Involvement of the Cytoplasmic Loop L6–7 in the Entry Mechanism for Transport of Ca2+ through the Sarcoplasmic Reticulum Ca2+-ATPase* 210

We previously found that mutants of conserved aspartate residues of sarcoplasmic reticulum Ca2+-ATPase in the cytosolic loop, connecting transmembrane segments M6 and M7 (L6–7 loop), exhibit a strongly reduced sensitivity toward Ca2+ activation of the transport process. In this study, yeast membranes, expressing wild type and mutant Ca2+-ATPases, were reacted with Cr·ATP and tested for their ability to occlude 45Ca2+ by HPLC analysis, after cation resin and C12E8treatment. We found that the D813A/D818A mutant that displays markedly low calcium affinity was capable of occluding Ca2+ to the same extent as wild type ATPase. Using NMR and mass spectrometry we have analyzed the conformational properties of the synthetic L6–7 loop and demonstrated the formation of specific 1:1 cation complexes of the peptide with calcium and lanthanum. All three aspartate Asp813/Asp815/Asp818 were required to coordinate the trivalent lanthanide ion. Overall these observations suggest a dual function of the loop: in addition to mediating contact between the intramembranous Ca2+-binding sites and the cytosolic phosphorylation site (Zhang, Z., Lewis, D., Sumbilla, C., Inesi G., and Toyoshima, C. (2001) J. Biol. Chem. 276, 15232–15239), the L6–7 loop, in a preceding step, participates in the formation of an entrance port, before subsequent high affinity binding of Ca2+ inside the membrane.

Combined with genetic modifications of the polypeptide chain this approach has allowed identification of membranous amino acid residues critical for binding and occlusion of Ca 2ϩ . Residues indispensable for intramembranous binding of Ca 2ϩ were initially found to be clustered in four transmembrane domains, M4 (Glu 309 ), M5 (Glu 771 ), M6 (Asn 796 , Thr 799 , and Asp 800 ), and M8 (Glu 908 ) (8,20). Subsequent experiments involving incubation of mutated Ca 2ϩ -ATPase with Cr⅐ATP showed that, while side chain conservation of five of these critical residues, localized to M4, M5, and M6, was also required to maintain calcium occlusion in the presence of Cr⅐ATP (16,21), this was not the case for Glu 908 whose side chain was therefore considered to play a less significant role than the other residues in the intramembranous binding of Ca 2ϩ (8). Direct calcium binding measurements of genetically modified Ca 2ϩ -ATPase were subsequently successfully carried out by Inesi and collaborators (22,23). Their data demonstrated the absence of both high affinity calcium-binding sites for the inactive E771Q, T799A, and D800N as well as for the partially active E908A mutants, but the retention of one of the two high affinity calcium-binding sites for N796A and E309Q, as first reported by Skerjanc et al. (24).
In previous studies we have obtained evidence for the involvement of extramembranous amino acid residues in the L6 -7 loop in the mechanism of calcium binding. This was initially suggested by the observation that, after limited proteolysis of the SR Ca 2ϩ -ATPase, changes in Ca 2ϩ -dependent rates of electrophoretic migration were observed for some of the C-terminal proteolytic peptides formed. The data could be rationalized on the basis that it is the first part of the L6 -7 loop (which comprises the amino acid sequence 808 -830) that interacts with Ca 2ϩ (25). This part of the loop contains three conserved aspartic residues (Asp 813 , Asp 815 , and Asp 818 ) whose cluster mutation generated the D813A/D818A and D813A/ D815A/D818A mutants that we found to display a marked reduction in the apparent affinity with which Ca 2ϩ controls ATPase phosphorylation and turn-over (12). We rationalized our findings in terms of a model in which the conserved aspartic residues of the L6 -7 loop mediate the interaction with calcium ions during the initial steps of Ca 2ϩ binding process in a manner that might be related to the involvement of the loop in the gating of calcium ions on the cytosolic side, as previously suggested (1,12,25). However, in light of the crystallographic structure (4) it remains unclear whether the acidic residues of the L6 -7 loop are in a conformation suitable for coordination of Ca 2ϩ binding (4). As an alternative it has been proposed that the decrease in Ca 2ϩ binding caused by mutation of the acidic residues is indirect and that the essential function of the L6 -7 loop is to mediate long-range contact between intramembranous Ca 2ϩ -binding sites and the catalytic site necessary for phosphorylation by ATP (23,26). It is of interest that the L6 -7 loop has been the focus of another recent mutagenesis study in which it was suggested that domain IB of phospholamban makes contact with Asp 813 of the loop (27).
To address these questions concerning the function of the L6 -7 loop we found it necessary to obtain more direct information on the role of the acidic residues of the loop for Ca 2ϩ binding and occlusion. To study Ca 2ϩ occlusion we proceeded in a similar way as described by Vilsen and Andersen (16) to isolate Cr⅐ATP-reacted and C 12 E 8 -solubilized wild type and mutated ATPase by HPLC chromatography. In addition, we have studied in detail by NMR and mass spectrometry the conformation of the synthetic L6 -7 loop peptide and its interaction with Ca 2ϩ and related lanthanides. Overall, our data support a role of the conserved aspartic residues in the initial interaction of calcium ions with Ca 2ϩ -ATPase, before their subsequent occlusion inside the membrane. Furthermore, our data throw some new light on the nature of the occlusion process.

Mutation and Expression of Ca 2ϩ -ATPase in Yeast-
The single mutants E309Q and E771Q and the cluster mutants D813A/D818A and D813A/D815A/D818A (referred to as ADA and AAA mutants, respectively) were obtained as previously described in Refs. 12 and 25. Wild type (28) and mutant cDNA were inserted into the yeast expression vector pYeDP 60 (29) (a gift from D. Pompon, C.N.R.S., Gif sur Yvette). The yeast strain used in this study was Saccharomyces cerevisiae W303.1B/Gal4 (␣, leu2, his3, trp1::TRP1-GAL10-GAL4, ura3, ade2-1, can r , cyr ϩ ) that had been genetically modified to overexpress Gal4p upon galactose addition (30). The yeast cells were transformed (31) and selected by ura3 complementation. The conditions used for large scale expression and preparation of membrane fractions are described elsewhere (30). In our Cr⅐ATP-induced calcium occlusion experiments (see below) we used the yeast light membrane fraction, which contains about 1% of fully active Ca 2ϩ -ATPase (Ref. 30, see also Refs. 12 and 25) and mainly consists of endoplasmic reticulum. However, we were confronted with the problem that this fraction has a high Ca 2ϩ content. By use of calcium indicators (Quin 2 and Fluo 3) and flame photometry we estimated a Ca 2ϩ content of 350 nmol/mg of protein (corresponding to a concentration of 3.5 mM for samples containing 10 mg/ml of total proteins). Ca 2ϩ transport by the expressed Ca 2ϩ -ATPase can only be held partially responsible for this phenomenon, since we also found a high Ca 2ϩ content in control yeast cell membranes, containing the empty plasmid (approximately 200 nmol/mg). The presence of substantial amounts of endogenous Ca 2ϩ associated with the membrane preparations necessitated a preliminary Bio-Rex treatment (see "Methods" below and Fig. 1) by which we were able to remove virtually all endogenous calcium before the start of the Cr⅐ATP-induced Ca 2ϩ occlusion experiments.
