The Cytoplasmic Loop Located between Transmembrane Segments 6 and 7 Controls Activation by Ca2+ of Sarcoplasmic Reticulum Ca2+-ATPase*

During active cation transport, sarcoplasmic reticulum Ca2+-ATPase, like other P-type ATPases, undergoes major conformational changes, some of which are dependent on Ca2+ binding to high affinity transport sites. We here report that, in addition to previously described residues of the transmembrane region (Clarke, D. M., Loo, T. W., Inesi, G., and MacLennan, D. H. (1989) Nature 339, 476–478), the region located in the cytosolic L6–7 loop connecting transmembrane segments M6 and M7 has a definite influence on the sensitivity of the Ca2+-ATPase to Ca2+, i.e. on the affinity of the ATPase for Ca2+. Cluster mutation of aspartic residues in this loop results in a strong reduction of the affinity for Ca2+, as shown by the Ca2+dependence of ATPase phosphorylation from either ATP or Pi. The reduction in Ca2+ affinity for phosphorylation from Pi is observed both at acidic and neutral pH, suggesting that these mutations interfere with binding of the first Ca2+, as proposed for some of the intramembranous residues essential for Ca2+ binding (Andersen, J. P. (1995)Biosci. Rep. 15, 243–261). Treatment of the mutated Ca2+-ATPase with proteinase K, in the absence or presence of various Ca2+ concentrations, leads to Ca2+-dependent changes in the proteolytic degradation pattern similar to those in the wild type but observed only at higher Ca2+ concentrations. This implies that these effects are not due to changes in the conformational state of Ca2+-free ATPase but that changes affecting the proteolytic digestion pattern require higher Ca2+ concentrations. We conclude that aspartic residues in the L6–7 loop might interact with Ca2+ during the initial steps of Ca2+binding.

Sarcoplasmic reticulum (SR) 1 Ca 2ϩ -ATPase belongs to the family of P-type cation-transporting ATPases that transport cations by an active mechanism involving the formation of a phosphorylated intermediate (1)(2)(3)(4). Members of this large family have been classified as type I or type II ATPases, corresponding to proteins transporting, respectively, heavy metals or cations such as H ϩ , Na ϩ , K ϩ , or Ca 2ϩ (5). Ca 2ϩ -ATPase in mature skeletal muscle contains 994 residues, as deduced from cloning and sequencing of the SERCA1a gene by MacLennan and co-workers (6). According to a topological model based on both sequence-derived predictions and experimental proteinchemical data, 70% of the polypeptide consists of two cytosolic domains connected to the membrane-embedded part by a stalk of putative helices. The membranous part comprises about 20% of the residues, arranged in 10 putative transmembrane spans, M1-10, with 10% additional residues that could form a small luminal domain. Structural three-dimensional data, obtained to date only at low resolution, have confirmed this description of the general organization of the protein (7), but the detailed structure of the Ca 2ϩ -ATPase remains an open question (8,9).
During Ca 2ϩ transport, Ca 2ϩ -ATPase probably undergoes several conformational changes, although only two main states were initially considered (1), which denote forms with a high and low affinity for Ca 2ϩ , respectively (see also Refs. 9 and 10). Binding of cytosolic Ca 2ϩ at the Ca 2ϩ activating and transport sites of the ATPase was measured directly under equilibrium conditions and found to occur with a stoichiometry of 2:1 and a positive cooperativity, which was interpreted on the basis of a sequential mechanism (11). It has been proposed to occur in a multistep process accompanied by conformational changes. Evidence has been presented that binding of only one Ca 2ϩ is sufficient to prevent phosphorylation from P i , while the binding of the second Ca 2ϩ is required to allow subsequent phosphorylation from ATP (12). Mainly based on the kinetics of Ca 2ϩ dissociation, Ca 2ϩ -binding sites have been suggested to be ar-ranged in a channel-like structure including two sites (13)(14)(15). After phosphorylation with ATP, the Ca 2ϩ ions can initially no longer dissociate from the high energy phosphoenzyme formed and are found in what has been described as an "occluded" state (16,17). The Ca 2ϩ ions are subsequently released toward the SR lumen because of reorganization of both sites and loss of their affinity for Ca 2ϩ (18 -20). It has been shown that after this stage, protons (or hydronium ions) may combine with and be countertransported by the enzyme (21,22). Chemical and/or genetic modification of the polypeptide chain has allowed us to examine the functional role of different domains (for reviews, see Refs. 5,10,23,and 24). Residues critical for Ca 2ϩ binding and transport were not found in the stalk sector, despite the fact that this region has a large content of acid residues (25). Instead, such residues are clustered in the transmembrane domain, in M4 (Glu-309), M5 (Glu-771), M6 (Asn-796, Thr-799, and Asp-800) and M8 (Glu-908) (26). Further experiments showed that the five residues in M4, M5, and M6 were also critical for occlusion in the presence of Cr-ATP (27), but not Glu-908, which was therefore considered to be involved in the initial recognition of the Ca 2ϩ ions but not in its final intramembranous binding (10,28).
In addition to the molecular biology approach, limited proteolysis of the SR Ca 2ϩ -ATPase has also been developed for identification of regions critical for Ca 2ϩ activation of the ATPase (29 -31). Experiments showed that the rate of electrophoretic migration of a small C-terminal ATPase peptide, p20C, was sensitive to Ca 2ϩ , while a slightly shorter peptide, p19C, was not. We found that p20C starts at Gly-808, at the beginning of the loop connecting putative transmembrane spans 6 and 7, while p19C starts at Asp-818, in the middle of loop. This led us to suggest that the L6 -7 loop might interact with Ca 2ϩ (32). The N-terminal part of the loop contains three aspartic residues (Asp-813, -815, and -818), which were mutated in a cluster, D813A/D818A and D813A/D815A/D818A. After expression in yeast, we found that these mutants had a low Ca 2ϩ -ATPase activity (32). It is noteworthy that in gastric H ϩ ,K ϩ -ATPase, residues Glu-837 and Asp-839, which correspond to Asp-813 and Asp-815 in Ca 2ϩ -ATPase, were found to render the ATPase unphosphorylatable by ATP when mutated to glutamine and asparagine residues, respectively (33).
