Calcium transport by sarcoplasmic reticulum Ca(2+)-ATPase. Role of the A domain and its C-terminal link with the transmembrane region.

After treatment of sarcoplasmic reticulum Ca(2+)-ATPase with proteinase K (PK) in the presence of Ca(2+) and a protecting non-phosphorylated ligand (e.g. adenosine 5'-(beta,gamma-methylenetriphosphate), we were able to prepare in high yield an ATPase species that only differs from intact ATPase because of excision of the MAATE(243) sequence from the loop linking the A domain with the third transmembrane segment. The PK-treated ATPase was unable to transport Ca(2+) and to catalyze ATP hydrolysis, but it could bind two calcium ions with high affinity and react with ATP to form a classical ADP-sensitive phosphoenzyme, Ca(2)E1P, with occluded Ca(2+). The ability of Ca(2)E1P to become converted to the Ca(2+)-free ADP-insensitive form, E2P, was strongly reduced, as was the ability of PK-treated ATPase to react with orthovanadate or to form an E2P intermediate from inorganic phosphate in the absence of Ca(2+). PK-treated ATPase also reacted with thapsigargin to form a complex with altered properties, and the tryptic cleavage "T2" site in the A domain was no longer protected in the absence of Ca(2+). It is probable that disrupting the C-terminal link of the A domain with the transmembrane region severely compromises reorientation of A and P domains and the functionally critical cross-talk of these domains with the membrane-bound Ca(2+) ions.

Sarco/endoplasmic reticulum Ca 2ϩ -ATPases (SERCA) 1 may be regarded as prototypes of P-type ATPases, a family of membrane proteins whose function is to actively transport inorganic cations like Ca 2ϩ , Na ϩ , K ϩ , and H ϩ (or even heavy metals), accompanied by the formation of a high energy aspartyl-phosphorylated intermediate during ATP hydrolysis. Recently, the details of the tertiary structure of SERCA1a (the adult skeletal muscle isoform) were obtained by x-ray diffraction analysis of three-dimensional membrane crystals produced from detergent-solubilized ATPase in the presence of Ca 2ϩ (1). Analysis of these crystals confirmed the predicted presence of 10 transmembrane-spanning segments in SERCA1a (2)(3)(4) and revealed the presence of three distinct cytosolic domains, widely separated from each other in the crystal. Two of these domains, the nucleotide binding domain (N) and the phosphorylation domain (P), are formed by the amino acid residues located between the membrane-spanning segments M4 and M5. The third domain (termed A) is formed from both the ATPase N-terminal end and the amino acids located between segments M2 and M3; the latter amino acids compose a central region, Gly 148 -Gln 202 , which is well conserved among P-type ATPases (5) and which had been correctly predicted to adopt a predominantly ␤-structure; at its N-terminal and C-terminal ends, this ␤-region (6) is connected to membrane segments M2 and M3 by extended loops.
The large differences between the "open" structure of the ATPase cytosolic domains deduced from the three-dimensional crystals (1) and the much more "compact" one previously deduced, at lower resolution, from two-dimensional crystals formed in the presence of EGTA and decavanadate (7) suggest that cytosolic domains undergo large scale movements during the transition from one intermediate form to the other during catalysis (1). However, the detailed functional role of these movements over long distances is still rather mysterious. Initial claims that the ␤-region had a special role in coupling hydrolytic activity to ion transport led to its designation as a "transduction domain," but because of the contradictory results of various proteolysis experiments, these claims have not been generally accepted (Ref. 8 versus Ref. 9). Nevertheless, the A domain is probably essential for ATP hydrolysis and cation transport, as suggested by proteolysis and mutagenesis studies, not only for Ca 2ϩ -ATPase (10 -15) but also for Na ϩ ,K ϩ -ATPase (16 -18) and, in part, H ϩ -ATPase (19 -20 and see the review in Ref. 5). Note that a soluble "relative" in the P-type ATPase family lacks not only the transmembrane portion of ion-transport pumps but also the A domain (21).
For SERCA1a Ca 2ϩ -ATPase, we previously found that treatment with V8 protease (a Glu-C endoproteinase) in the presence of Ca 2ϩ cleaved the ATPase closely after the end of the ␤-region, between Glu 231 and Ile 232 (13). Although cleavage at this particular site did not affect the ability of the ATPase to undergo Ca 2ϩ -dependent phosphorylation from ATP, it did block or strongly inhibit the subsequent release of inorganic phosphate after ATPase phosphorylation. When the ATPase was treated with another proteolytic enzyme, proteinase K (PK), many peptides were formed, but again ATPase fragments retained the ability to become phosphorylated from ATP if cleavage had only occurred in the N-terminal region, either between Leu 119 and Lys 120 or between Thr 242 and Glu 243 , i.e. closely before the beginning or closely after the end of the conserved ␤-region (15). The partial loss and partial retention of functional properties after cleavage at these locations suggest that specific steps in the catalytic cycle may be blocked by disruption of the links that the A domain forms with both the transmembrane portion of the ATPase and the phosphorylation domain.
The use of proteolytic dissection as an analytical tool can be an effective way of studying the function and interaction of different Ca 2ϩ -ATPase domains. However, a frequent problem encountered in the use of proteolytic enzymes is the formation of a large variety of cleavage products, resulting from the presence, on different parts of the ATPase molecule, of sites with approximately equal probabilities for proteolytic attack; therefore, it is seldom possible to produce sufficiently large amounts of well defined peptide fragments for detailed investigation of the functional consequences of proteolysis. In a previous study, we nevertheless noticed that the reaction of Ca 2ϩ -ATPase with CrATP not only slowed down proteolysis but also resulted in the formation of fewer cleavage products (4), suggesting that the C-terminal part and central portion of the ATPase had acquired protection against proteolytic attack. We therefore examined the effect of other non-hydrolyzable ATP analogues on Ca 2ϩ -ATPase proteolysis by PK. We show here that after proper adjustment of medium composition, especially of the divalent cation concentrations, it is possible in the presence of non-phosphorylating ATP analogues to restrict the effect of PK mainly to cleavage at a few adjacent peptide bonds located between Gln 238 and Gln 244 at the end of the A domain, thus cutting out the small MAATE 243 sequence from the ATPase; for the sake of simplicity, we will refer to this excision as "cleavage at Glu 243 ," to relate it to the proteinase K cleavage site between Thr 242 and Glu 243 that we described previously (15). It was possible in the present study to produce large amounts of this cleaved ATPase, which allowed detailed characterization of the perturbation of the ATPase catalytic cycle resulting from this cleavage. This enabled us to delineate more precisely the functional and structural role of the A domain and its C-terminal extension for the intermediary processes involved in Ca 2ϩ transport.

MATERIALS AND METHODS
Membrane Preparation and Proteolysis by PK-Ca 2ϩ -transporting sarcoplasmic reticulum (SR) vesicles were isolated from rabbit skeletal muscle according to established procedures (22,23). From these preparations, purified Ca 2ϩ -ATPase membranes were obtained by extraction of extrinsic proteins with a low concentration of deoxycholate (24). For controlled proteolysis, membranes were generally suspended at a concentration of 2 mg of protein/ml in a 100 mM MOPS-NaOH medium (pH 6.5) containing 0.3 mM Ca 2ϩ and 0.5 mM AMP-PCP (Sigma). The preparations were then treated with PK, usually at a protein concentration of 0.03 mg/ml at 20°C for 1-2 h when SR vesicles were used, and with 0.06 mg of PK/ml at 23°C for 15-45 min when purified Ca 2ϩ -ATPase membranes were used. Proteolysis was stopped by adding 0.5 or 1 mM PMSF and storing on ice for 10 min. At this stage the preparation was sometimes used immediately, e.g. to determine residual ATPase activity, but usually it was further processed by pelleting the membranes and resuspending them in an appropriate medium to remove PK, AMP-PCP, and any peptide material released from the membrane by proteolytic treatment. Sometimes the PK-treated and pelleted membranes (referred to as PK-SR) were resuspended in 0.3 M sucrose and 100 mM KCl (pH 7.4), frozen in liquid nitrogen, and stored at Ϫ80°C for subsequent use. Otherwise, experiments were conducted within a few hours of PK treatment. The typical appearance of PK-SR after SDS-PAGE is shown in Fig. 4A below, which illustrates the dominant presence of two proteolytic cleavage products, labeled p83C and p28N, plus a few percentage of intact ATPase (as well as the expected calsequestrin and M55 bands when SR vesicles were used instead of purified ATPase).