Protein Concentrations, Electrophoresis, Blotting, and Proteolysis-Protein concentrations were measured by a modification of the bicinchoninic assay (32) in the presence of 0.5% SDS, using bovine serum albumin as a standard. The amount of Ca 2ϩ -ATPase was quantified after immunodetection on Western blot as described (12). For this 20-l samples were added to 22 l of urea-SDS denaturing buffer (33), heated for 1 min at 100°C, and aliquots of the denatured proteins were separated by SDS-PAGE. After transfer to polyvinylidene difluoride membranes (34), the Ca 2ϩ -ATPase present in the sample was reacted with polyclonal 79B Ab (a kind gift of A-M. Lompré) or sequence-specific Ab (34) and revealed with the ECL kit (Amersham Biosciences, Inc.). Quantitative analysis of the amount of Ca 2ϩ -ATPase present on the blots was performed with a Bio-Rad densitometer using as standard SR vesicles which were assumed to have a Ca 2ϩ -ATPase content corresponding to 75% of the total protein content. Sarcoplasmic reticulum vesicles were prepared from rabbit skeletal muscle as previously described (35). Proteinase K digestion of SR to prepare peptide p20C and p19C was performed in 100 mM bis-Tris, pH 6.5, and 0.3 mM Ca 2ϩ , as described in Ref. 25. To follow the change of electrophoretic mobility of p20C relative to that of p19C, SDS-PAGE gels were prepared with inclusion of variable amount of Ca 2ϩ or 0.02 mM EGTA in the stacking and separation gels (25,36).
Occlusion Experiments on Cr⅐ATP-reacted and C 12 E 8 -solubilized Ca 2ϩ -ATPase-Occlusion experiments were performed by incubation of membrane samples with Cr⅐ATP and 45 Ca 2ϩ , followed by C 12 E 8 solubilization and HPLC analysis. The different steps in the procedure are presented in outline form in Fig. 1. To remove contaminating metal cations (mainly calcium), light membrane fractions (600 l at a concentration of 10 mg/ml protein in a 100 mM MOPS buffer, pH 6.5) were first treated with 200 l of Bio-Rex beads (Bio-Rex 70 resin with a wet bead size of 150 -300 m from Bio-Rad) for 20 min at room temperature on a rotating wheel. After pelleting of the beads by a brief centrifugation, 500 l of the supernatant was collected. To this sample (which contained a total of about 5 mg of proteins), Ca 2ϩ was added at a final concentration of 120 M constituted by 100 M 40 Ca 2ϩ and 20 M 45 Ca 2ϩ tracer from a stock solution together with Cr⅐ATP (at a final concentration of 0.8 mM). The samples were then incubated for 16 h at 20°C. Following this incubation, the membranes were solubilized by addition of C 12 E 8 (at a detergent/protein (w/w) ratio of 1.5:1) in the presence of 0.2 mM phenylmethylsulfonyl fluoride. After 15 min at 4°C, insoluble material was removed by centrifugation at 110,000 ϫ g for 30 min in a TL100.3 rotor. Most of the calcium was removed from the samples by 40 min incubation with Bio-Rex beads as described above. Occluded cal-cium was detected as previously described in Ref. 16, by submitting 200 l of solubilized proteins to molecular sieve HPLC on a Beckman TSK G3000 SW (0.75 ϫ 30 cm) column, equilibrated and eluted with 50 mM Tes, pH 7, 0.1 M KCl, 0.1 mM Ca 2ϩ , and 1 mg/ml C 12 E 8 . The flow rate was 1 ml/min and, in general, fractions of 500 l were collected between 3 and 13 min after sample application and analyzed for 45 Ca 2ϩ content by scintillation counting and for Ca 2ϩ -ATPase content by Western blot. In a few experiments, 250-l fractions were collected to define more precisely the pattern of 45 Ca 2ϩ and ATPase elution. For determination of the Ca 2ϩ -ATPase contents, 10 l of each collected fraction was loaded onto a SDS-PAGE gel to estimate the amount of Ca 2ϩ -ATPase by Western blotting. For analysis of the 45 Ca 2ϩ contents, the radioactivity in each collected fractions was counted in 4 ml of scintillation liquid (Packard, Pico-fluor 40). Ten l of the initial occlusion medium containing 120 M calcium was used as a standard, to convert measured counts/min values to nanomoles of 45 Ca 2ϩ . Finally, the ratio of nanomoles of occluded calcium per mg of Ca 2ϩ -ATPase was calculated. This calculation included a correction for background calcium radioactivity in the fractions arising from other phenomena than specific calcium occlusion by Ca 2ϩ -ATPase. Estimation of the extent of nonspecific binding of Ca 2ϩ was approached in a number of different ways. In one type of experiments, backgrounds were measured by incubation of control membranes without expressed Ca 2ϩ -ATPase (i.e. light membrane fractions obtained from yeasts cells plasmid pYeDP 60) under the same conditions in the presence of Ca 2ϩ with added radioactive calcium and presence or absence of Cr⅐ATP. In another type of control experiments, backgrounds were estimated by first performing an overnight incubation of membrane samples in the presence of 0.12 mM nonradioactive 40 Ca 2ϩ , while addition of radioactive calcium (20 M 45 Ca 2ϩ ) was delayed until after the membranes had been solubilized by C 12 E 8 . We also compared Cr⅐ATP-induced calcium occlusion for the expressed wild type and mutated Ca 2ϩ -ATPases with the occlusion of calcium by SR vesicles alone or by SR in the presence of yeast control membranes. In these controls, we took care to add rabbit SR Ca 2ϩ -ATPase to give an ATPase content similar to that present in light membranes containing expressed Ca 2ϩ -ATPases (ϳ1% of total protein).
Peptide Material and Ca 2ϩ -binding Measurements-The SERCA L6 -7 loop peptides used in this study were synthesized by Syntem (Montpellier), HPLC purified, and their purity and composition checked by mass spectrometry and two-dimensional NMR analysis. The peptides comprised not only SERCA 1 residues 808 -827 (G 808 FNPPDLDIMDRPPRSPKEP 827 ), but also an 808 -827 variant, with alanine substituted for each of the 3 aspartates (Asp 813 , Asp 815 , and Asp 818 ) and for Glu 826 and a shorter sequence of the L6 -7 loop, comprising residues 811-827. The hexapeptide, RETDYY, blocked at both termini, was used as a control when monitoring the spectral effects resulting from carboxylate side chain ionization and the binding of metal ions. The purity and composition of this hexapeptide, synthesized by Alta Biosciences (Birmingham), was also confirmed by mass spectrometry and NMR analysis.
Other peptides used corresponded to SERCA 1a residues 808 -847 and 818 -847, which represent a long and short version, designated as M7-L and M7-S, respectively, of the L6 -7 loop together with the Nterminal anchor of membrane span M7. These were the same peptides as previously used in a peptide study on micelle insertion as a model of membrane interaction (37). Ca 2ϩ binding of these and the L6 -7 peptides was measured with the aid of an Aminco double-beam spectrophotometer, using murexide as an indicator of the concentration of free Ca 2ϩ . The formation of the Ca 2ϩ -murexide complex was recorded from the increase in absorption of 0.2 mM murexide at 544 nm, relative to that at 495 nm. To measure Ca 2ϩ binding by M7-L and M7-S, the peptides (1 mg/ml) were first solubilized in SDS (10 mg/ml; these peptides are known to be monomeric in SDS (37)). Furthermore, peptide binding data were corrected from the decrease in free concentration of Ca 2ϩ resulting from the interaction with the SDS micelles as measured in the absence of peptide. These measurements were supplemented by estimations of Ca 2ϩ binding to a larger and C-terminal proteolytic fragment (p20C, obtained by treatment with proteinase K, and representing the 808 -994 sequence in SERCA 1a Ca 2ϩ -ATPase). In this case measurements of the interaction of the polypeptide fragment with Ca 2ϩ was based on changes in electrophoretic migration rate of this proteolytic fragment by Laemmli SDS-PAGE, induced by addition of Ca 2ϩ to the gel (25,36), which were quantitatively evaluated as a function of Ca 2ϩ concentration. The 10-residue shorter C-terminal proteolytic fragment (p19C, also obtained by treatment with proteinase K, and representing the 818 -994 sequence in SERCA 1a) is not affected by addition of Ca 2ϩ to the gel (25).
