selected mutants.
Fluoroaluminate Phosphoenzyme Analog and Ca 2ϩ Occlusion-The importance of Ca 2ϩ binding is related to its requirement for enzyme activation, which is transmitted to the extramembranous domains through a long range intramolecular linkage triggered by Ca 2ϩ occupancy of site II (16). This conformational state is identified as E1⅐Ca 2 . The structural changes produced by the final complex are quite extensive, as direct interaction of Ca 2ϩ with M4, M5, and M6 affects other transmembrane helices and, in turn, results in separation of the three headpiece domains (2,21). Opening of the headpiece domains in the Ca 2ϩ -activated enzyme (E1⅐Ca 2 ) then allows binding of ATP, whereby closure and cross-linking of the headpiece is produced by interactions of the adenosine moiety with the N domain and of the ␥-phosphate with the P domain (13,19,27). Phosphoryl transfer and formation of phosphoenzyme intermediates with high and low affinity for Ca 2ϩ occur as outlined in Scheme I.
Equilibrium measurements of Ca 2ϩ binding in the presence of fluoroaluminate are consistent with distribution of the enzyme into a fraction sustaining no high affinity Ca 2ϩ binding (i.e. E2-P analog), and a fraction sustaining high affinity Ca 2ϩ binding. Kinetic measurements then demonstrate that the latter fraction contains E1⅐Ca 2 and E1-P⅐Ca 2 , which undergo fast or slow dissociation upon addition of EGTA ( Fig. 5 and Table III). It is interesting that ADP or AMP-PCP produce strong stabilization of the E1-P⅐Ca 2 analog and great enhancement of Ca 2ϩ occlusion (Fig. 6). This effect is likely because of approximation of P and N domains through cross-linking by the nucleotide (13). The associated A domain rotation (19) is then transmitted to transmembrane helices that affect the Ca 2ϩ binding sites.
Considering that the combination of fluoroaluminate (i.e. analog of phosphate) and ADP produce strong occlusion, we wondered why we did not observe occlusion with AMP-PCP or AMP-PNP. In fact, both in our experiments with fluoroaluminate and in those by Vilsen and Andersen (28) with Cr-ATP, Ca 2ϩ occlusion occurred slowly and in parallel with engagement of Asp 351 by the ATP terminal phosphate or its analog (AlF 3 ). An explanation may be found in the tendency of the nucleotide substrate to bind with the phosphate chain either in a folded (29) or an extended (13) configuration. The folded configuration keeps the two terminal phosphates and bound Mg 2ϩ near Thr 441 , and does not cross-link effectively the N and P domains. On the other hand the extended configuration, which is required for the conformational change producing Ca 2ϩ occlusion, is subjected to unfavorable statistics because of repulsion of the terminal phosphate by Asp 351 (Table II) and is only stabilized by the covalent phosphorylation reaction. It is possible that the conformational change producing occlusion may also occur under special conditions that favor the extended configuration even in the absence of covalent interaction (30).
A point of great interest is that while both Ca 2ϩ bound to the E1-P⅐Ca 2 analog are strongly occluded by fluoroaluminate and ADP in the WT enzyme (Fig. 7), rapid dissociation is observed with the E309Q mutant. Therefore, Glu 309 has a determinant role not only in retardation of Ca 2ϩ dissociation from site I through occupancy of site II (i.e. Ca 2ϩ exchange), but also in the occlusion of both Ca 2ϩ sites that are produced by the long range conformational change triggered by formation of the ADP⅐E1-P⅐Ca 2 intermediate. It is also interesting that occlusion by fluoroaluminate and ADP is not produced efficiently even in the N796A mutant, and significant Ca 2ϩ dissociation from this mutant occurs upon addition of EGTA. It was previously shown that occurrence of a similar long range linkage (i.e. catalytic activation by Ca 2ϩ binding) requires engagement of all site II residues by Ca 2ϩ , including Asn 796 (16,31). It is then apparent that, in the reverse direction, a full effect of the long range signal also requires engagement of Glu 309 as well as other site II residues.
It should be understood that as opposed to retardation of site I [ 45 Ca]Ca 2ϩ exchange in E1⅐Ca 2 (Fig. 1), which is because of local engagement of the Glu 309 side chain by [ 40 Ca]Ca 2ϩ bound to site II, occlusion of both Ca 2ϩ (Figs. 6 and 7) is observed in the dissociation experiments upon addition of EGTA to the E1-P⅐Ca 2 analog. In this case, occlusion is produced by locking the Glu 309 side chain in the closed position through an extensive and long range conformational change that is triggered at the catalytic site and includes headpiece domains and transmembrane helices. Detailed structural analysis (30) shows that the binding cavity of E1⅐Ca 2 is capped by Glu 309 , but a waterfilled space still allows movements of the Glu 309 side chain. The consequent Ca 2ϩ dissociation from site II then leads to exchange with medium Ca 2ϩ (Fig. 1). On the other hand, follow- ing the long range conformational change, this aqueous space is filled by M1 (30), which now locks the Glu 309 side chain (compare Figs. 7 and 8) through a suitable switch of local hydrogen bonds and van der Waals interactions. The functional relevance of M1 has been also shown by mutational analysis (32). Gating of the occluded state must be then considered a specific function of Glu 309 (33), in addition to its prominent role in determining the high affinity of site II for Ca 2ϩ . A general discussion of the gating role of single amino acids in ion transport was recently published by Gadsby (34).
Conclusions-Our findings are consistent with a sequential binding mechanism, beginning with Ca 2ϩ entry through an aqueous cavity lined with polar residues (Lys 47 , Glu 55 , Glu 58 , Asp 59 , Glu 109 , Lys 246 , Lys 254 , and Gln 250 ) on the cytosolic side of the ATPase near the membrane interface (21). Progression through the cavity leads to Ca 2ϩ stacking in sites I and II. Formation of an enzyme intermediate (ADP⅐E1-P⅐Ca 2 ) analog with fluoroaluminate produces tight occlusion of both bound Ca 2ϩ . This phenomenon corresponds to the mechanism of Ca 2ϩ occlusion that occurs early in the catalytic cycle, in parallel with phosphoenzyme formation by utilization of ATP. Engagement of Glu 309 at the gateway of the binding cavity serves as an important gating device for occlusion of bound Ca 2ϩ . Mutational analysis indicates that the amino acid residues involved in stabilization of fluoraluminate at the SERCA phosphorylation site are apparently identical to the corresponding residues in the high resolution structure of the fluoroaluminate intermediate analog of PSPase (20).