Formation of the Stable Structural Analog of ADP-sensitive Phosphoenzyme of Ca2+-ATPase with Occluded Ca2+ by Beryllium Fluoride

As a stable analog for ADP-sensitive phosphorylated intermediate of sarcoplasmic reticulum Ca2+-ATPase E1PCa2·Mg, a complex of E1Ca2·BeFx, was successfully developed by addition of beryllium fluoride and Mg2+ to the Ca2+-bound state, E1Ca2. In E1Ca2·BeFx, most probably E1Ca2·BeF3−, two Ca2+ are occluded at high affinity transport sites, its formation required Mg2+ binding at the catalytic site, and ADP decomposed it to E1Ca2, as in E1PCa2·Mg. Organization of cytoplasmic domains in E1Ca2·BeFx was revealed to be intermediate between those in E1Ca2·AlF4− ADP (transition state of E1PCa2 formation) and E2·BeF3−·(ADP-insensitive phosphorylated intermediate E2P·Mg). Trinitrophenyl-AMP (TNP-AMP) formed a very fluorescent (superfluorescent) complex with E1Ca2·BeFx in contrast to no superfluorescence of TNP-AMP bound to E1Ca2·AlFx. E1Ca2·BeFx with bound TNP-AMP slowly decayed to E1Ca2, being distinct from the superfluorescent complex of TNP-AMP with E2·BeF3−, which was stable. Tryptophan fluorescence revealed that the transmembrane structure of E1Ca2·BeFx mimics E1PCa2·Mg, and between those of E1Ca2·AlF4−·ADP and E2·BeF3−. E1Ca2·BeFx at low 50–100 μm Ca2+ was converted slowly to E2·BeF3− releasing Ca2+, mimicking E1PCa2·Mg → E2P·Mg + 2Ca2+. Ca2+ replacement of Mg2+ at the catalytic site at approximately millimolar high Ca2+ decomposed E1Ca2·BeFx to E1Ca2. Notably, E1Ca2·BeFx was perfectly stabilized for at least 12 days by 0.7 mm lumenal Ca2+ with 15 mm Mg2+. Also, stable E1Ca2·BeFx was produced from E2·BeF3− at 0.7 mm lumenal Ca2+ by binding two Ca2+ to lumenally oriented low affinity transport sites, as mimicking the reverse conversion E2P· Mg + 2Ca2+ → E1PCa2·Mg.

Despite these atomic structures, yet unsolved is the structure of E1PCa 2 ⅐Mg, the genuine physiological intermediate E1PCa 2 with bound Mg 2ϩ at the catalytic site without the nucleotide. Its stable structural analog has yet to be developed. E1PCa 2 ⅐Mg is the major intermediate accumulating almost exclusively at steady state under physiological conditions. Its rate-limiting isomerization results in Ca 2ϩ deocclusion/release producing E2P⅐Mg as a key event for Ca 2ϩ transport. In E1Ca 2 ⅐CaAMP-PCP, E1Ca 2 ⅐AlF 4 Ϫ ⅐ADP, and E1PCa 2 ⅐Ca⅐AMP-PN, the N and P domains are cross-linked and strongly stabilized by the bound nucleotide and/or Ca 2ϩ at the catalytic site, thus they are crystallized (17,18,22). Kinetically, E1PCa 2 ⅐Ca formed with CaATP is markedly stabilized due to Ca 2ϩ binding at the catalytic Mg 2ϩ site, and its isomerization to E2P is strongly retarded in contrast to E1PCa 2 ⅐Mg (26,27). Thus, the bound Ca 2ϩ at the catalytic Mg 2ϩ site likely produces a significantly different structural state from that with bound Mg 2ϩ . Therefore, it is now essential to develop a genuine E1PCa 2 ⅐Mg analog without bound nucleotide and thereby gain further insight into the structural mechanism in the Ca 2ϩ transport process. It is also crucial to further clarify the structural importance of Mg 2ϩ as the physiological catalytic cation. In this study, we successfully developed the complex E1Ca 2 ⅐BeF x , most probably E1Ca 2 ⅐BeF 3 Ϫ , as the E1PCa 2 ⅐Mg analog by adding beryllium fluoride (BeF x ) to the E1Ca 2 state without any nucleotides. For its formation, Mg 2ϩ binding at the catalytic site was required and Ca 2ϩ substitution for Mg 2ϩ was absolutely unfavorable, revealing a likely structural reason for its preference as the physiological cofactor. In E1Ca 2 ⅐BeF 3 Ϫ , two Ca 2ϩ ions bound at the high affinity transport sites are occluded. It was also produced from E2⅐BeF 3 Ϫ by lumenal Ca 2ϩ binding at the lumenally oriented low affinity transport sites, mimicking E2P⅐Mg ϩ 2Ca 2ϩ 3 E1PCa 2 ⅐Mg. All properties of the newly developed E1Ca 2 ⅐BeF 3 Ϫ fulfilled the requirements as the E1PCa 2 ⅐Mg analog, and hence we were able to uncover the hitherto unknown nature of E1PCa 2 ⅐Mg as well as structural events occurring in the phosphorylation and isomerization processes. Also, we successfully found the conditions that perfectly stabilize the E1Ca 2 ⅐BeF 3 Ϫ complex.

EXPERIMENTAL PROCEDURES
Preparation of SR Vesicles and Treatment with BeF x , AlF x , and AlF x ⅐ADP-SR vesicles were prepared from rabbit skeletal muscle as described (28). The content of the phosphorylation site in the vesicles determined according to Barrabin et al. (29) was 4.49 Ϯ 0.22 nmol/mg of vesicle protein (n ϭ 5). The Ca 2ϩdependent ATPase activity determined at 25°C in a mixture containing 5 g/ml microsomal protein, 1 mM ATP, 1.7 M A23187, 7 mM MgCl 2 , 0.1 M KCl, 50 mM MOPS/Tris (pH 7.0), and 0.6 mM CaCl 2 with 0.5 mM EGTA (or 2 mM EGTA without added CaCl 2 ) was 1.87 Ϯ 0.14 mol/min/mg of vesicle protein (n ϭ 3). The Ca 2ϩ -ATPase was purified from the vesicles by deoxycholate as described (30,31). The E1Ca 2 state ATPase was incubated with fluoride compounds, 2 mM potassium fluoride and 100 M BeSO 4 or AlCl 3 , at 25°C for 30 min in the presence of 0.1 mM Ca 2ϩ , 15 mM MgCl 2 , 0.1 M KCl, 30 mM MOPS/Tris buffer (pH 7.0), unless stated otherwise. E1Ca 2 ⅐ AlF 4 Ϫ ⅐ADP was formed by including 50 M ADP in the above AlF x incubation mixture as described (32). E2⅐BeF 3 Ϫ , E2⅐AlF 4 Ϫ , and E2⅐MgF 4 2Ϫ were produced as described (23)(24)(25). Determination of EP-EP formation was performed with 3 M [␥-32 P]ATP in 100 (or 50) M Ca 2ϩ at 0°C for 3 s, and terminated by trichloroacetic acid containing carrier P i . The amount of EP formed was determined as described previously (28). The background level determined with excess 5 mM EGTA was less than 0.5% of the phosphorylation sites.
Ca 2ϩ Binding and Occlusion- 45 Ca 2ϩ binding and occlusion at the transport sites was determined at 25°C with 2 ml of the SR vesicle mixture (0.2 mg/ml protein) with a 0.45-m nitrocellulose membrane filter (Millipore) as described (31). In some cases, the vesicles on the filter were washed for 10 s by perfusion with 2 ml of a washing solution containing 5 mM EGTA. The amount of Ca 2ϩ bound at the transport sites was obtained by subtracting the nonspecific Ca 2ϩ binding level determined as described in the figure legends.
Fluorescence Measurements-The TNP-AMP fluorescence and intrinsic tryptophan fluorescence of Ca 2ϩ -ATPase (0.06 mg/ml protein) were measured on a RF-5300PC spectrofluorophotometer (Shimadzu, Kyoto, Japan) with excitation and emission wavelengths 408 and 540 nm for TNP-AMP (with band widths 5 and 10 nm), and 290 and 338.4 nm for tryptophan (with bandwidth 1.5 and 5 nm), unless otherwise described (28).