Mass Spectrometry-Electrospray ionization (ESI) mass spectrome-try measurements were made using a 9.4T Bruker Daltonics Fourier transform ion cyclotron resonance instrument as previously described (38). The external ESI source was equipped with a Pyrex capillary coated with platinum paint at both ends and using typical nozzle and skimmer voltages of 60 and 3 V, respectively. The background pressure in the ICR analyzer cell was maintained at less than 2 ϫ 10 Ϫ10 bar. The spectra were also used to confirm the purity and composition of the different peptide preparations. Microliter aliquots of lanthanum or calcium chloride stock solutions (100 mM) were added to 20 to 50 M solutions of the L6 -7 loop peptide (residues 808 -827) in 5 mM ammonium acetate buffer, pH 5.9, to achieve the required cation concentrations. The hexapeptide, RETDYY, blocked at both termini was used as a control when monitoring the extent of complex formation under these conditions. Negligible proportions of bound cation were detected using this dicarboxylate-containing peptide. NMR Methods-All peptides were freeze-dried prior to dissolution in the appropriate buffer. The pH values quoted are direct meter readings adjusted using concentrated stock solutions of acid/base. The proton one-dimensional spectra of each peptide were invariant over the concentration range 0.1 to 4 mM and the narrow line widths observed were taken to indicate the absence of aggregation. For each peptide used resonance assignment was achieved by a combination of two-dimensional total correlation spectroscopy (TOCSY) and nuclear Overhauser effect spectroscopy (NOESY) experiments (typically with 2K points in the F 2 dimension and 256 points in the F 1 dimension) which were carried out in 90% H 2 O, 10% 2 H 2 O at 500 MHz on a Bruker instrument at a sample temperature of 12°C at a typical peptide concentration of 2 mM. NOESY data were acquired using mixing times of 200 and 600 ms.
The cation binding properties of the Gly 808 -Pro 827 loop peptide were probed using calcium as well as lanthanide cations as calcium analogues. Lanthanum (La 3ϩ ) is diamagnetic, has an ionic radius similar to Ca 2ϩ (1.061 and 0.99 Å, respectively) and, like the other lanthanides, has been shown to bind to characteristic protein calcium-binding sites with similar stereochemistry (39) but with affinity higher than that displayed by Ca 2ϩ (40). Gadolinium (Gd 3ϩ ; ionic radius: 0.938Å), another lanthanide, is paramagnetic and causes isotropic relaxation (ϰ r Ϫ6 dependence where r is the distance from the probe (41)). It can therefore be used to map out groups in its vicinity to complement and extend the NOE data for the cation complex. On the other hand, praseodymium (Pr 3ϩ ; ionic radius: 1.013 Å) has an anisotropic paramagnetic effect, causing pseudocontact shifts whose magnitude and sign vary with the orientation of the corresponding group with respect to the carboxylate ligand metal complex (41). Lanthanum (La 3ϩ ), gadolinium (Gd 3ϩ ), and praseodymium (Pr 3ϩ ) nitrate were purchased from Aldrich (99.9% pure) and 0.5 M stock solutions were prepared in either H 2 O or 2 H 2 O, pH 5.6. The cation concentrations were determined by EDTA titration using xylene orange. Binding titrations were carried out by addition of 1-10-l aliquots of these solutions into 0.5 ml of 2 mM Gly 808 -Pro 827 , Pro 811 -Pro 827 , or RETDYY at pH 5.6 and 22°C. The spectral effects resulting from cation coordination were monitored by one-dimensional proton NMR spectra acquired with solvent presaturation, using difference spectroscopy to highlight the changes resulting from cation coordination.
NOE distance constraints for the calculation of the conformation of the metal-bound Gly 808 -Pro 827 peptide were obtained from the NOESY spectra acquired in the presence of a 2:1 molar ratio of lanthanum to peptide to ensure maximum complex formation. TOCSY spectral data were used to distinguish between inter-residue and intra-residue assignments. Interproton distance constraints were classified into four groups with upper band distances of 2.8, 3.3, 4.0, and 4.5 Å, respectively. The XPLOR program was then used to analyze the NOE data obtained for peptide residues Pro 813 -Pro 820 (Fig. 4), with pseudo-atom corrections added where appropriate and averaging applied to NOEs involving non-stereospecifically assigned protons. Forty-four unambiguous sequential and medium-range (i-j Ͼ 1) interproton distance restraints were calibrated according to cross-peak intensity and used to generate a family of 100 structures that were reduced to 64 possible solutions that satisfied the checks provided within the QUANTA Protein Health routine. Final structures were subjected to the X-PLOR Accept routine with a violation threshold for NOEs of 0.5 Å and dihedral angles of 5°. The structures gave four clusters with cluster 1 containing 26 of the 64 structures calculated and 15 of the calculated structures in cluster 2. Fifty steps of CHARMM minimization were run for the lowest energy members of the first two clusters and consistency sought with the lanthanide shift and relaxation data.

Cr⅐ATP-induced Calcium Occlusion by the Native SR or
Yeast-expressed Wild Type or Mutated Ca 2ϩ -ATPase-To measure Cr⅐ATP-induced calcium occlusion by Ca 2ϩ -ATPase, we combined pretreatment with Cr⅐ATP in the presence of 45 Ca 2ϩ with subsequent solubilization by C 12 E 8 and molecular sieve chromatography (see the summary of protocol shown in Fig. 1). This is essentially the same approach as that taken by Vilsen and Andersen (16,42) to study calcium occlusion by Ca 2ϩ -ATPase expressed in COS cells, but our procedure includes additional steps to reduce contamination by non-occluded Ca 2ϩ before and after the overnight preincubation period. Thus, immediately after C 12 E 8 solubilization and centrifugation to remove nonsolubilized membrane material, we subjected the solubilized membranes to a second Bio-Rex treatment. By this treatment, we were able to remove 99.9% of the calcium in the preparation. Then, we injected 200 l of Bio-Rex-treated samples onto a silica gel column equilibrated and eluted with 100 M Ca 2ϩ , and 1 mg of C 12 E 8 /ml, pH 7.0.
Representative elution profiles obtained in these experiments on wild type or mutated Ca 2ϩ -ATPases are shown in Fig.  2 which compares the elution of 45 Ca 2ϩ with quantitative data on the Ca 2ϩ -ATPase content in the eluted fractions (panels A-D), obtained by Western blot analysis (panel E). Due to the removal of bulk calcium by the Bio-Rex treatment, the signal to noise ratio for 45 Ca 2ϩ occlusion is very good both for wild type Ca 2ϩ -ATPase ( Fig. 2A) and SR control (Fig. 2C), and significantly better than previously reported (16). Furthermore, control membranes, prepared from yeast cells with empty vector, only gives rise to elution of a minor peak of protein bound 45 Ca 2ϩ in the chromatogram both in the presence (Fig. 2C, closed squares) and absence (Fig. 2C, open squares) of Cr⅐ATP. The same profile is present whether the membrane is pretreated with Cr⅐ATP overnight or if Cr⅐ATP is added after C 12 E 8 solubilization (data not shown). The extent of nonspecific binding was also estimated by overnight preincubation with Cr⅐ATP and cold Ca 2ϩ , postponing the addition of 45 Ca 2ϩ until after the solubilization of the membranes with C 12 E 8 ( Fig. 2A, closed triangles). In this situation only cold Ca 2ϩ will be occluded during the preincubation with Cr⅐ATP. As can be seen, the elution profile of 45 Ca 2ϩ under these conditions is even lower than observed with the control membranes without Ca 2ϩ -ATPase expression (compare the curve labeled "wild type 40 Ca 2ϩ " in chromatogram 2A with the curve labeled "Control Mb"). In the experiments reported below we chose to correct our data for nonspecific binding by subtracting the radioactivity obtained with the control membrane preparation, preincubated in exactly the same way as the Ca 2ϩ -ATPase expressing membranes with 45 Ca 2ϩ and Cr⅐ATP.
