Stable Structural Analog of Ca2+-ATPase ADP-insensitive Phosphoenzyme with Occluded Ca2+ Formed by Elongation of A-domain/M1′-linker and Beryllium Fluoride Binding*

We have developed a stable analog for the ADP-insensitive phosphoenzyme intermediate with two occluded Ca2+ at the transport sites (E2PCa2) of sarcoplasmic reticulum Ca2+-ATPase. This is normally a transient intermediate state during phosphoenzyme isomerization from the ADP-sensitive to ADP-insensitive form and Ca2+ deocclusion/release to the lumen; E1PCa2 → E2PCa2 → E2P + 2Ca2+. Stabilization was achieved by elongation of the Glu40-Ser48 loop linking the Actuator domain and M1 (1st transmembrane helix) with four glycine insertions at Gly46/Lys47 and by binding of beryllium fluoride (BeFx) to the phosphorylation site of the Ca2+-bound ATPase (E1Ca2). The complex E2Ca2·BeF3− was also produced by lumenal Ca2+ binding to E2·BeF3− (E2P ground state analog) of the elongated linker mutant. The complex was stable for at least 1 week at 25 °C. Only BeFx, but not AlFx or MgFx, produced the E2PCa2 structural analog. Complex formation required binding of Mg2+, Mn2+, or Ca2+ at the catalytic Mg2+ site. Results reveal that the phosphorylation product E1PCa2 and the E2P ground state (but not the transition states) become competent to produce the E2PCa2 transient state during forward and reverse phosphoenzyme isomerization. Thus, isomerization and lumenal Ca2+ release processes are strictly coupled with the formation of the acylphosphate covalent bond at the catalytic site. Results also demonstrate the critical structural roles of the Glu40-Ser48 linker and of Mg2+ at the catalytic site in these processes.

phosphoenzyme (E1P) that can react with ADP to regenerate ATP (steps 1-3). E1P formation results in Ca 2ϩ occlusion at the transport sites (E1PCa 2 ). Subsequent isomeric transition to an ADP-insensitive form (E2P), i.e. loss of ADP-sensitivity, results in Ca 2ϩ deocclusion and release into the lumen (steps 4 and 5). This Ca 2ϩ -release process is very rapid so that an E2PCa 2 intermediate state does not accumulate and in fact had never been found until we recently established its existence (10 -13) and successfully trapped it for the first time (14). The Ca 2ϩ -free E2P is finally hydrolyzed to the inactive E2 state (steps 6 and 7). Mg 2ϩ as the physiological catalytic cofactor is required for both phosphorylation and hydrolysis. The transport cycle is reversible. Thus, E2P can be formed from P i in the presence of Mg 2ϩ and absence of Ca 2ϩ . Subsequent Ca 2ϩ binding to lumenal-oriented low affinity transport sites reverses the Ca 2ϩ -releasing step and the E1P to E2P isomerization.
We have recently developed an E1Ca 2 ⅐BeF 3 Ϫ complex as a stable analog of E1PCa 2 ⅐Mg 2ϩ (E1PCa 2 with bound Mg 2ϩ at the catalytic site) (27). Structural analysis of the analog and intermediate states suggests that formation of native E1PCa 2 ⅐Mg 2ϩ results in structural changes in the cytoplasmic and transmembrane domains due to configuration and ligation changes of the phosphate moiety (27). The Mg 2ϩ bound at the catalytic site contributes to these structural changes (27). In fact, Ca 2ϩ could not substitute for Mg 2ϩ for formation of E1Ca 2 ⅐BeF 3 Ϫ , and an attempt to substitute Ca 2ϩ for Mg 2ϩ destroyed the complex (27). It is well known that Ca 2ϩ substitution of Mg 2ϩ at the catalytic site markedly retards E1PCa 2 ⅐Ca isomerization (28,29), a step that includes rotation of the A domain.
Further understanding of the mechanism of EP processing via the transient E2PCa 2 and of the critical roles of the A/M1Ј-linker and catalytic Mg 2ϩ requires detailed characterization of the development of E2PCa 2 and of factors contributing to its possible stabilization. A great advance would be the finding of an analog stable enough for crystallographic studies.
In this study we employed the mutant 4Gi-46/47 in which the A/M1Ј-linker is elongated by four glycine insertions at Gly 46 / Lys 47 (14) and explored the formation of a stable structural analog of E2PCa 2 using various configuration analogs of phosphate (BeF x /AlF x /MgF x ) and catalytic cations (Mg 2ϩ /Mn 2ϩ / Ca 2ϩ ). We found that BeF x is uniquely efficacious and that both mutant E1Ca 2 ⅐BeF 3 Ϫ and mutant E2⅐BeF 3 Ϫ are capable of producing mutant E2Ca 2 ⅐BeF x , most probably E2Ca 2 ⅐BeF 3 Ϫ , and that Ca 2ϩ can replace the catalytic Mg 2ϩ when coming from the former species. The mutant complex E2Ca 2 ⅐BeF 3 Ϫ is extremely stable even at 25°C.