Miscellaneous-Trypsin (L-1-tosylamido-2-phenylethyl chloromethyl ketone-treated) and proteinase K were obtained from Sigma. TNP-AMP was synthesized according to Hiratsuka (34). Protein concentrations were determined by the method of Lowry et al. (35) with bovine serum albumin as a standard. Free Ca 2ϩ concentrations were calculated by the Calcon program. Data were analyzed by nonlinear regression using the program Origin (Microcal Software, Inc., Northampton,

Formation of E1PCa 2 Analogs by
Fluoride Compounds-The Ca 2ϩ -ATPase of SR vesicles in 100 M Ca 2ϩ and 15 mM Mg 2ϩ was incubated with beryllium fluoride (BeF x ) or aluminum fluoride (AlF x ) for 30 min at 25°C. The ability of EP formation from ATP was almost completely lost (actually within 1 min) (Fig. 3), showing the stable complex formation with BeF x and AlF x . No inhibition occurred with Mg 2ϩ and fluoride without beryllium and aluminum (E1Ca 2 ϩ(MgF x )). Thus, MgF x (MgF 4 2Ϫ ) was not able to produce a complex with E1Ca 2 , in contrast to E2⅐MgF 4 2Ϫ formation from E2, the E2⅐P i product analog in E2P hydrolysis (19,25). This finding agrees with the in-line phosphorylation of E1Ca 2 to E1PCa 2 (37), in which there is no state with non-covalently bound P i .
The binding of Mg 2ϩ at the catalytic site as a physiological cation is nevertheless required for EP formation. Actually in Fig. 3, Mg 2ϩ was required for complex formation with BeF x in 100 M Ca 2ϩ . The apparent Mg 2ϩ affinity (K 0.5 of 5 mM, supplemental Fig. S1) was consistent with that of the catalytic site in phosphorylation from ATP or P i (e.g. Refs. 11,12,[38][39][40][41]. The BeF x -induced inhibition also occurred with Mn 2ϩ with apparent affinity (K 0.5 of 0.6 mM) significantly higher than that of Mg 2ϩ , as found in E1PCa 2 formation and ATP hydrolysis with Mn 2ϩ (40,42).
When Ca 2ϩ over millimolar amounts was added in place of Mg 2ϩ , the EP formation was not inhibited by BeF x (BeF x (5 mM Ca)). Therefore, Ca 2ϩ substitution probably at the catalytic Mg 2ϩ site abolished complex formation with BeF x . Although CaATP as a substrate and Ca 2ϩ bound at the catalytic Mg 2ϩ site are able to function for E1PCa 2 formation (26,27), Ca 2ϩ bound at the catalytic site likely produces different structure from the Mg 2ϩ bound structure (as in fact found, see below). The complex formation of E1Ca 2 by AlF x in 100 M Ca 2ϩ also required Mg 2ϩ or Mn 2ϩ with somewhat higher apparent affinities than those for BeF x -induced complex formation, and was abolished by Ca 2ϩ binding at the catalytic site.
Ca 2ϩ Binding and Occlusion at Transport Sites-In Fig. 4, Ca 2ϩ binding and occlusion in formation of the Ca 2ϩ -ATPase complexes with fluoride compounds were determined in 15 mM Mg 2ϩ with and without a 10-s EGTA filter washing. In all the cases without washing, Ca 2ϩ was bound to the Ca 2ϩ -ATPase with high affinities; K 0.5 at sub-micromolar to millimolar ranges, Hill coefficient ϳ2, and maximum levels of 9 -10 nmol/mg of protein, i.e. the stoichiometry of two Ca 2ϩ per phosphorylation site (inset at 50 M Ca 2ϩ ). Therefore, E1Ca 2 ⅐BeF x and E1Ca 2 ⅐AlF 4 Ϫ ⅐ADP/E1Ca 2 ⅐AlF x were produced by cooperative binding of two Ca 2ϩ ions at high affinity transport sites. This finding agrees with the property of the sites for Ca 2ϩ binding and the resulting enzyme activation for phospho-rylation by ATP as nicely demonstrated for the first time by Inesi et al. (14). Upon the EGTA washing of E1Ca 2 that is complexed with BeF x , the two bound Ca 2ϩ were not removed, and therefore occluded in the complex as "E1Ca 2 ⅐BeF x ." The two Ca 2ϩ are occluded also in E1Ca 2 ⅐AlF 4 Ϫ ⅐ADP and less strongly in Note that the Ca 2ϩ affinity became 2-3-fold lower for BeF x and AlF x . This may be because the Ca 2ϩ -free E2 state produces E2⅐BeF 3 Ϫ and E2⅐AlF 4 Ϫ (25), and therefore competes with Ca 2ϩ binding for formation of E1Ca 2 ⅐BeF x and E1Ca 2 ⅐AlF x . On the other hand, the observed ϳ3-fold Ca 2ϩ -affinity increase in formation of E1Ca 2 ⅐AlF 4 Ϫ ⅐ADP is probably brought about by the fact that ADP together with AlF x strongly stabilizes the crosslinked N-P domains (17,18), which is unfavorable for formation of the Ca 2ϩ -free E2 and E2⅐AlF 4 Ϫ , because for these structures the A domain should rotate into the opened space between the N and P domains and associate with them (19,23,24).
Cytoplasmic Structure in E1Ca 2 ⅐BeF x Is Intermediate between Those in E1Ca 2 ⅐AlF 4 Ϫ ⅐ADP and E2⅐BeF 3 Ϫ -Proteolytic analysis was made to reveal the organization state of the cytoplasmic domains in the newly developed E1PCa 2 ⅐Mg analog E1Ca 2 ⅐BeF x , and to compare with E1Ca 2 ⅐AlF 4 Ϫ ⅐ADP/ E1Ca 2 ⅐AlF x and E2⅐BeF 3 Ϫ (E2P⅐Mg) (see the typical cleavage in supplemental Fig. S2). The initial rate of the "A1" appearance upon cleavage at the T2 site (Arg 198 on the Val 200 loop of the A domain) in E1Ca 2 ⅐BeF x was substantially slower than the rapid cleavage of E1Ca 2 ⅐AlF 4 Ϫ ⅐ADP and E1Ca 2 ⅐AlF x as well as E1Ca 2 ( Table 1). The slowed T2 cleavage was also observed when E1Ca 2 ⅐BeF x was formed with 3 mM Mn 2ϩ in place of 15 mM Mg 2ϩ (data not shown). Also important was the slow but definitely occurring T2 cleavage in E1Ca 2 ⅐BeF x , in sharp contrast to its complete resistance in E2⅐BeF 3 Ϫ . Therefore A-P domain organization at the Val 200 loop in E1Ca 2 ⅐BeF x is intermediate between those in E1Ca 2 ⅐AlF 4 Ϫ ⅐ADP/E1Ca 2 ⅐AlF x and E2⅐BeF 3 Ϫ . All complexes were almost completely resistant to proteinase K at the major site of Thr 242 on the A/M3-linker that produces the "p83" fragment. Therefore, in E1Ca 2 ⅐BeF x , the A domain is rotated perpendicular to the membrane plane from its position in E1Ca 2 thereby causing the A/M3-linker strain, as in E1Ca 2 ⅐CaAMP-PCP, E1Ca 2 ⅐AlF 4 Ϫ ⅐ADP (18,19,24), and E1Ca 2 ⅐AlF x . These analyses revealed that in the change E1Ca 2 ⅐AlF 4 Ϫ ⅐ ADP 3 E1Ca 2 ⅐BeF x (i.e. upon the ADP release from the transition state), the A domain moves, i.e. probably rotates to some extent parallel to the membrane plane likely due to the A/M3linker strain, and thereby Arg 198 on the Val 200 loop comes close to the P domain. In the subsequent change, E1Ca 2 ⅐BeF x 3 E2⅐BeF 3 Ϫ , the A domain rotates further (by the A/M3-linker strain as predicted to be motive force (18,19,43,44)) and pro-  duces its tight association with the P domain at the Val 200 loop, mimicking E1PCa 2 ⅐Mg 3 E2P⅐Mg ϩ 2Ca 2ϩ .