Detection of Proteolytic Fragments and Estimation of Protein Concentrations-After proteolysis arrest, aliquots were added to a ureacontaining denaturation buffer (essentially as described in Ref. 25), boiled, and loaded onto 12% SDS-PAGE Laemmli gels generally prepared in the presence of 1 mM Ca 2ϩ (about 4 g of protein per lane, unless otherwise noted). The Coomassie Blue staining intensity of individual peptidic bands was deduced from gel scanning (Molecular Analyst). N-terminal and C-terminal fragments of Ca 2ϩ -ATPase were identified on Western blots after transfer to PVDF membranes as described previously (15). Total protein concentrations were estimated either by the Lowry method, using bovine serum albumin as standard, or from absorbance measurements at 280 nm after membrane solubilization with 1% SDS. In the latter case we used a value for specific protein absorption by PK-SR membranes that was slightly higher than that used for absorption by intact SR vesicles (1.15 instead of 1.0 for the absorption measured for a protein concentration of 1.0 mg/ml in a cuvette with a 1-cm path length). The correction factor was deduced from parallel A 280 nm and Lowry determinations of the protein content of proteolyzed and non-proteolyzed preparations. The reason for the slightly higher absorption coefficient of proteolyzed membrane preparations is that, in addition to causing selective cleavage at Glu 243 of most of the ATPase, proteolysis also causes extensive cleavage of a fraction of the ATPase, with release of a number of cytosolic peptides to the supernatant, whereas the rest of the peptidic material, mainly consisting of transmembrane segments with a preponderance of tryptophan and other aromatic residues, remains associated with the membranes (see below).
N-terminal Sequencing and Mass Spectrometry-The N terminus of the longest proteolytic fragment (p83C) produced in large quantity by treatment with PK was sequenced with an Applied Biosystem 473-A gas phase sequencer after SDS-PAGE separation and blotting onto PVDF membranes (15). No amino acid was detected when this protocol was applied to p28N, in agreement with the fact that this peptide starts with a blocked (acetylated) N-terminal methionine residue corresponding to the N terminus of Ca 2ϩ -ATPase. For identification of the C-terminal end of the p28N peptide, we analyzed by MALDI-TOF mass spectrometry the peptide fragments produced by further cleavage of the p28N peptide, with trypsin, at the well known T2 tryptic cleavage site, between Arg 198 and Ala 199 . This additional step was included in the protocol in order to diminish the size of the fragments to be identified (15). Thus, 10 mg/ml PK-SR membranes were treated with 0.2 mg/ml trypsin for 60 min in ice-cold medium containing 130 mM sucrose, 40 mM KCl, 9 mM MOPS-Tris (pH 7.3), 0.3 mM Ca 2ϩ , and 0.5 mM AMP-PCP. The reaction was stopped by adding 0.4 mg/ml trypsin inhibitor and 0.5 mM PMSF, and the medium was acidified with 10% formic acid to facilitate dissociation of the C-terminal p4 peptide expected to result from trypsin treatment of p28N. 20 l of this sample was then centrifuged in a TLA 100 rotor (55,000 rpm, 3 min, 8°C), and the supernatant was examined by MALDI-TOF mass spectrometry with or without prior Ziptip desalting, using a PerkinElmer Life Sciences Applied Biosystems Voyager System 4080 (e.g. Ref. 26).
Test Media and ATPase Measurements-In many experiments, proteolyzed as well as untreated membranes were diluted in a medium referred to as buffer A, containing 100 mM KCl, 5 mM Mg 2ϩ , 50 mM MOPS/Tris (adjusted to pH 7.0 at either 4 or 20°C), and various concentrations of Ca 2ϩ . Low concentrations of free Ca 2ϩ were buffered with either EGTA or MgEDTA using the following apparent dissociation constants for Ca 2ϩ : K d(Ca-EGTA) ϭ 0.45 M and K d(Ca-MgEDTA) ϭ 62 M. To facilitate study of the reactions involving the ADP-insensitive E2P species, the medium (now referred to as buffer B) was modified to contain dimethyl sulfoxide (Me 2 SO, at concentrations of 20 -40% v/v), no KCl, 100 mM MOPS-NaOH (pH 7.0, 20°C), and 10 mM Mg 2ϩ , as well as various concentrations of 32 P-labeled P i when phosphorylation from P i was measured. ATPase activity was usually measured spectrophotometrically, with a coupled enzyme assay, in a medium (referred to as buffer C) containing 100 mM KCl, 1 mM Mg 2ϩ , 50 45 Ca 2ϩ content in the wet volume by taking into account either the counts of 3 H present in the same filter or the counts of 45 Ca 2ϩ present on the lower, protein-free filter. In other experiments the rate of 45 Ca 2ϩ dissociation was measured by filtration through similar Millipore filters with the aid of the rapid filtration biologic equipment, as described previously (27).
Phosphorylation-ATP-dependent phosphorylation was generally measured at 4°C, by incubating the membranes for 10 -30 s with buffer A (adjusted to pH 7.0 on ice) supplemented with 10 -25 M [␥-32 P]ATP and various concentrations of Ca 2ϩ . In some cases, samples were then loaded onto Millipore HA filters and immediately flushed twice with 3 ml of a solution containing 10% (w/v) trichloroacetic acid and 10 mM pyrophosphate, and finally with 20 ml of a solution containing 0.1% trichloroacetic acid and 10 mM phosphate. In other experiments, ATPase phosphorylation was quenched in the test tube by adding 0.5 M perchloric acid, 15 mM carrier P i , and 0.5 mM carrier ATP to the sample, and denatured phosphoenzyme was recovered by filtration on a Gelman A/E glass fiber filter and thorough rinsing (28). In some cases, phosphorylation from ATP was also measured at 20°C in buffer B supplemented with 40% Me 2 SO, 100 M Ca 2ϩ , and [␥-32 P]ATP. Phosphorylation from inorganic phosphate was measured at 20°C by incubating the membranes in buffer B supplemented with 20 -40% Me 2 SO, 0.2 mM 32 P i , and 1 mM EGTA, followed by acid quenching.
Reconstitution Procedure for Ca 2ϩ Uptake Measurements-C 12 E 8solubilized Ca 2ϩ -ATPase was reconstituted with cholate-solubilized dioleoylphosphatidylcholine (DOPC) by a dialysis procedure (29). Reconstitution was carried out at a protein-to-lipid weight ratio of 1:100, with inclusion of 0.1 M phosphate in the dialysis buffer to permit precipitation of Ca 2ϩ inside the proteoliposomes during the subsequent 45 Ca 2ϩ transport assays. The reconstituted vesicles were fractionated by sucrose density centrifugation on a 3-12% (w/v) linear gradient, and the fraction that remained on top was used for these 45 Ca 2ϩ transport assays, performed before and after treatment with PK.
Fluorescence Measurements-Tryptophan fluorescence was measured as described previously (e.g. Refs. 30 and 31). Time-resolved measurements of the Ca 2ϩ -induced changes in tryptophan fluorescence were performed in stopped-flow Biologic equipment by mixing 1:1 (v/v mixture) membranes suspended at 0.4 mg/ml in the presence of 0.5 mM EGTA, with buffer containing 1 mM Ca 2ϩ . The dead time was 3 ms. The excitation beam was at 290 nm, and emitted light was collected through a wide band combined filter (MTO J324 ϩ MTO A340). Changes in fluorescence of bound fluorescein isothiocyanate (FITC) were also measured. For this purpose, labeling with FITC was performed by incubating control SR or PK-SR membranes at about 2 mg/ml protein with substoichiometric amounts of FITC (either 4 or 2 nmol of FITC/mg of protein) for 40 min at 20°C and pH 8, followed by return to neutral pH and storage on ice (32). We found previously (33) that in the isolated water-soluble p29/30 fragments formed after extensive ATPase proteolysis with PK, FITC specifically reacted with Lys 515 , as in intact ATPase. We therefore anticipated that in the p83C-28N complex resulting from milder proteolysis, FITC would also react with Lys 515 , again in the same manner as with intact ATPase. In agreement with this expectation, we found that upon covalent binding to ATPase cleaved at Glu 243 , FITC exhibited the same typical shift in its excitation spectrum as that of intact ATPase (data not shown). Because the concentration of residual free FITC is expected to be negligible under conditions of sub-stoichiometric labeling, the fluorometric properties of bound FITC were examined after simply diluting the FITC-treated sample 20-fold in buffer A at 20°C. Excitation and emission wavelengths were 495 and 520 nm, respectively.

