Ca2+ translocation across sarcoplasmic reticulum ATPase randomizes the two transported ions.

Cytoplasmic Ca2+ dissociation is sequential, and the Ca2+ ions bound to the nonphosphorylated ATPase are commonly represented as superimposed on each other, so that the superficial Ca2+ is freely exchangeable from the cytoplasm, whereas the deeper Ca2+ is not. Under conditions where ADP-sensitive phosphoenzyme accumulates (leaky vesicles, 5°C, pH 8, 300 mM K+), luminal Ca2+ dissociation is sequential as well, so that the representation of two superimposed Ca2+ ions still holds on the phosphoenzyme, with the superficial Ca2+ facing the lumen freely exchangeable and the deeper Ca2+ blocked by the superficial Ca2+. Under the same conditions, we have investigated whether a prebuilt Ca2+ order is maintained during membrane translocation. Starting from a prebuilt order on the cytoplasmic side, we showed that the Ca2+ ions cannot be identified after translocation to the luminal side. The same result was obtained starting from a prebuilt order on the luminal side and following the luminal to cytoplasmic translocation. We conclude that the two Ca2+ ions are mixed during ATP-induced phosphorylation as well as during ADP-induced dephosphorylation.

as two sites being sequentially accessible from the cytoplasm by the two Ca 2ϩ ions in a crevice with a deep site and a superficial site (Fig. 2). The first ion must reach the deep site to leave the superficial site vacant for the second ion to bind. The Ca 2ϩ bound to the superficial site is freely exchangeable with free Ca 2ϩ in the outer medium, whereas the Ca 2ϩ bound to the deep site is not. Inesi took advantage of the possibility of selectively placing a 40 Ca 2ϩ on top of a 45 Ca 2ϩ , to determine whether their dissociation toward the lumen is sequential. He concluded that this was the case and that the first Ca 2ϩ bound to E was the first to be internalized by monitoring the internalization of the Ca 2ϩ ions after phosphorylation. This would correspond to a channel-like structure with a first-in-first-out mechanism for membrane crossing.
The question of whether the dissociation of the Ca 2ϩ ions toward the lumen is sequential or not has been reinvestigated by Hanel and Jencks (5) and Orlowski and Champeil (6). Both groups concluded that the two ions cannot be kinetically distinguished, probably because Ca 2ϩ dissociation from the phosphorylated ATPase starts with a rate-limiting deocclusion step. More recently, Forge et al. (7) have shown that under appropriate conditions (pH 8, 300 mM K ϩ , 5°C), the two ions dissociate sequentially from the phosphorylated ATPase, suggesting that Ca 2ϩ dissociation was intrinsically sequential on both sides of the membrane.
As with the nonphosphorylated ATPase, the Ca 2ϩ sites of the phosphorylated ATPase may be represented by a crevice with a deep site and a superficial site accessible from the lumen (Fig.  3). The initial suggestion arising from these sketches is that the first ion to bind to the ATPase is the first to dissociate after transport, as originally proposed by Inesi (4). Because the conditions chosen by Forge et al. (7) have revealed the sequentiality of the luminal dissociation, they appear to be particularly appropriate for reinvestigating whether the Ca 2ϩ ion order is kept during transport. We show that there is mixing of the two Ca 2ϩ ions in the Ca 2ϩ -bound phosphoenzyme; i.e. the first Ca 2ϩ bound to the nonphosphorylated ATPase cannot be identified as being the first or the second to dissociate toward the lumen. This is also shown for the reverse step, i.e. during the ADP-induced dephosphorylation of Ca 2 E-PMg.

MATERIALS AND METHODS
SR vesicles were prepared and tested as described by Forge et al. (2). All experiments were carried out at 3°C in a cold room. The buffer was 100 mM Tes-Tris, pH 8, plus 1 or 3 mM Mg 2ϩ and 0 or 300 mM K ϩ , as specified in the figure legends. It was prepared with water filtered through a Milli-Q Water Purification System (Millipore Corp., Milford, MA). All salts were added as chlorides. Vesicles were made leaky by an incubation of at least 1 h at 2 mg/ml in 50 mM Tris, 10 mM K ϩ , 2 mM EDTA at room temperature.
