Effects of Inhibitors on Luminal Opening of Ca2+ Binding Sites in an E2P-like Complex of Sarcoplasmic Reticulum Ca22+-ATPase with Be22+-fluoride*

We document here the intrinsic fluorescence and 45Ca2+ binding properties of putative “E2P-related” complexes of Ca2+-free ATPase with fluoride, formed in the presence of magnesium, aluminum, or beryllium. Intrinsic fluorescence measurements suggest that in the absence of inhibitors, the ATPase complex with beryllium fluoride (but not those with magnesium or aluminum fluoride) does constitute an appropriate analog of the “ADP-insensitive” phosphorylated form of Ca2+-ATPase, the so-called “E2P” state. 45Ca2+ binding measurements, performed in the presence of 100 mm KCl, 5 mm Mg2+, and 20% Me2SO at pH 8, demonstrate that this ATPase complex with beryllium fluoride (but again not those with magnesium or aluminum fluoride) has its Ca2+ binding sites accessible for rapid, low affinity (submillimolar) binding of Ca2+ from the luminal side of SR. In addition, we specifically demonstrate that in this E2P-like form of ATPase, the presence of thapsigargin, 2,5-di-tert-butyl-1,4-dihydroxybenzene, or cyclopiazonic acid prevents 45Ca2+ binding (i.e. presumably prevents opening of the 45Ca2+ binding sites on the SR luminal side). Since crystals of E2P-related forms of ATPase have up to now been described in the presence of thapsigargin only, these results suggest that crystallizing an inhibitor-free E2P-like form of ATPase (like its complex with beryllium fluoride) would be highly desirable, to unambiguously confirm previous predictions about the exit pathway from the ATPase transmembrane Ca2+ binding sites to the SR luminal medium.

the conclusions derived from crystals that failed to reveal transport sites open toward the lumen.
Previous experimental results do not clearly suggest one answer or another. On the one hand, proteolysis experiments previously suggested that the presence of TG affects only little the overall ATPase conformation in its cytosolic region; for instance, the almost complete protection against proteolytic cleavage afforded by binding of vanadate or metal-fluoride complexes to the ATPase was observed to occur both in the presence and absence of TG (23). Vanadate and fluoride are both thought to allow formation of complexes that are more or less closely related analogs of the ADP-insensitive phosphorylated state of the enzyme or of the transition state for dephosphorylation; they are here dubbed "E2P-related" states, in a loose sense. TG also did not greatly affect the superfluorescent state of TNP-AMP bound to the ATPase complex with beryllium fluoride (24). On the other hand, it was previously shown that TG altered the affinity of Ca 2ϩ -free ATPase for ATP (16), and it was suggested that the interaction of ATPase with thapsigargin (or BHQ) might result in formation of modified E2 species (25)(26)(27). More specifically, TG was also shown to induce pretty large structural changes in two-dimensional crystals of the E2P-related E2⅐VO 4 species (28), and TG was suspected from Trp fluorescence experiments (but with no direct evidence) to close, in E2P-related states, the postulated Ca 2ϩ release pathway from the ATPase sites toward the SR lumen (24).
Relevant to these issues, it should also be mentioned that the so-called "E2P-related" complexes of Ca 2ϩ -free ATPase with fluoride described so far did not all appear to be equally close to the "true" E2P form; all fluoride forms tested were inactive, but reactivation by exposure to a high luminal Ca 2ϩ concentration was faster for the ATPase complex with beryllium fluoride than for those with aluminum fluoride or magnesium fluoride. Thus, the ATPase complex with beryllium fluoride was suggested to have its transport sites most widely open to luminal Ca 2ϩ (24). But again, in the absence of direct 45 Ca 2ϩ binding measurements, it is not possible to exclude the possibility that faster reactivation was due to reduced stability of the ATPase-fluoride complex in the presence of beryllium, compared with the two other metal ions.
Here, to address these questions, critical ones indeed for the sensible interpretation of presently available or future x-ray structures of Ca 2ϩfree ATPase forms, we establish a robust assay of 45 Ca 2ϩ binding to the luminal sites of Ca 2ϩ -ATPase forms where these sites appear, which gives results somewhat different from those previously reported by Vieyra et al. (7). Simultaneously, we study the effect of TG and other inhibitors on this binding. The properties of Ca 2ϩ -free ATPase-fluoride complexes formed in the presence of magnesium, aluminum, or beryllium and in the absence or presence of TG or other inhibitors (namely BHQ or cyclopiazonic acid (CPA)) are documented here. Intrinsic fluorescence measurements and 45 Ca 2ϩ binding measurements both demonstrate that in the absence of inhibitors, the ATPase complex with beryllium fluoride (but not that with magnesium or aluminum fluoride) constitutes an appropriate analog of the E2P state of Ca 2ϩ -ATPase, with its Ca 2ϩ binding sites readily accessible from the SR lumen. In addition, we specifically demonstrate that in this E2P-like form of ATPase, the presence of TG, BHQ, or CPA prevents binding of 45 Ca 2ϩ (i.e. presumably prevents opening of the luminally oriented 45 Ca 2ϩ binding sites).
As an unfortunate result of our study, there is therefore little hope that crystals grown in the presence of any of these inhibitors could reveal in an open form the pathway of Ca 2ϩ ions from their binding sites (within the bundle of ATPase transmembrane segments) toward the SR lumen (where these ions are released as a result of their transport). On the more positive side, our results establish a new tool for monitoring low affinity binding of Ca 2ϩ to the luminally oriented low affinity sites of E2P-like forms of Ca 2ϩ -ATPase. They simultaneously suggest that crystallizing an inhibitor-free E2⅐BeF x complex would indeed be of utmost interest, to confirm previous predictions about the Ca 2ϩ internalization pathway from transmembrane binding sites in Ca 2ϩ -ATPase toward the SR luminal medium (13).