In the present work, we have carried out more detailed experiments to investigate the role of the L6 -7 loop in Ca 2ϩ -ATPase. The experiments were designed to characterize the functional consequences of mutations of aspartic residues, D813A/D818A and D813A/D815A/D818A, as well as those of mutations of the proline residues in the same loop, Pro-811, -812, -820, and -821. We found that cluster mutations of aspartate residues led to a clear reduction in the apparent affinity with which Ca 2ϩ controlled ATPase phosphorylation or dephosphorylation. Although Asp to Ala mutations are often thought of as nonconservative, susceptibility to proteolytic digestion suggested that our mutation of the L6 -7 loop did not cause any major conformational change. Thus, the present findings are consistent with an interaction of the aspartic residues of the L6 -7 loop with the first Ca 2ϩ ion that binds to Ca 2ϩ -ATPase during the transport process and suggest that the L6 -7 loop contributes to the control of the activation of Ca 2ϩ -ATPase by Ca 2ϩ during the initial steps of Ca 2ϩ binding. A model is proposed describing one possible mechanism by which such a control might occur.

Mutation and Expression of Ca 2ϩ -ATPase in Yeast-
The single mutants E309Q and E771Q and the cluster mutants D813A/D818A (referred to later as ADA) and D813A/D815A/D818A (referred to later as AAA) were obtained as in Ref. 32. Using Ca 2ϩ -ATPase SERCA1a cDNA (34), single mutants P812A and P821A and cluster mutants P811A/ P812A and P820A/P821A were obtained by site-directed mutagenesis performed with the pAlter kit (Promega). The presence of the desired mutations and the absence of unexpected ones resulting from the polymerase chain reaction were verified by DNA sequencing. Wild type and mutant DNA were inserted into a yeast expression vector and expressed as described previously (32). A light membrane fraction obtained by differential centrifugation was prepared for each mutated ATPase and stored at 10 mg of proteins/ml in 10 mM Hepes, pH 7.4 (Tris), 0.3 M sucrose, and 0.1 mM CaCl 2 , as described in Ref. 32. Protein concentrations were measured by the bicinchoninic assay (35) in the presence of 0.5% SDS. The amount of Ca 2ϩ -ATPase was quantified by Western blotting as in Ref. 34, using the polyclonal antibody 577-588 (31) and a GS-700 imaging densitometer with Molecular Analyst software (Bio-Rad); for reference, we used native SR membranes in which 75% of the protein content is assumed to be Ca 2ϩ -ATPase. Typically, light membranes expressed wild type or mutant Ca 2ϩ -ATPase at about 0.75 mg of Ca 2ϩ -ATPase/100 mg of membrane proteins.
ATPase Assay-ATP hydrolysis was assayed spectrophotometrically at 30°C as in Ref. 32, generally with 50 g of yeast light membranes/ml in the assay and with a final Ca 2ϩ concentration of 0.1 mM. The reaction was started by the addition of 1 mM Na 2 ATP. Thapsigargin (1 g/ml; see Ref. 41) was added after 100 s, and the rate of hydrolysis was followed for an additional 100 s. The Ca 2ϩ -ATPase activity was calculated as the difference between the slopes obtained in the presence and in the absence of thapsigargin and was corrected for the (very small) background of thapsigargin-dependent activity obtained with light membranes of control yeasts. with various amounts of EGTA) were calculated using Maxchelator software, taking into account the endogenous Ca 2ϩ . Five mM NaN 3 , 0.05 g/ml bafilomycine A, and 0.1 mM ammonium molybdate were also included in the assay medium to inhibit other ATPases. Fifty g of light yeast membrane proteins were added to a final volume of 50 l. The reaction was started by adding 2 M [␥-32 P]ATP (1 Ci/ mmol) and quenched after 15 s by adding 2 ml of cold quenching solution containing 150 mM perchloric acid and 15 mM NaH 2 PO 4 . Aciddenatured proteins were retained on a glass fiber filter (Gelman, A/E type) and washed five times with 4 ml of quenching solution (36). The radioactivity in the filter was then counted in 4 ml of scintillation liquid (Packard, Filter-Count). A filter containing 100 pmol of [␥-32 P]ATP was used as a standard, to convert measured cpm to pmol of 32 P.
Ca 2ϩ Dependence of Ca 2ϩ -ATPase Phosphorylation from [␥-32 P]ATP, as Measured after Electrophoretic Separation-Ca 2ϩ -ATPase phosphorylation was carried out as in the case of the filtration experiments except that proteins were diluted 10 times more (i.e. into a final volume of 500 l with the same buffer), and the various free Ca 2ϩ concentrations were established more reliably, by mixing Ca 2ϩ with EGTA and MgEDTA (37). In this case, the reaction was stopped by adding 1 ml of a cold solution containing 9% trichloroacetic acid and 27 mM NaH 2 PO 4 . Acid-denatured proteins were treated according to Ref. 38 and sedimented by centrifugation for 15 min at 18,000 rpm (25,000 ϫ g av ) at 4°C. The pellet was neutralized by the addition of 1 l of 1M Tris-base and then suspended in 50 l of 150 mM Tris-Cl, pH 6.8, 10 mM EDTA, 2% SDS, 16% glycerol (v/v), 0.04% bromphenol blue (v/v), and 0.84 M ␤-mercaptoethanol. After 10 min at room temperature, proteins corresponding to 150 ng of Ca 2ϩ -ATPase were loaded on 7% polyacrylamide gels and run for 90 min at 120 V in 170 mM MOPS-Tris, pH 6.0, 0.1% SDS. After electrophoresis, the gels were fixed in 45% methanol and 10% acetic acid for 25 min and then dried overnight between two sheets of cellophane paper. Radioactivity was revealed with a PhosphorImager (Molecular Dynamics, Inc.) and a Biomax film (Amersham Pharmacia Biotech) and quantified by comparison with known amounts of [␥-32 P]ATP.