Overnight preincubation with Cr⅐ATP and 45 Ca 2ϩ of samples containing expressed or SR Ca 2ϩ -ATPase gives rise to a peak of radioactive calcium at around 6 ml (black circles in Fig. 2, A and B). In chromatogram A we see that elution of the peak of occluded Ca 2ϩ coincides with that of the expressed wild type Ca 2ϩ -ATPase as revealed by Western blots analysis. The elution position corresponds to the elution volume of ATPase monomers (43). However, it is also apparent from the chromatogram that the radioactivity profile is broader than that of immunochemically detected Ca 2ϩ -ATPase, stretching toward the peak of residual non-occluded 45 Ca 2ϩ that was not removed by the preceding Bio-Rex treatment and which elutes in the column total volume. This finding could mean that occluded Ca 2ϩ from the ATPase slowly leaks out from the protein during passage through the column (cf. chromatogram A); alternatively, that the trailing edge represents calcium slowly bound in an occluded state to other proteins than Ca 2ϩ -ATPase during the preincubation with Cr⅐ATP. In support of the first possibility we show that the same delay in Ca 2ϩ elution profile is obtained when Ca 2ϩ -ATPase from SR vesicles is added to the yeast control membranes and subjected to the same occlusion procedure as yeast expressed Ca 2ϩ -ATPase, while control membranes do not occlude to any significant extend (cf. chromatogram C). Notice that in the experiments with control membranes alone (shown in Fig. 2, A-C), 45 Ca 2ϩ elutes as a small peak at the same position as the Ca 2ϩ -ATPase monomer while virtually no calcium is present between this peak and the total volume. It therefore appears more probable that the trailing edge of the calcium occluded peak originates from the applied Ca 2ϩ -ATPase with little contribution from calcium firmly bound by other components than Ca 2ϩ -ATPase in the yeast membranes.
In Fig. 2B, it is seen that with the D813A/D818A mutant in the L6 -7 loop, the radioactivity profile is slightly lower than that observed for the wild type ATPase, but the Ca 2ϩ -ATPase content is also lower (the reason for this difference probably is unrelated to differences in the level of expression, see below). However, the position of the peaks for the mutated ATPase is the same as that of the wild type (compare chromatograms A and B, and also Western blots in E) and there is again a tendency for delayed elution of radioactivity (note that in this experiment we collected 250 l rather than 500-l fractions).
Finally, the chromatogram in Fig. 2D shows the result of an experiment performed with the E309Q mutated Ca 2ϩ -ATPase that was previously shown to be incapable of occluding Ca 2ϩ (21). In agreement with this previous report we find that the radioactive peak associated with this mutant does not exceed the background level. But it came as a surprise to us that the elution profile of the solubilized E309Q ATPase indicated an elution volume close to the V 0 of the HPLC column (approximately 4.5 ml). Thus, it appears that the mutated ATPase is more unstable than wild type ATPase, resulting in elution in FIG. 1. Summary of the occlusion protocol. In the solubilization tests, the calcium concentration (0.12 or 1 mM) was adjusted in step 2 or calcium was omitted and replaced by EGTA, 2 mM. After the solubilization/centrifugation step, Ca 2ϩ -ATPase content in the total and supernatant was analyzed by Western blot.
an aggregated state after detergent solubilization.
In Table I we summarize numerical estimates of Ca 2ϩ occlusion resulting from such experiments. To correct for nonspecific binding we subtracted the elution of 45 Ca 2ϩ observed in control membranes from our data. Furthermore, to obtain a minimum estimate of the occlusion, we only included the Ca 2ϩ eluted in the fractions where Ca 2ϩ -ATPase was immunochemically detectable (i.e. the sum of radioactivity eluting between 5 and 7 ml, Fig. 2). To obtain an upper estimate of the occlusion we also included the 45 Ca 2ϩ present in the following fractions up to (but excluding) the peak of unbound Ca 2ϩ associated with the total volume of the column (i.e. the sum of radioactivity eluting between 5 and 11 ml, Fig. 2). Both values have been given in Table I to define a range of numerical values for Ca 2ϩ occlusion. It can be seen that there is absolutely no difference between the Ca 2ϩ occluding capacity of the ADA mutant and that of wild type ATPase. Both kinds of expressed Ca 2ϩ -ATPase occlude Ca 2ϩ to the same extent as native SR Ca 2ϩ -ATPase for which the ratio of occluded Ca 2ϩ to that which can be maximally phosphorylated approaches a ratio of (1-2):1. The occlusion is thus close to the maximal level that can be expected on the basis of the concentration of active enzyme present in the preparation. For the ADA mutant, it is remarkable that this maximal theoretical level is approached already at a Ca 2ϩ concentration of 0.12 mM, although in previous experiments full phosphorylation from ATP required higher Ca 2ϩ concentrations (5-10 mM) (12). This suggests that complexation of the ADA mutant with Cr⅐ATP provides the ATPase with an efficient mechanism for trapping Ca 2ϩ bound inside the protein.
Solubilization by C 12 E 8 of Various Cr⅐ATP-reacted Ca 2ϩ -ATPases as a Function of Calcium-We made several attempts to measure Cr⅐ATP-induced Ca 2ϩ occlusion by the E771Q (previously reported to display no calcium occlusion (8)) and AAA mutants, by the same procedures. However, these experiments were unsuccessful in the sense that as a result of poor solubil-ity, we did not recover any Ca 2ϩ -ATPase after the HPLC elution, and therefore it was not possible to test their Ca 2ϩ occluding properties. There are two major ways to account for this result: either the mutated Ca 2ϩ -ATPase after expression forms large non-soluble (by C 12 E 8 ) aggregates in the membrane; or, alternatively, the Ca 2ϩ -ATPase, while being in a native-like conformation in the membranous state, when solubilized by C 12 E 8 , it is in such a labile state that it readily undergoes rapid denaturation and aggregation under these conditions. Note that we have stressed previously the increased susceptibility of this E771Q mutant to proteolytic digestion (12). This may indicate that loss of Ca 2ϩ occlusion, whenever it occurs, is associated with structural changes that not only decrease Ca 2ϩ binding, but also affect the structural integrity of the protein (cf. also "Discussion").
Cation Binding Characteristics of the Synthetic L6 -7 Peptide Analyzed by Mass Spectrometry-ESI mass spectrometry has become an established technique to follow noncovalent interactions such as cation binding (see, e.g. Ref. 44). The retained ability of the ADA mutant to occlude Ca 2ϩ demonstrates that the two intramembranous high affinity Ca 2ϩ -binding sites are still intact. The strongly reduced affinity for Ca 2ϩ of this and other aspartate mutant of the loop, as shown by the Ca 2ϩ dependence of ATPase phosphorylation from either ATP (12) or P i (12,25,26), or by direct equilibrium binding of Ca 2ϩ (23,26), thus suggests that an initial step in the calcium binding process has been affected. We therefore analyzed the cation binding characteristics of the synthetic L6 -7 loop by mass spectrometry so as to determine whether the acidic residues of the loop are capable of acting as suitable ligands. The high resolution ESI-Fourier transform ion cyclotron resonance mass spectra of the loop peptide 808 -827 in the absence and presence of calcium (Fig. 3A) and lanthanum ions (Fig. 3B)  complex and, at 1 mM Ca 2ϩ , the concurrent generation of smaller amounts of the 2:1 Ca 2ϩ -peptide complex even though calcium-free peptide was still readily detectable (Fig. 3A). Exact masses showed that the binding of one Ca 2ϩ (⌬M ϭ 38 instead of 40) displaced two H ϩ from the protein. By comparison, tighter cation binding by the peptide was observed in the presence of La 3ϩ as indicated by the distribution of the mass spectral peaks detected at 0.1 and 1 mM of added cation (Fig.  3B). The spectra indicated the relative absence of the metalfree peptide and the prevalence of a stable La 3ϩ -peptide complex with a 1:1 stoichiometry. The binding of one La 3ϩ (⌬M ϭ 135.9 instead of 138.9) displaced three H ϩ from the protein.