Ca 2ϩ Ligation at the Catalytic Mg 2ϩ Site-The proteolysis further revealed that E1Ca 2 ⅐BeF x was not produced from E1Ca 2 in 5 mM Ca 2ϩ without Mg 2ϩ (E1Ca 2 ϩ 5 mM Ca 2ϩ ϩ BeF x in Table 1), and that E1Ca 2 ⅐BeF x produced in 15 mM Mg 2ϩ and 50 -100 M Ca 2ϩ was decomposed to the E1Ca 2 by 5 mM Ca 2ϩ (E1Ca 2 ⅐BeF x ϩ 5 mM Ca 2ϩ ), as shown by the rapid cleavage rates at the T2 and proteinase K sites. In E1PCa 2 ⅐Ca and E1Ca 2 ⅐ CaAMP-PCP formed in 5 mM Ca 2ϩ without Mg 2ϩ (Table 1), 3 the T2 site was also rapidly cleaved, in contrast to its substantially slowed cleavage in E1Ca 2 ⅐BeF x formed with Mg 2ϩ . Thus, for organization of the cytoplasmic domains at the T2 site (Arg 198 ), E1Ca 2 ⅐CaAMP-PCP and E1PCa 2 ⅐Ca are very similar to E1Ca 2 ⅐AlF 4 Ϫ ⅐ADP, but differ from E1Ca 2 ⅐BeF x . The close similarity between E1Ca 2 ⅐CaAMP-PCP and E1Ca 2 ⅐AlF 4 Ϫ ⅐ADP is in agreement with their nearly same atomic structures and previous observations (17,18,45). Also notably, structure E1PCa 2 ⅐Ca⅐AMP-PN formed by CaAMP-PNP in 10 mM Ca 2ϩ without Mg 2ϩ (22) is almost identical with those of E1Ca 2 ⅐CaAMP-PCP and E1Ca 2 ⅐AlF 4 Ϫ ⅐ADP (see also Table 1).
In E1Ca 2 ⅐CaAMP-PCP and E1PCa 2 ⅐Ca⅐AMP-PN, the N-P domain cross-linked state is stabilized by Ca 2ϩ bound at catalytic Mg 2ϩ site I (Asp 351 /Thr 353 /Asp 703 and the phosphate) and by the nucleotide to be nearly identical to the state stabilized with AlF 4 Ϫ plus ADP in E1Ca 2 ⅐AlF 4 Ϫ ⅐ADP (17,18,22). The results on E1PCa 2 ⅐Ca further indicated that such an N-P domain closed state is stabilized solely by site I Ca 2ϩ ligation without the nucleotide. The stabilization of this state in E1PCa 2 ⅐Ca is consistent with its markedly retarded isomerization to E2P (27), because isomerization requires the A domain rotation into the space between the N and P domains. In E1Ca 2 ⅐BeF x formed with Mg 2ϩ at the catalytic site (site I), such a Ca 2ϩ ligation effect is obviously not present. Therefore, the N and P domains are probably more easily separated from each other, and the A domain can rotate into the space between the N and P domains to some extent thus resulting in partial T2 resistance (but not yet as completely as in E2⅐BeF 3 Ϫ ). As the cause of the Ca 2ϩ -induced E1Ca 2 ⅐BeF x to E1Ca 2 decomposition, Ca 2ϩ replacement of Mg 2ϩ at site I altered the domain organization state and made the BeF x ligation unfavorable (see "Discussion").
TNP-AMP Superfluorescence-TNP-AMP binds to the nucleotide binding site with an extremely high affinity (46 -48), and in the E2P ground state and its analog E2⅐BeF 3 Ϫ , the bound TNP-AMP develops its extremely high fluorescence "superfluorescence" (25), which reflects a strongly hydrophobic atmosphere around Asp 351 (49,50). On the other hand, it has been controversial whether E1PCa 2 develops TNP-AMP superfluorescence, mostly because its tight binding to the nucleotide binding site prevents phosphorylation to form E1PCa 2 ⅐Mg, so the TNP-AMP⅐E1PCa 2 ⅐Mg complex is not formed in significant amounts. Nakamoto and Inesi (47), nevertheless, predicted the development of superfluorescence in E1PCa 2 .
Here, with E1Ca 2 ⅐BeF x as the E1PCa 2 ⅐Mg analog, we examined the superfluorescence without ATP. In Fig. 5, A and B, we first formed E1Ca 2 ⅐BeF x in 15 mM Mg 2ϩ and 50 M Ca 2ϩ , then TNP-AMP was added to give a saturating level of 4 M. E1Ca 2 ⅐BeF x rapidly developed superfluorescence, and then the fluorescence decreased slowly (much more slowly and extensively at 4°C). 4 The proteolysis after the loss of superfluores-cence revealed that E1Ca 2 ⅐BeF x was decomposed to E1Ca 2 (data not shown). Thus, E1Ca 2 ⅐BeF x develops superfluorescence, and the TNP-AMP binding per se causes its decomposition to E1Ca 2 , in sharp contrast to the completely stable E2⅐BeF 3 Ϫ even after TNP-AMP binding. The maximum superfluorescence level of E1Ca 2 ⅐BeF x was slightly lower than that of E2⅐BeF 3 Ϫ (Fig. 5, A-C), which is the same as that of E2P⅐Mg formed from P i (25). The results clearly revealed that the atmosphere around Asp 351 in E1Ca 2 ⅐BeF x is strongly hydrophobic, similar to E2⅐BeF 3 Ϫ , although the cytoplasmic domain organization in E1Ca 2 ⅐BeF x distinctly differs from and did not yet reach the most compactly organized state in E2⅐BeF 3 Ϫ . E1Ca 2 ⅐AlF x as well as E2 and E1Ca 2 did not develop superfluorescence despite high affinity TNP-AMP binding. Therefore the catalytic site in E1Ca 2 ⅐AlF x is hydrophilic and differs critically from the strongly hydrophobic site in E1Ca 2 ⅐BeF x . Note also that E2⅐AlF 4 Ϫ (E2-P ‡ ) and E2⅐MgF 4 2Ϫ (E2⅐P i ) do not develop TNP-AMP superfluorescence (25). The superfluorescence therefore develops solely with Ca 2ϩ -ATPase complexed with BeF x ; E1Ca 2 ⅐BeF x and E2⅐BeF 3 Ϫ . The superfluorescence development of E1Ca 2 ⅐BeF x in the presence of 15 mM Mg 2ϩ and 50 M Ca 2ϩ was rapidly diminished with increasing Ca 2ϩ over millimolar concentrations of 4 The fluorescence level after the decrease of the transient superfluorescence of E1Ca 2 ⅐BeF x was somewhat higher than the non-superfluorescent low level of TNP-AMP bound to E1Ca 2 especially at 25°C. We obtained the results indicating that a small fraction of E2⅐BeF 3 Ϫ was produced even in the presence of 50 M Ca 2ϩ (more at 25 than at 4°C) after the TNP-AMP-induced E1Ca 2 ⅐BeF x decomposition to E1Ca 2 , and therefore remained somewhat superfluorescence (data not shown).