Ca 2ϩ Dependence of Ca 2ϩ -ATPase Phosphorylation from [ 32 P]P i -Phosphorylation from P i was assayed both at pH 6.0 and at pH 7.0. The reaction was carried out at 20°C in a total volume of 500 l of 100 mM MES-Tris, pH 6.0, or 100 mM MOPS-Tris, pH 7.0, 5 mM MgCl 2 , 20% dimethyl sulfoxide, 5 mM NaN 3 , 0.05 g/ml bafilomycine A, and 0.1 mM ammonium molybdate, also containing various mixtures of Ca 2ϩ and EGTA or MgEDTA to obtain the free Ca 2ϩ concentrations indicated in Figs. 5 and 6. The reaction was started by adding 50 g of light membranes (corresponding to 375 ng of Ca 2ϩ -ATPase) and 0.1 mM [ 32 P]P i and stopped after 15 s by the addition of 2 ml of cold quenching medium (9% trichloroacetic acid, 27 mM NaH 2 PO 4 ). Quantification of phosphoenzyme formed was carried out by the electrophoretic method as described above for phosphorylation from [␥-32 P]ATP.
Proteolysis, Electrophoresis, and Blotting-Proteinase K digestion of wild type and mutated Ca 2ϩ -ATPases was carried out in 100 mM Tes-Tris, pH 7.0, 5 mM MgCl 2 , at various free Ca 2ϩ concentrations. For each membrane sample, containing either wild type or mutated ATPase, 20 g of proteins were rapidly thawed and diluted in 200 l of proteolysis buffer. Proteolysis was started by the addition of 0.36 g of proteinase K and was performed for 30 min at 20°C. The reaction was stopped by the addition of 1 mM phenylmethylsulfonyl fluoride, and samples were left on ice for 10 min and then centrifuged at 75,000 rpm (200,000 ϫ g av ) for 60 min at 4°C in a TLA100 rotor and a Beckman TL100 ultracentrifuge. The pellet was suspended in 42 l of urea-SDS denaturing buffer (39) and heated for 1 min at 100°C, and aliquots corresponding to 150 ng of Ca 2ϩ -ATPase were loaded on a 11.4% polyacrylamide gel containing 1 mM Ca 2ϩ . After electrophoresis, proteins were transferred to PVDF membranes for immunodetection (31). Peptides were first immunodetected with polyclonal 79B Ab and then revealed by an ECL kit (Amersham Pharmacia Biotech). Polyclonal 79B Ab is a kind gift of A-M. Lompré (40); it reacts with a main epitope located in the Nterminal portion of the Ca 2ϩ -ATPase, as previously discussed for 78(7) Ab in Ref. 31. The membrane was stripped, and peptides were then immunodetected with 577-588 Ab (31).

Choice of Mutations in the L6 -7 Loop-
The primary structure of the L6 -7 loop is shown in Fig. 1. The N-terminal region of the loop, which is highly conserved among SERCA ATPases, presents two remarkable features; it contains three aspartic residues, Asp-813, Asp-815, and Asp-818 (note that conservative replacements with Glu or Asn residues are found for Asp-818 among SERCA ATPases) (triangles in Fig. 1), and these residues are located between two couples of proline residues, Pro-811-Pro-812 and Pro-820-Pro-821 (squares in Fig. 1). We initially wondered whether the acidic residues could be part of an initial Ca 2ϩ -binding site, located at the membrane-cytosol interface and structured by the proline residues. Therefore, these residues were mutated to alanine residues to give the following mutants: D813A/D818A (referred to as ADA), D813A/ D815A/D818A (referred to as AAA), P811A/P812A, P812A, P820A/P821A, and P821A. Cluster replacement of the negatively charged Asp residues by the small neutral Ala residues was performed in preference to the a priori more conservative Asp 3 Asn or Asp 3 Glu substitutions with the aim of suppressing all potential calcium-liganding oxygen atoms in the side chain. A few single mutations (D813A and D815A) were also tested (see "Ca 2ϩ -ATPase Activities of Mutants").
All mutants were prepared by site-directed mutagenesis and expressed in yeast as described under "Materials and Methods." The results presented below were obtained using a yeast light membrane fraction in which wild type and each mutated Ca 2ϩ -ATPases were expressed at about 0.75% (w/w) of the membrane proteins. No significant difference in expression level for the various mutants was observed (data not shown).
Ca 2ϩ -ATPase Activities of Mutants- Fig. 2 shows the overall ATPase activity of native, wild type, and mutant enzymes in the presence of Ca 2ϩ . Examples of typical traces are displayed in Fig. 2A, which shows ATP hydrolysis as measured by the rate of NADH oxidation recorded at 340 nm. In the presence of

FIG. 2. ATPase activity of wild type and mutated ATPases.
Ca 2ϩ -dependent ATP hydrolysis was assayed as described under "Materials and Methods." The results for single or cluster mutants of the L6 -7 loop (P812A, P811A/P812A, P821A, P820A/P821A, ADA, and AAA) were compared with those for the WT or native (SR) ATPase and also with those for two previously described mutants, E309Q and E771Q (10,26). The assay with control membranes refers to light membranes of yeast transformed with the expression vector alone. A, typical traces for ATPase activity assay. The activity of the SERCA enzyme was estimated from the difference in slope before and after the addition of the specific inhibitor thapsigargin (Tg). B, specific activities of mutant and wild type Ca 2ϩ -ATPases are indicated with S.D. values (3-5 independent experiments).
several ATPases inhibitors, even control membranes displayed measurable ATPase activity. However, this activity was virtually insensitive to thapsigargin, a specific inhibitor of SERCA ATPases (41), while the additional ATPase activity of expressed wild type Ca 2ϩ -ATPase (WT assay), or that of an equivalent amount of native SR Ca 2ϩ -ATPase added to control membranes (SR ϩ control membrane assay) was specifically inhibited by thapsigargin.