Ca 2ϩ Binding and Phosphoenzyme Measurements-Ca 2ϩ binding and phosphoenzyme levels were measured by the filtration technique. Kinetic measurements involving 45 Ca 2ϩ or [␥-32 P]ATP all started with the same incubation and rinsing steps. Vesicles (0.2 mg/ml) were first * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Double Filtration System-Two rapid filtration systems from Bio-Logic (8) were associated to ensure that the phosphorylation reaction would last precisely for a short period of time and therefore be reproducible. Indeed, when manual phosphorylation is performed before kinetics perfusion by a single system, there can be 5-10 s between the beginning of the manual filtration of the phosphorylation medium and the beginning of the perfusion driven by the filtration system; for some experiments, this time lag is too long. To avoid this difficulty, we have coupled two rapid filtration systems face to face and linked them by rails on which the filter holder slides from one system to the other ( Fig.  1). At the end of the filtration by the first system, the filter holder is manually moved from the first system to the second system rapidly enough to ensure that there are Ͻ1.5 s between the end of the first perfusion and the beginning of the second.
For the experiments described here, the first filtration system was used to deliver phosphorylation substrate during 0.4 s, and the second filtration system was used to perfuse the various experimental media precisely 2 s after the phosphorylation perfusion had started. This ensured that each kinetic measurement started from the same initial state. The experiments shown in Figs. 4 and 5 were carried out with this combined system. All other steps preceding the use of the double filtration system were done as described above when using a single filtration system.

RESULTS
The Ca 2ϩ Ions Bound at the Transport Sites Are Distinguishable-As mentioned above, the two Ca 2ϩ ions bound on the cytoplasmic side of ATPase are kinetically distinguishable. This is illustrated in Fig. 2 under the conditions used below to test whether the Ca 2ϩ order is kept during transport: leaky vesicles, pH 8, 3°C. The experiment started with 45 Ca 2ϩ -bound ATPase. 45 Ca 2 E was perfused either with 1 mM EGTA, which induced rapid dissociation of both 45 Ca 2ϩ ions, or with 100 M 40 Ca 2ϩ which induced rapid dissociation of only one of the 45 Ca 2ϩ ions. This can be visualized as an isotopic exchange at the superficial site of a crevice (Fig. 2). K ϩ is known to increase the rates of Ca 2ϩ dissociation and exchange (9), so that in this experiment, it was omitted to ensure that the so-called deep 45 Ca 2ϩ ion was still blocked after a 10-s perfusion. Recently, Forge et al. (7) demonstrated similar behavior for SCHEME 1 FIG. 1. Double filtration system. F, filter holder on rails; S1, electronically controlled syringe belonging to the first rapid filtration system; S2, electronically controlled syringe belonging to the second rapid filtration system. When the filter holder is in position P1, the filter (with the previously adsorbed SR vesicles) can be perfused by the first system (for phosphorylation here). The filter holder is then pulled toward the second system. In position P2, the filter can be perfused by the second system (kinetics were obtained here by perfusing for various times with the second system). Ca 2ϩ dissociation on the luminal side, i.e. from the phosphorylated ATPase. The experiment required high K ϩ to accumulate Ca 2 E-PMg, the ADP-sensitive phosphoenzyme species (10) and low temperatures to stabilize this phosphoenzyme long enough to start the luminal dissociation perfusion before it has spontaneously turned over. Finally, according to de Meis et al. (11), pH 8 and low temperatures induce high affinity for the luminal Ca 2ϩ sites making it possible to block the deep Ca 2ϩ . 45 Ca 2 E-PMg was perfused either with Ca 2ϩ -poor buffer (see "Materials and Methods") or with 40 Ca 2ϩ . The Ca 2ϩ -poor buffer induced complete dissociation of 45 Ca 2ϩ and ATPase dephosphorylation ( Fig. 3A), whereas 10 mM 40 Ca 2ϩ induced partial dissociation of the 45 Ca 2ϩ bound to the phosphoenzyme without affecting the phosphoenzyme level (Fig. 3B). As with the cytoplasmic side, this can be visualized as an isotopic Ca 2ϩ exchange at the most protruding site on the luminal side (Fig. 3).