Proteinase K Treatment of Ca 2ϩ -ATPase in the Presence of Ca 2ϩ and Non-phosphorylating ATP Analogues Mainly Results in Excision of the MAATE 243 Short Sequence
We found previously (15) that treatment of SR vesicles with proteinase K in the presence or absence of Ca 2ϩ leads to fast proteolysis at several exposed regions of Ca 2ϩ -ATPase, result-ing in the formation of several well characterized proteolytic fragments. For instance, the main fragments transiently formed during proteolysis in the presence of Ca 2ϩ (see lanes 2-5 in the left part of Fig. 1A) include peptides derived from the N-terminal part (p81N and p28N) or the C-terminal part (p83C, p27/28C, and p19C) of the ATPase, together with watersoluble p29/30 fragments relatively resistant to protease and revealing the central nucleotide binding domain (33). It was recently reported that adding the non-hydrolyzable ATP analogue AMP-PCP in the presence of Ca 2ϩ and Mg 2ϩ slows down FIG. 1. PK treatment in the presence of Ca 2؉ and AMP-PCP mainly cleaves Ca 2؉ -ATPase to p28N (residues 1-238) and p83C (residues 244 -994). A, SR membranes (2 mg of protein/ml) were treated with PK (0.03 mg/ml) at 20°C in proteolysis buffer (100 mM MOPS-NaOH (pH 6.5), 0.3 mM Ca 2ϩ , and 15 mM residual sucrose) in the absence (lanes 2-5) or presence (lanes 6 -9) of 0.5 mM AMP-PCP. Lanes 1 and 10 contain molecular mass markers (Amersham Biosciences kit). B, SR membranes were similarly treated for 1 h and 45 min in the presence of AMP-PCP, centrifuged after proteolysis arrest, resuspended, and frozen in liquid nitrogen for storage (PK-SR). Later, these PK-treated membranes, together with intact membranes, underwent SDS-PAGE (with only 0.8 g per lane in this case). The peptides were transferred to PVDF membranes and immuno-characterized with antibodies against the N terminus (1-15 residues, left) or the C terminus (985-994 residues, right) of the ATPase, as described in Ref. 15. On the original Western blots, small amounts of other fragments are visible, e.g. p27/28C and p19C (as revealed by the C-terminal antibody) corresponding to cleavage at Val 734 /Val 747 and Asp 818 (cf. Ref. 15). C, to identify the C terminus residue of p28N, PK-treated and resuspended SR membranes underwent further proteolysis, this time in the presence of trypsin, to obtain cleavage at both the T1 (Arg 505 -Ala 506 ) and T2 (Arg 198 -Ala 199 ) sites, located in p83C and p28N, respectively (see "Materials and Methods"). The supernatant of an acidified aliquot containing the putative fragment extending from the trypsin T2 site to the p28N C terminus (p4) was examined by MALDI-TOF, together with known peptides used for calibration purposes, namely ACTH fragment 7-38 (molecular mass 3659.96 Da) and bovine insulin (molecular mass 5737.77 Da).
ATPase cleavage by PK to a very large extent (34), and we fully confirm this (data not shown).
We have now found that in the absence of Mg 2ϩ and K ϩ (but the presence of Ca 2ϩ ), i.e. under conditions where cleavage at Glu 243 is faster than in the presence of Mg 2ϩ (15), adding AMP-PCP still reduces the rate of overall ATPase degradation (see lanes 6 -9 in the right part of Fig. 1A), but simultaneously reduces to an even greater extent secondary cuts at other positions, resulting in the prominent accumulation of two polypeptide fragments with apparent molecular masses of 83 and 28 kDa. Only minor amounts of other fragments of appreciable length are formed under these conditions (on these gels, short transmembrane segments were left undetected). Probing with N-and C-terminal specific antisera (Fig. 1B) reveals that the larger polypeptide fragment is derived from the C-terminal part of SERCA1a and the smaller fragment from its N-terminal part, suggesting that these two fragments might correspond to the p83C and p28N fragments previously demonstrated to result from cleavage of intact Ca 2ϩ -ATPase in the absence of AMP-PCP at the peptide bond between Thr 242 and Glu 243 (15).
To establish the exact identity of these fragments, the p83C peptide formed after 2 h of proteolysis in the presence of AMP-PCP was submitted to N-terminal sequencing following SDS-PAGE separation and transfer to PVDF membranes. The first N-terminal amino acids found were 244 QDKT and not 243 EQDKT as reported previously (15) for the p83C formed after only a few minutes of proteolysis in the absence of nucleotide. This shift by one amino acid probably corresponds to further cleavage of the N-terminal residue of p83C, formed after an initial cut at Thr 242 -Glu 243 (in this connection, a peptide starting at Gln 244 was also previously detected after extensive cleavage by V8 protease, see Ref. 35). We then attempted to establish whether the complementary peptide, p28N, did end at Thr 242 or not. For this purpose, ATPase first underwent 2 h of cleavage with PK, followed by centrifugation, and the resuspended membranous p83C-28N complex (PK-SR) was subsequently treated with trypsin under conditions expected to result in cleavage of p28N at the T2 site (between Arg 198 and Ala 199 ) and therefore formation of a short peptide, "p4," corresponding to the portion of p28N C-terminal to this cleavage site. A single short p4 peptide was indeed found by MALDI-TOF mass spectrometry (Fig. 1C), but with an M r ϭ 4119.6 unambiguously showing that it ended at Gln 238 , i.e. a few residues before Thr 242 . Our proteolysis conditions therefore mainly result in the excision of the short intervening MAATE 243 sequence from the ATPase. In the rest of this paper, for the sake of simplicity, this excision will usually be simply referred to as "cleavage at Glu 243 ." In our procedure, the use of 0.5 mM AMP-PCP as the standard protective ligand is expected to saturate the nucleotidebinding site (36 -38), and AMP-PCP concentrations of a few 10s of M were indeed sufficient to promote significant accumulation of p83C and p28N in the presence of Ca 2ϩ (data not shown; see also Fig. VI in the Supplemental Material). As an alternative to the addition of AMP-PCP, we found that the addition of other non-phosphorylating ATP analogues like AMP-PNP and TNP-ATP to the Ca 2ϩ -containing proteolysis medium also induced p83C and p28N accumulation (Fig. IA in the Supplemental Material). A qualitatively similar effect was observed after the reaction of Ca 2ϩ -ATPase with CrATP, whereas guanosineor cytosine-containing trinucleotides (5 mM) and acetylphosphate (10 mM) failed to exert the same effect (data not shown). The presence of moderate concentrations of Ca 2ϩ (here 0.3 mM) was essential for accumulation of p83C and p28N after long proteolysis periods; in the presence of EGTA, AMP-PCP (and other non-phosphorylating ATP analogues, data not shown) hardly had any effect on ATPase degradation (Fig. IB in the Supplemental Material). Note that, at variance with the results that can be obtained in the presence of Mg 2ϩ and K ϩ (Ref. 34 and data not shown), we found that in the absence of Mg 2ϩ and K ϩ (but in the presence of Ca 2ϩ ), ADP was as effective as AMP-PCP for partially protecting Ca 2ϩ -ATPase from initial cleavage as well as for protecting p83C and p28N from further degradation (lanes 5 and 7 in Fig. IB in the Supplemental Material). This shows that in the absence of Mg 2ϩ and K ϩ , the adenosine moiety of the nucleotides plays a significant role in protection against extensive proteolysis, but a ␥-phosphate is not required (nor is phosphorylation).
After proteolytic cleavage close to Glu 243 , the p83C and p28N fragments remain associated, and the cleaved Ca 2ϩ -ATPase maintains at least some of its structural features. This was indicated to us by the fact that after solubilization of PKtreated membranes with the mild non-ionic detergent C 12 E 8 , size exclusion chromatography in the presence of this detergent revealed that the N-terminal and C-terminal parts of the ATPase eluted together at the same position and retained exactly the same elution profile as that of the uncleaved, monomeric Ca 2ϩ -ATPase. By contrast, denatured ATPase fragments were eluted from the column in an aggregated state, as a broad band between the void volume and the monomeric peak

Catalytic and Transport Activity of Cleaved Ca 2ϩ -ATPase; Overall Loss of Ca 2ϩ Transport and ATP Hydrolysis Activity
We first examined the overall functional consequences of PK-dependent cleavage at Glu 243 . We found that this cleavage led to the virtually complete disappearance of ATPase hydrolytic activity, both for SR vesicles made leaky by addition of a Ca 2ϩ ionophore and for ATPase solubilized by detergent ( Fig.  2A). The ability of Ca 2ϩ -ATPase to become phosphorylated from P i in the absence of Ca 2ϩ was also drastically reduced after proteolysis (Fig. 2B, circles). In contrast, most of the Ca 2ϩ -dependent phosphorylation from ATP and of the ability of the ATPase to bind Ca 2ϩ in the absence of ATP were preserved (ϳ50 -70% after 2 h, triangles and squares in Fig. 2B) to an extent roughly corresponding to the sum of p83C-28N complex plus remaining intact ATPase (Fig. 2C, squares). On the other hand, the residual ATP hydrolysis rate observed after extensive proteolysis (about 5-10% of the original rate), like the residual ability to be phosphorylated from P i , roughly corresponded to the proportion of ATPase that remained intact (circles in Fig. 2C). These observations suggest that the p83C-28N complex has lost most of its hydrolytic activity but has nevertheless retained the ability to bind Ca 2ϩ and to become phosphorylated by ATP, properties that are completely lost after further breakdown into smaller products (15).
To account for the moderate reduction in the ATPase ability to bind Ca 2ϩ and become phosphorylated by ATP, it should be kept in mind that stabilization of the p83C-28N complex is of course not absolute, even in our empirically optimized proteolysis medium in the presence of relatively high concentrations of AMP-PCP. After thorough proteolysis (1-2 h), this complex amounts to not more than 50 -70% of the original ATPase content (Fig. 2C), because part of the ATPase has been further degraded into smaller peptides (including the water-soluble p29/30 fragments and the membranous p19C fragments mentioned above; the latter degradation products are faintly visible in lane 9 of the gel shown in Fig. 1A, and are represented as a schematic in the inset to Fig. 2C).