Competition by calcium was investigated using an equimolar concentration of Ca 2ϩ and La 3ϩ (1 mM each). In this case the predominant species observed was still the 1:1 La 3ϩ -peptide complex, with a 1:1 Ca 2ϩ -peptide and La 3ϩ :Ca 2ϩ :peptide components detectable, but at much lower concentration (data not shown). Mg 2ϩ and K ϩ binding were also investigated and K ϩloop complexes were absent even at 10 mM KCl while Mg 2ϩloop complexes were always much less abundant than Ca 2ϩ -loop complexes (data not shown). Negligible proportions of bound Ca 2ϩ and La 3ϩ were detected using the loop peptide 808 -827 containing alanine substituted for each of the carboxylates and for the dicarboxylate-containing peptide RET-DYY (data not shown). In conclusion, these observations demonstrate the ability of the loop peptide 808 -827 to bind both Ca 2ϩ and trivalent lanthanide attributable to the presence of acidic amino acid residues in the loop peptide and that the isolated L6 -7 loop was able to coordinate Ca 2ϩ , although binding of the divalent cation was weaker than that of lanthanum.  The ratio of occlusion for each set of experiments is given in terms of nmol of occluded calcium per mg of Ca 2ϩ -ATPase. Radioactivity due to calcium binding to yeast membranes was subtracted. The lower estimates do not take into account the calcium delayed elution but the upper estimates do. Previous data reported 6 nmol occluded Ca 2ϩ /mg of SR Ca 2ϩ -ATPase (13) and 3 nmol of occluded Ca 2ϩ /mg of expressed Ca 2ϩ -ATPase for the COS microsomal fractions (16). The last column reports the aggregation state of Cr.ATP-reacted solubilized Ca 2ϩ -ATPase proteins as deduced from the elution position in size exclusion chromatography. helical turn involving residues 816 -819 (4). Furthermore, we found that the conformation of the heptapeptide segment of the isolated L6 -7 loop was sensitive to alanine substitution of the aspartate residues (Asp 813 , Asp 815 , and Asp 818 ). Thus, the alanine-substituted peptide lacked the stretch of d ␣N (i, iϩ2) NOE proximities in the residue sequence 812-819 and was instead characterized by new inter-residue cross-peaks for the sequence 815-818 indicative of a different conformational preference adopted by these residues (Fig. 4). Consistent with the backbone proton chemical shift changes of Leu 814 , Ile 816 , Met 817 , and Arg 819 (Supplementary Material) these observations are indicative of a conformational rearrangement upon alanine substitution of Asp 813 , Asp 815 , and Asp 818 characterized by increased backbone flexibility of residues 813-819, resulting in a less extensively stabilized conformation of the L6 -7 loop.

Conformational Changes of the Loop Peptide upon Cation
Binding-To investigate any dependence of the conformation adopted by residues 813-819 on carboxylate ionization and to provide a comparative indicator for possible metal-ion complex formation the spectral consequences of proton binding by the loop peptide residues 808 -827 were monitored. The shifts of the ␤CH 2 peaks of the three aspartic acid residues resulting from ionization reflect the same apparent pK a ϭ 3.85 Ϯ 0.16 (Fig. 5A). A pK a of 3.79 Ϯ 0.10 was similarly obtained for Glu 826 . The slope, n ϭ 1, derived from the titration data of aspartic acid residues indicated that there is no apparent cooperativity in proton binding to these groups. Since spatial proximity of the aspartic acid side chains would have introduced higher pK a values, the observed pH dependence suggest that the ionized groups are oriented away from each other.
Titration with the lanthanum (La 3ϩ ) cation led to upfield shifts of several of the peptide resonances that showed saturation at an equimolar concentration of added La 3ϩ (Fig. 5B). Consistent with the mass spectroscopic analysis (see Fig. 3B), this indicated that only one lanthanum ion is bound per peptide and, given the concentrations of peptide and metal ion used, that the affinity of the complex was greater than 10 3 M Ϫ1 . Since no spectral changes were detected for any of the resonances of Glu 826 upon titration with La 3ϩ , it can be concluded that the 1:1 complex involves the tridentate coordination of the cation by all three aspartate side chains of the L6 -7 loop.
The simultaneous coordination of lanthanum by the three carboxylates would require the reorganization of the peptide conformation so as to enable the proximity and correct aspartate side chain orientation. Evidence of this was obtained by measurements of the 3 J NHC␣ coupling constant for those backbone -NH signals that are resolvable in the one-dimensional spectra. Indicative of a change toward a more helical conformation (47), the most significant change (Ͼ0.4Hz) was found for Ile 816 . The observed decrease in 3 J NHC␣ is consistent with a cation-dependent change of the Asp 815 -Ile 816 peptide linkage as reflected also by a change in backbone amide and C ␣ H chemical shift position of Leu 814 and Ile 816 (Supplementary Material).
The geometry of the tridentate 1:1 complex was further explored by use of the paramagnetic lanthanide cations, gadolinium (Gd 3ϩ ) and praseodymium (Pr 3ϩ ). Binding of Gd 3ϩ results in isotropic distance dependent relaxation of residues proportional to r Ϫ6 from the bound cation (41). Under the conditions of fast exchange between metal-free and metal-bound peptide that characterize complex formation here, the peptide line widths were therefore expected to show a monotonic increase with increasing Gd 3ϩ concentration. The relaxation effects induced by the added Gd 3ϩ indicated that this cation competed for the same coordination site as the bound lanthanum. Line width changes were observed only for residues in the Asp 813 -Arg 819 region with equally marked relaxation effects for each of the ␤CH 2 signals of the three Asp residues and lesser broadening effects were observed for the side chain and NH signals of Leu 814 , Ile 816 , and Met 817 (Supplementary Material). Negligible relaxation occurred for the side chain resonances of Glu 826 indicating that gadolinium is only bound to the site comprised of the three aspartates. Competition experiments were also carried out by titrating Ca 2ϩ into a solution of the peptide in the presence of Gd 3ϩ . Increasing concentrations of Ca 2ϩ led to markedly reduced relaxation effects, indicating progressive displacement of the bound lanthanide that again suggests isomorphous coordination by the calcium ion. Mapping of the peptide conformation about the cation bound to the 3 aspartates was also obtained during titration with Pr 3ϩ , an anisotropic paramagnetic probe that affects residue signals in a manner dependent upon their orientation with respect to the magnetic susceptibility axis of the bound cation (41). The ␤CH 2 signals of three aspartate residues were shifted by an equal amount in the same direction with increasing concentration of Pr 3ϩ (data not shown). This indicates that their carboxylate groups are symmetrically oriented with respect to the bound cation.
Complementary NOE data was obtained in the presence of La 3ϩ (Supplementary Material) and a solution structure was generated, consistent with the paramagnetic probe data. An initial set of 64 structures that complied with the NOE constraints were clustered into four superimposable groups based on the backbone of the sequence Asp 813 -Arg 819 . Subsequent relative distance restraints derived from the gadolinium titration provided further constraints that enabled the four structure clusters to be compared. Only cluster 1, containing 26 solutions, gave good agreement with this data as judged by the coordination geometry required to form a tridentate complex. The average conformation representative of this cluster is shown in Fig. 6A, where it is compared with the structure of the loop in the Ca 2ϩ -ATPase structure of Toyoshima et al. (4). Significantly there is good agreement between the conformation of the loop as observed in the crystal structure and that derived from the solution studies with bound cation at the N-terminal and C-terminal part of the isolated peptide as indicated by the overlay of these segments in Fig. 6B. However, the overlay shows that a reorientation in the middle segment accompanies cation binding by the isolated L6 -7 peptide so as to generate the coordination of the cation by all 3 aspartate residues. Overall the solution studies therefore show that the loop sequence is an independently folded substructure suitably composed to be involved in binding of cations.