EXPERIMENTAL PROCEDURES
Mutagenesis and Expression-The pMT2 expression vector (30) carrying the mutant rabbit SERCA1a cDNA with four glycine residues inserted between Gly 46 and Lys 47 (4Gi-46/47) was constructed as described previously (14). Transfection of pMT2 DNA into COS-1 cells and preparation of microsomes from the cells were performed as described previously (31). The amount of expressed SERCA1a was quantified by a sandwich enzyme-linked immunosorbent assay (32). Expression levels of wild type SERCA1a and the mutants were 2-3% that of total microsomal proteins.
Formation of EP-Phosphorylation of SERCA1a in microsomes with [␥-32 P]ATP was performed under conditions described in the legends for Figs. 3-8. The reactions were quenched with ice-cold trichloroacetic acid containing P i . Precipitated proteins were separated by 5% SDS-PAGE at pH 6.0 according to Weber and Osborn (37). The radioactivity associated with the separated Ca 2ϩ -ATPase was quantified by digital autoradiography as described (38).
Ca 2ϩ Occlusion in SERCA1a-Microsomes treated with metal fluoride were diluted with "washing solution" contain-ing excess EGTA and then immediately filtered through a 0.45-m nitrocellulose membrane filter (Millipore). The filter was washed extensively with the washing solution, and 45 Ca 2ϩ remaining on the filter was quantified. The amount of Ca 2ϩ specifically bound to the transport sites of EP in the expressed SERCA1a was obtained by subtracting the amount of nonspecific Ca 2ϩ -binding, which was determined as described in the legends for Figs. 8 and 9. The Ca 2ϩ occluded/mg of expressed SERCA1a protein was calculated from the amount of expressed SERCA1a and the amount of occluded Ca 2ϩ .
Limited Proteolysis and Western Blot Analysis-Major intermediates of the Ca 2ϩ -ATPase and their stable analogs were produced and subjected to structural analysis by limited proteolysis with trypsin and proteinase K (prtK) as described in the legends for supplemental Figs. S3 and S4. Proteolysis was terminated by 2.5% (v/v) trichloroacetic acid. The digests were subjected to SDS-PAGE (39) followed by Western blot analysis with IIH11 monoclonal antibody to the rabbit SERCA1a (Affinity Bioreagents), which recognizes an epitope between Ala 199 and Arg 505 as described (14).
Miscellaneous-Protein concentrations were determined by the method of Lowry et al. (40) with bovine serum albumin as a standard. Data were analyzed by nonlinear regression using the program Origin (Microcal Software, Inc., Northampton, MA). Three-dimensional models of the enzyme were reproduced by the program VMD (41).

Inhibition of EP Formation by Metal
Fluoride-The E1Ca 2 state of wild type and mutant 4Gi-46/47 SERCA1a in 10 M Ca 2ϩ was treated with BeF x or AlF x and functionally analyzed. The ability to form EP from ATP (Fig. 3, A and C) and from P i (data not shown) is almost completely lost in the presence of 15 mM Mg 2ϩ but not in its absence. EP formation is not inhibited when F Ϫ treatment in 15 mM Mg 2ϩ is made without Be 2ϩ or Al 3ϩ . The results show that the E1Ca 2 state of the mutant as well as of wild type forms stable complexes with BeF x and AlF x in the presence of Mg 2ϩ but not with MgF x .
When the E2 state of wild type and mutant 4Gi-46/47 in the absence of Ca 2ϩ was treated with BeF x , AlF x , and MgF x (in the absence of Be 2ϩ and Al 3ϩ ), the complexes E2⅐BeF 3 Ϫ , E2⅐AlF 4 Ϫ , and E2⅐MgF 4 2Ϫ , respectively, are produced (14,25), and EP formation from ATP (Fig. 3, B and D, open bars) and from P i (data not shown) is almost completely inhibited. These complexes were then treated with 10 mM Ca 2ϩ for 1 h in the presence of Ca 2ϩ ionophore A23187 (black bars in Fig. 3, B and D). In the case of wild type, the ability to form EP is restored, consistent with the previous observation (25,36)    ATP-induced phosphorylation. Therefore, the complex produced in the mutant with BeF x is likely E2Ca 2 ⅐BeF 3 Ϫ , an analog of E2PCa 2 (as is in fact shown later in the Ca 2ϩ binding and structural analyses in Fig. 8 and supplemental Figs. S3 and S4).
Kinetic Analysis of BeF x -induced Complex Formation-The E1Ca 2 state of mutant 4Gi-46/47 was treated with various concentrations of Be 2ϩ and 1 mM F Ϫ in 10 M Ca 2ϩ and 15 mM Mg 2ϩ , and the resulting species was analyzed (Fig. 4A). The Ϫ without TG, the cluster structure is rather loose (as the side chains of Leu 119 /Tyr 122 are pointing away from the hydrophobic cluster), but Leu 119 /Tyr 122 produce a more extended interaction network involving Thr 430 of the N domain and the hydrophobic cluster (see more details in supplemental Fig. S5). b, the catalytic site is enlarged, and the residues involved in the Mg 2ϩ (site I) are depicted. The Val 679 -Lys 686 region of the P domain is not depicted for simplicity (because it is positioned over the region of interest).
presence of both Be 2ϩ and F Ϫ (BeF x ) but not F Ϫ without Be 2ϩ or Be 2ϩ (20 M) without F Ϫ inhibits EP formation. The time courses of BeF x -induced inhibition follow first order kinetics. A plot of the inhibition rate constants versus Be 2ϩ (BeF x ) concentration is a straight line with no evidence of saturation within the experimental range, indicating that BeF x binding is the ratedetermining step in the inhibition process (Fig. 4B). BeF x inhibits wild type at nearly the same rate as it does the mutant as seen at a representative 20 M Be 2ϩ with 1 mM F Ϫ .