EXPERIMENTAL PROCEDURES
Membrane Preparation and Stock Solutions for Chemicals-SR membrane vesicles were prepared from rabbits as previously described (29), after 2-day fasting of the animals and inclusion of 1 g/ml ␣-amylase in the initial homogenization buffer (to avoid SR contamination with muscle phosphorylase; see, for example, Ref. 30 45 Ca 2ϩ binding measurements; 1 mM KF was added, either alone (i.e. simply with Mg 2ϩ ) or together with 50 M BeCl 2 or 50 M AlCl 3 (and Mg 2ϩ , present from the start). When Trp fluorescence was to be measured after the addition of fluoride, SR was present at 0.1 mg of protein/ml from the start; when 45 Ca 2ϩ binding was to be measured after preincubation, SR was generally added to the complete fluoride-containing reaction medium, at either 0.3 or 3 mg of protein/ml. To measure residual ATPase activity, aliquots were taken after preincubation for various periods and diluted into a standard ATPase assay medium containing 0.5 mg/ml octaethylene glycol monododecyl ether.
Intrinsic Fluorescence Measurements-SR membrane vesicles were suspended at 0.1 mg/ml in the thermostatted and continuously stirred cuvette of a SPEX Fluorolog fluorometer (e.g. see Ref. 31), and Trp fluorescence was monitored using an excitation wavelength of 295 nm and an emission wavelength of either 315, 330, or 355 nm. Bandwidths were 2 and 10 nm, respectively. 45 Ca 2ϩ Binding Measurements- 45 Ca 2ϩ binding to Ca 2ϩ -ATPase was measured as previously described (32), using nitrocellulose Millipore filters to adsorb the ATPase membranes and including [ 3 H]glucose in the 45 Ca 2ϩ -containing buffer to serve as a marker of the amount of fluid trapped in the filter (and therefore of the amount of nonbound 45 33). Note that "nonspecific" sites (with respect to Ca 2ϩ -ATPase), including those inside the SR compartment if the vesicles are made leaky (e.g. calsequestrin), also contribute to the measured amount of bound 45 Ca 2ϩ , especially at pH 8; this was minimized by working at 5 mM MgCl 2 and in the presence of 100 mM KCl. Experiments in which 0.3 or 0.6 mg of SR protein was to be retained by the filter were performed with Millipore HA or GS filters, respectively. In the latter case, the smaller pore diameter (0.22 m instead of 0.45 m) results in a larger surface available for adsorption of SR membranes. Filters were not rinsed, and they were counted for both 3 H and 45 Ca 2ϩ radioactivity.

RESULTS
Background-As already mentioned in the Introduction, one of the ATPase intermediate forms during its catalytic cycle must have its Ca 2ϩ binding sites opened toward the luminal side and simultaneously endowed with poor affinity for Ca 2ϩ , to permit Ca 2ϩ release into an SR lumen already containing a high concentration of free Ca 2ϩ (and therefore active transport of Ca 2ϩ ). This ATPase form is usually described as the ATPase "E2P" form, and a major aim of the present work was to measure binding of 45 Ca 2ϩ to such a form.
A first prerequisite for such measurement is availability of a membrane preparation in which the SR luminal side is made accessible to 45 Ca 2ϩ added to the outer medium. For this purpose, we used a Ca 2ϩ ionophore (calcymicin, A23187) in most of our experiments (but not in all of them, to make it possible to discriminate between binding sites accessible from the cytosolic side and binding sites accessible from the luminal side).
A second prerequisite is the availability of a tool allowing discrimination between binding of 45 Ca 2ϩ to specific sites on the ATPase and "nonspecific" binding of 45 Ca 2ϩ to other sites in the SR preparation. We chose to perform experiments in the absence or presence of TG, as TG is known, at least as concerns Ca 2ϩ binding to an ATPase initially in the E2 form, to block or dramatically slow down Ca 2ϩ binding to the ATPase high affinity cytosolic sites (15)(16)(17)(18)25).
Third, demonstrating Ca 2ϩ binding to E2P is made difficult both by the transient nature of the E2P form and by its above-mentioned poor affinity for Ca 2ϩ . Thus, we chose to increase the stability of the E2P form by including Me 2 SO in the solution (34) and to increase its affinity for Ca 2ϩ by making the pH in the solution more alkaline (35). A recent report already made use of such conditions to try and measure the binding of Ca 2ϩ to E2P (7). In our hands, however, even such conditions were not sufficient to make the experiments reliable, as judged from dephosphorylation experiments (see below).
Thus, in addition to the above general conditions, but instead of using a genuine E2P form, we chose to use more stable ATPase forms, namely ATPase complexes with fluoride, which are thought to be potential E2P analogs and which can be formed in the absence of Ca 2ϩ but in the presence of various other metals: either magnesium alone or, in addition, aluminum or beryllium (these complexes will be referred to below as E2⅐MgF 4 , E2⅐AlF 4 , and E2⅐BeF x ).
A few properties of these complexes have already been described, and two of them (E2⅐MgF 4 , E2⅐AlF 4 ) have in fact already been crystallized (13-14, 24, 36 -40). According to previous indirect evidence, it is in the fluoride complex formed in the presence of beryllium, E2⅐BeF x that the ATPase conformation is most similar to that of the E2P form and that its Ca 2ϩ sites are most accessible from the luminal medium, as judged, in particular, from the highest efficiency of luminal Ca 2ϩ for reversing the inhibition of this particular complex (24).
Trp Fluorescence Properties of Complexes of Ca 2ϩ -free ATPase with Fluoride (Especially E2⅐BeF x ) in the Absence or Presence of TG or BHQ-To start with, the conformational properties of the various ATPase complexes with fluoride were examined with intrinsic fluorescence. It had been shown previously that the drop in intrinsic fluorescence typically observed upon binding of TG to Ca 2ϩ -free ATPase (17) was significantly larger for E2P or the E2⅐BeF x forms than for either E2 or the other ATPase-fluoride forms (24). Here, this point was investigated further. Fig. 1A first confirms the finding in Ref. 38 that the addition of fluoroaluminate to Ca 2ϩ -free ATPase, leading to formation of an E2⅐AlF 4 species, only induces a slow decrease in the overall fluorescence intensity of the ATPase and the finding in Ref. 24 that the subsequent addition of TG quenches the fluorescence of this species only moderately, as initially described for E2 (17). Fig. 1B also confirms that the same addition of TG to E2⅐BeF x induces a much larger fluorescence decrease, as previously described (24).