A Double Filtration System for Phosphorylating Having Bound the Ca 2ϩ Ions in a Specific Order-To follow either Ca 2ϩ ion during its translocation across the membrane, we had to first place a 40 Ca 2ϩ on top of a 45 Ca 2ϩ , or vice versa, on the cytoplasmic side; then phosphorylate ATPase; and finally induce luminal Ca 2ϩ dissociation from the phosphoenzyme by perfusing either 40 Ca 2ϩ or EGTA. This required that the luminal perfusion started during the first cycle of ATPase; otherwise, the prebuilt Ca 2ϩ order on the cytoplasmic side would have been lost due to turnover before the dissociation experiments had started. Therefore, starting with a 40 Ca 2ϩ on top of a 45 Ca 2ϩ , we first studied the turnover of the deep 45 Ca 2ϩ while perfusing a phosphorylating medium during various times to determine an optimal duration for phosphorylation. The level of bound 45 Ca 2ϩ slowly decreased during perfusion of the phosphorylation substrate, showing that there would be Ͻ20% loss of bound Ca 2ϩ if a luminal dissociation experiment was started 2 s after the phosphorylation perfusion had started (data not shown).
To control the decrease in bound Ca 2ϩ during preliminary phosphorylation, we had to switch from manual to electronically controlled phosphorylation. Therefore, we associated two rapid filtration systems in a so-called double filtration system, with the first system perfusing the phosphorylation medium and the second starting the luminal perfusion 2 s after phosphorylation perfusion was triggered (see "Materials and Methods"). This ensured starting the dissociation kinetics from the same initial state repeatedly. Covalent phosphorylation was checked upon phosphorylation with [␥-32 P]ATP followed by acid quenching by the second system (120 mM perchloric acid, 15 mM P i , data not shown).
We also checked that the phosphoenzyme formed within 2 s was ADP-sensitive by measuring the dephosphorylation kinetics induced by a mixture of ADP and EGTA, as shown in Fig. 4. The ratio of the initial level of bound 45 Ca 2ϩ to the initial phosphoenzyme is compatible with all ATPase being in Ca 2 E-PMg. The ADP plus EGTA mixture induced complete Ca 2ϩ dissociation and dephosphorylation confirming that the phosphoenzyme formed was entirely the Ca 2 E-PMg species. In addition, ADP-induced dephosphorylation was too fast to be accurately measured by the filtration technique, even at 3°C, in agreement with other quench flow measurements (12). Ca 2ϩ dissociation that followed ADP-induced dephosphorylation can be compared to cytoplasmic Ca 2ϩ dissociation (Fig. 2). In Fig. 4, Ca 2ϩ dissociation is faster than in Fig. 2 because of the presence of K ϩ and ADP.
The Order Created on the Cytoplasmic Side Is Lost after Translocation-Suppose that we have superimposed a 40 Ca 2ϩ on a 45 Ca 2ϩ , as in Fig. 2. If during transport, the Ca 2ϩ ions crossed the membrane in a single file, we would expect to have after membrane crossing, a 45 Ca 2ϩ at the superficial luminal site on top of a 40 Ca 2ϩ . Because this superficial 45 Ca 2ϩ is freely exchangeable with the medium, it should display the same dissociation kinetics toward the lumen, whether its dissociation is induced by EGTA or by 40 Ca 2ϩ . Conversely, if during transport the Ca 2ϩ order was inverted, the 45 Ca 2ϩ would be blocked by the 40 Ca 2ϩ at the superficial luminal site and thus perfusion with 40 Ca 2ϩ would not induce noticeable radioactive Ca 2ϩ dissociation, at variance with perfusion with EGTA. The same reasoning holds starting with a 45 Ca 2ϩ superimposed on a 40 Ca 2ϩ on the cytoplasmic side. Therefore, if during transport the Ca 2ϩ order was kept or inverted, we would expect to observe superimposed EGTA-and 40 Ca 2ϩ -induced dissociation kinetics in one of the two cases described above. We demonstrate below that this is not the case.