Affinity of the L6 -7 Loop for Ca 2ϩ -We previously found that the electrophoretic migration rate of an SDS solubilized Cterminal proteolytic fragment of Ca 2ϩ -ATPase, containing the L6 -7 loop (p808 -994, referred to as p20C), is affected by the presence of Ca 2ϩ (1 mM) in the gel buffer. This contrasts with the behavior of a slightly shorter version, p818 -994, referred to as p19C, which lacks the N-terminal aspartate residues of the L6 -7 loop, and whose migration rate is not affected by Ca 2ϩ (25). Actually, this was the first evidence of a role of the 3 N-terminal aspartate residues in the L6 -7 loop for Ca 2ϩ bind-  ing. We have now systematically tested the dependence of this effect on Ca 2ϩ concentration in the range of 0.01-2 mM, and as can be seen from Fig. 7 (triangles), a Ca 2ϩ concentration of around 0.05 mM suffices to produce a half-maximal effect on migration rate which levels off at 0.8 -2 mM. In Fig. 7 these data are compared with results on Ca 2ϩ binding obtained with murexide on the chemically synthesized smaller peptides. As can be seen, both the water soluble L6 -7 peptide (closed circles) as well as the slightly longer M7-L (p808 -847) peptide (closed squares) bind appreciable amounts of Ca 2ϩ at a free concentration of 1 mM, in contrast to the L6 -7 peptide with alanine substitution of the acidic residues and the M7-S (p818 -847) peptide, lacking the N-terminal sequence of L6 -7 (open circles and squares, respectively). For the water-soluble unmodified L6 -7 peptide a dissociation constant (half-maximal binding) at ϳ1 mM can be estimated on the basis of one Ca 2ϩ -binding site per peptide molecule, whereas the affinity of the SDS-solubilized M7-L peptide is between that of p20C and L6 -7. Thus there seems to be a trend toward a higher affinity for the longer and presumably more organized peptide fragments. Note that, by comparison with the above data, the mass spectrometry data on L6 -7 (Fig. 3A) suggests half-maximum binding of Ca 2ϩ within this range (50% Ca 2ϩ -bound peptide for about 0.4 mM free calcium). The relatively low affinity of the L6 -7 peptide may reflect that the free peptide in solution has access to a larger range of conformations, only some of which bind Ca 2ϩ ; but as pointed out under "Discussion" even an affinity as low as 1 mM does not rule out an important role of Ca 2ϩ binding by the loop as an entrance port mediating contact with the intramembranous residues as part of the transport mechanism for Ca 2ϩ through the membrane.

Ca 2ϩ
Occlusion by the L6 -7 Loop-Functional characteristics of the aspartate mutants of the L6 -7 loop and of other calcium binding mutants have been described in a number of earlier papers (7,8,12,19,(22)(23)(24)(25)(26)48). Unlike the D813A/D818 (ADA) or the E908A mutations, the other substitutions that cause drastic reduction of Ca 2ϩ binding also result in an absence of ATPase activity and Ca 2ϩ transport. The D813A/D818 (ADA) and the E908A mutants, however, still retain some ATPase and transport activity. In our previous work we have demonstrated the probable involvement of the Asp 813 , Asp 815 , and Asp 818 of the cytoplasmic L6 -7 loop of SERCA 1a in the initial steps of the calcium transport cycle (12,25). Although we previously suggested that the L6 -7 loop could function as a pre-calcium-binding site, we also considered that by acting as a cytosolic gate the L6 -7 loop could contribute to the high affinity binding and occlusion of the calcium embedded at the membrane transport sites. It was therefore important to explore the properties of these mutants further. To test if the occlusion properties of the L6 -7 aspartate mutant were retained or lost we established an improved assay for quantitative estimation of Cr⅐ATP-induced 45 Ca 2ϩ -occlusion of the Ca 2ϩ -ATPase expressed in yeast. The occlusion measurements were performed on Bio-Rex-treated yeast membranes that had been preincubated with 45 Ca 2ϩ and Cr⅐ATP, before being subjected to C 12 E 8 solubilization and size exclusion chromatography. First, we confirmed that with this procedure we could demonstrate that the wild type is able to occlude Ca 2ϩ , while the classical mutant E309Q is not. Applying the procedure to the ADA mutant we then found that the mutated and C 12 E 8 -solubilized protein can occlude calcium to nearly the same level as the wild type protein. The view that we are dealing with a Ca 2ϩ affinity mutant is supported by the recent studies of Zhang et al. (23,26) who, consistent with our earlier observations (12,25), found that alanine or asparagine double and triple mutants of Asp 813 , Asp 815 , and Asp 818 hydrolyze ATP at a slower rate than wild type ATPase and, at least in the case of the double mutants (23), with a reduced Ca 2ϩ affinity, but apparently with retention of Ca 2ϩ cooperativity. Furthermore, in our experiments appreciable phosphorylation of the ADA mutant from ATP required Ca 2ϩ concentrations higher than 1 mM (12). A low degree of phosphorylation was also a characteristic feature in the study of Zhang et al. (23), both for these mutants as well as for mutants of basic amino acid residues (Arg 819 and Lys 823 ) localized in the C-terminal part of the L6 -7 loop (26). The latter mutations did not, however, result in markedly reduced Ca 2ϩ affinity. Thus, in contrast to the classical intramembranous mutants it seems that mutation of Asp 813 and Asp 818 primarily changes the initial interaction and affinity of the Ca 2ϩ -ATPase for Ca 2ϩ and reaction with ATP, but otherwise leaves intact the structure of the intramembranous Ca 2ϩ transport-binding sites.
L6 -7 Loop in the Structure of SERCA 1a-The recent x-ray diffraction analysis of the calcium-bound form of SERCA 1a (4) has provided a number of insights into the mechanistic aspects of the Ca 2ϩ transport process. In particular, the structural data confirm the involvement in high affinity calcium binding of all the key intramembranous residues previously characterized by site-directed mutagenesis, including that of Glu 908 whose mutation to alanine, as described here for the ADA mutant, did not impair Cr⅐ATP-based Ca 2ϩ occlusion (8,16). The likelihood of a central role of the L6 -7 loop in the reaction mechanism is supported by the occurrence of potential hydrogen bonds that link a number of amino acid residues in the L6 -7 loop to Arg 751 , Asn 755 , and Asn 756 on the M5 helix and to key residues in the cytosolic P-domain (Fig. 8). Accordingly, Zhang et al. (26) FIG . 7. Ca 2؉ binding by the L6 -7 loop and other C-terminal peptides. Ca 2ϩ binding of the L6 -7 loop and other chemically synthetized peptides were measured on the basis of changes in the absorption of murexide (0.2 mM) at 544 nm, relative to that of 495 nm, on an Aminco Double Beam spectrophotometer. Closed circles, binding isoterm of the L6 -7 (p808 -827) peptide at 1 mg/ml in 20 mM Tris, pH 8.0, and 100 mM NaCl buffer; open circles, corresponding data for binding of the L6 -7 peptide in which all acidic amino acids (3 Asp and 1 Glu) were substituted by alanine; closed squares, binding isoterm for M7-L (p808 -847) at 1 mg/ml after solubilization with SDS (10 mg/ml) in 20 mM Tris, pH 8.0, and 100 mM NaCl; open squares, corresponding data for the shorter peptide M7-S (p818 -847). In the latter two cases the data were corrected for interaction of Ca 2ϩ by the SDS micelles (70 Ϯ 2%), measured over the same range of Ca 2ϩ concentrations (50 -1000 M) and in the same Tris buffer and detergent concentration (10 mg/ml) as in the presence of peptide. Ca 2ϩ binding of the longer peptide p20C (residues 808 -994) was deduced from Ca 2ϩ -induced changes in electrophoretic migration rate of the proteolytic fragment (triangles) on Laemmli SDS-PAGE, induced by addition of Ca 2ϩ to the gel. The changes were quantitatively evaluated as a function of Ca 2ϩ concentration with respect to standard proteins not affected by the presence of calcium in the gel and with respect of the 10-residue shorter C-terminal proteolytic fragment (p19C) which is not affected either (25,36). As above, the data were corrected for interaction of Ca 2ϩ by the SDS micelles (30 Ϯ 1%), measured over the same range of Ca 2ϩ concentrations but this time in the Tris buffer of the Laemmli gel which has a SDS concentration of 1 mg/ml. have interpreted the reduced phosphorylation level that results from mutations in the L6 -7 loop primarily as the consequence of a disruption of a network of hydrogen bonds that mediate contact between the intramembranous Ca 2ϩ -binding sites and the phosphorylation domain. 2 Furthermore, the pronounced decrease in Ca 2ϩ binding affinity observed for the aspartate mutation in L6 -7 (ADA and AAA) was presumed to be indirect, caused by an allosteric effect, since in the Ca 2ϩ -bound crystals the carboxylate groups of Asp 813 and Asp 818 are too far apart to form a Ca 2ϩ -binding site (4). However, the x-ray structure provides a static picture of the ATPase in the E1Ca 2 state and does not rule out the possibility that the L6 -7 loop transiently forms a Ca 2ϩ -binding site, associated with the conformational changes accompanying the reaction of the Ca 2ϩ -deprived enzyme with Ca 2ϩ . The lower resolution structure available for the calcium-free form of Ca 2ϩ -ATPase (3) indeed led to the suggestion that the L6 -7 loop undergoes substantial reorientation upon high affinity calcium binding with the disposition of L6 -7 being much higher above the membrane surface in the calcium-free conformation (Ref. 4, see also discussion below). Interestingly, such movement has also been invoked by Asahi et al. (27) to rationalize their observation that in the E2 state Asp 813 becomes sufficiently exposed to participate in specific interaction with phospholamban that functions to lower the apparent affinity for Ca 2ϩ of SERCA 2a. To test the potential of the aspartate residues to form a Ca 2ϩ -binding site we therefore examined the conformation and Ca 2ϩ binding ability of the isolated loop peptide. Although conclusions based on studies of an isolated portion of the protein need to be drawn with caution, we were encouraged in this approach by the outcome an earlier analysis undertaken on the isolated M6 transmembrane fragment by which we had been able to predict the presence of a nonhelical region in the middle of this segment (50) which subsequently was confirmed by the 2.6-Å crystal structure (4).