But in addition, Fig. 1B reveals that formation of E2⅐BeF x itself results in significant Trp fluorescence enhancement, at a rate consistent with the rate of inhibition of ATPase under identical conditions (half-time of about 2 min in the absence of Me 2 SO; data not shown). After the formation of this E2⅐BeF x , complex, ATPase fluorescence was no longer sensitive to the addition of cytosolic Ca 2ϩ , as expected (see supplemental Fig. S1B). The fact that beryllium fluoride enhances ATPase fluorescence in the absence of Ca 2ϩ contrasts with the fact that in the presence of Ca 2ϩ and ADP, no such rise is detectable (see supplemental Fig. S1D), again despite the fact that aluminum fluoride, as previously described (31), now triggers formation of an E1⅐ADP⅐AlF x form of high fluorescence (supplemental Fig. S1C). The rise in fluorescence observed upon the formation of E2⅐BeF x (Fig. 1B) was only observed in the presence of . The excitation wavelength was 295 nm, and the emission wavelength was 315 nm (bandwidths were 2 and 10 nm, respectively). For the experiments in C and D, 0.4 g/ml A23187 had been added to SR before EGTA. In B, the first addition of TG saturates the ATPase (1 g/ml TG corresponds to about 1.5 M TG, whereas 0.1 mg/ml SR corresponds to 0.5-0.7 M ATPase), whereas further additions induce much smaller further quenching (see also Refs. 17 and 24). All traces were corrected for the (small) artifacts induced by dilution. Mg 2ϩ (data not shown), as expected from the Mg 2ϩ -dependence of ATPase inhibition in the presence of beryllium and fluoride (37) (data not shown). Note that a rise in ATPase intrinsic fluorescence is also observed upon formation of the E2P (29,41).
Remarkably enough, this rise in Trp fluorescence was converted into a fluorescence drop in the presence of A23187 (calcimycin), used as an hydrophobic fluorescence quencher (Fig. 1D), whereas the large drop observed upon the addition of TG was wiped off. The drop in Trp fluorescence observed in the presence of A23187 upon formation of the E2⅐BeF x complex is reminiscent of the fact that a drop in the presence of A23187 or other hydrophobic quenchers was also observed upon formation of the genuine E2P phosphoenzyme (42,43). Thus, both in the absence and presence of quencher, these fluorescence results support the previous claim by Danko et al. (24) that the E2⅐BeF x complex (but not the E2⅐AlF 4 complex) is a fair analog of genuine E2P.
We also tried to add beryllium fluoride to Ca 2ϩ -free ATPase previously cleaved at Thr 242 -Glu 243 by proteinase K, under conditions resulting in a fairly homogeneous "p28N-p83C" complex, which is no longer phophorylatable by P i although it remains phophorylatable by ATP (44). Fluorescence experiments with such a p28N-p83C complex showed that the cleaved ATPase was not able to form a high fluorescence E2⅐BeF x complex, although it did form the Ca 2 ⅐E1⅐ADP⅐AlF x complex (data not shown). This again supports a close resemblance between E2⅐BeF x and genuine E2P.
Interestingly, a second inhibitor, BHQ, also thought to react specifically with E2 forms of Ca 2ϩ -ATPase, behaves like TG as regards its effect on ATPase intrinsic fluorescence (Fig. 2). BHQ has already been shown to quench the fluorescence of the E2 form of ATPase (27). We found that, as for TG, this quenching is larger for BHQ binding to E2BeF x than for its binding to E2⅐AlF 4 (with binding to E2 being intermediate) and only minimal after the addition of BHQ to E1 forms (supplemental Fig. S2). Moreover, both for BHQ and for TG, this quenching depends on the wavelength used for observing fluorescence, being largest at the shortest emission wavelengths (Fig. 2).
It was already known that going to the blue edge of Trp fluorescence (315 nm) increases the amplitude of the fluorescence rise induced by Mg 2ϩ addition, whereas going to the red edge (355 nm) results in a signal of opposite sign for Mg 2ϩ (45), and this was confirmed here. As a corollary, we found that the E2⅐BeF x signal itself is even more prominent at 355 nm ( Fig. 2, C and F). These data therefore provide spectroscopic characterization of those Trp residues that respond to E2⅐BeF x formation or TG binding. In contrast, parallel experiments did not reveal any clear dependence on observation wavelength of the signal associated with formation of the Ca 2 ⅐E1⅐ADP⅐AlF x complex (data not shown).
Measurement of Rapid Binding of 45 Ca 2ϩ to E2⅐BeF x , an E2P-like Form of Ca 2ϩ -ATPase-We then performed 45 Ca 2ϩ binding experiments (Fig. 3A). For these experiments, Ca 2ϩ -free SR vesicles made leaky by ionophore A23187 were preincubated with fluoride and beryllium (here in the presence of 20% Me 2 SO), leading to formation of the E2⅐BeF x species (and inhibition of ATPase activity) (e.g. see supplemental Fig. S3). Control vesicles were prepared similarly but in the absence of beryllium fluoride (i.e. in the E2 form). Then, preincubated samples were loaded onto a nitrocellulose filter and immediately perfused for various periods manually, with a 45 Ca 2ϩ -containing binding buffer at pH 8 and in the continued presence of 20% Me 2 SO. The final free 45 Ca 2ϩ concentration in the perfusion buffer was 200 M, to render possible detection of 45 Ca 2ϩ binding with relatively poor affinity. Filters were counted without washing. In some cases, preincubated samples had been supplemented with TG immediately before loading onto the filter and 45 Ca 2ϩ perfusion.
Starting with measurements performed with ATPase initially in the E2 form (circles), Fig. 3A shows that the presence of TG together with E2 reduced the amount of 45 Ca 2ϩ found on the filter by about 10 nmol of 45 Ca 2ϩ /mg of protein, from 18 -19 nmol/mg in the absence of TG (open circles) to 8 -9 nmol/mg in its presence (closed circles). Since TG prevents rapid 45 Ca 2ϩ binding to ATPase in the E2 state (17), this difference (i.e. the TG-sensitive fraction of 45 Ca 2ϩ binding, 10 nmol/mg) is to be attributed to 45 Ca 2ϩ binding to the ATPase, and this is indeed the level expected for binding with a stoichiometry of two binding sites per ATPase monomer (e.g. Refs. 32 and 46). The open circles were labeled E2 in A because the ATPase initially was in its E2 form, but in this case the ATPase of course adopts its Ca 2 E1 state after Ca 2ϩ binding, with Ca 2ϩ bound to the classical high affinity sites on the ATPase cytosolic side.