In Fig. 5, the mutant E1Ca 2 state in 10 M Ca 2ϩ was incubated with BeF x at various Mg 2ϩ concentrations, and the level of inhibition of EP formation was determined. BeF x -induced inhibition is markedly accelerated with increasing Mg 2ϩ , giving a K 0.5 value of 4.9 mM. The observed apparent Mg 2ϩ affinity is consistent with those values obtained through phosphorylation of native Ca 2ϩ -ATPase (42)(43)(44)(45)(46)(47) and for the formation of E1Ca 2 ⅐BeF 3 Ϫ (E1PCa 2 ⅐Mg 2ϩ analog) (27), i.e. the Mg 2ϩ binding affinity at the catalytic Mg 2ϩ site (site I composed of Asp 351 / Thr 353 /Asp 703 and the phosphate moiety (BeF 3 Ϫ )). Therefore, Mg 2ϩ binding at site I is likely a prerequisite for BeF x binding and complex formation.
In Figs. 6 and 7, we further observed that the BeF x -induced complex formation from E1Ca 2 in the mutant occurs with Mn 2ϩ or Ca 2ϩ in place of Mg 2ϩ . The K 0.5 values are 1.4 mM for Mn 2ϩ and 0.76 mM for Ca 2ϩ (supplemental Figs. S1 and S2) and are consistent with such values for binding to the catalytic Mg 2ϩ site (46,48). In wild type the BeF x -induced E1Ca 2 ⅐BeF 3 Ϫ formation, which inhibits EP formation occurs with Mn 2ϩ but not with 10 mM Ca 2ϩ in place of Mg 2ϩ (Figs. 6 and 7). Thus, the complex formed from E1Ca 2 with BeF x in the mutant 4Gi-46/47 (i.e. Ϫ of wild type. Interestingly, the Hill coefficient for the Mg 2ϩ as well as Mn 2ϩ and Ca 2ϩ dependence for complex formation with BeF x (E2Ca 2 ⅐BeF 3 Ϫ ) in the mutant is nearly 2 ( Fig. 5 and supplemental Figs. S1 and S2), suggesting the involvement of more  than one metal ion. This is in contrast to the value 1 for E1Ca 2 ⅐BeF 3 Ϫ formation with Mg 2ϩ and Mn 2ϩ in wild type (see supplemental Fig. 1 in Ref. 27).
AlF x produces the complex with the E1Ca 2 state of the mutant 4Gi-46/47 as well as of wild type (E1Ca 2 ⅐AlF x ) with Mg 2ϩ and Mn 2ϩ but not with Ca 2ϩ at the catalytic Mg 2ϩ site (Figs. 3, 6, and 7). Therefore, in the mutant the complex with AlF x (E1Ca 2 ⅐AlF x ) is distinct from that with BeF x (E2Ca 2 ⅐BeF 3 Ϫ ) with respect to the strict preference of the divalent cation at the catalytic Mg 2ϩ site. The stoichiometry of the occluded Ca 2ϩ is nearly 2 per phosphorylation site, which is 4.3 nmol/mg as determined from the intercept on the ordinate. Therefore, the complex formed with BeF x has two occluded Ca 2ϩ . When the mutant was incubated for 15 min with BeF x and 1.5 mM Mn 2ϩ in place of Mg 2ϩ under otherwise identical conditions, EP formation was completely inhibited, and the amount of occluded 45 Ca 2ϩ was 8.3 nmol/mg of expressed SERCA1a mutant protein, giving a stoichiometry of 2 per phosphorylation site (data not shown).
Ϫ was first formed in the mutant in the absence of Ca 2ϩ   and then incubated for 1 min at 25°C with various concentrations of 45 Ca 2ϩ in the presence of Ca 2ϩ ionophore A23187. The amount of occluded 45 Ca 2ϩ was determined after a large dilution followed by filtration and extensive EGTA washing. The maximum amount of occluded 45 Ca 2ϩ is 7.7 nmol/mg of mutant SERCA1a protein and 1.8 times that of the phosphorylation site (4.3 nmol/mg), giving a stoichiometry of nearly 2. Thus, mutant E2Ca 2 ⅐BeF 3 Ϫ is produced from mutant E2⅐BeF 3 Ϫ by the addition of Ca 2ϩ in the presence of A23187. K 0.5 and the Hill coefficient observed in Fig. 8B are 0.1 mM and ϳ2, respectively, i.e. very similar values to those observed during E2PCa 2 formation from E2P and Ca 2ϩ in the mutant (14). The observed low Ca 2ϩ affinity is in agreement with the wild type property (25,49) that E2⅐BeF 3 Ϫ as well as E2P have low affinity Ca 2ϩ binding sites; that is, the lumenal-oriented transport sites. Importantly, E2Ca 2 ⅐BeF 3 Ϫ /E2PCa 2 formed in the mutant (either from E1Ca 2 or from E2⅐BeF 3 Ϫ /E2P) is remarkably stable and virtually not in equilibrium with E1Ca 2 ⅐BeF 3 Ϫ / E1PCa 2 or E2⅐BeF 3 Ϫ /E2P, i.e. their formation is almost irreversible, as shown previously (14) and in this study. When Ca 2ϩ comes from the cytoplasmic side for E2PCa 2 formation from E1Ca 2 with ATP (via E2 3 E1Ca 2 3 E1PCa 2 3 E2PCa 2 ) in the mutant, the apparent Ca 2ϩ affinity is very high, with K 0.5 ϭ 0.14 M (14), equal to the value for cytoplasmic Ca 2ϩ binding at the transport sites in wild type. Also in the case of mutant E2Ca 2 ⅐BeF 3 Ϫ formation from E1Ca 2 with BeF x in Fig. 8A, 10 M Ca 2ϩ is obviously enough to saturate (even 1 M Ca 2ϩ saturates (data not shown)), suggesting a similar high Ca 2ϩ affinity as in E2PCa 2 formation from E1 ϩ 2Ca 2ϩ .