TABLE 1 Summary of fluorescence changes and proteolysis rates
Maximal TNP-AMP fluorescence intensity at saturating 4 M TNP-AMP is given as % value of that of E2⅐BeF 3 Ϫ without A23187. Tryptophan fluorescence change upon complex formation with the ligand from E1Ca 2 (or from other state when indicated in parentheses) is shown as % value of the intensity of E1Ca 2 (see also supplemental Fig.  S3E). The cleavage rate at the T2 site (Arg 198 ) with trypsin and the digestion rate of the 110-kDa ATPase chain with proteinase K were obtained by the detailed time course analysis in the initial 1 (T2) and 30 min (proteinase K), and shown as % values of those determined with E1Ca 2 in 0.1 mM Ca 2ϩ . Some of these experiments were done at 0.05 mM Ca 2ϩ instead of 0.1 mM, but the results were virtually the same and therefore are represented with 0.1 mM Ca 2ϩ for simplicity. It should also be mentioned that the digestion rates in E1Ca 2 were not altered by 5 or 0.7 mM Ca 2ϩ or by A23187 (being 97-101% of the rates of E1Ca 2 in 0.1 mM Ca 2ϩ without A23187), and that the ligand-free E2 state was also rapidly digested by trypsin and proteinase K, and TG binding to E2 and A23187 did not alter essentially the rapid cleavage rates (Refs. [23][24][25]. The other Ca 2ϩ -ATPase complexes were produced under the same buffer conditions as those for E1Ca 2 ⅐BeF x formation, otherwise as follows and noted below: E1Ca 2 ⅐MgAMP-PCP by 5 mM MgAMP-PCP; E1Ca 2 ⅐CaAMP-PCP by 5 mM CaAMP-PCP; E1PCa 2 ⅐Ca⅐AMP-PN by 5 mM CaAMP-PNP; E1PCa 2 ⅐Ca by 5 mM CaATP. E1Ca 2 ⅐BeF x ϩ 5 mM Ca 2ϩ , the E1Ca 2 ⅐BeF x complex was formed in 15 mM Mg 2ϩ and 100 M Ca 2ϩ and then incubated with the subsequently added 5 mM Ca 2ϩ for 3 h; E1Ca 2 ϩ 5 mM Ca 2ϩ ϩ BeF x , the E1Ca 2 state ATPase was incubated with BeF x for 10 min in the presence of 5 mM Ca 2ϩ without Mg 2ϩ .  38 16 Structural Analog of E1PCa 2 ⅐Mg Intermediate of Ca 2؉ -ATPase AUGUST 21, 2009 • VOLUME 284 • NUMBER 34

JOURNAL OF BIOLOGICAL CHEMISTRY 22727
Ca 2ϩ (see Fig. 8 and Table 1). Also, inclusion of 5 mM Ca 2ϩ without Mg 2ϩ in the E1Ca 2 ⅐BeF x formation mixture abolished the superfluorescence ( Table 1). The results agree with the above findings that E1Ca 2 ⅐BeF x is not produced from and decomposed to E1Ca 2 by Ca 2ϩ ligation at the catalytic Mg 2ϩ site (site I).
Transmembrane Domain Structure-The 12 tryptophan residues among 13 in the Ca 2ϩ -ATPase are located at the transmembrane region. The tryptophan fluorescence changes in fact reflect the transmembrane domain structural changes, i.e. rearrangements of the transmembrane helices upon Ca 2ϩ binding to the high affinity transport sites and during the transport cycle (28,51,52) as found originally by Dupont and Leigh (53). As summarized in Table 1 with typical fluorescence traces in supplemental Fig. S3, the fluorescence changes were determined at 4°C upon formation of the E1PCa 2 analogs by the addition of fluoride compounds to E1Ca 2 in 15 mM Mg 2ϩ and 100 M Ca 2ϩ . E1Ca 2 ⅐BeF x formation decreased fluorescence by 1.3% very similar to the decrease in E1PCa 2 ⅐Mg formation from E1Ca 2 by MgATP, i.e. in E1Ca 2 ⅐MgATP 3 E1PCa 2 ⅐Mg (52). In contrast, E1Ca 2 ⅐ AlF x formation did not cause any change. The E1Ca 2 ⅐AlF 4 Ϫ ⅐ADP formation increased the fluorescence by 0.8%. (F Ϫ alone and ADP alone did not cause any change, except the dilution (F Ϫ ) and absorption of excitation light (ADP).) Thus the transmembrane structure of E1Ca 2 ⅐BeF x mimics that of E1PCa 2 ⅐ Mg, but those of E1Ca 2 ⅐AlF 4 Ϫ ⅐ADP and E1Ca 2 ⅐AlF x differ substantially although the Ca 2ϩ ions are occluded at the transport sites (or less strongly in E1Ca 2 ⅐AlF x , Fig. 4). This observation is consistent with the finding in proteolysis and TNP-AMP superfluorescence that organization of the cytoplasmic domains and structure at the catalytic site in E1Ca 2 ⅐BeF x substantially differ from those in E1Ca 2 ⅐AlF 4 Ϫ ⅐ADP and E1Ca 2 ⅐AlF x . It is concluded that the transmembrane structure with the occluded Ca 2ϩ adopts not simply one state, but changes with the change in the cytoplasmic region during phosphoryl transfer and ADP release (see the diagram of tryptophan fluorescence change in supplemental Fig. S3E (with Ref. 54)).
Upon the exclusive accumulation of E1PCa 2 ⅐Ca by CaATP without Mg 2ϩ , tryptophan fluorescence decreased by 0.9%, being slightly less than that by formation of E1PCa 2 ⅐Mg and E1Ca 2 ⅐BeF x ( Table 1). Thus in the overall structure, Ϫ produced with BeF x in the absence of Ca 2ϩ , the E1Ca 2 and E2 states without the fluoride compounds, and without SR vesicles (no SRV) were also followed. C, the TNP-AMP fluorescence intensities were measured at various concentrations of TNP-AMP at 25°C, otherwise as described for A. For E1Ca 2 ⅐BeF x , the maximum level of transient superfluorescence was determined by extrapolating its decrease to the time of TNP-AMP addition. In the lower panel in C, the low fluorescence was replotted on the expanded scale. The maximum fluorescence intensities at saturating 4 M TNP-AMP were obtained by subtracting the background level without TNP-AMP and the level of 4 M TNP-AMP without the SR vesicles, and given as the relative values in Table 1.
E1PCa 2 ⅐Ca may be between E1Ca 2 ⅐CaAMP-PCP and E1PCa 2 ⅐Mg (E1Ca 2 ⅐BeF x ), and closer to the latter state. Although Ca 2ϩ ligation at catalytic Mg 2ϩ site I in E1PCa 2 ⅐Ca favors the N-P domain closed state, similar to E1Ca 2 ⅐CaAMP-PCP, the absence of the N-P domain cross-linking nucleotide in E1PCa 2 ⅐Ca likely altered the overall structure slightly.
Upon formation of E2⅐BeF 3 Ϫ from E2 by BeF x and Mg 2ϩ without Ca 2ϩ , the fluorescence increased by 0.7%, mimicking the change upon E2P⅐Mg formation from E2 with P i and Mg 2ϩ , and reflecting the opening of the lumenal gate from the closed state (25). As a consequence, the fluorescence of E1Ca 2 ⅐BeF x was definitely higher by ϳ1.3% than that of E2⅐BeF 3 Ϫ , showing their distinct difference in the transmembrane structure. In agreement, the previous kinetic analysis have shown (28) that tryptophan fluorescence decreases by ϳ1% in the isomerization/ Ca 2ϩ release, E1PCa 2 ⅐Mg 3 E2P⅐Mg ϩ 2Ca 2ϩ , reflecting the transmembrane structural change from the Ca 2ϩ -occluded state to the Ca 2ϩ -released and lumenally opened state.
Upon the addition of thapsigargin (TG) to E1Ca 2 ⅐BeF x and E2⅐BeF 3 Ϫ , tryptophan fluorescence decreased rapidly by 5.4 and 4.6%, respectively, and reached the level of E2⅐BeF 3 Ϫ with bound TG (E2⅐BeF 3 Ϫ (TG), see Table 1). TNP-AMP superfluorescence (supplemental Fig. S5, A and B) and proteolysis (Table 1) also demonstrated that E1Ca 2 ⅐BeF x was converted by TG to E2⅐BeF 3 Ϫ (TG). Importantly, as described under supplemental Fig. S4, two Ca 2ϩ occluded in E1Ca 2 ⅐BeF x are most likely released into the lumen by the TG-induced structural perturbation and trapped in the lumen by the bound TG, as TG fixes the lumenal gate in the closed state and suppresses Ca 2ϩ leakage (16,55).