The 45 Ca 2ϩ dissociation experiments starting with a 40 Ca 2ϩ on a 45 Ca 2ϩ are shown in Fig. 5A and those starting with a 45 Ca 2ϩ on a 40 Ca 2ϩ in Fig. 5B. The phosphoenzyme level measured in a parallel experiment without radioactive Ca 2ϩ ions is shown in Fig. 5C.
In Fig. 5C, the initial phosphoenzyme level corresponded to full phosphorylation; it was stable during the 40 Ca 2ϩ perfusion, whereas EGTA perfusion induced dephosphorylation of ATPase. Prior to start of the kinetics or after 10 s of perfusion, ATPase was manually perfused with a mixture of ADP plus EGTA to check that there was negligible ADP-insensitive phosphoenzyme. These measurements confirmed that during both EGTA and 40 Ca 2ϩ perfusions, the phosphoenzyme was ADP sensitive thus, in the Ca 2 E-PMg form, exchanging its superficial Ca 2ϩ with that of the medium during the 40 Ca 2ϩ perfusion. Fig. 5A shows luminal Ca 2ϩ dissociation from phosphorylated ATPase, having previously bound one 40 Ca 2ϩ on top of one 45 Ca 2ϩ on the cytoplasmic side. EGTA induced complete dissociation of 45 Ca 2ϩ with a rate constant in good agreement with Fig. 3 and with corresponding dephosphorylation shown in Fig.  5C. In contrast, 40 Ca 2ϩ induced biphasic dissociation of the bound 45 Ca 2ϩ , whereas the phosphoenzyme level remained constant (Fig. 5C). 40 Ca 2ϩ -induced 45 Ca 2ϩ dissociation did not follow the first-in-first-out (FIFO) or the first-in-last-out (FILO) pattern described above. 45 Ca 2ϩ kinetics were not superimposed with the EGTA-induced dissociation, as expected if the transport mechanism had kept the Ca 2ϩ order (FIFO mechanism), nor were the kinetics stable, as expected if the transport mechanism had reversed the Ca 2ϩ order (FILO mechanism). Instead, part of the initially bound Ca 2ϩ was exchanged, as if there had been mixing during transport: part of the ATPases had lost their 45 Ca 2ϩ , as in a FIFO process and the other part had kept their 45 Ca 2ϩ , as in a FILO process. This mixing is confirmed by the symmetrical experiment described below (Fig. 5B).
The experiment was repeated (Fig. 5B), inverting 45 Ca 2ϩ and 40 Ca 2ϩ during the cytoplasmic superimposing of the two Ca 2ϩ and perfusing 45 Ca 2ϩ during phosphorylation. Again the EGTA-and 40 Ca 2ϩ -induced 45 Ca 2ϩ dissociations were not superimposed, and 45 Ca 2ϩ was not blocked by 10 mM 40 Ca 2ϩ . Again, this does not correspond to a pure FIFO or to a pure FILO mechanism. Apart from an additional stoichiometry in Fig. 5B, the experiments shown in Fig. 5, A and B show similar results, suggesting that whatever the order created on the cytoplasmic side, it is lost during Ca 2ϩ transport toward the lumen (Scheme 2).