To address the question to what extent the low Ca 2ϩ -depend-ent ATPase activity that persists after long proteolysis periods is attributable to the residual intact ATPase, rather than to a slow but significant turnover activity of the p83C-28N complex, we measured the residual ATPase activity of samples subjected to prolonged PK treatment in the absence of AMP-PCP. Under these conditions, as expected, p28N and p83C peptides were initially formed but were subsequently degraded into smaller fragments; yet a small fraction of ATPase remained uncleaved, almost the same as in the presence of AMP-PCP. Samples treated this way were able to hydrolyze ATP at about the same residual rate (0.6 mol/mg/min) as when proteolysis had been performed in the presence of AMP-PCP (0.8 mol/min/mg). This suggests that most of the residual ATP hydrolysis activity present after extensive proteolysis ( Fig. 2A) actually originates from residual intact ATPase in the PK-treated preparation, which for some unclarified reason resists proteolysis (perhaps because it is present in the membrane as aggregates). Further evidence in favor of a vanishingly low activity of the p83C-28N complex will be provided below, in connection with our study of partial ATPase reactions. We also investigated the passive permeability of SR membranes after PK treatment. Ca 2ϩ accumulation by tight intact vesicles is known to result in a very strong inhibition of ATPase activity, revealed by a stimulatory effect of ionophore or detergent on ATPase activity. After 15 or 30 min proteolysis, i.e. after cleavage of more than half the ATPase molecules in the vesicle preparation, significant stimulation of ATPase activity by ionophore or detergent persisted ( Fig. 2A), suggesting that the PK-treated vesicles had not become completely leaky. Independent passive permeability measurements indicated that the permeability of SR vesicles to Ca 2ϩ did increase during proteolysis, but only moderately. After cleavage of more than 80% of the ATPase, the rate of Ca 2ϩ efflux from passively loaded SR vesicles was still measurable, although it had increased 30-fold (data not shown). It is, however, not clear whether this increased permeability was attributable to the formation of the p83C-28N complex per se or, alternatively, to its subsequent cleavage into smaller transmembrane segments (or else, to increased permeability through other PK-sensitive pathways present in the SR membrane).
To test the ability of PK-treated ATPase to transport Ca 2ϩ , we therefore used Ca 2ϩ -ATPase reconstituted into phosphatecontaining proteoliposomes with a very low protein-to-lipid ratio. This is a system in which the effect of permeability changes on Ca 2ϩ accumulation can be expected to be minimal, both because of the presence of Ca 2ϩ -precipitating phosphate inside the vesicles and because the phospholipid bilayer component of the proteoliposomes plays a dominant role for their permeability. In one series of experiments, reconstituted proteoliposomes were first treated with PK for various periods. As Fig. 3 shows, Ca 2ϩ accumulation into proteoliposomes was significantly reduced after 30 min of proteolysis, and uptake was almost stopped after 60 min of proteolysis; after longer periods Ca 2ϩ uptake was completely blocked. This striking effect of PK treatment could not be ascribed to an increase in Ca 2ϩ permeability of the proteoliposomes, because in a separate series of experiments, when proteoliposomes had first been allowed to accumulate 45 Ca 2ϩ and only then treated with PK, it appeared that even after 1 h of PK treatment previously accumulated 45 Ca 2ϩ was not released, even if EGTA was included in the medium to assay permeability (data not shown). SDS-PAGE analysis indicated that during PK treatment of these proteoliposomes only 40 -50% of ATPase was converted into p28N and p83C, i.e. ATPase degradation was much less extensive than after PK treatment of SR vesicles. The presence of an appreciable fraction of undegraded but non-transporting molecules of reconstituted ATPase can be easily understood, as after asymmetric incorporation of ATPase into the proteoliposomes, only those ATPase molecules that have been reconstituted with their cytosolic head facing the medium will be able to transport Ca 2ϩ and be sensitive to PK (4). Thus, the drastic drop in Ca 2ϩ accumulation by these proteoliposomes after partial ATPase conversion into the p83C-28N complex is further evidence in favor of the inability of the complex to transport Ca 2ϩ .
FIG. 2. PK treatment of ATPase in the presence of Ca 2؉ and AMP-PCP results in loss of ATP hydrolysis but retention of the ATPase ability to bind Ca 2؉ and be phosphorylated from ATP, although not from P i . SR membranes (2 mg of protein/ml) were treated with PK (0.03 mg/ml) at 20°C in the presence of 0.5 mM AMP-PCP, as described in the legend to Fig. 1A, and the reaction was quenched after various periods. A, Samples were tested for their residual ATPase activity after 100-fold dilution in the assay medium (buffer C, see "Materials and Methods"), under control conditions (tight SR, closed circles), or in the presence of 0.5 mg/ml C 12 E 8 detergent (solubilized SR, squares) or of 0.5 g/ml A23187 ionophore (permeabilized SR, open circles). Inset, the same samples were run on a Laemmli gel, prepared here without Ca 2ϩ . B, samples were tested for their ability to undergo phosphorylation after 10-fold dilution in buffer B (see "Materials and Methods") supplemented with 40% Me 2 SO and either 1 mM EGTA and 0.2 mM 32 P i (circles) or 0.1 mM Ca 2ϩ and 0.02 mM [␥-32 P]ATP (triangles). EP levels were plotted as % of their initial values (4.7 Ϯ 0.3 nmol/mg for EP formed from P i in the absence of Ca 2ϩ , and 4.1 Ϯ 0.2 nmol/mg for EP formed from ATP in its presence). Samples were also tested for their ability to bind Ca 2ϩ (squares) in an assay medium (buffer A) containing 50 M (total) 45 Ca 2ϩ and 50 M [ 3 H]glucose. Ca 2ϩ binding was plotted as % of its initial value (11.9 Ϯ 0.6 nmol/mg). Data points are the means of triplicate measurements. C, SDS-PAGE gels were stained with Coomassie Blue and scanned. The relative intensities of the residual uncleaved ATPase (circles), the p83C and p28N fragments (triangles and diamonds, respectively), and the sum of these three major bands (squares) are plotted. Inset, shows the approximate proportions (after 2 h of cleavage) of residual uncleaved ATPase (5-10%, far right), p83C-28N complex (50 -70%, center), and of ATPase fully cleaved into small transmembrane fragments (30 -40%, left).

Detailed Study of Partial ATPase Reactions
Ca 2ϩ Binding or Dissociation and Ca 2ϩ Binding-Induced Transitions Are Quasi-normal after ATPase Cleavage-To study the functional properties of the p83C-28N complex in more detail, we first compared, at equilibrium (Fig. 4B), the 45 Ca 2ϩ binding properties of PK-treated and resuspended SR membranes with those of intact SR membranes (the corresponding SDS-PAGE gels are shown in Fig. 4A). The p83C-28N complex retained the ability to bind Ca 2ϩ with high affinity both at 4°C (the temperature chosen for the experiment illustrated in Fig. 4B) and at room temperature, as well as in media with pH values ranging from 7.0 to 7.5 and Mg 2ϩ concentrations ranging from 1 to 5 mM (data not shown). Ca 2ϩ binding to PK-treated SR membranes in fact occurred with a marginally higher affinity than for intact SR membranes; thus, at 4°C (pH 7.0) and 5 mM Mg 2ϩ , the half-saturation pCa value for high affinity Ca 2ϩ binding to PK-SR was 5.66 Ϯ 0.05, compared with 5.47 Ϯ 0.05 for intact SR (Fig. 4B). When expressed as nanomoles of Ca 2ϩ bound per mg of total protein, the maximal binding capacity was slightly lower for PK-treated than for intact membranes (10.4 Ϯ 1.4 versus 13.2 Ϯ 1.5 nmol/mg), but this difference is entirely attributable to the above-mentioned presence, in addition to the p83C-28N complex, of small transmembrane proteolytic fragments resulting from extensive degradation of part of the ATPase. 45 Ca 2ϩ binding was drastically reduced by thapsigargin (TG) for both PK-treated and intact membranes (see the diamond and the square at time 0 in Fig. 4, D and C, respectively), which is evidence that TG binds to the Ca 2ϩ -free p83C-28N complex, although perhaps less efficiently than to intact ATPase (this will be further discussed below). The similarity between the p28N-p83 complex and intact ATPase as regards high affinity Ca 2ϩ binding was further documented by the fact that the tryptophan fluorescence level of PK-treated SR membranes exhibited a similar sensitivity to Ca 2ϩ as that of intact membranes (again with a slightly higher affinity for PK-SR). The relative amplitude of Ca 2ϩ -induced fluorescence changes was smaller in the PK-treated membranes (data not shown here,  40 Ca 2ϩ (closed symbols). Zero time measurements (equilibrium binding) were obtained without rinsing. Similar equilibrium binding was measured in the presence of thapsigargin (the TG to ATPase ratio was 1% w/w) and resulted in 45 Ca 2ϩ binding levels illustrated by the square and the diamond in C and D. E and F, kinetics of the Ca 2ϩ -induced changes in tryptophan (Trp) fluorescence at two different pH. SR (E) or PK-SR (F) membranes, initially at 0.4 mg/ml in the presence of 0.5 mM EGTA, were mixed 1:1 in a stopped-flow machine with buffer containing Ca 2ϩ (the final free Ca 2ϩ concentration was 250 M), and Trp fluorescence was then recorded. The medium (at 20°C) was buffered either at pH 6 (it then contained 100 mM MES-Tris and no Mg 2ϩ , in the absence (bottom traces) or presence (middle traces) of 10 M ADP, present from the start) or at pH 7 (top traces; this was buffer A, containing 5 mM Mg 2ϩ ). At pH 6, the fits to single exponential processes gave rate constants of 1.5 and 6.9 s Ϫ1 for SR in the absence and presence of ADP, respectively, and 2.0 and 5.8 s Ϫ1 for PK-SR in the absence and presence of ADP, respectively. At pH 7, the fits to double exponential processes gave rate constants of 7.5 Ϯ 0.5 and 31 Ϯ 4 s Ϫ1 for SR, and 6.9 Ϯ 1 and 28 Ϯ 5 s Ϫ1 for PK-SR. The ratio of fast to slow amplitudes was 1.4 (Ϯ 0.4) for SR and 2.2 (Ϯ 0.5) for PK-SR.