The residual 8 -9 nmol/mg 45 Ca 2ϩ bound in the presence of TG (closed circles in Fig. 1A) must be attributed to binding to other sites, "nonspecific" with respect to Ca 2ϩ -ATPase, probably mostly (see below) Ca 2ϩ binding sites inside the SR lumen, like calsequestrin, made . The excitation wavelength was 295 nm, and the emission wavelength was 315 nm for A and D, 330 nm for B and E, and 355 nm for C and F. Bandwidths were 2 and 10 nm, respectively, for excitation and emission. BHQ induces an inner filter effect due to its small absorbance at 295 nm (about 0.6 -0.8% for 1 M BHQ); the apparent fluorescence quenching due to this inner filter effect is slightly more prominent at 355 nm than at 315 or 330 nm, because BHQ also emits some fluorescence at 315 or 330 nm, but very small at 355 nm, so that at 315 or 330 nm the real inner filter effect is partially compensated for by the BHQ fluorescence (at either 315 or 330 nm, about 0.3-0.5% of the corresponding fluorescence of 0.1 mg/ml SR). All traces were corrected for the (small) artifacts induced by dilution.
accessible by the presence of ionophore (note that under these conditions, binding to ionophore itself is hardly detectable), or perhaps sites on the external surface of the vesicles, which can bind Ca 2ϩ at the alkaline pH used here (pH 8). This residual binding of 45 Ca 2ϩ to permeabilized SR vesicles with ATPase in its E2⅐TG state was relatively fast, confirming that the added ionophore allows fast access to the SR lumen. This fast binding also fits with what can be expected for binding to sites of poor affinity. Since it has been suggested that TG does not fully block Ca 2ϩ binding to Ca 2ϩ -ATPase but only reduces the ATPase affinity for Ca 2ϩ (18,25), it might be argued that part of the bound 45 Ca 2ϩ measured in the presence of TG might reside on the Ca 2ϩ -ATPase. However, the dramatic slowing down, over minutes, of the kinetics of 45 Ca 2ϩ binding in the presence of TG simultaneously reported by the same authors (18,25) makes that alternative interpretation of our results unlikely, since in our experiments 45 Ca 2ϩ binding in the presence of TG was already complete after 2-3 s (more also below).
In the absence of TG, binding of 45 Ca 2ϩ to ATPase initially in its E2⅐BeF x state (Fig. 3A, open triangles) was lower than binding to E2 (open circles) but quite significant. This reduced binding is consistent with low affinity binding to E2⅐BeF x (see below). Per se, it also proves that binding of 45 Ca 2ϩ did not kick beryllium and fluoride out of the ATPase catalytic site within the time period (60 s) of these relatively rapid binding measurements, despite the fact that perfusion washed away fluoride or beryllium from the filter. This binding of 45 Ca 2ϩ to E2⅐BeF x was completed within a few seconds too, which again fits with what can be expected for binding to sites of poor affinity but disagrees with the putative very slow binding of 45 Ca 2ϩ to E2P recently suggested in Ref. 7.
Importantly, in the presence of TG, binding of 45 Ca 2ϩ to ATPase initially in its (E2⅐BeF x ϩ TG) state (Fig. 3A, closed triangles) dropped to the same low level as when ATPase was in its E2⅐TG state. This shows that the low affinity Ca 2ϩ binding sites present in E2⅐BeF x are no longer available in the presence of TG (see also subsequent figures below). Fig. 3B shows an experiment similar to that in Fig. 3A A and D, circles). At the end of this preincubation period (which, when fluoride was present, was more than sufficient to lead to complete inactivation; see supplemental Fig. S3), each of the three samples was separated in four aliquots (1 ml each, typically). Calcimycin (ionophore A23187) at 0.03 mg/ml (i.e. 1% w/w with respect to SR protein) was added to two of these aliquots, and TG (also at 0.03 mg/ml) was also added to two of the four aliquots, one with ionophore and one without (closed symbols correspond to the aliquots supplemented with TG, and open symbols correspond to control aliquots in the absence of TG). Each aliquot was then diluted 10-fold (e.g. 0.9 ml in a 9-ml final volume) into a final 45 Ca 2ϩ binding medium (0.3 mg/ml SR protein and 25 M EGTA from the preincubation medium as well as 225 M total 45 Ca 2ϩ , 225 M [ 3 H]glucose, 100 mM KCl, 5 mM Mg 2ϩ , 20% Me 2 SO, and 90 mM Tes-Tris (and 5 mM Mops-Tris from the preincubation medium) at pH 8 and 20°C), and after various periods of incubation, 2-ml subaliquots were filtered through a Millipore GS filter, and the filter was counted. A and D are labeled E2, because for these panels the ATPase initially is in the E2 form, but of course, after Ca 2ϩ binding to the ATPase high affinity sites, the ATPase is in its classical Ca 2 E1 state.
time-dependent over this 60-s period of time, which we interpret as revealing Ca 2ϩ -induced destabilization of E2⅐VO 4 forms within this period; this was confirmed by measuring the rate of Ca 2ϩ -induced recovery from ATPase inhibition under similar conditions (supplemental Fig. S4).
Binding of 45 Ca 2ϩ to ATPase initially in its E2P form (Fig. 3B, squares) almost immediately reached the same level as the one found after binding of 45 Ca 2ϩ to E2 (Fig. 3A, circles); however, classical dephosphorylation assays (not shown) showed that this was mainly due to the fact that 20% Me 2 SO (in the presence of 100 mM KCl) does not slow down E2P dephosphorylation sufficiently. Thus, under our experimental conditions, only the E2⅐BeF x form (Fig. 3A) was stable enough to make analysis of the low affinity 45 Ca 2ϩ binding curves significant.