E1Ca 3 ⅐BeF x Is ADP-sensitive-In Fig. 6, two 45 Ca 2ϩ occluded in E1Ca 2 ⅐BeF x were rapidly removed by washing with 1 mM ADP, whereas the occluded 45 Ca 2ϩ remained completely without ADP. Thus ADP caused the loss of Ca 2ϩ occlusion. In agreement, ADP binding to E1Ca 2 ⅐BeF x increased tryptophan fluorescence to the E1Ca 2 level, and resulted in the tryptic T2 site cleavage as E1Ca 2 with bound ADP (data not shown). By contrast, ADP binding to E2⅐BeF 3 Ϫ did not alter its structure (data not shown). The ADP-induced decomposition of E1Ca 2 ⅐BeF x to E1Ca 2 was also demonstrated with the ADPinduced loss of TNP-AMP superfluorescence, in contrast to normal superfluorescence development in E2⅐BeF 3 Ϫ after ADP incubation (data not shown). Thus E1Ca 2 ⅐BeF x is ADP-sensitive as E1PCa 2 ⅐Mg, and E2⅐BeF 3 Ϫ is ADP-insensitive as E2P⅐Mg. Conversion of E1Ca 2 ⅐BeF x to E2⅐BeF 3 Ϫ at 50 M Ca 2ϩ -In Fig.  7, E1Ca 2 ⅐BeF x was first formed in SR vesicles with BeF x at 25°C in 50 M Ca 2ϩ and 15 mM Mg 2ϩ , then further incubated at 25 and 4°C in the presence of these ligands. The amount of bound and occluded Ca 2ϩ was lost slowly (t1 ⁄ 2 ϭ ϳ2 h at 25°C and ϳ7 h at 4°C). TNP-AMP superfluorescence (Fig. 8) and tryptic and proteinase K proteolyses (data not shown) revealed that E1Ca 2 ⅐BeF x turned to E2⅐BeF 3 Ϫ with Ca 2ϩ loss. Thus E1Ca 2 ⅐ BeF x proceeded its spontaneous slow conversion to E2⅐BeF 3 Ϫ , as the autoisomerization of E1PCa 2 ⅐Mg to E2P⅐Mg. The Ca 2ϩ ions released into the lumen may leak out during such long periods. In E1Ca 2 ⅐AlF x and E1Ca 2 ⅐AlF 4 Ϫ ⅐ADP, the amount of bound (occluded) Ca 2ϩ was not decreased during the above 10-h incubation at 25°C (data not shown). The proteolysis showed that these complexes were not converted to the Ca 2ϩ -released forms, E2⅐AlF 4 Ϫ (E2⅐AlF x ) with and without ADP (data not shown). The results indicate that the product state E1PCa 2 ⅐Mg in the phosphoryl transfer acquires the structure ready for autoisomerization to E2P⅐Mg releasing Ca 2ϩ , whereas the transition state structure is not yet fully prepared for autoisomerization to the Ca 2ϩ -released E2P form. Interestingly, as described in supplemental Figs. S4 and S5 (with Refs. 56 and 57), the conversion E1Ca 2 ⅐BeF x 3 E2⅐BeF 3 Ϫ was markedly accelerated by the transmembrane structural perturbation with hydrophobic reagents such as A23187, lasalocid, and C 12 E 8 , as  well as TG. In contrast, E1Ca 2 ⅐AlF x and E1Ca 2 ⅐AlF 4 Ϫ ⅐ADP were resistant against these reagents. E1Ca 2 ⅐BeF x Is Perfectly Stabilized at 0.7 mM Ca 2ϩ -As found here, Ca 2ϩ binding at high affinity transport sites in E1Ca 2 is obligatorily required for E1Ca 2 ⅐BeF x formation, whereas millimolar high Ca 2ϩ (Ca 2ϩ ligation at the catalytic Mg 2ϩ site I) decomposes this complex to E1Ca 2 . Furthermore, E1Ca 2 ⅐BeF x at 50 M Ca 2ϩ is spontaneously and slowly converted to E2⅐BeF x releasing Ca 2ϩ , and the conversion is markedly accelerated by transmembrane perturbation with hydrophobic reagents such as C 12 E 8 and A23187 (see supplemental materials). The results showed that E1Ca 2 ⅐BeF x as the E1PCa 2 ⅐Mg analog possesses the structure prepared for its isomerization to E2⅐BeF 3 Ϫ with Ca 2ϩ release as E1PCa 2 ⅐Mg 3 E2P⅐Mg ϩ 2Ca 2ϩ . On the other hand, it is essential for crystallographic studies to find conditions to perfectly stabilize the E1Ca 2 ⅐BeF x complex. In Fig. 8A, E1Ca 2 ⅐BeF x was first formed in 0.1 mM Ca 2ϩ and 15 mM Mg 2ϩ , then further incubated with various concentrations of Ca 2ϩ with and without A23187. The structural state was monitored by TNP-AMP superfluorescence (Fig. 8), proteolysis, and tryptophan fluorescence (see Table 1 for representative data). Then we successfully found that Ca 2ϩ at a very narrow concentration range, 0.7 mM, perfectly stabilizes E1Ca 2 ⅐BeF x and maintains this complex for at least 12 days at 25°C (and 4°C) even in the presence of A23187. The 45 Ca 2ϩ binding measurements on E1Ca 2 ⅐BeF x in 0.7 mM 45 Ca 2ϩ demonstrated that two Ca 2ϩ ions are bound and occluded in this complex (Fig. 9A).
The perfectly stable E1Ca 2 ⅐BeF x was produced from E1Ca 2 even in the presence of A23187 if 0.7 mM Ca 2ϩ was included before BeF x addition (Fig. 8, B and F, and Table 1). Also, E1Ca 2 ⅐BeF x was successfully produced with the Ca 2ϩ -ATPase purified from SR vesicles by delipidation with deoxycholate (30); in this case again, by including 0.7 mM Ca 2ϩ before BeF x addition. E1Ca 2 ⅐BeF x thus produced with the purified and delipidated Ca 2ϩ -ATPase was perfectly stable at least for 12 days at 4 and 25°C in 0.7 mM Ca 2ϩ (Fig. 8, B and G, at 25°C). E1Ca 2 ⅐BeF 3 Ϫ Is Produced from E2⅐BeF 3 Ϫ by Lumenal Ca 2ϩ Binding-We successfully found also that E1Ca 2 ⅐BeF x (E1Ca 2 ⅐BeF 3 Ϫ ) can be produced from E2⅐BeF 3 Ϫ by lumenal Ca 2ϩ binding, as mimicking the lumenal Ca 2ϩ -induced reverse transition, E2P⅐Mg ϩ 2Ca 2ϩ 3 E1PCa 2 ⅐Mg. In Fig. 10, we added Ϫ formed without Ca 2ϩ , E1Ca 2 in 0.1-10 mM Ca 2ϩ , and E2 without Ca 2ϩ were also incubated. B, in the presence of 0.7 mM Ca 2ϩ , E1Ca 2 in SR vesicles was first incubated with and without 1.2 M A23187 (A23), then BeF x was added (E1Ca 2 ⅐BeF x ). E1Ca 2 ⅐BeF x (DOC-E1Ca 2 ⅐BeF x ) was also produced from E1Ca 2 in 0.7 mM Ca 2ϩ of the Ca 2ϩ -ATPase purified and delipidated from SR vesicles by deoxycholate (DOC) treatment (30). The incubation was continued for 12 days, otherwise as indicated in A. E2⅐BeF 3 Ϫ (E2⅐BeF 3 Ϫ , DOC-E2⅐BeF 3 Ϫ ) without Ca 2ϩ and E1Ca 2 in 0.7 mM Ca 2ϩ (E1Ca 2 , DOC-E1Ca 2 ) and E2 without Ca 2ϩ (E2, DOC-E2) were also incubated. C-G, the fluorescence traces upon TNP-AMP addition were shown for the representative samples with incubation periods and Ca 2ϩ concentration (mM). Note that at 0.7 mM Ca 2ϩ , both with and without A23187, the development of the E1Ca 2 ⅐BeF x characteristic transient superfluorescence remained perfectly the same for 12 days. By contrast, the transient superfluorescence was converted to the stable and higher superfluorescence characteristic of E2⅐BeF 3 Ϫ at 0.1 and 0.4 mM Ca 2ϩ , and it was markedly reduced by 3 and 10 mM Ca 2ϩ due to decomposition to E1Ca 2 .