Proteolytic Excision of MAATE 243 from SERCA1a Ca 2ϩ -ATPase but see the difference in fluorescence levels in the stopped-flow experiments shown below, Fig. 4, E and F, or the EGTAinduced drop in equilibrium fluorescence illustrated in Fig. VI in the Supplemental Material), presumably because of the presence of Ca 2ϩ -insensitive tryptophan residues in the transmembrane segments of the fraction of ATPase extensively proteolyzed. In complementary experiments, we also labeled intact and PK-treated SR with FITC, an extrinsic label specifically bound to Lys 515 , and we found that the fluorescence level of bound FITC varied in about the same range of concentrations of free Ca 2ϩ for the two types of membranes (Fig. III in the Supplemental Material).
Rapid filtration experiments (Fig. 4, C and D) revealed also that the kinetics of 45 Ca 2ϩ dissociation from PK-treated SR membranes exhibited properties similar to those found for intact SR; thus, when 40 Ca 2ϩ was added to the dissociation medium, only one of the two previously bound 45 Ca 2ϩ ions (the "superficially" bound 45 Ca 2ϩ ion) was released quickly in exchange with 40 Ca 2ϩ , whereas the other one (the "deeply" bound 45 Ca 2ϩ ion) only exchanged very slowly. This observation indicates retention by PK-treated ATPase of the sequential dissociation mechanism for Ca 2ϩ (see Ref. 39; see other references in Ref. 27). When the dissociation medium contained EGTA to chelate Ca 2ϩ , both 45 Ca 2ϩ ions dissociated rapidly, with a halftime that was slightly increased from 40 -60 to 80 -100 ms after PK treatment. This suggests that the slight increase in equilibrium binding affinity observed after PK treatment (Fig.  4B), if significant, mainly reflects a slightly reduced dissociation rate, whereas the kinetics of binding probably remain essentially unchanged.
In agreement with the above conclusions, we found that the rate at which Ca 2ϩ binding induces changes in the ATPase intrinsic fluorescence was also very similar for PK-treated (Fig.  4F) or intact SR membranes (Fig. 4E) under several different conditions; this was the case at pH 7 as well as at pH 6 and in the presence or absence of Mg 2ϩ . At pH 6, the Ca 2ϩ -induced rise in Trp fluorescence was slow and monophasic for both types of membranes. At pH 7 in the presence of Mg 2ϩ , it was faster and biphasic, again for both types of membranes, and the PK treatment only had marginal effects on either the relative amplitudes or the rate constants of these two phases (see Refs. 23 and 40 -44 for interpretation of these pH-dependent changes, as well as additional explanations in the Supplemental Material). When intact or PK-treated SR membranes were labeled with FITC, the kinetics of the Ca 2ϩ -induced changes in FITC fluorescence were also fairly similar in both cases (data not shown).
Stopped-flow Trp fluorescence experiments at pH 6 also provided a straightforward way to test whether the calcium binding induced transition was still modulated by low concentrations of nucleotide after PK treatment, as it is in intact ATPase (e.g. see Ref. 45). As shown by the two lower traces in Fig. 4F (and compare with the two lower traces in Fig. 4E), this was definitely the case, at least qualitatively, since 10 M ADP accelerated the rate of the observed spectroscopic transition for both PK-treated and intact SR membranes. Thus, ATPase cleavage at Glu 243 has not ruined the ability of nucleotides to modulate the rate of the Ca 2ϩ -induced transition.

Phosphorylation from ATP in the Presence of Ca 2ϩ and Ca 2ϩ Occlusion by This Phosphoenzyme Also Are Quasi-normal after
ATPase Cleavage-In further experiments we compared the ability of PK-treated and intact SR membranes to become phosphorylated from [␥-32 P]ATP in the presence of various concentrations of Ca 2ϩ . The data obtained revealed a slightly stronger apparent affinity and cooperativity for PK-treated than for intact membranes (half-saturation pCa values were 6.85 Ϯ 0.05 and 6.59 Ϯ 0.05, respectively, Fig. 5A). There was almost no difference in the maximal amount of EP that could be formed by the two different membrane preparations, despite the previously described fact that maximal 45 Ca 2ϩ binding was lower in PK-treated membranes (Fig. 4B). The reason for all these facts presumably is that phosphorylation from ATP of proteolyzed ATPase is more complete than it is for intact ATPase, consistent with the slower turnover at steady state indicated by the activity data in Fig. 2A. The fact that steadystate phosphorylation profiles for both intact and PK-treated membranes (Fig. 5A) exhibit apparent affinities for Ca 2ϩ about 1 pCa unit higher than those found for the equilibrium Ca 2ϩ binding profiles (Fig. 4B) is consistent with the well known fact that phosphoenzyme turnover is rate-limiting for the catalytic cycle in both cases.
The kinetic properties of the phosphoenzyme formed from ATP at 4°C were examined in detail. As shown by the open symbols in Fig. 5, B and C, this phosphoenzyme was fully sensitive to ADP both in intact and in PK-treated SR membranes, being rapidly and completely dephosphorylated in the presence of ADP ϩ EGTA (open diamonds). This result clearly demonstrates that the proteolytic cut at Glu 243 does not prevent the formation of the Ca 2ϩ -dependent ADP-sensitive Ca 2 E1P form of phosphorylated ATPase. By contrast, the apparent rate of dephosphorylation of PK-treated phosphorylated ATPase observed after addition of EGTA alone, i.e. in the "forward" direction (open squares), was slowed down, consistent with the reduction in ATPase activity observed at 20°C.
The true forward rates of dephosphorylation at 4°C probably are even slower than the rates suggested by these curves, because the small amount of ADP that has been generated before EGTA addition makes a significant contribution to the observed dephosphorylation rates. Thus, in additional experiments, we measured the rate of dephosphorylation after diluting the phosphoenzyme (formed at 4°C) with a large volume (at 20°C) of the same assay medium as the one used for measuring ATPase activity in the experiment illustrated in Fig. 2A, a medium that also contained 2 mM MgATP (likely to chase any newly formed ADP) and an ATP-regenerating system. This dilution protocol at 20°C resulted in rapid forward dephosphorylation of the phosphoenzyme formed from intact ATPase, too rapid to be measured reliably (closed squares in Fig. 5B); in contrast, phosphoenzyme formed from PK-treated ATPase decayed much more slowly, with a half-time of 20 -40 s (closed squares in Fig. 5C) corresponding to an estimated hydrolysis rate of not more than 0.003-0.006 mol⅐mg Ϫ1 ⅐min Ϫ1 . Note that this decay rate implies that the catalytic activity is about 1000-fold lower for the p83C-28N complex than it is for intact ATPase. These findings make it clear that in the ATPase activity measurements illustrated in Fig. 2A, it is the intact residual ATPase and not the p83C-28N complex that is the main contributor to the 5-10% activity remaining after prolonged PK treatment.
Under conditions similar to those used above to measure ATP-derived phosphorylation at 4°C in Fig. 5, B and C, we also attempted to establish how far the p83C-28N complex retained the characteristic ability of the ATP-derived phosphoenzyme species in intact SR to occlude Ca 2ϩ , i.e. to reduce dramatically the rate of Ca 2ϩ release from the phosphoenzyme toward one or the other side of the ATPase. To reveal 45 Ca 2ϩ occlusion, we essentially followed a previously published protocol (Fig. 5 of Ref. 46). Use of this protocol in our experiments clearly demonstrated that PK-treated membranes were still able to occlude 5-6 nmol/mg of 45 Ca 2ϩ (as deduced from Fig. 5D Fig. 5C)). This was the case irrespective of the presence or absence of ionophore (open or closed symbols in Fig. 5D), presumably because vesicles have already been made permeable to Ca 2ϩ by the extensive PK treatment (see above); in control experiments performed with intact SR, demonstration of Ca 2ϩ occlusion did require permeabilization with ionophore (data not shown), as previously reported by Dupont (46). We conclude that, just like intact ATPase, the p83C-28N complex is fully able to occlude Ca 2ϩ after formation of the ADP-sensitive phosphoenzyme. In separate experiments (data not shown), we also found that 45 Ca 2ϩ became occluded after reaction of the p83C-28N complex with CrATP, as described previously (47) for intact ATPase.