Specific Ca 2ϩ -ATPase activities were estimated for each mutant, after immunoquantification of the Ca 2ϩ -ATPase. The result is shown in Fig. 2B. As previously reported (42), a change of Pro-812 to alanine did not significantly affect the hydrolytic activity of Ca 2ϩ -ATPase, while cluster mutation P811A/P812A reduced this activity to about 50%. More pronounced effects were observed by changing the second couple of proline to alanine residues; the single mutation P821A reduced the ATPase activity by 50%, while mutation to alanine of the two proline residues 820 and 821 resulted in a Ca 2ϩ -ATPase with only 10% activity. Mutation to alanine of aspartic residues in loop 6 -7 (ADA and AAA mutants) also reduced the ATPase activity to about 10% or less, as previously reported (32), confirming that these aspartic residues are important. Mutants E309Q and E771Q, previously shown to be defective in Ca 2ϩdependent control (26,43), were also prepared as negative controls and found to be virtually devoid of ATPase activity.
Ca 2ϩ Dependence of Phosphoenzyme Formation-Mutants were then tested for their ability to become phosphorylated from [␥-32 P]ATP, a reaction that requires binding of two Ca 2ϩ ions to the protein. These experiments were first carried out by using a method based on filtration of perchloric acid-precipitated protein (Fig. 3) and were subsequently confirmed using a method based on electrophoretic separation of the phosphorylated intermediate (Fig. 4). Fig. 3 shows the amount of protein phosphorylated from 2 M [␥-32 P]ATP at various Ca 2ϩ concentrations. A basal phosphorylation level of about 0.5 pmol/50 g of protein was observed in all assays, including membranes of yeasts that did not express Ca 2ϩ -ATPase. This is attributable to phosphorylation of proteins other than Ca 2ϩ -ATPase in the membrane fraction. For WT Ca 2ϩ -ATPase, phosphorylation was distinctly dependent on pCa. The maximal level of Ca 2ϩ -dependent phosphorylation was about 1.5 pmol of phosphoenzyme/50 g of total proteins, obtained for saturating Ca 2ϩ concentrations (i.e. about 4 nmol of phosphoenzyme/mg of Ca 2ϩ -ATPase, corresponding to 0.45 mol of EP/mol Ca 2ϩ -ATPase, a reasonable value compared with results previously reported for intact SR (e.g. Refs. 11 and 16)). Activation by Ca 2ϩ of phosphoenzyme formation occurred in two steps. The first step, in the micromolar range (10 Ϫ7 Ϫ 10 Ϫ4 M), corresponds to the high affinity binding of Ca 2ϩ to the ATPase. The subsequent increase in phosphorylation, observed in the millimolar range (10 Ϫ3 Ϫ 10 Ϫ2 M) of Ca 2ϩ concentrations, is probably due to the stabilization of phosphoenzyme arising from replacement of Mg 2ϩ at the hydrolytic site by Ca 2ϩ (see Refs. 45-48 and "Discussion").
Similar results were obtained when phosphoenzyme formation was measured after electrophoretic separation, as shown in Fig. 4. In that case, phosphorylated proteins contributing to the basal phosphorylation level were separated from phosphorylated Ca 2ϩ -ATPase. The autoradiograms displayed in Fig. 4A allow us to visualize phosphorylated wild type Ca 2ϩ -ATPase as a band migrating with an apparent molecular mass of about 100 kDa, like the native Ca 2ϩ -ATPase in SR (for reference, SR was added to control membranes at pCa 4; see right lanes in panel A); the amount of radioactivity in this region was very low at pCa 7, but it became prominent at higher free Ca 2ϩ concentrations. Due to phosphoenzyme hydrolysis during electrophoretic separation, the maximal level of phosphorylated protein was lower than that measured in the filtration experiments, but radioactivity was retained to a significant extent, with a maximal value of 0.14 pmol of 32 P/150 ng of Ca 2ϩ -ATPase, i.e. close to 25% of the value before electrophoretic separation.
Similar experiments were performed with the Ca 2ϩ -ATPase mutants. Figs. 3A and 4, A and B, show that the ADA and AAA mutants required much higher Ca 2ϩ concentrations for phosphorylation than WT ATPase; at 10 mM Ca 2ϩ , this resulted in full phosphorylation of the ADA mutant and partial phosphorylation of the AAA mutant. From these data, the apparent Ca 2ϩ affinity for Ca 2ϩ activation of the ADA mutant Ca 2ϩ -ATPase was estimated to be 2-5 mM. In contrast, proline mutants did not reveal any change in the apparent affinity for Ca 2ϩ deduced from phosphoenzyme formation from [␥-32 P]ATP. This is shown for P811A/P812A and P820A/P821A in Fig. 3B, and for P821A in Figs. 3B and 4C. However, the maximum level of phosphoenzyme formed was significantly reduced for the cluster mutant P811A/P812A and even more for P820A/P821A. In the latter case, the effect of the mutation seems to be mainly related to the P821A mutation, since the residual Ca 2ϩ -dependent EP level was similar for P821A and for P820A/P821A. For the P812A mutant, phosphoenzyme formation was similar to that of WT ATPase at all pCa values (not shown). As a control for these measurements, phosphorylation from [␥-32 P]ATP of the E309Q mutant was also measured at pCa 4; as described in Ref. 26, no phosphorylation was observed (Fig. 4C).