NMR and Mass Spectrometry Studies of the Isolated L6 -7 Loop Peptide and Its Cation Binding-NMR spectroscopy and mass spectrometry data confirmed the ability of the L6 -7 loop peptide to interact with both Ca 2ϩ and lanthanides. Furthermore, the NMR data permitted us to conclude that in solution the central part of the peptide enclosed between the Pro 811 -Pro 812 and Pro 820 -Pro 821 trans-peptide bonds has a unique structure, consistent with the x-ray diffraction data of Ca 2ϩ -ATPase (4). Our data suggest a decisive role of the aspartate residues in the maintenance of this structure, since the conformation of the related peptide in which the acidic residues had been replaced by alanine approached that of a random coil. Furthermore, our pH titration data indicated that the different carboxylate ionizations are independent, implying that the acidic residues individually (non-cooperatively) bind protons. Nevertheless, the dependence of the spectrum on the concentration of added cation indicated the formation of a 1:1 complex with all three aspartate residues simultaneously being involved in the coordination of the metal ion. As can be seen from Fig. 6 this cation-bound conformation only differs from the loop topology in the x-ray based structure mainly in the orientation of the central (LDIM). Such a change of structure, permitting Ca 2ϩ binding by the three aspartate residues, would account for the changes in Ca 2ϩ -dependent rates of electrophoretic migration that were observed for some of the C-terminal proteolytic peptides formed (25). The possibility that the central segment of the L6 -7 loop undergoes a conformational rearrangement during the E2-E1 transition is also suggested by the apparently strained conformation of the side chain of Leu 814 that packs within a cluster of residues (Asp 815 , Gln 920 , Met 923 , Arg 924 ; PDB number 1eul, Ref. 4). It should also be pointed out that the position of the L6 -7 loop in the threedimensional crystal of the Ca 2 E1 form is different from the tentative structural model for E2P (PDB accession number 1fqu; obtained by fitting the three-dimensional atomic model of the Ca 2 E1 form to the maps derived from two-dimensional crystals of membranous ATPase prepared in the presence of vanadate, EGTA, and thapsigargin). In this E2P model the loop appears to be lifted up toward the cytoplasm and Asp 813 (as well as Asp 815 and Asp 818 ) faces the exterior of the protein.
The L6 -7 Loop as a Potential Entrance Port for P-type ATPases-Consistent with an important role of the L6 -7 loop in the function of type II P-ATPases it was previously reported that mutations of one glutamate or one aspartate in the comparable L6 -7 sequence of gastric H ϩ ,K ϩ -ATPase (corresponding to Asp 813 and Asp 815 in SERCA 1a) prevent phosphoenzyme formation from ATP (49). Previously, Karlish and co-workers (51)(52)(53) have proposed that in Na ϩ ,K ϩ -ATPase the homologous L6 -7 loop serves as an entrance port and pre-binding site for Na ϩ , essential for cation approach toward the high affinity binding pocket inside the membrane. There is a conspicuous conservation of carboxylates in this region of H ϩ ,K ϩ -ATPase, Na ϩ ,K ϩ -ATPase, and SERCA (1,25). Therefore, at this point it is pertinent to consider to what extent similar considerations apply to Ca 2ϩ -ATPase and how they accord with the present SERCA x-ray diffraction model. 2 Note that the simulation of steady state kinetics yielding overall ATPase velocity for the triple N mutant (Fig. 9 of Ref. 26), while it certainly will lead to a decrease in turnover rate and a decrease in ATP phosphorylation, does not account for previous evidence of a decreased affinity for Ca 2ϩ during turnover (23) nor for direct observations of a decreased affinity for Ca 2ϩ binding (23, 26) or for phosphorylation from P i (12,25,26). Therefore this simulation cannot be considered as contradictory to the idea that an initial step in calcium binding is affected in the case of the triple N mutant.
FIG. 8. Diagrammatic representation of a subset of the many contacts that link residues of the cytoplasmic L6 -7 loop (shown in bold) with residues of the Ca 2؉ -ATPase that are also located sequentially close to the calcium binding ligands of the intramembrane helices (M4-M9). The network of contacts include potential H-bonds as well as numerous non-directional side chain-side chain packing interactions that would enable the ready transmission of conformational adjustment (either as part of the enzyme cycle or as a result of introduced mutations). Note, for example, the contacts that straddle the cytoplasmic ends of M4 and M5, linkages that would account for cooperative changes resulting from reorientation of the L6 -7 loop upon calcium binding. The mesh of contacts of the L6 -7 loop are consistent with its acting a potential entrance port. The coordinates for the crystal structure are those of the PDB deposition 1eul (4).
The possibility that, in an early step of the reaction mechanism, the L6 -7 aspartate residues are part of a Ca 2ϩ prebinding site is consistent with the exposed location of residues 808 -818 as indicated both by the x-ray structure and susceptibility of this region to undergo proteolytic cleavage by proteinase K (34). It is also consistent with the fact that this region represents a highly immunoreactive part of the membraneous sector of intact Ca 2ϩ -ATPase using anti-peptide antibody (54). From the loop-binding site, the bound Ca 2ϩ may gain access to the intramembranous site 1 (formed by Ca 2ϩ liganding groups within M5, M6, and M8) via a slit between the L6 -7 and the more deeply situated L8 -9 loop. If dissociation of Ca 2ϩ from L6 -7 preferentially occurs toward the inside by a flip-like movement of the loop this will result in an overall increase in the binding affinity at site 1 which can be represented schematically by a two-step process.
where EЈCa L6 -7 represents loop binding associated with the conformation of ATPase in an EЈ conformation, while EЉCa (1) represents the binding of Ca 2ϩ at the intramembranous site 1 with ATPase (by an isomerization) in the EЉ conformation. Provided that reactions (Reactions 1 and 2) are strictly coupled, i.e. that formation of EЉCa(1) (as shown) only proceeds via formation of EЈCa L6 -7 , the overall dissociation constant (K d ) for the process E ϩ Ca 7 EЉCa(1), will be given by where K d Ј and K d Љ denote the dissociation constant for Reactions 1 and 2, respectively (note that K d Љ for an isomerization process has no unit). If K d Ј is in the range of 0.05-1 mM as indicated from our study on the Ca 2ϩ binding on the chemically synthetized peptides and p20C, abolition of the binding ability of the L6 -7 loop by mutation of the aspartate residues will lead to a decrease in binding affinity at site 1 of the order 10 3 -10 4 , more than sufficient to account for the decreased dependence of enzyme activity on Ca 2ϩ concentration by the ADA mutant (12). Note also that a single aspartate residue has a much lower affinity for Ca 2ϩ than what we have determined here for the L6 -7 peptides (K d ϭ 25 mM Ref. 55), which suggests that there is a simultaneous binding of calcium to all three asparte residues of the loop, in agreement with the NMR results.