In Fig. 5B, the initial level was higher than in Fig. 5A. It should be pointed out here that from a procedural point of view, the experiments shown in Fig. 5, A and B, are slightly different. Namely, in the experiment described in A, radioactive Ca 2ϩ was added at the beginning and rinsed twice before the luminal perfusion has started, whereas in the experiment described in B, radioactive Ca 2ϩ was added during the isotopic exchange and the phosphorylation steps, and thus there was no rinsing before the luminal perfusion started. This could induce nonspecific binding which would appear in Fig. 5B.
The Order Created on the Luminal Side Is Lost When Reversing the Cycle toward the Cytoplasmic Side-In the results presented above, the important step for Ca 2ϩ reorganization was the phosphorylation step, when the transport sites change their orientation from the cytoplasmic to the luminal side. We next examined the dephosphorylation step during the reverse cycle, when the transport sites change from luminal to cytoplasmic orientation. The experimental procedure was to pro-  Fig. 2. Phosphorylation was then induced by perfusing 100 M 40 Ca 2ϩ plus 90 M EGTA, 100 M ATP, 300 mM K ϩ with the first filtration system. In B, the prebuilt cytoplasmic order was a 45 Ca 2ϩ on top of a 40 Ca 2ϩ . Cytoplasmic order and phosphorylation were achieved as in A, except for the presence of 45 Ca 2ϩ instead of 40 Ca 2ϩ . In C, cytoplasmic sites were saturated with 40 Ca 2ϩ , and the cytoplasmic exchange was simulated with 40 Ca 2ϩ to reproduce the same protocol as in A and B. Phosphorylation was then achieved as in A, except for the presence of [␥-32 P]ATP. In A, B, and C, once the cytoplasmic order was built and 2 s after phosphorylation has started, the second filtration system perfused ATPase either with 10 mM EGTA (E, Ⅺ, q, f) or 10 mM 40 Ca 2ϩ (É, छ, ç, ࡗ) in the presence of 1 mM Mg 2ϩ and 300 mM K ϩ for various times. Immediately after the phosphorylation perfusion or immediately after 40 Ca 2ϩ or EGTA had been perfused for 10 s, the sensitivity to ADP was tested by manually perfusing a mixture of ADP and EGTA (q, ç, f, ࡗ). E, É, q, ç, bound 45 Ca 2ϩ ; Ⅺ, छ, f, ࡗ, phosphoenzyme. For all these curves the standard error of the mean is figured by the size of the symbols.
duce a specific order on the phosphorylated ATPase and then to perfuse ADP plus EGTA to dephosphorylate and dissociate Ca 2ϩ during the same perfusion, or to perfuse ADP plus 40 Ca 2ϩ to dephosphorylate and exchange Ca 2ϩ on the cytoplasmic side. We first checked that mixtures of ADP plus 40 Ca 2ϩ or ADP plus EGTA induced the same rapid dephosphorylation of Ca 2 E-PMg, too fast to be measured by the filtration technique (data not shown).
Cytoplasmic Ca 2ϩ dissociation kinetics were induced by perfusing 1.5 mM ADP plus 1.5 mM 40 Ca 2ϩ , i.e. 1 mM free ADP plus 1 mM free 40 Ca 2ϩ , or 1 mM ADP plus 1 mM EGTA (Fig. 6). When two 45 Ca 2ϩ ions were initially bound to the phosphorylated ATPase, the ADP plus EGTA mixture induced dissociation of both 45 Ca 2ϩ ions and the ADP plus 40 Ca 2ϩ mixture induced dissociation of only one 45 Ca 2ϩ ion (Fig. 6A). The difference between 40 Ca 2ϩ -or EGTA-induced Ca 2ϩ dissociations was observed whether the dissociation experiments were started from Ca 2 E-PMg in presence of ADP or from Ca 2 E (Figs. 2 and 6A).