Inhibition by Ca 2ϩ of Phosphorylation from [ 32 P]P i -The filtration method used in Fig. 3 could not be used to measure phosphorylation from P i , due to the fact that the P i concentration that had to be used (100 M) resulted in a much larger background than in the experiments with [␥-32 P]ATP (2 M ATP). Thus, phosphorylation from P i was quantified by our second method only, i.e. after separation on SDS-PAGE of the phosphorylated Ca 2ϩ -ATPase. The experiments were carried out both at pH 6.0 and 7.0, under conditions previously described (26 -28, 49). Fig. 5A shows examples of autoradiograms from dried gels after electrophoresis at pH 6.0. No signal was observed with control membranes. Panel B shows that the expressed wild type Ca 2ϩ -ATPase (open circles) was phosphorylated from P i in the absence of Ca 2ϩ to the same extent as the equivalent amount (as deduced from Western blotting) of Ca 2ϩ -ATPase in native SR (closed circles); the maximal level of phosphorylation from P i , obtained at low Ca 2ϩ concentration, was in the same range as the maximal level of phosphorylation from ATP (see Fig. 4). As expected, the level of phosphorylated protein formed decreased as the Ca 2ϩ -free concentration increased; Ca 2ϩ inhibited phosphorylation from P i with an apparent Ca 2ϩ affinity of 40 M under these conditions (pH 6.0, 20% Me 2 SO). For the ADA mutant (open triangles in panel B), the maximal extent of phosphorylation from P i observed in the absence of Ca 2ϩ was the same as that of the WT, but phosphorylation was much less sensitive to the presence of Ca 2ϩ , since the addition of 100 M free Ca 2ϩ was not sufficient to produce any inhibition; the K 0.5 value for Ca 2ϩ inhibition of phosphorylation from P i of the ADA mutant was estimated to be around 4 mM. In contrast to these results, Fig. 5C shows that in the absence of Ca 2ϩ , the proline mutants P821A and P820A/P821A were phosphorylated to only 25% of the WT level, while the apparent affinity of these mutants for Ca 2ϩ appeared not to be altered, as previously deduced also from the measurements of phosphorylation from ATP in Figs. 3B and 4C.
Inhibition by Ca 2ϩ of Ca 2ϩ -ATPase phosphorylation from P i was also measured at pH 7.0, and the result is shown in Fig. 6. Fig. 6A displays results obtained for wild type and ADA mutated Ca 2ϩ -ATPases. It clearly appears that at pH 7.0 the curve for the ADA mutant is again strongly shifted toward high Ca 2ϩ concentrations, as compared with that of wild type Ca 2ϩ -ATPase. For comparison, Fig. 6B shows the result of similar experiments performed with mutants E309Q and E771Q, previously characterized under the same conditions (27,28). The E771Q mutant has a behavior that, from a phenomenological point of view, is similar to that of the ADA mutant, while the E309Q mutant, as previously found, has an apparent affinity for Ca 2ϩ close to that of WT ATPase.
Ca 2ϩ -dependent Pattern of Proteolysis by Proteinase K of Wild Type and Mutated Ca 2ϩ -ATPases-The above described phosphorylation experiments were supplemented with a completely different type of experiments, in which the ability of various mutants to bind Ca 2ϩ was deduced from the Ca 2ϩ dependence of their proteolysis pattern. Wild type Ca 2ϩ -ATPase as well as E309Q, E771Q, and ADA mutants were submitted to proteolytic attack by proteinase K in the presence of various Ca 2ϩ concentrations at neutral pH. The fragments were separated by SDS-PAGE, followed by Western blot. Among the fragments produced, peptides p95 and p83C (see their locations in Fig. 7D) were recognized by immunodetection with 577-588 Ab (Fig. 7A), and peptide p28N was recognized with 79B Ab (Fig. 7B). As illustrated in Fig. 7D and found previously for native Ca 2ϩ -ATPase treated under similar conditions (31), p95 is produced by a proteolytic cleavage between residues Leu-119 and Lys-120, while p28N and p83C are both produced by a proteolytic cleavage between residues Thr-242 and Glu-243.
In Fig. 7A, the left lanes corresponding to proteinase K treatment of WT ATPase show that, at pCa 7, more than 50% of the Ca 2ϩ -ATPase was proteolyzed, mainly leading to production of p95 peptide as well as of a small amount of p83C (corresponding to 10% of starting material). Thus, the cleavage sites at Leu-119-Lys-120 and Thr-242-Glu-243 were both accessible, although cleavage at the former site was predominant. The presence of Ca 2ϩ (lanes pCa 5.8 and 4) resulted in cleavage of about 70% of the Ca 2ϩ -ATPase, with a modified cleavage pattern where p95 was no longer present, while p83C and p28N both accumulated to a higher extent. Open circles in Fig. 7C illustrate the Ca 2ϩ dependence of p28N formation in WT ATPase.
For the ADA mutated Ca 2ϩ -ATPase (third group of lanes from the left in Fig. 7, panels A and B), proteinase K treatment at pCa 7 produced the same amount of p95 as for the wild type. However, at pCa 5.8, p95 was still detectable, while p83C and p28N were present to a lesser extent than in the wild type. At pCa 4, the amount of p28N formed remained lower than that of the wild type (see Fig. 7C, open triangles compared with open circles), while p95 disappeared completely and p83C started to accumulate. Thus, the apparent affinity with which Ca 2ϩ modified the proteolysis pattern of the ADA mutant was reduced compared with that for WT ATPase, in agreement with the phosphorylation results in Figs. 3-7. Nevertheless, since in the absence of Ca 2ϩ the ADA mutant has the same pattern of proteolysis as WT ATPase, the conformational state of the Ca 2ϩ -free form of this mutated ATPase seems to be similar to that of WT ATPase.