Formation of Ca 2ϩ -free Phosphoenzyme Is Compromised after ATPase Cleavage, as Well as Tight Binding of Orthovanadate in the Absence of Ca 2ϩ and Proper Binding of Thapsigargin-According to the usual description of the catalytic cycle of intact ATPase, the formation of an ADP-sensitive ("Ca 2 E1P") phosphoenzyme is immediately followed by its transformation into another phosphoenzyme form (sometimes designated "Ca 2 E2P"), from which Ca 2ϩ ions rapidly dissociate toward the lumen before phosphoenzyme hydrolysis takes place (48). This Ca 2ϩ -free phosphoenzyme, now ADP-insensitive ("E2P"), may accumulate under certain steady-state conditions in permeabilized SR vesicles, e.g. in the presence of Me 2 SO or at low (or very high) pH, resulting in release of Ca 2ϩ ions previously bound to the ATPase transport sites (49 -51). The fact that the p83C-28N complex is characterized by a severely inhibited ATP hydrolysis rate, with retention of an ADP-sensitive phosphoenzyme, suggests that the phosphorylated complex has a reduced ability to form this Ca 2ϩ -free phosphoenzyme from the initial ADP-sensitive phosphoenzyme. Accordingly, we found that for PK-treated SR membranes the release of previously occluded Ca 2ϩ was slow (see Fig. IV in the Supplemental Material), while it is known to occur rapidly, on a subsecond time scale, in intact membranes (51).
In the absence of Ca 2ϩ , the Ca 2ϩ -free and ADP-insensitive (E2P) phosphoenzyme can also be formed from P i , by the "back door" or "reverse" reaction (52). Following our initial observations (Fig. 2B), we investigated further the respective abilities of intact and PK-treated SR membranes to form this phosphoenzyme from P i . We first checked that for intact membranes, full ATPase phosphorylation (more than 5 nmol/mg, see Fig. 6A) was possible in the presence of P i and EGTA in our pH 7.0 potassium-free buffer B medium supplemented with a large amount of Me 2 SO (here, 40% v/v), as described previously (23,52). However, even under these rather extreme conditions, PK-treated membranes were only phosphorylated to a very low extent (ϳ1 nmol/mg, Fig. 6B). Considering that PK-treated membranes contain 5-10% residual uncleaved ATPase, part of this low phosphorylation level is accounted for by the presence of residual intact ATPase, and the portion of E2P that can be attributed to the p83C-28N complex must therefore be even lower, 0.5-0.8 nmol/mg at most. From the time course seen from the figure, this phosphoenzyme also seems to be formed more slowly than is the case for intact ATPase. We therefore conclude that in Ca 2ϩ -free p83C-28N complex, formation of E2P from P i through the back door reaction is severely perturbed.
As orthovanadate has been shown previously (53) to act as a transition state analogue for ATPase phosphorylation from P i , even in the absence of Me 2 SO, we investigated further the properties of the Ca 2ϩ -free p83C-28N complex by using orthovanadate and, specifically, FITC-labeled preparations. In our standard buffer A medium in the absence of Ca 2ϩ , and using FITC-labeled intact membranes, we could easily record  Fig. 5. D, 45 Ca 2ϩ occlusion experiments after PK treatment. PK-SR membranes were suspended at 0.5 mg/ml in buffer A supplemented with 40 M 45 Ca 2ϩ and [ 3 H]glucose at 4°C (pH 7.3), in the presence of 5 g/ml A23187 ionophore (open symbols) or in its absence (closed symbols). Passively bound 45 Ca 2ϩ was measured, without rinsing (time 0). Then 50 M MgATP was added (single arrow), and samples (triangles) were taken at various intervals and filtered without rinsing to monitor the amount of 45 Ca 2ϩ that remained bound to the ATPase. In separate series, 30 s after ATP addition, samples (0.5 ml) were diluted (double arrow) 10-fold at 4°C in buffer A supplemented with 1.1 mM EGTA in the presence of 1.1 mM ADP (diamonds) or in its absence (squares); these samples were further incubated for various periods and then filtered to determine the residual amount of bound 45 Ca 2ϩ . Similar dilution measurements (double arrow, again) were performed at time 0, i.e. in the absence of ATP. the formation of the tight ATPase-orthovanadate complex as a marked increase in the fluorescence of bound FITC (Fig. 6C), as described previously (53). In contrast, for PK-treated membranes, we found that orthovanadate was essentially unable to form a similar high fluorescence complex (Fig. 6D), despite the above-mentioned fact that the fluorescence of the FITC-labeled PK-treated ATPase had remained sensitive to Ca 2ϩ removal from the medium. In contrast to orthovanadate, addition of decavanadate (data not shown) to FITC-labeled PK-treated SR membranes induced essentially the same fluorescence drop as that described previously (54) for intact SR. We thus conclude that the absence of effect of orthovanadate on FITC-labeled PK-treated membranes (Fig. 6, C vs. D) reflects the poor ability of the p83C-28N complex to enter an E2P-like conformation.
The question can be raised whether the apparent inability of the p83C-28N complex to enter an E2P-like conformation is strictly due to a deficiency in the phosphorylation step itself or whether this inability also reflects a difference between the conformations of the non-phosphorylated Ca 2ϩ -free forms of intact and cleaved ATPase. In fact, regarding the fluorescence properties of FITC, we noted significant differences between intact and PK-treated non-phosphorylated membranes. First, although as previously mentioned, the fluorescence level of FITC bound to PK-treated membranes responded to Ca 2ϩ removal from the medium with about the same dependence on free Ca 2ϩ concentration as it did for FITC bound to intact ATPase, the amplitudes of the corresponding changes were distinctly smaller for PK-treated membranes than for intact SR membranes (compare the traces in Fig. 6, C and D; the same fact is also apparent from Fig. III in the Supplemental Material). An even more prominent difference between intact and PK-treated membranes was observed in relation to the interaction of these two types of membranes with TG. When added to Ca 2ϩ -free FITC-labeled intact membranes, TG raised the fluorescence of bound FITC significantly and blocked, as expected, the effect of subsequent addition of Ca 2ϩ (Fig. 6E). In contrast, the change observed when TG was added to PKtreated FITC-labeled membranes was distinctly smaller (Fig.  6F), even though the lack of effect of subsequent addition of Ca 2ϩ indicated that interaction of TG with ATPase had taken place in this case also (as evidenced by the inhibition of 45 Ca 2ϩ binding previously illustrated in Fig. 4, C and D). Thus, the TG-inhibited p28N-p83C complex does not adopt exactly the same conformation as the TG-inhibited intact ATPase; transmission of conformational effects from the membrane-bound TG, presumably bound close to the M3/S3 region (14), to the distantly located FITC probe, is compromised after the cut at Glu 243 .

Proteolytic Evidence That after Cleavage at Glu 243 Movements of the A Domain Are Partially Deconnected from Those in the Rest of the Polypeptide Chain, Especially in the Absence of Ca 2ϩ
To address further the question of the ability of the p83C-28N complex to adopt a conformation similar to that of intact ATPase, especially in the absence of Ca 2ϩ , we also examined whether the susceptibility of the p28N-83C complex to proteases retained the same dependence on free Ca 2ϩ as that of intact ATPase. We initially found that when Ca 2ϩ -ATPase was first submitted to a first round of PK treatment in the presence of Ca 2ϩ and AMP-PCP, to produce the p83C-28N complex, then washed free of AMP-PCP by repeated centrifugation, and finally submitted to a second round of PK treatment in the absence of AMP-PCP but in the presence or absence of Ca 2ϩ , a p14 fragment distinctly showed up only when Ca 2ϩ was absent, suggesting that the bond at Leu 119 -Lys 120 (15) in PK-SR (in p28N) became exposed as a result of the Ca 2ϩ depletion, as in intact ATPase (data not shown).
However, proteolysis experiments performed with trypsin provided more specific information. The relevance of trypsin proteolysis experiments for studying calcium-dependent movements of the A domain derives from the fact that one of the tryptic cleavage sites, the T2 site, located between Arg 198 and Ala 199 in the A domain, has been described to be fully accessible in the presence of Ca 2ϩ , but less so in its absence (34,55). When PK-treated SR membranes were used for such an experiment, the relative protection of the Arg 198 -Ala 199 bond observed in the absence of Ca 2ϩ with intact ATPase was no longer apparent (the SDS-PAGE gels corresponding to these experiments may be seen as Fig. V in the Supplemental Material). This suggests that the p83C-28N complex is poorly able to stabilize the orientation of the A domain in which the T2 site is protected.