Structures of Complexes Formed from E1Ca 2 with Metal Fluoride-During the Ca 2ϩ transport cycle, the A, P, and N domains move and reorganize substantially. These changes can be monitored by proteolytic patterns and resistance against trypsin and prtK (23,24). Therefore, we applied proteolytic analyses to mutant E2Ca 2 ⅐BeF 3 Ϫ to reveal the position of the domains and to establish whether it is a true structural E2PCa 2 analog (supplemental Figs. S3 and S4 and Tables S1 and S2 and Refs. 54 and 55). All the various major intermediates and their analogs were formed from E1Ca 2 in the mutant and wild type and then subjected to proteolyses. The results show that mutant E2Ca 2 ⅐BeF 3 Ϫ has the same structure as that of mutant   Fig. 3C. The sample was then further diluted 20-fold at 0°C with the washing solution, immediately filtered as above, and washed rapidly with ice-cold trichloroacetic acid containing P i . The EP level was not changed during the above sample handling because the decay of EP (E2PCa 2 ) is almost completely blocked in the mutant (14). The amount of 45  description and reasons for protection in supplemental Fig. S5 and Ref. 56). By contrast, in mutant E2Ca 2 ⅐BeF 3 Ϫ and mutant E2PCa 2 , the prtK-site Leu 119 is rapidly cleaved and, thus, exposed. Evidently Leu 119 /Tyr 122 on M2 in mutant E2Ca 2 ⅐BeF 3 Ϫ and mutant E2PCa 2 have moved from their hidden position in E1PCa 2 (E1Ca 2 ⅐BeF 3 Ϫ ) but are not yet buried again through interaction with the A and P domains as in E2P (E2⅐BeF 3 Ϫ ), suggesting an intermediate structure. The results also reveal how critical the native length of the A/M1Ј-linker is for moving M2 and the A and P domains to realize the Ca 2ϩ -free state E2P (E2⅐BeF 3 Ϫ ). The proteolyses also reveal that wild type and mutant E1Ca 2 ⅐AlF x are not structurally similar to wild type E1Ca 2 ⅐BeF 3 Ϫ (E1PCa 2 ) and mutant E2Ca 2 ⅐BeF 3 Ϫ (E2PCa 2 ). Interestingly, the rate of cleavage at the T2 site of mutant E1Ca 2 ⅐AlF x is intermediate between that of wild type transition state (E1Ca 2 ⅐AlF x / E1Ca 2 ⅐AlF 4 Ϫ ⅐ADP) and that of the E1PCa 2 product state (E1Ca 2 ⅐BeF 3 Ϫ ), suggesting that the structure is also intermediate. Thus, elongation of the A/M1Ј-linker brought the E1Ca 2 ⅐ AlF x structure closer to that of wild type E1Ca 2 ⅐BeF 3 Ϫ . Only BeF x , and not AlF x , produces a species analogous to the E2PCa 2 structural state (E2Ca 2 ⅐BeF 3 Ϫ via E1Ca 2 ⅐BeF 3 Ϫ ). This means that the phosphorylation reaction must have passed through the transition state to progress to the isomerization step.