The other groups of lanes in Fig. 7, A and B, show the proteolysis pattern for the E309Q and E771Q mutants. The behavior of the E309Q mutant was remarkable; at pCa 7, p95 formation was not observed at all, while distinct amounts of p28N (see Fig. 7C, closed squares) and p83C were present, suggesting a change in the proteolytic cleavage pattern in the absence of Ca 2ϩ . At higher Ca 2ϩ concentrations, still larger amounts of p83C and p28N were present; as shown in Fig. 7C, the amount of p28N (closed squares) at pCa 4 was as high as that of WT (open circles). This indicates that the E309Q mutant has definitely retained sensitivity to the presence of Ca 2ϩ , in agreement with the phosphorylation results in Fig. 6B, but that its conformation in the Ca 2ϩ -free state differs from that of WT ATPase. The situation was also remarkable with the E771Q mutated ATPase, which was more strongly proteolyzed than WT ATPase and the other mutants. At pCa 7, p28N was again present, in an amount equivalent to that for E309Q (open diamonds in Fig. 7C compared with closed squares). At higher pCa, p28N did not accumulate (open diamonds in Fig. 7); this was also the case for p83C, which even at higher Ca 2ϩ levels was only present in a low amount. Again in agreement with the results of the phosphorylation experiments, this suggests that the mutated E771Q Ca 2ϩ -ATPase is indeed only very poorly sensitive to the presence of Ca 2ϩ (although fragments may be degraded too quickly under the conditions tested to detect a difference). In addition, for E771Q as for E309Q, the conformational state of the Ca 2ϩ -free ATPase differs from that of WT ATPase.

DISCUSSION
Functional Properties of the L6 -7 Mutants-The N-terminal side of the L6 -7 loop in Ca 2ϩ -ATPase (see Fig. 1) is characterized by the presence of three aspartic residues, surrounded by two couples of proline residues. An important outcome of our investigation is that cluster mutations of the aspartic residues shifted the Ca 2ϩ -ATPase affinity for Ca 2ϩ to much higher concentrations than those required to activate wild type ATPase; the K 0.5 for Ca 2ϩ activation of phosphorylation from ATP of the D813A/D818A mutated ATPase was shifted from the micromolar to the millimolar range, while the maximal level of phos- phorylation remained the same as that of wild type ATPase (Figs. 3 and 4). It can therefore be concluded that the ADA mutant has an intact phosphorylation site and can still bind two Ca 2ϩ ions, although at higher Ca 2ϩ concentrations than WT ATPase. In agreement with this, the double and triple mutants D813A/D818A and D813A/D815A/D818A only displayed a very low ATPase activity when they were tested at pCa 4, (Fig. 2). Note that the single mutant D815A gave a fully active ATPase, while mutation of D813A alone resulted in only a moderate loss of activity of 56%. 2 An unexpected feature of our phosphorylation experiments was that Ca 2ϩ stimulation of phosphoenzyme formation from ATP by WT ATPase revealed two steps, separated by a plateau around pCa 5.5-3.5 (Figs. 3A and 4B). However, a comparable behavior has been observed previously for purified ATPase or leaky SR vesicles under similar conditions (see, for example, Fig. 7 in Ref. 45). Stimulation of phosphorylation by micromolar Ca 2ϩ is caused by binding of Ca 2ϩ to the high affinity transport sites, but saturation of these sites is not necessarily accompanied by complete phosphorylation, since the latter is dependent on the balance between the overall rates of phosphorylation and dephosphorylation. Increasing the Ca 2ϩ concentration to millimolar values slows down the rate at which phosphoenzyme is hydrolyzed, presumably because of substitution of Ca 2ϩ for Mg 2ϩ at the catalytic site, and thereby increases the steady state level of phosphorylation (see Refs. [45][46][47][48]. Conceivably, for the ADA mutant, the same substitution of Ca 2ϩ for Mg 2ϩ could also contribute to the rise of the steady state EP level at millimolar Ca 2ϩ concentrations. Consequently, the "apparent" affinity for Ca 2ϩ of the ADA mutant deduced from the phosphorylation data probably represents a "mixed" affinity, partly reflecting the usual exchange of Mg 2ϩ for Ca 2ϩ at the ATPase catalytic site and partly reflecting the binding of Ca 2ϩ at the transport sites with reduced affinity. Mutation of both couples of proline residues to alanine affected the Ca 2ϩ -ATPase in different ways. For the P811A/ P812A double mutant, the resulting Ca 2ϩ -ATPase activity (Fig.  2) and the ability to become phosphorylated from ATP (Fig. 3B) were not drastically affected, in good agreement with similar data previously reported for proline residue 812 (42). This suggests that the proline residues of the first doublet are not essential for Ca 2ϩ -ATPase function. On the other hand, proline 2 T. Menguy, unpublished results. residues 820 -821 seem to play a more important role; we found that the Ca 2ϩ -ATPase activity at pCa 4 of the corresponding double alanine mutant was only 10% that of wild type ATPase (Fig. 2B), and maximal phosphorylation from ATP and P i was low. Comparable effects were obtained after a single mutation of Pro-821 to alanine, suggesting that this residue is more critical than Pro-820. Note that Pro-821 is well conserved in all transport ATPases (5). Nevertheless, a clear difference from the effect of mutation of the aspartic residues in the L6 -7 loop is that after mutations of proline residues the apparent affinity of the ATPase for Ca 2ϩ was not modified, as judged from the Ca 2ϩ -dependence of ATPase phosphorylation from either ATP or P i (Figs. 3-5). This shows that the proline residues of loop L6 -7 do not participate in the formation of the Ca 2ϩ -binding site, even indirectly by structuring it. One possible explanation for the effects observed on V max and EP max is that the mutation of Pro-821 stabilizes the nonphosphorylated E 1 species of the ATPase and that the integrity of this residue may be required to overcome rate-limiting step(s) in the cycle, associated with the E 2 /E 1 conformational conversion.
The Relationship between Phosphorylation and Ca 2ϩ Binding-Because of the above mentioned biphasic Ca 2ϩ dependence of ATPase phosphorylation from ATP, the complementary experiments in which we tested the inhibition by Ca 2ϩ of P iderived phosphorylation are of critical importance. In agreement with the low affinity for Ca 2ϩ of the ADA mutant suggested by the ATP-derived phosphorylation experiments, we observed that in this mutant P i -derived phosphorylation was also less sensitive to Ca 2ϩ inhibition than in WT ATPase, both under acidic and under neutral conditions (Figs. 5A and 6A). Some of the "classical" mutations have been reported to make the ATPase less sensitive to Ca 2ϩ irrespective of pH in P iderived phosphorylation experiments (e.g. the well known E771Q, T799A, D800N, and E908A mutations), while for another subset of the "classical" mutants (e.g. E309Q and N796A) Ca 2ϩ inhibits P i -derived phosphorylation at pH 7 with an affinity that is not very different from that of wild type ATPase (10,28,50) (it should be mentioned that in acid media, the behavior of the N796A mutant becomes different from that of the E309Q mutant (26,51). Our results in Fig. 6B for E771Q and E309Q mutants fully confirm the different behavior reported for these mutants at pH 7. In addition, it is clear from the results shown in Fig. 6A that, from a phenomenological point of view, the behavior of the ADA mutant is more closely related to the subclass of "classical" mutants that have a reduced affinity for Ca 2ϩ irrespective of pH.