Structural Analog of E1PCa 2 ⅐Mg Intermediate of Ca 2؉ -ATPase
various concentrations of Ca 2ϩ to E2⅐BeF 3 Ϫ formed in SR vesicles in 15 mM Mg 2ϩ without Ca 2ϩ in the presence and absence of A23187, then at 10 s after Ca 2ϩ addition the structural state was examined by TNP-AMP superfluorescence. With increasing Ca 2ϩ to 0.7 mM in the presence of A23187, the stable superfluorescence of E2⅐BeF 3 Ϫ was converted to the transient and slightly lower superfluorescence characteristic of E1Ca 2 ⅐BeF x with K 0.5 of 0.4 mM Ca 2ϩ and a Hill coefficient of 4 (Fig. 10, A  and D). A further Ca 2ϩ increase in the millimolar range caused the marked loss of superfluorescence with K 0.5 of 1.7 mM and a Hill coefficient of 1 (Fig. 10, B and D). The proteolysis also clearly showed that E2⅐BeF 3 Ϫ was converted to E1Ca 2 ⅐BeF x by 0.7 mM Ca 2ϩ in A23187 (Table 1), and this complex was further decomposed to E1Ca 2 by 10 mM Ca 2ϩ (data not shown). In Fig.  9B, two 45 Ca 2ϩ were shown to be bound producing E1Ca 2 ⅐BeF x , when 0.7 mM 45 Ca 2ϩ was added to E2⅐BeF 3 Ϫ in the presence of A23187. In contrast, in the absence of A23187, E2⅐BeF 3 Ϫ was neither converted to E1Ca 2 ⅐BeF x nor decomposed to E1Ca 2 even at 10 mM Ca 2ϩ (Figs. 9B and 10, C and E, and Table 1 (proteolysis at 0.7 mM Ca 2ϩ )).
The results demonstrated that E1Ca 2 ⅐BeF x , most probably E1Ca 2 ⅐BeF 3 Ϫ , was produced from E2⅐BeF 3 Ϫ by the lumenal Ca 2ϩ binding at the lumenally oriented low affinity transport sites, and further that Ca 2ϩ substitution of Mg 2ϩ at the catalytic site in E1Ca 2 ⅐BeF 3 Ϫ produced from E2⅐BeF 3 Ϫ caused its decomposition to E1Ca 2 , therefore as the change E2⅐BeF 3 Ϫ ϩ 2Ca 2ϩ 3 E1Ca 2 ⅐BeF 3 Ϫ 3 E1Ca 2 . Note that Mg 2ϩ bound at the catalytic site in E2P⅐Mg is occluded, whereas it is not and therefore is exchangeable in E1PCa 2 ⅐Mg (42). Thus, these two distinctly different states of the ligated Mg 2ϩ at the catalytic site (site I) in E2P⅐Mg and E1PCa 2 ⅐Mg are obviously mimicked here by the respective analogs E2⅐BeF 3 Ϫ and E1Ca 2 ⅐BeF 3 Ϫ . The perfect stabilization of E1Ca 2 ⅐BeF 3 Ϫ achieved by 0.7 mM Ca 2ϩ (Fig. 10) obviously involves lumenal Ca 2ϩ binding and prevention of the Ca 2ϩ release into the lumen. The stabilization by 0.7 mM Ca 2ϩ in the absence of A23187 is probably due to Ca 2ϩ moved passively into the vesicle lumen during the long incubation periods. All these findings show that the forward and reverse transition, E1PCa 2 ⅐Mg 7 E2P⅐Mg ϩ 2Ca 2ϩ , is mimicked by the forward and reverse conversion, E1Ca 2 ⅐BeF 3 Ϫ 7 E2⅐BeF 3 Ϫ ϩ 2Ca 2ϩ . It is of interest to note the Hill coefficient of 4 in the lumenal Ca 2ϩ -induced reverse conversion, E2⅐BeF 3 Ϫ ϩ 2Ca 2ϩ 3 E1Ca 2 ⅐BeF 3 Ϫ at 0.1-0.7 mM Ca 2ϩ in Fig. 10A. This might be indicative of the existence of lumenal Ca 2ϩ access sites in addition to transport sites and their possible cooperative involvement in lumenal Ca 2ϩ access to the transport sites. In fact, two such sites besides the two transport sites have been suggested by the kinetics and protein-chemical study on the lumenal loops (58,59).

Formation of E1Ca 2 ⅐BeF 3
Ϫ -As a structural analog of the physiological intermediate E1PCa 2 ⅐Mg, the E1Ca 2 ⅐BeF x complex was successfully produced by BeF x binding to the E1Ca 2 state Ca 2ϩ -ATPase and from E2⅐BeF 3 Ϫ by lumenal Ca 2ϩ binding to the lumenally oriented low affinity transport sites. All the revealed properties of E1Ca 2 ⅐BeF x met the requirements for the E1PCa 2 ⅐Mg analog; i.e. two Ca 2ϩ occluded at the transport sites, Mg 2ϩ bound (but not occluded) at the catalytic site, the ADP-released but still ADP-sensitive state, and its isomerization to the ADP-insensitive Ca 2ϩ -released state E2P⅐Mg (E2⅐BeF 3 Ϫ ) and reversal by lumenal Ca 2ϩ binding, E1PCa 2 ⅐Mg 7 E2P⅐Mg ϩ 2Ca 2ϩ .
Furthermore, the coordination chemistry of beryllium fluoride, actually BeF 3 Ϫ , fulfills the requirement of E1Ca 2 ⅐BeF x as the E1PCa 2 ⅐Mg analog. In chemistry, beryllium fluoride compounds are known to adopt tetrahedral geometry with the Be-F 1.55-Å bond length, thereby making them strictly isomorphous to the tetrahedral phosphate group (60). Moreover, because of the high charge density due to the small size, beryllium is able to coordinate the aspartate-oxygen in addition to F Ϫ . The -O-BeF 3 Ϫ thus produced with Asp 351 -oxygen in fact possesses the tetrahedral geometry superimposable with the covalently bound phosphate at the aspartate, as actually seen in E2⅐BeF 3 Ϫ , the E2P⅐Mg ground-state analog (21,22,25). MgF 4 2Ϫ also possesses the tetrahedral geometry, but magnesium is not able to be coordinated directly with the Asp 351 -oxygen, as seen in E2⅐MgF 4 2Ϫ , the E2⅐P i analog. AlF 4 Ϫ in E1Ca 2 ⅐AlF 4 Ϫ ⅐ADP and E2⅐AlF 4 Ϫ (17, 18, 20) (or AlF 3 in some cases in the haloacid dehalogenase superfamily (61)) possesses planar geometry, in which Asp 351 -oxygen and ADP ␤-phosphate or the hydrolytic water (E2⅐AlF 4 Ϫ ) coordinate the aluminum at apical positions producing the bipyramidal structure superimposable to the penta-coordinated phosphorus in the transition state of in-line phosphoryl transfer E1PCa 2 ⅐ADP⅐Mg ‡ and acylphosphate hydrolysis E2-P⅐Mg ‡ . Thus all chemical properties of the P i ana-  Fig. 4. The amount of bound Ca 2ϩ was obtained with subtraction of the background level (10.1 Ϯ 0.5 nmol/mg (n ϭ 6)) determined by EGTA washing the vesicles incubated without BeF x and A23187. The occluded Ca 2ϩ was determined by EGTA washing in the absence of A23187 and by subtraction of the background level. (Here the determination of occluded Ca 2ϩ in A23187 by EGTA washing was not feasible, because in the absence of (or even in 0.1 mM) Ca 2ϩ , A23187 converts E1Ca 2 ⅐BeF x very rapidly to E2⅐BeF 3 Ϫ releasing Ca 2ϩ (supplemental Figs. 4 and 5). B, E2⅐BeF 3 Ϫ was first produced in SR vesicles without A23187 and Ca 2ϩ . Subsequently, the samples were diluted 10 times with the buffer containing 45 CaCl 2 and BeF x with and without 5 M A23187, to give 0.7 mM 45 Ca 2ϩ and the same buffer conditions as in A. At 15 s after dilution, the amount of bound 45 Ca 2ϩ was determined without EGTA washing and by subtracting the nonspecific Ca 2ϩ binding (1.0 Ϯ 0.1 nmol/mg (n ϭ 6)) determined by EGTA washing the sample incubated without BeF x and A23187.