In addition, the A2 fragment no longer showed up distinctly, FIG. 6. In the absence of Ca 2؉ , the p83C-28N complex hardly forms any phosphoenzyme from P i or any tight complex with ortho-VO 4 ; FITC bound to Ca 2؉ -free complex is less sensitive to addition of Ca 2؉ or TG than FITC bound to intact ATPase. A and B, phosphorylation from 32 P i in the absence of Ca 2ϩ . SR membranes and PK-treated membranes (PK-SR) were incubated at about 0.2 mg/ml in buffer B supplemented with 40% Me 2 SO, 0.2 mM 32 P i , and either 1 mM EGTA (open symbols) or 0.1 mM Ca 2ϩ for control (closed symbols). Samples were acid-quenched after various periods, filtered, and thoroughly rinsed. C and D, FITC fluorescence measurements. After partial labeling by incubation with 2 nmol of FITC per mg of protein, FITClabeled SR (C) or PK-SR (D) was diluted to 100 g/ml in buffer A (with contaminating Ca 2ϩ present, at a few M). Subsequent additions were as follows: 20 M Ca 2ϩ first, followed by 200 M EGTA, 200 M orthovanadate, and lastly 700 M Ca 2ϩ . E and F, SR or PK-SR membranes were here partially labeled by incubation with 2 nmol of FITC per mg of protein and then suspended at 0.1 mg/ml in buffer A at 20°C. Additions were as follows: 20 M Ca 2ϩ first, followed by 200 M EGTA, 1 g/ml thapsigargin, and finally 700 M Ca 2ϩ . irrespective of the presence or absence of Ca 2ϩ during tryptic cleavage (Fig. V in the Supplemental Material); instead, under different pH conditions, a weak band of smaller size transiently appeared, suggesting that cleavage at Glu 243 had considerably increased the susceptibility of the A domain (as well as other parts of the ATPase) to proteolysis. Consistent with these findings, after trypsin cleavage of ATPase at Lys 234 or Arg 236 , Imamura and Kawakita (10,11) were only able to stabilize an "A1 b ϩB" complex which was devoid of A2 (and which, like our p83C-28N complex, remained phosphorylatable from ATP but was unable to rapidly release P i ). Experiments with another protease, V8 protease, also revealed enhanced susceptibility of the N terminus (including the A domain) after PK-induced cleavage at Glu 243 (see Supplemental Material). All these results indicate special vulnerability to proteolytic attack of the p28N part of the p83C-28N complex, compared with intact ATPase, especially (but not only) in the absence of Ca 2ϩ , and presumably due to increased "floppiness" of the N-terminal region after proteolysis at Glu 243 . DISCUSSION We have shown here that after proteolysis under specific conditions, it is possible to obtain large amounts of cleaved ATPase (50 -70% of the initial ATPase) in which only a few amino acids around Glu 243 (the MAATE 243 sequence) have been excised with no cuts in other ATPase regions, resulting in what we call the "p83C-28N complex." We have then used this complex to study the functional consequences of such excision in detail, and a clear-cut conclusion is that cleavage at Glu 243 prevents ATPase from entering what is often termed the E2P conformation, resulting in a block of Ca 2ϩ transport. Our discussion will therefore focus on how an intact link between the A domain and the M3 transmembrane helix can be a critical requirement for ATPase entry into the E2P state and opening of the lumenal gate for Ca 2ϩ , two processes that are intrinsically coupled in normal Ca 2ϩ -ATPase function. This will be done after first discussing the susceptibility of Ca 2ϩ -ATPase to proteolytic enzyme and the protection afforded by ligands in relation to the conformational state of the enzyme.
Significance, in Terms of ATPase Structure, of the Protecting Effect of Ligands against Proteolysis-A remarkable feature of the three-dimensional structural model of Ca 2ϩ -ATPase in its Ca 2ϩ -bound and unphosphorylated state (1) is the open structure of the three cytosolic domains. Although this structure is consistent with the ability of the ATPase to be easily cleaved by proteinase K in the presence of Ca 2ϩ , it cannot represent the conformation required for phosphoryl transfer after the binding of ATP. For this transfer to occur, there must be apposition between the N and P domains, presumably favored by ATP binding. In accordance with this prediction, we have shown here that in the presence of Ca 2ϩ (but not in its absence), strong protection of ATPase against proteolytic degradation beyond cleavage in the Glu 243 region is afforded by non-phosphorylating ATP analogues. We therefore suggest that nucleotide binding by itself, without phosphorylation, significantly alters the average conformation of the cytosolic part of the ATPase and stabilizes a more compact structure, by protecting, in particular, the central and C-terminal cytosolic regions against cleavage. In other studies, based on Fe 2ϩ -induced cleavage of the polypeptide chain, a related effect of bound nucleotide on Na ϩ ,K ϩ -ATPase and H ϩ ,K ϩ -ATPase was also deduced from the fact that the mere binding of ferrous nucleotide salts to these ATPases, without phosphorylation, leads to the formation of a species in which there is evidence for interactions between the N and P domains (56,57). A protective effect of nucleotide and Ca 2ϩ on proteolytic degradation of Ca 2ϩ -ATPase by proteinase K or V8 protease (the latter enzyme primarily cleaves the ATPase between Glu 231 and Ile 232 ) has also been recently reported by Danko et al. (34) in the presence of Mg 2ϩ . Our results confirm these previous results, but our use of a proteolysis medium devoid of Mg 2ϩ and K ϩ , probably in part by accelerating overall proteolysis, made it easier to reveal that secondary cleavage sites were protected by nucleotide binding to an extent even greater than the primary cleavage site at Glu 243 .
Despite the above evidence for ligand binding-induced stabilization against proteolytic cleavage, it should be realized that among all intermediate states of the intact Ca 2ϩ -ATPase, "E2P-like" conformations are unique in the sense that in these conformations the entire Ca 2ϩ -ATPase, including the A domain and the adjoining region around Glu 231 and Glu 243 , now becomes completely protected from proteolytic attack. This was shown to be the case after reaction with vanadate (55), but during the course of the present experiments we have also been able to observe complete resistance against proteolysis after the formation of an authentic E2P species formed from P i in the presence of Me 2 SO (data not shown). This extreme degree of protection means that upon entering the E2P conformation, Ca 2ϩ -ATPase experiences a further overall change in its structure (or dynamics), so that it now has a very compact (or tense) conformation in which it is even less reactive to proteases than after mere nucleotide and Ca 2ϩ binding. In particular, it is remarkable that the extended loop connecting the A domain with transmembrane segment M3, which has a distinctly peripheral location in the Ca 2ϩ -bound form of ATPase, as well as a high mobility (according to B factors in the PDB 1EUL structure), now becomes fully protected from proteolysis by PK (at Glu 243 ), V8 (at Glu 231 ), or trypsin (at Arg 198 ). We shall return later to the possible implications of this feature for structural aspects of the energy transduction mechanism.
Evidence That Cleavage at Glu 243 (with MAATE 243 Excision) Blocks Entry of the Proteolyzed ATPase into the Conformation Involved in Lumenal Ca 2ϩ Release and P i Liberation-Clearcut results that can be deduced from our study of the partial reactions of proteolyzed Ca 2ϩ -ATPase can be briefly summarized as follows: after cleavage at Glu 243 followed by MAATE excision, both binding of Ca 2ϩ to the cleaved ATPase and phosphorylation from ATP in the presence of Ca 2ϩ remain possible, but transformation of the ADP-sensitive phosphoenzyme thus formed into a phosphoenzyme species competent for P i liberation is almost totally blocked. We obtained evidence for Ca 2ϩ occlusion by the p83C-28N complex after phosphorylation from ATP but not for transport of Ca 2ϩ , and Ca 2ϩ release from the p83C-28N complex phosphorylated from ATP was reduced and proceeded slowly. Simultaneously, we found that the cleaved ATPase was characterized by an inability to become phosphorylated to any significant extent by the back door reaction after addition of inorganic phosphate in the absence of Ca 2ϩ , as well as by an inability to form a transition complex after reaction with vanadate. Thus, after cleavage at Glu 243 , it seems that the ATPase cannot enter what is known as the E2P state (nor form a tight E2-orthovanadate complex, an analogue of the transition state), at least not at a rate sufficient to maintain even a modest turnover rate. The Glu 243 -containing loop, connecting domain A and M3 in the transmembrane sector, therefore appears to be critically involved in the conformational changes that are transmitted from the cytosolic regions to the membranous sector and vice versa and which are responsible for the coupled ATP hydrolysis and Ca 2ϩ translocation, via formation of E2P-like states.