In the mutant and wild type, the prtK-site Thr 242 on the A/M3-linker is completely resistant in all the states E1Ca 2 ⅐ AlF 4 Ϫ ⅐ADP/E1Ca 2 ⅐AlF x , E1Ca 2 ⅐BeF 3 Ϫ (E1PCa 2 ), E2Ca 2 ⅐BeF 3 Ϫ and E2PCa 2 , and E2⅐BeF 3 Ϫ /E2⅐AlF 4 Ϫ /E2⅐MgF 4 2Ϫ (as shown previously with sarcoplasmic reticulum Ca 2ϩ -ATPase (23,24)). The result indicates that in both mutant and wild type, the A/M3-linker is strained by the A-domain rotation perpendicular to the membrane plane upon E1PCa 2 formation from E1Ca 2 and remains taut during EP processing. E2Ca 2 ⅐BeF 3 Ϫ Formation from E2⅐BeF 3 Ϫ by Lumenal Ca 2ϩ Binding-The Ca 2ϩ -free complexes E2⅐BeF 3 Ϫ , E2⅐AlF 4 Ϫ , and E2⅐MgF 4 2Ϫ (the analogs of the E2P ground state, transition state, and product complex of E2P hydrolysis, respectively (25)) were first formed in mutant 4Gi-46/47 and wild type, with Mg 2ϩ bound at the catalytic site, and subsequent proteolyses were performed with and without a 10 mM Ca 2ϩ treatment in the presence of ionophore A23187 (supplemental Fig. S4 and Table S2). Under these conditions Ca 2ϩ -treated mutant E2⅐BeF 3 Ϫ exhibits complete resistance at the tryptic T2 site Arg 198 and a fairly rapid prtK cleavage at Leu 119 on the top of M2, exactly as in mutant E2PCa 2 and E2Ca 2 ⅐BeF 3 Ϫ produced from E1Ca 2 . These results agree with those in Fig. 3D where it is found that the ability to form EP is not restored by Ca 2ϩ treatment of E2⅐BeF 3 Ϫ . Thus, E2Ca 2 ⅐BeF 3 Ϫ , as the E2PCa 2 analog, is produced from both E2⅐BeF 3 Ϫ and from E1Ca 2 (mimicking lumenal Ca 2ϩ binding to E2P in the reverse direction of the pump cycle and the forward ATP-induced EP formation and isomerization, respectively). On the other hand, mutant and wild type complexes E2⅐AlF 4 Ϫ and E2⅐MgF 4 2Ϫ and wild type E2⅐BeF 3 Ϫ are destroyed by Ca 2ϩ treatment as found previously with sarcoplasmic reticulum Ca 2ϩ -ATPase (25,27).

Stability of Complex E2Ca 2 ⅐BeF 3
Ϫ -In Fig. 9, E2Ca 2 ⅐BeF 3 Ϫ was first produced from mutant E1Ca 2 with BeF x in 50 M 45 Ca 2ϩ and 15 mM Mg 2ϩ , then further incubated at 25°C in the pres-ence of these ligands, and the amount of occluded 45 Ca 2ϩ was determined. The results show that the complex E2Ca 2 ⅐BeF 3 Ϫ of the mutant is perfectly stable even after 1 week. Proteolysis confirms that the structure remains unchanged during the incubation (data not shown). The stability of the complex was further tested by diluting into an EGTA-containing solution without BeF x , and the incubation was continued at 25°C (see the inset). Ca 2ϩ is slowly released with a rate constant of 7.0 h Ϫ1 . The addition of thapsigargin (TG) to the diluent only doubles the rate of release, indicating that the transmembrane domain is fairly resistant to TG-induced structural perturbation. These decay rates are very similar to those of mutant E2PCa 2 without and with TG addition, 9.7 and 27.3 h Ϫ1 , respectively (14). Thus, in this respect also, mutant E2Ca 2 ⅐BeF 3 Ϫ is analogous to mutant E2PCa 2 .

Mutant E2Ca 2 ⅐BeF 3
Ϫ as an Analog of Native Transient State E2PCa 2 -Using our elongated A/M1Ј-linker mutant, we have developed the complex E2Ca 2 ⅐BeF x , most probably E2Ca 2 ⅐ BeF 3 Ϫ , as a stable structural analog of the native transient state E2PCa 2 (ADP-insensitive EP with two Ca 2ϩ at the transport sites), an intermediate in EP isomerization and Ca 2ϩ deocclusion/release. The complex E2Ca 2 ⅐BeF 3 Ϫ has two occluded Ca 2ϩ and is produced from both mutant E1Ca 2 and mutant E2⅐BeF 3 Ϫ , mimicking native E2PCa 2 formation from E1Ca 2 after ATPinduced forward phosphorylation via E1PCa 2 isomerization  Fig. 8A. Then a small volume of A23187 was added to give 1 M, and the incubation was further continued at 25°C. At various times, the amount of 45 Ca 2ϩ specifically bound and occluded in the mutant was measured after an EGTA wash and by subtracting the background levels determined in the absence of F Ϫ in the incubation mixture and otherwise as in Fig. 8A. Inset, after the formation of E2Ca 2 ⅐BeF 3 Ϫ as above, the sample was diluted 100-fold at 25°C with a solution containing 1 M A23187, 0.1 M KCl, 7 mM MgCl 2 , 2 mM EGTA, and 50 mM MOPS/Tris (pH 7.0) (without BeF x ) in the absence (E) or presence (F) of 1 M TG and incubated for various periods, and the amount of 45 Ca 2ϩ specifically bound and occluded in the mutant was obtained as above. The values presented are the mean Ϯ S.D. (n ϭ 7). Solid lines in the inset show the least squares fit to a single exponential, and the decay rate constants thus obtained are 7.0 (E) and 14.0 (F) h Ϫ1 without and with TG, respectively. In the main panel and inset, the amount of Ca 2ϩ occluded in the complex E2Ca 2 ⅐BeF 3 Ϫ at time 0 (immediately before starting the long incubation or the dilution) was normalized to 100%. and in the reverse direction from E2P after lumenal Ca 2ϩ binding. Mutant E2Ca 2 ⅐BeF 3 Ϫ formation requires Mg 2ϩ at the catalytic site as in native ATP-and P i -induced EP formation. The disposition of the cytoplasmic domains in mutant E2Ca 2 ⅐BeF 3 Ϫ is equivalent to that in E2PCa 2 trapped with the mutant and intermediate between native E1PCa 2 ⅐Mg 2ϩ (E1Ca 2 ⅐BeF 3 Ϫ of wild type) and native E2P⅐Mg 2ϩ (E2⅐BeF 3 Ϫ of wild type and mutant). All these properties of mutant E2Ca 2 ⅐BeF 3 Ϫ meet the requirements of a native E2PCa 2 analog. Importantly, AlF x and MgF x are not able to produce this E2PCa 2 analog either from mutant E1Ca 2 or from mutant E2⅐AlF 4 Ϫ and E2⅐MgF 4