It has previously been emphasized (10) that the effect of mutations on sensitivity to Ca 2ϩ , as determined in phosphorylation assays, may be accounted for by any one of the following reasons: (a) one or both Ca 2ϩ sites are disrupted; (b) the conformational equilibrium of ATPase is displaced in favor of an "E 2 -like" conformation that can be phosphorylated from P i but does not bind Ca 2ϩ efficiently; and (c) the signal transduction from the Ca 2ϩ -binding sites to the phosphorylation site is disrupted so that, irrespective of the ability to bind Ca 2ϩ , there is no concomitant change in the chemical specificity at the phosphorylation site. The last possibility can be excluded in the case of the ADA mutation, since the presence of Ca 2ϩ was found to result in the expected phosphorylation or dephosphorylation events, although with a reduced affinity. It is worth recalling that binding of a single Sr 2ϩ is believed to be sufficient to prevent P i -derived phosphorylation, while occupation of the two Sr 2ϩ -binding sites is required to permit phosphorylation from ATP (12). On this basis, and assuming that the same is true for the binding of Ca 2ϩ , Andersen (10) originally proposed that the classical mutants in which Ca 2ϩ is able to prevent P i -derived phosphorylation at neutral pH presumably have the first binding site intact, while mutants in which Ca 2ϩ is unable to prevent P i -derived phosphorylation have the first binding site altered by the mutation. Applying the same rationale to the results obtained with the ADA mutant would suggest that in the ADA mutant the initial steps in Ca 2ϩ binding are also severely altered, consistent with a role of the L6 -7 aspartic residues in controlling at least the first of the two Ca 2ϩ -binding sites. Whether such control is exerted through the first or second mechanism described above is the next question to be addressed.
Proteolytic Degradation of ATPase Mutants by Proteinase K-In the absence of a highly resolved three-dimensional structure, mild proteolysis is a powerful tool with which to investigate the spatial organization of transport ATPases (29 -31, 52). Since several conformational changes probably occur upon Ca 2ϩ binding to the Ca 2ϩ -ATPase, it was of interest to study by this method these conformational changes in wild type and mutated Ca 2ϩ -ATPase, and the dependence of these changes on pCa. In the absence of bound Ca 2ϩ , proteinase K predominantly cleaves Ca 2ϩ -ATPase in SR vesicles at two previously identified sites (31). The most N-terminal one of these two sites produces a short N-terminal peptide as well as peptide p95 (Lys-120-COOH terminus), while the other site gives rise to p28N (Thr-242-NH 2 terminus) and p83C (Glu-243-COOH terminus). The differential accessibility of these sites in the absence of Ca 2ϩ can be assumed to reflect the conformational equilibrium between the various forms of Ca 2ϩ -free ATPase. On the other hand, binding of Ca 2ϩ results in masking of the most N-terminal proteolytic site (Lys-120) and a higher degree of exposure of the second proteolytic site (Glu-243) in the ␤-strand region.
The extension of this technique to the study of the degradation of relatively small amounts of Ca 2ϩ -ATPase, produced by heterologous expression in yeast, was made feasible by the availability of highly specific antibodies reacting with various regions of the Ca 2ϩ -ATPase and allowing us to reveal p28N, p83C, and p95 with the expressed WT or mutant Ca 2ϩ -ATPases (Fig. 7). We found that when proteolysis was performed in the presence of a low Ca 2ϩ concentration (at pCa 7), mutant E309Q, in contrast to WT ATPase, did not produce p95 peptides in detectable amounts. The same fact was observed with the E771Q ATPase, which was also more degraded than the wild type under the same conditions. This suggests that the conformational state of these two Ca 2ϩ -free mutants is different from that of the wild type enzyme. Nevertheless, when proteolysis was performed in the presence of high Ca 2ϩ concentrations, the experiments revealed that the proteolytic degradation pattern of the E309Q mutant was still sensitive to Ca 2ϩ , in agreement with the functional evidence for Ca 2ϩ binding deduced from inhibition by Ca 2ϩ of P i -derived phosphorylation and the proposal that E309Q has the first Ca 2ϩ -binding site intact. Concerning the E771Q mutant, its degradation pattern remained almost unaffected by the addition of Ca 2ϩ at the concentrations used here (up to pCa 4), again consistent with the suggested disruption of the first Ca 2ϩ -binding site in this mutant. Conversely, at a low Ca 2ϩ concentration (pCa 7), the ADA mutant displayed the same pattern of proteolysis as that of the wild type, suggesting that the conformational state of the Ca 2ϩ -free ATPase is not altered in the ADA mutant. The only distinct characteristic of the ADA mutant was that a higher concentration of Ca 2ϩ than in wild type ATPase was required to obtain the disappearance of p95 and appearance of p83C and p28N. Consequently, by the use of limited proteinase K proteolysis, we obtain evidence that the ADA mutation directly affects the binding affinity of one or both of the Ca 2ϩ -binding sites, ex-cluding the second possibility mentioned above. These findings clearly establish the status of the ADA version as a "pure" Ca 2ϩ affinity mutant, which undergoes the same conformational changes as the wild type ATPase but with a reduced Ca 2ϩ binding affinity.