As an alternative possibility, suggested by Zhang et al. (26), mutation of the L6 -7 aspartate groups may induce allosteric effects which result in a decrease of the binding affinity for Ca 2ϩ by the intramembranous sites. Such an effect is, however, unlikely to result simply from the disruption of hydrogen bonds involving the aspartate residues. Inspection of the x-ray structure shows that the hydrogen bonds involving Asp 813 and Asp 818 derive only from side chain contacts of Asp 813 with the side chains of Asn 755 , Ser 917 , and Gln 920 (see Fig. 8). Alanine mutation of Ser 917 or Gln 920 did not, however, alter Ca 2ϩ affinity and ATPase activity (26) and mutant N755A/N756A retained full transport function (56).
These observations therefore raise the possibility that it is changes in non-directional Van der Waals and hydrophobic contacts (e.g. the side chain-side chain contacts between Arg 751 and Pro 821 and, Arg 325 and Pro 812 ), rather than hydrogen bonds, that are likely to be particularly significant in the context of the conformational transitions of the enzyme involving the L6 -7 loop since small side chain variations could be readily translated into a significant structural alteration.
Variations in Ca 2ϩ Binding by Isoforms of Ca 2ϩ -ATPase-With respect to isoform variations of the L6 -7 loop it may be noted that the L6 -7 loop region is 100% conserved among the SERCA 1b (neonatal) as compared with the SERCA 1a (adult) skeletal muscle isoform, whereas there is up to 85% identity with the SERCA 2 (slow-twitch cardiac muscle, and other tissues) and 3 (non-muscle tissue) isoforms. In all forms the putative Ca 2ϩ -binding region, flanked by extended Pro-Pro segments, is characterized by the presence of hydrophobic amino acid residues between the cation liganding groups. In this connection the organization of the hydrophobic side chains Leu 814 , Ile 816 , and Met 817 in the vicinity of the cation bound to the SERCA 1a loop peptide is a noteworthy finding. It may be that the function of these hydrophobic spacers between the metal ion liganding groups is to reduce the rate of rehydration of the bound cation and thereby to increase binding affinity and delivery to the inside of the ATPase. Poch et al. (57) have noted that one of the variant of human SERCA 3 isoforms, where Met 817 is substituted for Ile 817 , displays a 10-fold lower ATPase or calcium transport activity (without a change in apparent calcium affinity). Similarly, Asahi et al. (27) report that the site-directed mutants L814A, I816A, or M817A have Ca 2ϩ transport activity so low that no functional interaction with phospholamban could be evaluated. This illustrates the influence of hydrophobic residues in this region of the protein. This influence is likely to remain also upon substitution of aspartate with asparagine at Asp 818 as it is a characteristic difference between the SERCA 1 and 2 isoforms. In conclusion, the central part of L6 -7 loop represents a prototype sequence which is readily adaptable for binding of calcium.
The Relation between Ca 2ϩ Occlusion and Ca 2ϩ Binding-As a consequence of our findings and the above considerations we here propose the involvement of the L6 -7 loop in both Ca 2ϩ binding and in the formation of an entrance port leading to Ca 2ϩ binding at site 1. By the term "entrance port" we refer to an access pathway controlled by a device (such as the L6 -7 loop suggested above) capable of triggering the transfer of the transported cation toward the interior of the protein. Note that the concept of an entrance port, where there is an interaction between the transported ion and a specific part of the protein, differs from that of "gates" of channel proteins which, at least in principle, open and close independently of the admitted cation/anion. As general characteristics of entrance ports we expect them to have low or intermediate affinity for the transported ion (otherwise admittance toward the inside would be obstructed at concentrations of the transported ion saturating the intramembranous binding sites). On the other hand (due to the low binding levels resulting from low affinities) we also expect them to have high outside/in transition rates which, while not approaching those of channel proteins (10 6 -10 7 s Ϫ1 ), nevertheless would be significantly higher (e.g. of the order of 10 3 -10 4 s Ϫ1 ) than the turnover rate for the complete transport cycle.
Previously, entrance ports for Ca 2ϩ -ATPase have been proposed to be localized at the N-terminal side, between M2, M4, and M6 (4), or along M3, possibly via an interaction with the C-terminal end of L6 -7 loop (58), or along the N-terminal hydrophilic part of the M1 helix, leading to Glu 309 (59). However, it should be noted that at the present stage all assignments of access pathways are tentative, due to the compactness and absence of clearly defined hydrophilic pathways within the Ca 2ϩ -ATPase structure. It may well be that a different access pathway (an N-terminal and a C-terminal one) exists for each of the two bound calcium ions. As pointed out previously (4,59) the compact structure of Ca 2ϩ -ATPase differs from channel proteins like the KscA K ϩ channel and porins where the hydrophilic pathways through the membrane are clearly discern-ible in the reported x-ray structural models. However, this also means that occlusion of Ca 2ϩ in the intramembranous sector can be seen as being a natural consequence of cation enclosure within a compact structure, from which release is only possible in connection with well defined conformational changes to expose the bound Ca 2ϩ . From this perspective the retention of occlusion by mutation of the solvent exposed aspartate residues in the L6 -7 loop is not surprising. However, a corollary of this view is that the lack of occlusion observed after mutation of critical liganding residues in M4, M5, and M6 cannot be attributed to a reduction in intrinsic Ca 2ϩ affinity alone. In addition, it may be supposed that such mutations are accompanied by more profound changes in the membranous sector of the enzyme, leading to the formation of leakage pathways from the intramembranous sites. In agreement with this view we have previously found that mutation drastically increases the susceptibility of the E771Q mutant to proteolysis by proteinase K (12). In the present study we found that treatment of this mutant with C 12 E 8 in the yeast expression system led to complete aggregation which prevented us from performing occlusion experiments with our technique. 3 In the case of the E309Q mutant we also noted some lability after C 12 E 8 treatment, but nevertheless, with this mutant we were able to confirm the absence of Ca 2ϩ occlusion. On the other hand, the E908A mutant, with reduced Ca 2ϩ binding affinity (22) has been reported to be perfectly capable of occluding Ca 2ϩ (16), but in this case distinct transport activity is retained at high Ca 2ϩ concentrations (8). These data can be rationalized in terms of a reduced intrinsic affinity for binding of Ca 2ϩ which allows retention of Ca 2ϩ occlusion within a relatively unperturbed native membrane structure. As a result this mutation results in a phenotype similar to that of the L6 -7 mutants.
Concluding Comments-The present study shows that the isolated peptide corresponding to the L6 -7 loop in solution adopts an independently folded structure, similar to that of the published x-ray structure in Ca 2ϩ -ATPase, but in which the aspartate groups Asp 813 , Asp 815 , Asp 818 by reorientation of the central peptide bonds readily form a Ca 2ϩ -binding site. The fact that mutations of these aspartate residues in Ca 2ϩ -ATPase to alanine do not affect Ca 2ϩ occlusion, while decreasing the turnover rate of the ATPase, is consistent with their participation in Ca 2ϩ binding in an early part of the Ca 2ϩ transport cycle. We propose that the loop may act as a switch by which Ca 2ϩ ions from the cytosol are guided toward intramembranous binding. Given its strategic position in the ATPase structure, the loop may also (4,23,26) at a later stage participate in the conformational changes transmitted to the P-domain to enable this domain to become phosphorylated by ATP.