2Ϫ
. Thus, BeF x is unique in this regard. The coordination chemistry of the beryllium in BeF x (BeF 3 Ϫ ) allows it to directly ligate the aspartyl oxygen, thereby producing the same tetrahedral geometry as the covalent Asp 351acylphosphate, as seen in the atomic structure of the E2P ground state analog E2⅐BeF 3 Ϫ (21, 22). On the other hand, AlF x (AlF 3 or AlF 4 Ϫ ) mimics the transition state of phosphorylation and dephosphorylation as seen in structures E1Ca 2 ⅐AlF 4 Ϫ ⅐ADP and E2⅐AlF 4 Ϫ (17,19,22). MgF 4 2Ϫ mimics P i in the product complex E2⅐P i after E2P hydrolysis as seen in structure E2⅐MgF 4 2Ϫ (19). Our results taken together with the coordination chemistry show that the structural changes for EP isomerization and Ca 2ϩ deocclusion/release in the forward and reverse reactions are strictly coupled with the particular configuration of the acylphosphate after formation of the covalent bond within the catalytic site. The product E1PCa 2 state and the E2P ground state are ready for the changes, but the transition state structures are not.
Roles of A/M1Ј-linker and Structural Changes during EP Formation and Processing-The transient E2PCa 2 state formed during EP processing and its analog E2Ca 2 ⅐BeF 3 Ϫ were trapped and stabilized by elongation of the A/M1Јlinker. As revealed by the proteolyses, in mutant E2Ca 2 ⅐BeF 3 Ϫ and mutant E2PCa 2 , the A domain has already rotated parallel to membrane from its position in E1Ca 2 ⅐BeF 3 Ϫ (E1PCa 2 ⅐Mg) and has associated with the P domain at the Val 200 loop. Because mutant E2PCa 2 is ADP-insensitive (14), the outermost loop TGES 184 of the A domain is most probably docked onto the Asp 351 region, thereby blocking ADP access to the Asp 351 acylphosphate (19). Thus, in mutant E2Ca 2 ⅐BeF 3 Ϫ and mutant E2PCa 2 , the A domain is positioned above the P domain. On the other hand, the proteolyses also show that the spatial relationship of the top part of M2 (Leu 119 /Tyr 122 ) with the P and A domains in mutant E2Ca 2 ⅐BeF 3 Ϫ (equivalent to native E2PCa 2 ⅐ Mg) is intermediate between those of the wild type E1Ca 2 ⅐BeF 3 Ϫ (native E1PCa 2 ⅐Mg) and the wild type and mutant E2⅐BeF 3 Ϫ (native E2P⅐Mg). Thus, Leu 119 (the prtK site) on the top part of M2 has broken its van der Waals contact with upper M4 seen in E1PCa 2 but has not yet reached the P and A domains to form their interaction network at Leu 119 /Tyr 122 , i.e. the Tyr 122 hydrophobic cluster has not formed (see supplemental Fig. S5 for its structure). This interaction network formed from Ile 179 / Leu 180 /Ile 232 of the A domain, Val 705 /Val 726 of the P domain, and Tyr 122 /Leu 119 of M2 is actually critical for the E2P structure (11)(12)(13). Therefore, in E2Ca 2 ⅐BeF 3 Ϫ and E2PCa 2 stabilized by elongation of the A/M1Ј-linker, the inclining motions of domains and helix are not yet advanced enough to reach the E2P structure.
Deletion of any single residue in the A/M1Ј-linker, i.e. shortening it, completely blocks E1PCa 2 isomerization to E2PCa 2 (26). By contrast, its elongation markedly accelerates the isomerization and greatly stabilizes E2PCa 2 blocking Ca 2ϩ deocclusion/release from this transient state (14). These findings suggest that formation of the transient E2PCa 2 state (mutant E2Ca 2 ⅐BeF 3 Ϫ ) from E1PCa 2 (E1Ca 2 ⅐BeF 3 Ϫ ), strains the A/M1Ј-linker with the wild type/native length due to rotation and positioning of the A domain above the P domain, which in turn causes further movements of the A and P domains facilitating Ca 2ϩ deocclusion/release (14) (see the schematic model in supplemental Fig. S6). The A and P domains incline more, as will M1/M2 and M4/M5 connected with these domains, favoring release of the Ca 2ϩ . This view agrees with the structural changes required for Ca 2ϩ release described by Toyoshima et al. (19); the bending and movement of M4/M5 by inclination of the P domain is predicted to destroy the Ca 2ϩ binding sites, and the inclination of M2 and M1 (as a V-shaped rigid body) will push the lower part of M4 via M1 and open the lumenal gate.