JOURNAL OF BIOLOGICAL CHEMISTRY 22731
logs agree with the conclusion that E1Ca 2 ⅐BeF x is the analog for E1PCa 2 ⅐Mg, and BeF x is most probably BeF 3 Ϫ , i.e. E1Ca 2 ⅐BeF 3 Ϫ . Here note that the replacement of phosphate with BeF 3 Ϫ produces stabilization of the E1PCa 2 ⅐Mg structure with the same geometry of BeF 3 Ϫ as phosphate, and therefore probably with the same binding residues for them within the catalytic site. The E1Ca 2 ⅐BeF 3 Ϫ stability is likely brought about by the specific chemical nature of fluoride. Namely, it possesses a significantly higher electronegativity than oxygen (actually the highest among all atoms) and a small size, therefore producing stronger BeF 3 Ϫ binding in the catalytic site and fixing the intermediate structure.
Structure of E1Ca 2 ⅐BeF 3 Ϫ -Then with the newly developed E1Ca 2 ⅐BeF 3 Ϫ , we explored its structural properties and uncovered the hitherto unknown nature of the physiological intermediate E1PCa 2 ⅐Mg and structural changes during the phosphoryl transfer/ADP release and subsequent EP isomerization/Ca 2ϩ release. The observed proteinase K resistance of Thr 242 on the A/M3-linker revealed that, in E1PCa 2 ⅐Mg (E1Ca 2 ⅐BeF 3 Ϫ ) the A domain is already rotated perpendicular to the membrane plane from the position in E1Ca 2 , thereby bringing up its junction with the A/M3-linker and imposing a strain on this linker, similarly to E1Ca 2 ⅐AlF 4 Ϫ ⅐ADP and E1Ca 2 ⅐CaAMP-PCP (17,18 5 Our results revealed that such a strained state is achieved even without the N-P domain cross-linking nucleotide but solely with BeF 3 Ϫ and Mg 2ϩ binding at the catalytic site, and therefore remains in E1PCa 2 ⅐Mg after ADP release. The strain of the A/M3-linker thus imposed has been predicted with the atomic structure (18,19) to function as a motive force for the A domain rotation parallel to the membrane in the E1P to E2P isomerization. The partial resistance at T2 site Arg 198 in E1Ca 2 ⅐BeF 3 Ϫ (as compared with the rapid cleavage in E1Ca 2 ⅐CaAMP-PCP/E1Ca 2 ⅐AlF 4 Ϫ ⅐ADP and E1Ca 2 ⅐AlF x ) further indicated that in E1PCa 2 ⅐Mg, the A domain is already likely 5   rotated parallel to the membrane to some extent from the position in E1Ca 2 ⅐MgATP * /E1PCa 2 ⅐ADP⅐Mg ‡ . Thus, the A/M3linker strain is likely functioning for this partial A domain rotation during the phosphoryl transfer/ADP release to produce E1PCa 2 ⅐Mg, and further for the large and complete rotation to achieve the tight A-P domain association at Arg 198 on the Val 200 loop in the Ca 2ϩ -released state E2P⅐Mg (E2⅐BeF 3 Ϫ ). The A-P domain interaction at the Val 200 loop is actually critical for formation of the proper Ca 2ϩ -released structure, E2P⅐Mg and its analog E2⅐BeF 3 Ϫ (25,62,63). Here, it is of interest to note that residues Asp 351 , Thr 353 , and Asp 703 ligating Mg 2ϩ and phosphate will come more proximate to each other during E1PCa 2 ⅐ADP⅐Mg ‡ 3 E1PCa 2 ⅐Mg ϩ ADP (E1Ca 2 ⅐AlF 4 Ϫ ⅐ADP/E1Ca 2 ⅐AlF x 3 E1Ca 2 ⅐BeF x ). As a consequence, a further P domain bending and more strain for the A/M3-linker will likely be induced by this coordination-chemical change, thereby contributing to inducing the A domain rotations during E1PCa 2 ⅐Mg formation and subsequent isomerization to E2PCa 2 ⅐Mg (besides the release of the N-P domain cross-linking nucleotide ADP). In any case, our results show that E1PCa 2 ⅐Mg (E1Ca 2 ⅐BeF 3 Ϫ ) as the product of the phosphorylation reaction acquires the structure ready for isomerization and Ca 2ϩ deocclusion/release (i.e. ready for the large A domain rotation to produce E2P⅐Mg (E2⅐BeF 3 Ϫ )), whereas the transition state structure in the phosphorylation (E1Ca 2 ⅐AlF 4 Ϫ ⅐ADP and E1Ca 2 ⅐AlF x ) is not yet fully prepared. Note again that the E1PCa 2 ⅐Mg structure before such motions for its isomerization to E2P⅐Mg is stabilized with replacement of phosphate with BeF 3 Ϫ in E1Ca 2 ⅐BeF 3 Ϫ . Important also, we found that the Ca 2ϩ -occluded transmembrane structure adopts not simply one state, but will proceed through changes during the phosphoryl transfer and ADP release to form E1PCa 2 ⅐Mg (see supplemental Fig. S3E). The structural change is probably coupled with the above described motions of the P and A domains (more bending and rotation) during this process. In the subsequent Ca 2ϩ deocclusion/release in E1PCa 2 ⅐Mg 3 E2P⅐Mg ϩ 2Ca 2ϩ , the transmembrane structure changes further (52), which was also clearly mimicked here in the change E1Ca  Ϫ formation, and why the Mg 2ϩ -coordinated structure E1Ca 2 ⅐BeF 3 Ϫ differs from Ca 2ϩ -coordinated E1PCa 2 ⅐Ca, E1Ca 2 ⅐CaAMP-PCP, and E1PCa 2 ⅐Ca⅐AMP-PN structures as well as from E1Ca 2 ⅐ AlF 4 Ϫ ⅐ADP/E1Ca 2 ⅐AlF x , especially in A domain positioning. These questions may be relevant to the questions of why forward isomerization of E1PCa 2 ⅐Ca to E2P is markedly retarded in contrast to E1PCa 2 ⅐Mg and thus why Mg 2ϩ is preferred as the catalytic cation. In stringent coordination chemistry, the coordination distance of Mg 2ϩ is shorter than Ca 2ϩ , typically by 0.2 Å (e.g. 2.1 versus 2.3 Å (64, 65)). As a consequence, in the case of E1Ca 2 ⅐CaAMP-PCP, the distance between the ␥-phos-phate and Asp 351 -oxygen becomes 3.24 Å, being greater by 0.3 Å than that predicted in E1Ca 2 ⅐MgAMP-PCP. Therefore MgAMP-PCP (MgATP) binding would result in steric clash, and E1Ca 2 ⅐CaAMP-PCP is more stable than E1Ca 2 ⅐MgAMP-PCP, and therefore has less tendency to decompose to E1Ca 2 (also in the forward direction to the EP formation and its decay in the case of E1Ca 2 ⅐CaATP) (45). In E1Ca 2 ⅐BeF 3 Ϫ formed here with Mg 2ϩ , the direct coordination between Asp 351 and the beryllium and their proximate positioning would probably favor the closely positioned ligand residues (Thr 353 /Asp 703 / Asp 351 ) for BeF 3 Ϫ and Mg 2ϩ but not for Ca 2ϩ . Therefore Ca 2ϩ substitution of Mg 2ϩ probably disrupted the precise geometry and decomposed the E1Ca 2 ⅐BeF 3 Ϫ complex. Also, a possible difference in the coordination number might be involved; Mg 2ϩ prefers definitely six, whereas Ca 2ϩ can accommodate seven or eight ligands (65)(66)(67)(68)(69).