Luminal Binding of 45 Ca 2ϩ to E2⅐BeF x and Lack of Binding to Other E2-Fluoride Forms-To discriminate between binding sites accessible from the cytosolic side and binding sites accessible from the luminal side, we repeated similar measurements in the absence or presence of the Ca 2ϩ ionophore A23187 (Fig. 4). These experiments were performed over a moderately longer time scale (up to 4 min) and with a slightly different protocol, to see how robust the previous results were. Here, intact SR vesicles, therefore mostly tight (however, see below), were first incubated (at a protein concentration 10-fold larger than previously) either in the absence of fluoride (A and D), or in its presence together with magnesium and beryllium (B and E) or together with magnesium alone (C and F); they were subsequently diluted 10-fold into a [ 3 H]sucrose-and 45 Ca 2ϩ -containing medium, and 45 Ca 2ϩ binding was allowed to occur during various periods of incubation. Some of these samples had been supplemented with ionophore and/or TG immediately before dilution into the 45 Ca 2ϩ binding medium. Only after dilution and 45 Ca 2ϩ binding were the samples loaded onto nitrocellulose filters, for subsequent double radioactivity counting of the filter without washing. The final free 45 Ca 2ϩ concentration during 45 Ca 2ϩ binding (after dilution) was kept at 200 M, taking into account the residual EGTA from the preincubation medium. Fig. 4, A-C, first shows results obtained with this new protocol when vesicles had been made leaky with ionophore, as previously; results identical to the previous ones were obtained for ATPase initially in its E2 state (Fig. 4A) or in its E2⅐BeF x state (Fig. 4B). In the absence of TG, a lower amount of bound 45 Ca 2ϩ was again found for ATPase initially in the E2⅐BeF x state (Fig. 4B, open triangles) than for ATPase initially in the E2 state (Fig. 4A, open circles), whereas in the presence of TG, the amount found was again similar in both cases (Fig. 4, A and B, closed  symbols). The TG-sensitive fraction of bound 45 Ca 2ϩ , although not very large, was definitely significant, considering the range of possible errors in the various subtractions (data corresponding to three independent experiments all fall on the same line, as will be shown in Fig. 6). The very fact that 45 Ca 2ϩ binding to E2⅐BeF x remained fairly stable again ensures that Ca 2ϩ -dependent destabilization of this fluoride form was not too much of a concern in these experiments over a few minutes at 200 M free Ca 2ϩ (see below for further discussion of the effect of Ca 2ϩ concentration on this destabilization).
When similar experiments in the presence of ionophore were repeated with ATPase initially in the E2⅐MgF 4 state (Fig. 6C), the measured amount of bound 45 Ca 2ϩ was found to be hardly larger in the absence of TG than in its presence (Fig. 6C, open and closed diamonds,  respectively). TG-resistant 45 Ca 2ϩ binding was identical to that in A and B, considered to represent binding to non-ATPase sites. These results are nicely consistent with the previous suggestion that luminal Ca 2ϩ sites are much less accessible in E2⅐MgF 4 than in E2⅐BeF x (24), and in fact they prove that the much slower recovery from inhibition observed in that paper for E2⅐MgF 4 (compared with E2⅐BeF x ) was not due to greater stability of this complex with fluoride but was indeed due to poorer exposure of its Ca 2ϩ binding sites. Fig. 4, D-F, then shows results obtained in the absence of ionophore. In the presence of TG (closed symbols), 45 Ca 2ϩ binding was in all cases much lower than in the presence of ionophore, supporting the previous interpretation that TG-resistant binding to SR vesicles represents 45 Ca 2ϩ binding to sites different from those on Ca 2ϩ -ATPase and residing mainly inside the SR lumen. In the absence of TG, 45 Ca 2ϩ binding to ATPase initially in its E2 state was again about 10 nmol/mg higher than in the presence of TG (compare open and closed circles in Fig. 4D), and this difference, the TG-sensitive fraction of bound 45 Ca 2ϩ , again represents the amount of 45 Ca 2ϩ rapidly and specifically bound to the cytosolically accessible high affinity Ca 2ϩ binding sites on the ATPase (32). 45 Ca 2ϩ binding to ATPase initially in its E2⅐MgF 4 state was again hardly different from that in the presence of TG (Fig. 4F), and the TG-sensitive fraction of 45 Ca 2ϩ bound to E2⅐BeF x (i.e. the difference between open and closed triangles) was definitely smaller in the absence of ionophore (Fig. 4E) than in its presence (Fig. 4B), demonstrating that Ca 2ϩ binds to sites on E2⅐BeF x mostly from the luminal side, as previously suggested.
It is fair to recognize that this TG-sensitive fraction of 45 Ca 2ϩ bound to E2⅐BeF x , although smaller in the absence of ionophore than in its presence, was not exactly zero in the absence of ionophore (Fig. 4E). Similarly, Ca 2ϩ was previously found to partially stimulate ATPase recovery in the absence of ionophore (24). These somewhat puzzling observations do not ruin the above claim of a luminal orientation for the Ca 2ϩ binding sites in E2⅐BeF x . We think they simply reflect the presence, in most SR vesicle preparations (including ours), of a fraction of vesicles that are not completely tight, for instance because of imperfect SR resealing during muscle homogenization (or after freezing and thawing), or because of partial denaturation of a few ATPases (opening leaks in the membrane (see, for example, Ref. 47). In fact, in the old days, selecting the subpopulation of vesicles that were completely tight and therefore could be actively loaded with Ca 2ϩ has been the purpose of a number of attempts (e.g. see Ref. 48).
Note that the amount of 45 Ca 2ϩ bound in the absence of ionophore to SR vesicles with ATPase initially in its E2 state slowly rose with time (over minutes), and this was the case both in the absence or presence of TG (open and closed circles in D) and also to some extent for fluoride complexes (Fig. 4, E and F). This slow rise most probably represents slow passive 45 Ca 2ϩ entry into the lumen of the tight vesicles (32). Conversely, the presence of a subpopulation of SR vesicles that are not tight accounts for part of the extrapolation to zero time of the amount of bound 45 Ca 2ϩ in the presence of TG and absence of ionophore (closed circles in Fig. 4D). This excludes further the hypothesis, alluded to above, that TG-resistant binding could represent binding to TG-inhibited ATPase. The similar levels found after preincubation with or without fluoride also exclude it (closed symbols in Fig. 4, A-C or D-F). Related measurements with E2⅐AlF 4 will be shown below.