These domain and segmental motions associated with Ca 2ϩ release will establish the interaction network at Leu 119 /Tyr 122 , the Tyr 122 hydrophobic cluster, and stabilize the E2P structure with the lumenal gate open (11)(12)(13). The position of the two A-P domain interaction networks, with Leu 119 /Tyr 122 at the lower part and Val 200 loop on the upper part of the interface, seems particularly appropriate to stabilize the inclined A and P domains and helices and, therefore, the gate in an open state.
These cluster formations are also critical for producing the E2P catalytic site with hydrolytic ability (11)(12)(13). Therefore, in this mechanism E2P hydrolysis can only occur after Ca 2ϩ release, ensuring energy coupling. The relative stability of native E2P may function as a brake to allow enough time for releasing Ca 2ϩ and for refining the catalytic site for subsequent hydrolysis, e.g. appropriate positioning of TGES 184 and Glu 183coordinated attacking water molecule.
Ca 2ϩ Substitution of Mg 2ϩ at the Catalytic Site-In the elongated A/M1Ј-linker mutant, Ca 2ϩ as well as Mg 2ϩ bound at the catalytic Mg 2ϩ site is able to produce E2Ca 2 ⅐BeF 3 Ϫ from E1Ca 2 via E1Ca 2 ⅐BeF 3 Ϫ . This binding of Ca 2ϩ is also found when mutant E2PCa 2 is formed from CaATP in the absence of Mg 2ϩ (14). This is in sharp contrast to the situation in the wild type, where Ca 2ϩ cannot substitute for Mg 2ϩ at the catalytic site for E1Ca 2 ⅐BeF 3 Ϫ formation. An attempt to substitute Ca 2ϩ for Mg 2ϩ actually destroys wild type E1Ca 2 ⅐BeF 3 Ϫ (27). The extremely rapid isomerization of EP with bound Ca 2ϩ at the Mg 2ϩ site in the elongated A/M1Ј-linker mutant (E1PCa 2 ⅐ Ca 3 E2PCa 2 ⅐Ca) is again very different to the markedly retarded E1PCa 2 ⅐Ca isomerization in wild type (14). The atomic structures provide insights into why elongation of the linker allows Ca 2ϩ to replace Mg 2ϩ at the catalytic site.
In the atomic structures of E1Ca 2 ⅐CaAMPPCP and E1Ca 2 ⅐ AlF 4 Ϫ ⅐ADP described by Toyoshima et al. (18,19), Mg 2ϩ or Ca 2ϩ ligation at the catalytic Mg 2ϩ site I (Asp 351 /Thr 353 /Asp 703 of the P domain and the phosphate moiety (or its analog); see  Fig. S6). This A-domain rotation raises its junctions with the A/M1Ј-linker and the A/M3-linker. The strain imposed on the A/M3-linker in E1PCa 2 probably drives the large horizontal rotation of the A domain during E1PCa 2 to E2P isomerization (18,19,50,51). In the stringent coordination chemistry, the ligation length is shorter in Mg 2ϩ than in Ca 2ϩ typically by 0.2 Å (e.g. 2.1 versus 2.3 Å (52, 53)). Therefore, Mg 2ϩ ligation probably induces more P-domain bending and in consequence more upward swinging of the A domain, leading to a stronger pull from the A/M3-linker to effect the horizontal rotation of the A domain (27). This is substantiated by the finding that in wild type, E1PCa 2 ⅐Mg 2ϩ is rapidly isomerized, whereas in E1PCa 2 ⅐ Ca it is markedly retarded (28,29).
The observed formation of E2Ca 2 ⅐BeF 3 Ϫ and E2PCa 2 (via very rapid E1PCa 2 isomerization) from mutant E1Ca 2 with Ca 2ϩ or Mg 2ϩ at the catalytic Mg 2ϩ site shows that the poor Ca 2ϩ effect on the A-domain upward rotation and subsequent horizontal rotation is relieved by elongation of the A/M1Ј-linker. Note again that the A-domain junction with the A/M1Ј-linker is raised by the upward movement of the A domain. It is, therefore, likely that in wild type, the A/M1Ј-linker is strained to some extent by this movement of the A domain on formation of E1PCa 2 . This possible strain is evidently not deleterious for wild type, but it becomes a serious energy barrier when the A/M1Јlinker is shortened by deletion of any single residue as the deletions completely block E1PCa 2 to E2PCa 2 isomerization (26). Strain in the wild type A/M1Ј-linker in E1PCa 2 is likely to be important as a build up to generating stronger strain during E1PCa 2 to E2PCa 2 isomerization. Thus, the strain of the A/M1Јlinker seems to be imposed increasingly during E1PCa 2 formation and the subsequent isomerization to E2PCa 2 , and this energy finally could be used for inducing structural changes for Ca 2ϩ deocclusion and release.