In Fig. 6B, a specific order was first produced on the luminal side via an isotopic exchange on the phosphoenzyme, as described in Fig. 3B, i.e. a 40 Ca 2ϩ on top of a 45 Ca 2ϩ . When Ca 2ϩ dissociation was induced by EGTA after this exchange, the kinetics started from one 45 Ca 2ϩ bound per ATPase and resulted in the loss of this 45 Ca 2ϩ ion. When the ADP plus 40 Ca 2ϩ perfusion was performed, the kinetics resulted in the loss of part of this 45 Ca 2ϩ ion. Note the different scales in Fig. 6, A and B. Thus, the radioactive 45 Ca 2ϩ , which was bound at the deepest site on the luminal side was neither completely dissociated nor completely blocked after ATPase dephosphorylation. As observed for the cytoplasmic-to-luminal transport, during reverse reorientation of the transport sites leading to luminal-tocytoplasmic transport, the prebuilt Ca 2ϩ order was lost (Scheme 3).

DISCUSSION
The aim of this work was to study a particular feature of Ca 2ϩ ion transport, namely the putative FIFO mechanism. Inesi (4) showed that two different Ca 2ϩ isotopes can be superimposed in the cytoplasmic sites and Forge et al. (7), using specific conditions, demonstrated that the same superimposition was possible in the luminal sites. There is some controversy about the luminal sequentiality, because Inesi (4) found that luminal dissociation was sequential, whereas Hanel and Jencks (5) and Orlowski and Champeil (6) found that the Ca 2ϩ ions were not distinguishable during luminal dissociation. This controversy indicates that under the usual conditions the sequentiality is difficult to investigate and may depend critically on a number of factors.
Starting from this observation, Forge et al. (7) defined a set of conditions that allow accumulation of Ca 2 E-PMg, the ADPsensitive phosphoenzyme, stabilizing this phosphoenzyme long enough to start the luminal dissociation perfusion before it has significantly turned over, and finally to ensure high affinity for the luminal Ca 2ϩ sites (300 mM K ϩ , pH 8, and 5°C). These conditions were used here to identify, during its luminal dissociation, a Ca 2ϩ ion that had been specifically bound to one or the other cytoplasmic site. We interpret the data presented here as resulting from a mixing of the two Ca 2ϩ ions occurring after binding and during translocation.
Mg 2ϩ at the Catalytic Site-In 1987, Inesi (4) proposed a FIFO mechanism for Ca 2ϩ translocation on the basis of the following evidence: phosphorylation of ATPase with two bound 45 Ca 2ϩ ions induced biphasic Ca 2ϩ internalization, suggesting that the two ions dissociate sequentially, whereas phosphorylation of ATPase with a 40 Ca 2ϩ bound on top of a 45 Ca 2ϩ induced fast monophasic luminal Ca 2ϩ dissociation, suggesting that the deep 45 Ca 2ϩ was the first to dissociate. However, these experiments were conducted under conditions that probably lead to phosphorylation by both CaATP and MgATP. Because a phosphoenzyme with Ca 2ϩ at its catalytic site releases Ca 2ϩ much more slowly than a phosphoenzyme with Mg 2ϩ at its catalytic site (13), the two phases could be attributed to Ca 2ϩ dissociation from Ca 2 E-PMg (rapid phase) and to Ca 2ϩ dissociation from Ca 2 E-PCa (slow phase), rather than to sequential dissociation of the two Ca 2ϩ ions from a unique phosphoenzyme.
Being aware of this difficulty, we used the filtration technique which allows, under our conditions, rinsing the phosphorylation substrates before perfusing high concentrations of Ca 2ϩ , thus avoiding the formation of CaATP complex. In some experiments, phosphorylation was performed in the presence of 100 M Ca 2ϩ and without EGTA which leads to a MgATP: CaATP ratio of ϳ20, instead of ϳ200 when there was 90 M EGTA. We verified that there was no significant difference in Ca 2ϩ dissociation kinetics from phosphoenzymes formed with or without EGTA, suggesting that even with a ratio of 20, 5% CaATP did not significantly perturb our measurements. The possibility that a direct substitution of Mg 2ϩ at the catalytic site for Ca 2ϩ occurred during luminal perfusion of Ca 2 E-PMg with 40 Ca 2ϩ has already been envisaged and ruled out by comparing luminal Ca 2ϩ dissociation kinetics induced with various Mg 2ϩ concentrations (7).