Submillimolar Affinity for 45 Ca 2ϩ Binding to E2⅐BeF x and Ca 2ϩ -induced Destabilization-The reduced binding of 45 Ca 2ϩ to E2⅐BeF x compared with E2 (in Figs. 3 and 4) was interpreted above as implying poor affinity for Ca 2ϩ of the luminal binding sites in E2⅐BeF x . This is because of the result of 45 Ca 2ϩ binding experiments similar to those shown in Fig. 4 but performed at a lower free Ca 2ϩ concentration, 100 M instead of 200 M (Fig. 5); in those experiments, whereas TG-resistant 45 Ca 2ϩ binding to leaky SR vesicles was reduced to about 5 nmol/mg (closed symbols in Fig. 5, A-C), the TG-sensitive fraction of (cytosolic) binding to E2 remained 9 -10 nmol/mg (difference between open and closed circles in Fig. 5A), but the TG-sensitive fraction of (luminal) bind-ing to E2⅐BeF x dropped to about 3 nmol/mg at 100 M free Ca 2ϩ (difference between open and closed triangles in Fig. 5B), compared with 4 -5 nmol/mg at 200 M free Ca 2ϩ (Figs. 3A and 4B). The affinity for Ca 2ϩ of the luminally oriented sites in E2⅐BeF x therefore appears to be of the order of 150 -200 M. Note that in this series of experiments, we included an ATPase sample prepared as the E2⅐AlF 4 state; we found ( Fig.  5C) that 45 Ca 2ϩ binding to that state (in the presence of ionophore) was much lower than binding to E2⅐BeF x , again demonstrating directly the previous suggestion that luminal binding sites are also much less accessible to Ca 2ϩ in the transition state analog E2⅐AlF 4 than in the E2P-like E2⅐BeF x (24).
Conversely, measuring binding of 45 Ca 2ϩ to the luminal sites of E2⅐BeF x at free Ca 2ϩ concentrations higher than 200 M would be desirable to confirm the stoichiometry of 45 Ca 2ϩ binding; however, this would be made very difficult by signal/noise problems, resulting from both the increased amount of 45 Ca 2ϩ trapped in the wetting volume of the filter (which has to be subtracted) and the increased nonspecific binding of 45 Ca 2ϩ to SR internal sites (corresponding to the closed symbols in Fig. 4, D-F). Thus, the free Ca 2ϩ concentration of 200 M chosen for the experiments illustrated in Figs. 3 and 4 was a fair compromise. In addition, 45 Ca 2ϩ binding experiments performed at higher free Ca 2ϩ concentrations would probably suffer from significant Ca 2ϩ -induced destabilization of the E2⅐BeF x state, as now discussed.
The latter prediction was confirmed by ATPase activity recovery measurements, similar to those in Ref. 24 and designed to more precisely evaluate under our own exact conditions the rate of this destabilization. Interpolation of measurements, performed at two different Ca 2ϩ concentrations (100 M and 1 mM), of the Ca 2ϩ -induced recovery of E2⅐BeF x from activity inhibition (supplemental Fig. S4) showed that under our experimental conditions over a few minutes, as in Fig. 4B, destabilization by 200 M Ca 2ϩ probably affects less than 10 -20% of the ATPase E2⅐BeF x complexes, whereas going to 1 mM Ca 2ϩ would result in marked (possibly cooperative) destabilization over the same period. Supplemental Fig. S4, D-F, also shows that the presence of 20% Me 2 SO in our 45 Ca 2ϩ binding buffer was essential for making E2⅐BeF x sufficiently stable in these activity recovery experiments and therefore also in our 45 Ca 2ϩ binding experiments at 100 or 200 M free Ca 2ϩ .
Even at 100 or 200 M, there is, in fact, a slight upward drift, over minutes, in the curves for 45 Ca 2ϩ binding to E2⅐BeF x . This can be made even more apparent in longer term 45 Ca 2ϩ binding experiments. For instance, at 200 M Ca 2ϩ , experiments similar to those in Fig. 4 were conducted over up to 30 min and did show a significant time-dependent rise at the end of this period. This is shown in supplemental Fig. S5.
Note that the latter experiment was performed with spontaneously leaky membranes of deoxycholate-purified ATPase (kindly provided by Prof. J. V. Møller) instead of SR vesicles made leaky with ionophore; the smaller amount of calsequestrin in these purified ATPase membranes results in a lower level of TG-resistant 45 Ca 2ϩ binding but a similar level of TG-sensitive 45 Ca 2ϩ binding to E2⅐BeF x over the first minutes, compared with calsequestrin-containing SR made leaky with ionophore (supplemental Fig. S5), also consistent with the above interpretation of our data.