An important outcome of the proteolysis assays carried out in the absence of calcium is that the two nonconservative substitutions Asp 3 Ala that have been introduced into the ADA mutant do not seem to alter the conformation of the Ca 2ϩ -free ATPase, whereas the replacement of a single Glu by a Gln at Glu-309 and Glu-771 in the transmembrane domain unexpectedly turns out to have a much more perturbing effect on the overall Ca 2ϩ -free ATPase structure. Interestingly, the same conclusions were deduced from measurements of P i -derived phosphorylation, since in the absence of bound Ca 2ϩ the phosphorylation level of the ADA mutant was similar to that in Ca 2ϩ -free WT ATPase (Fig. 5B), whereas that of Ca 2ϩ -free E309Q and E771Q mutants was different (see legend to Fig. 6), possibly due to modified proton binding (10). All this suggests that cluster and "nonconservative" mutations from Asp to Ala can probably be tolerated more easily in a region exposed to the cytosol like the L6 -7 loop, which is readily cleaved by proteinase K at Gly-808 and Asp-818 (32), than in the transmembrane section of the ATPase.
A Model for Control by the L6 -7 Loop of Binding of Ca 2ϩ by Ca 2ϩ -ATPase-As a starting point, our results with the ADA mutant could be interpreted by assuming that the double Asp 3 Ala mutation within the L6 -7 loop produces a reduction in the overall affinity of the ATPase for Ca 2ϩ because of an indirect structural perturbation exerted by these mutations on the position of the Ca 2ϩ liganding residues in the contiguous M6 segment. Since M6 is suspected to contribute 2 (51) or 3 (26) residues participating in Ca 2ϩ complexation (with other residues in M4, M5, and M8), it is indeed likely that any displacement of M6 in the Ca 2ϩ binding cluster would reduce the affinity of the site for Ca 2ϩ . The fact that long range perturbations are possible is demonstrated by the effect of L6 -7 proline mutations on the catalytic site.
Arguing somewhat against this possibility, however, are the facts that mutation of the proline residues, in contrast to the aspartic acid residues, did not change the apparent affinity of ATPase for Ca 2ϩ and that mutation of the Pro-820-Pro-821 doublet was more deleterious than mutation of the Pro-811-Pro-812 doublet, although the latter might have been expected to influence to a larger extent the conformation of the nearby M6 segment. In addition, as discussed above, the proteolysis experiments displayed in Fig. 7 allow us to conclude that in the absence of calcium the conformation of the ADA mutant is similar to that of WT ATPase.
Thus, as an alternative model, the L6 -7 loop could be a direct contributor in the control of Ca 2ϩ binding to Ca 2ϩ -ATPase, as deduced from (a) our previous demonstration that Ca 2ϩ has some affinity for the Gly-808-Gly-994 peptide but not for the shorter Asp-818-Gly-994 peptide (p20 and p19, respectively; Ref. 32) and (b) the present data that show that the removal of some of the oxygen atoms (Asp 3 Ala) located in the region from Gly-808 to Asp-818 of the L6 -7 loop decreases the affinity for calcium by almost 3 orders of magnitude. Taking into account the recent proposal of a side-by-side location of the two Ca 2ϩ -binding sites (9) and the cytosolic location of the L6 -7 loop, Fig. 8 depicts a possible mechanism by which the aspartic residues of the L6 -7 loop could participate in the initial steps of high affinity binding of Ca 2ϩ to Ca 2ϩ -ATPase; interaction of Ca 2ϩ with the L6 -7 loop could direct the first FIG. 8. Hypothesis for the initial steps in Ca 2؉ binding to the Ca 2؉ -ATPase. This scheme displays Ca 2ϩ -binding sites 1 and 2 with the membranous residues involved in these sites, Glu-771, Thr-799, Asp-800 (and possibly Glu-908), indicated in shaded areas for site 1 and Glu-309, Asn-796, and again Asp-800 for site 2, as identified by Clarke et al. (26) and separated in two classes by Andersen (10). The two intramembranous Ca 2ϩ -binding sites are assumed to have a side-by-side location (see also Ref. 9). Aspartic acid residues 813, 815, and 818 of the L6 -7 loop are pointing toward the cytosol. In the E(nH ϩ ) state the two Ca 2ϩ -binding sites are occupied by protons; site 2 is not directly accessible, possibly because it is masked by the L6 -7 loop. The first Ca 2ϩ ion interacts with aspartic residues 813, 815, and 818 of the L6 -7 loop, which are the most accessible residues at this step. This initial interaction favors binding of Ca 2ϩ in a subsequent step to the first membranous site (site 1), after reorientation of the loop, leading to the E(Ca1) state. This reorientation results in an increased accessibility of Ca 2ϩ for interaction with the critical residues at the second site (site 2) and/or optimization of the positions of these residues, and thus binding of Ca 2ϩ to site 2. At the same time, the loop is locked in position to function as a gate for the first bound Ca 2ϩ . This results in an E(Ca 2ϩ 2)[Ca 2ϩ 1] state from which Ca 2ϩ bound to site 1 cannot be released unless Ca 2ϩ bound to site 2 is removed. This model is a possible alternative to the "single file" model, designed to explain evidence of sequential Ca 2ϩ dissociation from the ATPase sites (13,14,53). In the single file model, the two Ca 2ϩ ions bind in a sequential manner on top of each other in an ATPase crevice, i.e. first at site 1 and thus at site 2. In the present model with Ca 2ϩ sites side-by-side, the sequential nature of Ca 2ϩ binding and dissociation is caused by changes in the position of the L6 -7 loop. Ca 2ϩ ion toward the membrane-embedded liganding residues of the first site and open the way for binding of Ca 2ϩ at the second site. In this model the sequential order during cytosolic Ca 2ϩ dissociation is maintained by the L6 -7 loop functioning as a cytosolic gate that prevents Ca 2ϩ dissociation of the first bound Ca 2ϩ before dissociation of the second bound Ca 2ϩ . In view of the emerging structure of the Ca 2ϩ -ATPase transmembrane domain (7), it is not impossible that the L6 -7 loop plays some kind of gating role at the entrance of the crevice leading to site 1.