Furthermore, the difference in A domain positioning between the Mg 2ϩ -coordinated state E1Ca 2 ⅐BeF 3 Ϫ and the Ca 2ϩ -coordinated states may be reasonably understood by the consequence of the stringent coordination chemistry. Namely, because the shorter coordination distance of Mg 2ϩ , P domain bending, and the resulting A domain rotation perpendicular to the membrane will be greater in the Mg 2ϩ -coordinated state. Therefore the strain of the A/M3-linker and A domain rotation parallel to the membrane will be more in the Mg 2ϩ state E1Ca 2 ⅐BeF x . In this context, it is also reasonable that E1PCa 2 ⅐Mg is more rapidly isomerized to E2P with less energy barrier for the large A domain rotation, in contrast to the retarded isomerization in E1PCa 2 ⅐Ca that is stabilized by the likely conformational inadequacy. Here note that the cause of the E1PCa 2 ⅐Ca stabilization is obviously different from that of E1PCa 2 ⅐Mg stabilization produced by replacement of phosphate with BeF 3 Ϫ (see the above discussion for E1Ca 2 ⅐BeF 3 Ϫ stabilization).
Previously it was documented (45,64,70) that destabilization of the non-covalent complex E1Ca 2 ⅐MgATP by Mg 2ϩ (as found with MgAMP-PCP versus CaAMP-PCP) together with stabilization of the transition state by Mg 2ϩ (as found with E1Ca 2 ⅐AlF 4 Ϫ ⅐ADP bound Mg 2ϩ at both sites I and II) leads to a decrease of the activation energy and a rapid phosphoryl transfer. As another critical reason for Mg 2ϩ preference for catalysis, we predict here by exploring the property of the E1Ca 2 ⅐BeF x that the Mg 2ϩ bound at the catalytic site produces the proper E1PCa 2 structure, which is ready for rapid transition to E2P in this rate-limiting process of the transport cycle.
Hydrophobic Catalytic Site in E1Ca 2 ⅐BeF 3 Ϫ -The microenvironment around Asp 351 in E1PCa 2 ⅐Mg was further predicted by TNP-AMP superfluorescence in E1Ca 2 ⅐BeF 3 Ϫ to be strongly hydrophobic and thus a closed state, and this will become even more in the change E1PCa 2 ⅐Mg 3 E2P⅐Mg ϩ 2Ca 2ϩ (E1Ca 2 ⅐BeF 3 Ϫ 3 E2⅐BeF 3 Ϫ ). The observed distinct difference between E1Ca 2 ⅐BeF 3 Ϫ and E2⅐BeF 3 Ϫ (transient versus stable and slightly higher superfluorescence) is probably ascribed to the distinct difference in the organization state of cytoplasmic domains. The superfluorescence, nevertheless, developed solely in the Ca 2ϩ -ATPase complexed with beryllium fluoride, E1Ca 2 ⅐BeF 3 Ϫ and E2⅐BeF 3 Ϫ (E2P⅐Mg), but no development in E1Ca 2 ⅐AlF x and in E2⅐AlF 4 Ϫ and E2⅐MgF 4 2Ϫ (25). Therefore, the hydrophobic closed catalytic site is accomplished by the direct coordination and close proximity of the beryllium with Asp 351oxygen and by the specific coordination of the tetrahedral -O-BeF 3 Ϫ , i.e. Asp 351 -acylphosphate within the catalytic site. This is obviously not the case in AlF 4 Ϫ (the penta-coordinated phosphorus of the transition states) and MgF 4 2Ϫ (non-covalently bound P i ), thus in these states, the catalytic site is more accessible to nonspecific water molecules.
Whether E1PCa 2 ⅐Mg develops the superfluorescence had been controversial. In addition to the obvious problem of TNP-AMP competition against ATP for phosphorylation, the observed TNP-AMP-induced decomposition of E1Ca 2 ⅐BeF 3 Ϫ further revealed that the E1PCa 2 ⅐Mg structure may be similarly disrupted rapidly by TNP-AMP binding, therefore making it virtually impossible to examine the superfluorescence development in E1PCa 2 ⅐Mg. The TNP-AMP-induced E1Ca 2 ⅐BeF 3 Ϫ decomposition might have occurred by means of a similar structural change as the ADP-induced one, i.e. disruption of the cytoplasmic domain organization and possible BeF 3 Ϫ release. The most important conclusion here is that the hydrophobic and closed property of the phosphorylated catalytic site both in E1PCa 2 ⅐Mg and E2P⅐Mg may be requisite to avoid a possible attack of nonspecific water molecules on the Asp 351 -acylphosphate thus accomplishing Ca 2ϩ release into the lumen and energy coupling.
Formation of E1Ca 2 ⅐BeF 3 Ϫ from E2⅐BeF 3 Ϫ and Perfect Stabilization of E1Ca 2 ⅐BeF 3 Ϫ -E1Ca 2 ⅐BeF 3 Ϫ was produced also from E2⅐BeF 3 Ϫ by binding two lumenal Ca 2ϩ to the lumenally oriented low affinity transport sites at 0.7 mM Ca 2ϩ and 15 mM Mg 2ϩ , as mimicking the reverse transition E2P⅐Mg ϩ 2Ca 2ϩ 3 E1PCa 2 ⅐Mg. At the critical concentration of 0.7 mM Ca 2ϩ in 15 mM Mg 2ϩ , E1Ca 2 ⅐BeF 3 Ϫ is perfectly stabilized without decomposition to E1Ca 2 or conversion to E2⅐BeF 3 Ϫ . The perfect E1Ca 2 ⅐BeF 3 Ϫ stabilization is obviously achieved by preventing Ca 2ϩ release into the lumen and by avoiding the absolutely unfavorable Ca 2ϩ substitution of Mg 2ϩ in site I at the most appropriately balanced concentrations of Ca 2ϩ and Mg 2ϩ . As noted in the last paragraph under "Results," stabilization of E1Ca 2 ⅐BeF 3 Ϫ might possibly involve lumenal Ca 2ϩ access at the putative lumenal gating sites besides the transport sites. If this is the case, the gate-opening and Ca 2ϩ release into the lumen takes place when the lumenal Ca 2ϩ is low enough to avoid the possible lumenal Ca 2ϩ access to the gate.
Integrated Picture of EP Processing-Recently, we successfully identified and trapped for the first time the intermediate state E2PCa 2 ⅐Mg, ADP-insensitive EP with two Ca 2ϩ occluded at transport sites, by elongating the A/M1Ј-linker (71), and revealed that the proper length of this linker is critical for inducing structural changes for Ca 2ϩ deocclusion and release from E2PCa 2 ⅐Mg. This dependence on the length of the linker is probably because the length controls the extent of strain between the A domain and M1Ј, which causes motions of the cytoplasmic A and P domains thereby transmitting the structural signal to the transmembrane transport sites. In trapped E2PCa 2 ⅐Mg, the A domain is already largely rotated, and A-P domain associations at Val 200 and TGES 184 loops are already produced, although the interaction network is not produced properly at the Tyr 122 -hydrophobic cluster (71), which is criti-cal for Ca 2ϩ deocclusion/release and E2P hydrolysis (72)(73)(74). In the Ca 2ϩ -released E2P⅐Mg, this cluster is formed from seven residues of the A (Ile 179 /Leu 180 /Ile 232 ) and P (Val 705 /Val 726 ) domains and the top part of M2 (Leu 119 /Tyr 122 ) (see Fig. 2).
The results indicated that the successive structural changes take place as follows: in E1PCa 2 ⅐Mg 3 E2PCa 2 ⅐Mg, the A domain rotates largely (further from the position in E1PCa 2 ⅐Mg) into the space between the N and P domains and docks onto the Asp 351 -acylphosphate of the P domain, thereby causing loss of ADP sensitivity and also the strain of the A/M1Јlinker (because the A domain is brought above the P domain). The strain thus imposed will cause inclinations of the A and P domains and the connected M2 and M4/M5 thereby rearranging the helices to destroy Ca 2ϩ sites and open the lumenal gate thus to release Ca 2ϩ . Upon these motions, the Tyr 122 -hydrophobic cluster is produced from the inclined A and P domains and M2. Hence, interactions at this cluster and at the Val 200 loop stabilize the Ca 2ϩ -released structure E2P⅐Mg, and also produce the catalytic site for the acylphosphate hydrolysis to occur after Ca 2ϩ release, ensuring energy coupling (63,(72)(73)(74). Atomic level structural studies of E1Ca 2 ⅐BeF 3 Ϫ as E1PCa 2 ⅐Mg and the trapped intermediate state E2PCa 2 ⅐Mg will contribute to further understanding of EP processing, Ca 2ϩ handling, and E2P hydrolysis.