BHQ and CPA, Too, Inhibit the Luminal Binding of 45 Ca 2ϩ to E2⅐BeF x -Lastly, additional experiments were performed to test for the effect, on luminal 45 Ca 2ϩ binding to E2⅐BeF x , of additional inhibitors of Ca 2ϩ -ATPase previously also described to interact with the E2 form: BHQ and CPA. E2 and E2⅐BeF x forms were again prepared, ionophore was again added, and aliquots were then supplemented with either TG as above, CPA, or BHQ. Fig. 6 makes clear that, just as with TG, ATPase prein-  cubation with CPA (inverted open triangles) inhibited both the (cytosolic) binding of 45 Ca 2ϩ to the E2 form (Fig. 6A), as well known, and the (luminal) binding of 45 Ca 2ϩ to the E2⅐BeF x form (Fig. 6B). BHQ (squares) also did so; however, under our conditions, BHQ appeared to bind with poorer affinity than TG or CPA, since a larger molar excess of BHQ over ATPase was required for almost complete inhibition (closed versus open squares). The poor apparent affinity with which BHQ binds to the ATPase under our conditions probably is an unfavorable consequence of the presence of Me 2 SO in our buffers. 3

DISCUSSION
Opening of the ATPase Ca 2ϩ Transport Sites toward the Luminal Side of SR-In crystals of Ca 2ϩ -ATPase (SERCA1a) containing bound Ca 2ϩ ions, the binding sites for Ca 2ϩ located in the transmembrane section of SERCA1a are found shielded from the aqueous medium (8,11,12). Based on functional measurements, Ca 2ϩ ions had previously also been described as being "occluded" in some conformations of the Ca 2ϩ pump, for instance in the so-called E1P state of ATPase (e.g. see Ref. 49 for a review). However, to make active transport of Ca 2ϩ possible, these binding sites for Ca 2ϩ must of course transiently open, first toward the cytosolic medium, to take up Ca 2ϩ with high affinity (a binding generally described to be associated with the so-called E1 form of ATPase) and then toward the luminal side of the membrane, after experiencing simultaneously transition to a state where they have lost most of their affinity for Ca 2ϩ , so that Ca 2ϩ can be released into the SR lumen. Previous studies have dubbed "E2P" this putative state in which the Ca 2ϩ -ATPase, after phosphorylation, has its transport sites accessible from the luminal side. In this state, a high Ca 2ϩ on the luminal side of SR vesicles has been found to promote backwards functioning of the ATPase cycle in the direction of ATP synthesis (e.g. see Refs. 1 and 35). A high luminal Ca 2ϩ has also been found to promote destabilization of putative E2P-like intermediates, formed in the absence of Ca 2ϩ from either vanadate or fluoride (in the latter case, in the presence of various metals) (e.g. see Refs. 24, 50, and 51). We have attempted here to directly reveal these luminally oriented binding sites. 45 Ca 2ϩ Binds from the Luminal Side to SR Ca 2ϩ -ATPase in Its E2⅐BeF x Form-In the present work, we were able to reveal these luminally oriented sites in a stable ATPase complex with fluoride formed in the presence of beryllium, the "E2⅐BeF x " form. On the basis of indirect measurements, this form had previously been suggested to be the ATPase-fluoride form in which Ca 2ϩ binding sites were the most accessible to luminal Ca 2ϩ . Our Trp fluorescence results confirm its resemblance to the genuine E2P form. They also provide spectroscopic characterization for it (although alternative explanations for some of the observed changes in fluorescence are conceivable, in line with previous results; see further discussion in supplemental material and Refs. 50 and 52). In addition, we were able to demonstrate the remarkable previous suggestion that Ca 2ϩ sites in the E2⅐BeF x form are readily accessible to luminal Ca 2ϩ by directly measuring rapid 45 Ca 2ϩ binding to these sites. Conversely, we found that 45 Ca 2ϩ does not bind well to other ATPasefluoride complexes, either E2⅐AlF 4 or E2⅐MgF 4 , and these results again are the proof of the previous suggestion that opening of the Ca 2ϩ sites in these forms is restricted (24). As concerns other ATPase forms, like E2⅐VO 4 or genuine E2P, it was nevertheless difficult under our experimental conditions (20% Me 2 SO and 100 mM KCl at pH 8) to obtain clear cut evidence for (or against) luminal binding of 45 Ca 2ϩ to these forms, because of their insufficient stability (see further discussion in supplemental material; Fig. S4).
Compared with the previously reported attempt to measure 45 Ca 2ϩ binding to a genuine E2P form (7), a remarkable feature of our results is that luminal binding of 45 Ca 2ϩ to E2BeF x was fast on a time scale of seconds (Fig. 3), which contrasts with the slow (tens of minutes) rate of 45 Ca 2ϩ binding to E2P suggested by that previous report (see further discussion in supplemental material). For binding to a low affinity site, however, our fast rate of binding sounds perfectly reasonable. Nevertheless, deciding whether the luminal sites in E2BeF x remain open permanently or whether they flicker from closed states to open states but spend a significant fraction of their time open cannot be deduced from the present data (even if the latter view sounds the most likely, based on general principles of protein dynamics). 45 Ca 2ϩ Binding to Its Luminal Sites Induces Time-delayed Destabilization of the E2-Fluoride Complex-The fact that binding of 45 Ca 2ϩ to these luminal sites was completed within a few seconds (Fig. 3), whereas Ca 2ϩ -dependent recovery from inhibition occurred on a much longer time scale (supplemental Fig. S4), has mechanistic implications. First, they bear on the long debate (e.g. see Ref. 53) about whether the species previously dubbed "E2P⅐Ca 2 " may exist. We show that at least as concerns the fluoride complex E2⅐BeF x , a complex of Ca 2ϩ with an E2P-like form does exist, resulting from fast, low affinity binding of Ca 2ϩ from the luminal side. Second, this fast binding of Ca 2ϩ to E2⅐BeF x reveals that for Ca 2ϩ -induced destabilization of the fluoride complex, Ca 2ϩ must of course bind, but binding per se is not rate-limiting for Ca 2ϩ -dependent recovery from inhibition. The rate-limiting step for recovery from inhibition from fluoride is much slower than Ca 2ϩ binding. Incidentally, this is why, in contrast with direct 45 Ca 2ϩ binding measurements, the observation of a slow rate of recovery for E2⅐MgF 4 or E2⅐AlF 4 , as in Ref. 24, may suggest but does not strictly prove accessibility or nonaccessibility of the binding sites (we had a similar difficulty in interpreting the experiments with E2⅐VO 4 forms).
Presumably, Ca 2ϩ -induced destabilization of the E2⅐fluoride complex will take place because fast binding of luminal Ca 2ϩ will affect the transmembrane helices harboring the Ca 2ϩ -liganding residues, this strain will be transmitted to the catalytic site (through a transition inverse of the so-called E1P to E2P transition, whereby phosphorylation at the catalytic site somehow triggers rotation of the A domain and reorganization of the Ca 2ϩ binding transmembrane segments), and the beryllium-fluoride complex bound at the catalytic site will now have a chance to dissociate, on a much slower time scale, however.