E1Ca 2 ⅐AlF x Formed from E1Ca 2 in the Elongated A/M1Јlinker Mutant-The proteolytic analyses reveal that in wild type organization of the cytoplasmic domains of the transition state analog E1Ca 2 ⅐AlF x is identical to that of E1Ca 2 ⅐AlF 4 Ϫ ⅐ADP and has obviously not yet reached the product E1PCa 2 state E1Ca 2 ⅐BeF 3 Ϫ . Namely, during the reaction E1Ca 2 ⅐AlF 4 Ϫ ⅐ADP/ E1Ca 2 ⅐AlF x 3 E1Ca 2 ⅐BeF 3 Ϫ , the A domain rotates partially in a horizontal direction and comes close to the P domain at tryptic T2 site Arg 198 but is not completely engaged, so that it cannot produce the E2Ca 2 ⅐BeF 3 Ϫ and E2⅐BeF 3 Ϫ states (Ref. 27 and see the schematic in supplemental Fig. S6). On the other hand, in the elongated A/M1Ј-linker mutant, the structure of E1Ca 2 ⅐ AlF x is intermediate between those of E1Ca 2 ⅐AlF 4 Ϫ ⅐ADP and E1Ca 2 ⅐BeF 3 Ϫ of wild type as judged from the intermediate tryptic cleavage rate at Arg 198 . Thus, elongation of the A/M1Јlinker partly relieves barriers to A-domain rotation, bringing the structure of E1Ca 2 ⅐AlF x closer to that of E1Ca 2 ⅐BeF 3 Ϫ . The finding agrees with our above postulate that the A/M1Ј-linker is strained by the A-domain upward movement during E1PCa 2 (E1Ca 2 ⅐BeF 3 Ϫ ) formation from the transition state (E1Ca 2 ⅐AlF x ). In fact, because the length of the Asp 351 O-phosphate bond in the transition state (as mimicked by AlF x ) is obviously longer than that of the covalent acylphosphate bond (as mimicked by BeF 3 Ϫ ), the transition state (AlF x ) must exhibit less P-domain bending. Ϫ (E2P ground state immediately before Ca 2ϩ binding). This is in contrast to the closed gate in E2⅐AlF 4 Ϫ and E2⅐MgF 4 2Ϫ (25). Thus, lumenal gating is strictly coupled with the configuration change in the phosphate during E2P hydrolysis, thereby avoiding possible Ca 2ϩ leakage (25). Note that in wild type, E2⅐BeF 3 Ϫ (open lumenal gate) formed with Mg 2ϩ is converted to E1Ca 2 ϩ BeF x by Ca 2ϩ , because cycle reversal and subsequent Ca 2ϩ substitution of Mg 2ϩ at the catalytic site destabilizes E1Ca 2 ⅐BeF 3 Ϫ as previously demonstrated (27). E2⅐AlF 4 Ϫ and E2⅐MgF 4 2Ϫ (gates closed) in wild type and mutant were also decomplexed to E1Ca 2 by Ca 2ϩ but probably by the high Ca 2ϩ concentration disrupting the lumenal and transmembrane regions, thereby destabilizing AlF 4 Ϫ and MgF 4 2Ϫ ligation at the catalytic site. Mg 2ϩ Dependence of E2Ca 2 ⅐BeF 3 Ϫ Formation from E1Ca 2 -The Mg 2ϩ as well as Mn 2ϩ or Ca 2ϩ dependence of E2Ca 2 ⅐BeF 3 Ϫ formation from mutant E1Ca 2 ( Fig. 5 and supplemental Figs. S1 and S2) exhibited a Hill coefficient of 2, which is in contrast to the value of 1 for wild type E1Ca 2 ⅐BeF 3 Ϫ formation from E1Ca 2 (27). The results suggest that one or more Mg 2ϩ besides the one at catalytic Mg 2ϩ site I is involved cooperatively in the E2Ca 2 ⅐BeF 3 Ϫ formation from E1Ca 2 . In the atomic structures of E1Ca 2 ⅐CaAMPPCP and E1Ca 2 ⅐AlF 4 Ϫ ⅐ADP, only one Mg 2ϩ (or Ca 2ϩ ) at site I is seen (in addition to the one coordinated with the nucleotide, which was predicted to aid phosphoryl transfer). Also, in the structures of E2⅐BeF 3 Ϫ , E2⅐AlF 4 Ϫ , and E2⅐MgF 4 2Ϫ , only one Mg 2ϩ is seen (at site I). Therefore, in E2Ca 2 ⅐BeF 3 Ϫ (E2PCa 2 ) formation a second (or more) Mg 2ϩ may possibly be required only transiently and, together with the catalytic ion, aids the motions of N, P, and A domains and their gathering during the E1PCa 2 isomerization to E2PCa 2 .
In summary, our previous (14,26) and present studies show that the A/M1Ј-linker should be appropriately long for the E1PCa 2 to E2PCa 2 isomerization then short enough for the Ca 2ϩ deocclusion/release from E2PCa 2 and again appropriately long for E2P hydrolysis. Thus, the length of the A/M1Ј-linker in wild type is naturally designed to induce successive structural changes and motions of the cytoplasmic and transmembrane domains for these processes. These functions of the A/M1Ј-linker act in concert with the changing configuration of the phosphate and catalytic Mg 2ϩ and the Asp 351 -phosphate bond length, with strength being critical in the formation of E2PCa 2 , a species poised to deliver Ca 2ϩ to the lumen. The stable analogs, E1Ca 2 ⅐BeF 3 Ϫ (27) and E2Ca 2 ⅐BeF 3 Ϫ (this study) with bound Mg 2ϩ could be critically important for obtaining atomic models of E1PCa 2 ⅐Mg 2ϩ and the hitherto elusive transient E2PCa 2 ⅐Mg 2ϩ intermediate for further understanding of the transport mechanism.