Transport Mechanism across the Membrane-Hanel and Jencks (5) and Orlowski and Champeil (6), in studying the luminal sequentiality could not distinguish the two Ca 2ϩ ions and concluded that this was possibly due to a rate-limiting step preceding Ca 2ϩ dissociation, i.e. a conformational change leading to Ca 2ϩ deocclusion. Our particular conditions probably changed the rate-limiting step from this deocclusion step to the true Ca 2ϩ dissociation step. Because we were able under these conditions to manipulate the phosphoenzyme and therefore the luminal oriented Ca 2ϩ sites, we studied the Ca 2ϩ order during transport from the cytoplasmic to the luminal side, as well as from the luminal to the cytoplasmic side. We followed each of the two Ca 2ϩ ions selectively labeled on the cytoplasmic side and found that the EGTA-and 40 Ca 2ϩ -induced dissociation kinetics were different for each ion. Therefore, any ordered mechanism such as FIFO or FILO was ruled out. In contrast, the data favor the occurrence of ion mixing during transport. The same mixing was observed during ADP-induced dephosphorylation.
Therefore, there seems to be an intermediate and transient species of the phosphoenzyme in which the Ca 2ϩ ions are randomized, even though they are ordered on both cytoplasmic and luminal sides (Schemes 2 and 3).
Bearing in mind the common assumption that there are two Ca 2ϩ sites located in a narrow channel that change their affinity and orientation during phosphorylation or dephosphorylation, our findings suggest that the randomization of the Ca 2ϩ ions occurs during these chemical reactions and probably during the so-called occlusion step. However, this is in contradiction with the idea of a narrow channel in the phosphoenzyme, since during this step the Ca 2ϩ ions are occluded and need to exchange positions, as if their binding to the amino acid residues and the local structure were becoming loose.
Site-directed mutagenesis results show six charged residues belonging to the putative membrane helices M4, M5, M6, and M8 as critical residues for Ca 2ϩ transport (14), and among them five were critical for Ca 2ϩ occlusion (15,16). The possibility that the two Ca 2ϩ ions can be superimposed in a channel formed by M4, M5, M6, and M8 is still discussed in the literature (17)(18)(19)(20).
Recently, Mészá ros and Bak (21,22) and Jencks and coworkers (23,24) proposed that the cytoplasmic and luminal sites could be distinct; i.e. there are four Ca 2ϩ sites on the ATPase. Making this assumption and given our results, there would be during phosphorylation and dephosphorylation a transfer of the Ca 2ϩ ions from one pair of sites to the other, during which the Ca 2ϩ ions would be randomized (Scheme 4A). Martonosi (25) proposed a structural model accounting for four possible sites with two channels, including the above described channel and an additional channel, formed by M2, M3, M4, and M5. Since each Ca 2ϩ ion in this model crosses the membrane in its own channel, it does not allow the mixing observed in our results. This model to be correct should include, at least during phosphorylation and dephosphorylation, a different arrangement of the six helices mentioned above, i.e. in a tubular hexagonal structure allowing the Ca 2ϩ ions to mix (Scheme 4B). In this context, note that the representation of the two Ca 2ϩ ions as being superimposed is a choice of one possible structure (the simplest one) among others to interpret the sequential dissociation. The only requirement derived from experiment is that the Ca 2ϩ sites are not independent.
Because the two Ca 2ϩ ions are thought to be randomized during their translocation, it appears necessary to include a major structural change involving the Ca 2ϩ sites. Starting from an ordered state for the Ca 2ϩ ions, the chemical reaction (phosphorylation or dephosphorylation) must lead to another ordered state for the Ca 2ϩ ions through a transient state where the Ca 2ϩ coordinations become loose and allow randomization. Because phosphorylation and dephosphorylation reactions are very rapid, this transient state may have not been seen in structural studies (for references, see Ref. 25). SCHEME 4