TG and Other Inhibitors Block Accessibility of the Binding Sites to the Luminal Medium-Back to the E2⅐BeF x form, in addition to making it possible to clearly reveal binding of 45 Ca 2ϩ to the luminal side of ATPase, our use of this very stable form as a mimic of the normal E2P ATPase form provided an additional advantage; it allowed us to demonstrate the effect of TG and other inhibitors on accessibility of the binding sites to the luminal medium. In this respect, measuring 45 Ca 2ϩ binding to E2P in the presence of TG would not be as easy, because the addition of TG has been shown to reduce significantly the amount of phosphoenzyme present (21,54), and this may occur relatively rapidly, as judged from dephosphorylation measurements under our conditions (data not shown). In contrast, the stability of the ATPase-fluoride forms is sufficient to allow TG binding to occur without promoting a concomitant chase of the fluoride ligand. This is demonstrated, for instance, by the fact that although the ATPase in its E2 form is susceptible to mild proteolysis by either proteinase K (at Lys 120 ) or trypsin (at the T2 site) in the absence or presence of TG, the ATPase susceptibility is reduced to nearly zero in E2-fluoride forms both in the absence and presence of TG; this implies that TG does not release fluoride (see Table 2 in Ref. 23 and Table 1 in Ref. 24). Thus, the results in Fig. 1 also prove that TG binding to E2⅐BeF x blocks the opening of the transport sites toward the luminal side of the ATPase nearly completely. The same is of course probably true for the other fluoride forms, which already in the absence of TG do not open easily.
The fact that TG not only prevents binding of cytosolic Ca 2ϩ to E2, as already found long ago (15)(16)(17)(18)25), but also binding of luminal Ca 2ϩ to E2⅐BeF x (as found in the present report), is probably due to TG gluing transmembrane segments together (as many people in the field think, based on the structure of the E2⅐TG forms in Ref. 9 and simply on its protecting effect). In fact, accessibility of the transport sites to one side of the membrane and accessibility to the other side probably share a common prerequisite, namely the possibility for the protein to "breathe" and thereby open cytosolic or luminal "gates" between the aqueous medium and the transmembrane binding sites. These breathing movements are probably prevented by TG (21). The fact that in E2P-related forms TG stabilizes a closed conformation of Ca 2ϩ binding sites that would otherwise open toward the luminal side might be the reason why, in previous electron microscopy experiments, the luminal region of ATPase proved to be significantly different in the presence or absence of TG in two-dimensional crystals of ATPase grown in the absence of Ca 2ϩ but the presence of vanadate (28).
Implications for the ATPase Three-dimensional Structures to be Derived from Crystals-Crystallization of an enzyme is a difficult task, for which the presence of strong inhibitors (among which are transition state analogs and other inhibitors) has generally been found to be favorable, presumably because the formation of dead end complexes of the enzyme with such inhibitors slows down protein dynamics and/or selects well defined conformations. In many cases, these "frozen" conformations are a great help for understanding the catalytic cycle of the normal enzyme; in less favorable cases, certain features of the inhibitorselected conformations might be more indicative of the enzyme-inhibitor complex than of the active enzyme itself. In the case of SR Ca 2ϩ -ATPase, the small number of crystalline forms available has provided immensely valuable insight into the functioning of the catalytic cycle, by revealing details we would not have ever known without crystallography. Conditions of crystallization might have, however, prevented in certain cases the acquisition of all desirable information.
From the present results, it appears that such was the case for the two crystals of Ca 2ϩ -free ATPase-fluoride forms that have already been described, E2⅐MgF 4 and E2⅐AlF 4 (13,14), in which the Ca 2ϩ release pathway from the ATPase transport sites toward the SR lumen was found closed. This can now most likely be ascribed to the fact that first, those ATPase-fluoride forms were prepared in the presence of either aluminum fluoride or magnesium fluoride, and second, they were prepared in the presence of TG. In relation to these crystals, it would have been desirable for us to be able to repeat our 45 Ca 2ϩ binding experiments under less alkaline conditions, since the presently available Ca 2ϩfree crystals of ATPase were all prepared at slightly acidic pH. Unfortunately, the low affinity with which, a priori, Ca 2ϩ binds to E2P luminal sites at acidic pH (34,35) makes the experiment hopeless.
A likely (and unhappy) consequence of our findings is that future crystals of E2P-like forms, if they are grown in the presence of TG (as would seem reasonable to compensate for the instability (e.g. see Ref. 55) of these Ca 2ϩ -free forms in detergent), will have only very little chance of ever being able to reveal in an open state the Ca 2ϩ release pathway from the ATPase occlusion sites toward the SR lumen. Unfortunately again, this conclusion can probably be extended to crystals that might be grown in the presence of BHQ or CPA, despite the fact that these two other inhibitors of the ATPase probably bind to sites different from the one to which TG binds (22,27,56). 4 Nevertheless, to be more positive, our present results show that the E2⅐BeF x form, if it can be stabilized in the absence of such a gluing agent, will hopefully provide the possibility of revealing the open state of this Ca 2ϩ release pathway. Our Trp fluorescence measurements further substantiate the similarity between E2⅐BeF x and genuine E2P. Based on the E2⅐MgF 4 structure (13) and judging from the Trp fluorescence results, we may anticipate that in genuine E2P and E2⅐BeF x , the luminal half of the M4 helix (which harbors Trp 288 ) and M1 helix (which harbors Trp 77 ) could be more inclined (i.e. horizontal) than what was seen in E2⅐MgF 4 , bringing those Trp residues into a hydrophobic environment and forming a larger space for Ca 2ϩ to access from the luminal side.
Note, finally, that we did not obtain any evidence (see final discussion in supplemental material) for the existence of more (7,57,58) than the two classical Ca 2ϩ binding sites per ATPase monomer (although it is fair to recognize that our data cannot exclude the existence of those additional sites). These two binding sites for Ca 2ϩ , endowed with relatively poor affinity in E2P or E2⅐BeF x states, can most probably be formed by reorganization of the residues that are responsible, in other Ca 2ϩ -ATPase states, for the high affinity binding of cytosolic Ca 2ϩ . Such sites are likely to be similar to those predicted for the K ϩ -binding sites in Na ϩ ,K ϩ -ATPase (59), because the only critical difference is that Asn 796 in Ca 2ϩ -ATPase is replaced with Asp in Na ϩ ,K ϩ -ATPase.