Conversely, it is remarkable that integrity of the loop connecting domain A and M3 in the transmembrane sector does not seem to be essential for a number of other steps in the catalytic cycle, associated with what is usually described as the "E1" state; binding of Ca 2ϩ , phosphorylation by ATP, and occlusion of Ca 2ϩ proceed in a largely undisturbed manner despite the fact that after excision of MAATE 243 the N terminus is severely perturbed and becomes rather floppy, as revealed by increased proteolytic susceptibility. This suggests that these steps depend mainly on the central and C-terminal parts of the Ca 2ϩ -ATPase (including their transmembrane domains). At least it appears that ATPase cleavage of the Glu 243 -containing extended loop does not perturb significantly those interactions of the A domain with the P and N domain that are necessary for maintaining an E1 state. Subsequent work will examine whether ATPase cleavage at Lys 120 , in the loop connecting the membrane with the ␤-domain on its N-terminal side, has similar effects to cleavage at Glu 243 , in the loop connecting the membrane with the ␤-domain on its C-terminal side. 2 Comparison with Previous Proteolysis, Mutagenesis, and Cross-linking Studies of Ca 2ϩ -ATPase, and with Previous Chymotryptic Studies of Na ϩ ,K ϩ -ATPase-Our present data confirm and extend previous results (8 -15, 58 -65) obtained after proteolysis, mutagenesis, and cross-linking of Ca 2ϩ -ATPase. This topic is dealt with in detail in the Supplemental Material, and it will suffice here to mention one of these results: mutation to Val of the conserved Gly 233 , located in the same Ca 2ϩ -ATPase region that is susceptible to proteolysis by V8 or PK (i.e. after the end of the A domain), leads to functional changes similar to those found here, namely both a slowing down of the transition from Ca 2 E1P to E2P and an increased difficulty in forming from P i significant amounts of E2P at steady state (Refs. 12 and 68 and see also Ref. 69 for related results obtained after mutation of the homologous residue in rat kidney Na ϩ ,K ϩ -ATPase, Gly 263 ). Note that this phenotype is relatively rarely observed after site-directed mutagenesis.
Regarding related previous studies of other P-type ATPases, it is especially relevant here to compare the properties of our p83C-28N complex, derived from Ca 2ϩ -ATPase, with those of the complex formed after chymotryptic cleavage of pig kidney Na ϩ ,K ϩ -ATPase at Leu 266 ; the latter residue is also located closely after the end of the A domain, in the region homologous to the V8 and PK sites in Ca 2ϩ -ATPase (Leu 266 in Na ϩ ,K ϩ -ATPase corresponds to Gln 238 in Ca 2ϩ -ATPase, located between the V8 (Glu 231 ) and PK (Glu 243 ) site). Like the p83C-28N complex, Na ϩ ,K ϩ -ATPase cleaved at Leu 266 was found to have a catalytic cycle blocked at the level of the E1P to E2P transition (16). Whether cleaved Na ϩ ,K ϩ -ATPase was able to form E2P from P i in the absence of Na ϩ , or not, was, to our knowledge, not measured, but binding of [ 48 V]orthovanadate to the cleaved enzyme (to form an E2P-like state) proved to be strongly reduced, in agreement with our own results with the p28N/p83C complex.
An additional remarkable finding with chymotryptically cleaved Na ϩ ,K ϩ -ATPase was that although cleaved Na ϩ ,K ϩ -ATPase was still able to occlude Rb ϩ , an ability generally considered to be typical of an E2 conformation, occluded Rb ϩ was no longer displaced by ATP binding, suggesting that after cleavage at Leu 266 events at the transport sites had become uncoupled from those at the nucleotide site (16 -18). In our experiments, we tested whether a similar uncoupling could be demonstrated after cleavage of Ca 2ϩ -ATPase at Glu 243 , but as shown in Fig. 4F we found by stopped-flow fluorimetry that acceleration of the E2 to E1 transition by binding of nucleotide, which is a prominent feature of intact Ca 2ϩ -ATPase, was retained to a very significant extent after cleavage at Glu 243 . As further evidence for the preservation of long range effects in the p83C-28N complex, we have found in Trp fluorescence experiments that Ca 2ϩ dissociation from the nonphosphorylated p83C-28N complex was still accompanied by changes in the affinity of this complex for AMP-PCP (as previously reported for intact SR, see Refs. 36 -38), with the K d value for AMP-PCP binding changing from less than 1 M in the presence of 0.3 mM Ca 2ϩ to 5-10 M in the presence of EGTA (Fig.  VI in the Supplemental Material). Thus, with respect to the influence of nucleotides on events occurring at the membranous ion transport sites, the unphosphorylated p83C-28N complex behaves almost normally, although apparently the cleaved Na ϩ ,K ϩ -ATPase does not. The difference between the two ATPases might be more quantitative than qualitative, as in Na ϩ ,K ϩ -ATPase the conformational equilibrium is more heavily poised toward E2 than it is in Ca 2ϩ -ATPase, but the issue remains puzzling.
Consequences of MAATE 243 Excision for the Description of the Non-phosphorylated Forms of p83C-28N Complex-Concerning the above puzzle, a reasonable view probably is that the various conformations adopted by our p28N/p83C complex in the presence and absence of Ca 2ϩ cannot be described in a fully meaningful way by using the same nomenclature as for intact ATPase. This will now be discussed.
For the PK-cleaved ATPase described here, a number of our results seem to favor the idea that the Ca 2ϩ -free p83C-28N complex can indeed adopt a conformation similar to that of intact Ca 2ϩ -free ATPase; interaction of the nonphosphorylated p83C-28N complex with Ca 2ϩ is only slightly perturbed with regard to both equilibrium and kinetic aspects, and many of the typical signals that normally accompany Ca 2ϩ binding to (or dissociation from) nonphosphorylated ATPase are still observed, e.g. changes in the fluorescence of both the intrinsic Trp residues (most of which are located at the membrane/water interphase) and the extrinsic FITC label (bound in the nucleotide binding domain); the same is true of changes in ATPase affinity and reactivity toward nucleotides (see above), as well as of changes in susceptibility to PK of the Leu 119 -Lys 120 peptidic bond. However, in some other aspects, such as the reduced amplitudes of the FITC fluorescence changes in response to removal of Ca 2ϩ or to interaction with thapsigargin, the relation of the structure of PK-treated ATPase to that of intact ATPase is less clear-cut; for instance, the p28N/p83C complex can probably not adopt the conformation that intact ATPase adopts in the presence of EGTA and TG, and as regards the effect of trypsin, it is also apparent that, at variance with the situation in intact ATPase, tryptic cleavage at the T2 site is no longer protected by Ca 2ϩ in the p28N/p83C complex.
It therefore appears that keeping the classical names, "E1" and "E2," to describe conformational states of ATPase molecules that have been modified by proteolytic cleavage may be misleading in some cases, because coupling between the various domains may have been severely perturbed. This may also apply to site-directed mutagenesis experiments in some cases; note that in their initial work with the G233V mutant of Ca 2ϩ -ATPase, Andersen et al. (12) also felt unable to decide to what extent the inability of this mutant to form an E2P species was related to a possible deficiency of the mutant in forming the Ca 2ϩ -free non-phosphorylated E2 form. In our case, the p83C-28N complex is still able to transmit long range information from the membranous Ca 2ϩ -binding sites toward the cytosolic P and N domains, presumably because contacts necessary for these interactions are left intact, such as those to which the L6 -7 loop and the M5/S5 and M4/S4 segments contribute (cf. Refs. 1 and 66). In contrast, the movements of the A domain that normally accompany Ca 2ϩ dissociation, and which, for instance, are critical for the proteolytic susceptibility of the Arg 198 -Ala 199 bond, are no longer efficiently controlled after proteolytic cleavage at Glu 243 .
Possible Structural Interpretation and Concluding Comments-At this step, to discuss the structural basis for the deficient Ca 2ϩ translocation observed here after N-terminal proteolytic cleavage at Glu 243 , it is worth comparing the structure of the Ca 2ϩ -bound form of ATPase (PDB 1EUL) to the structure recently deduced from analysis of a crystal prepared from Ca 2ϩ -free nonphosphorylated ATPase in the presence of TG (70). In this new structure (Fig. 7), an important feature is that the three cytoplasmic domains of Ca 2ϩ -ATPase now form a single headpiece, i.e. the P and N domains have bent forward and the A domain has indeed been rotated by about 90°(as compared with the Ca 2ϩ -bound structure) to interact with the P and N domains (which in the presence of Ca 2ϩ , as discussed above, may already become closely apposed after interaction with nucleotide). The TGES 184 motif is now located close to the phosphorylation site in the P domain, similar to what has been suggested for Na ϩ ,K ϩ -ATPase in the E2P state (56), whereas the T2 cleavage site (Arg 198 -Ala 199 ) has become less accessible to trypsin than it is in the Ca 2ϩ -bound structure. At the same time the extended loop that connects the A domain with the third transmembrane helix M3 has moved from a peripheral to a more central position where it has become apposed to the C-terminal conserved part of the P domain (previously referred to as a junctional region, which is one of the most conserved regions in the P-type ATPase family, see Ref. 5). Thus, the Glu 243 -containing loop in this E2⅐TG structure is located very close to helix P6 in the C-terminal part of the P domain, and the structure around Glu 243 seems to be stabilized by hydrogen bonds between Lys 712 and three residues, namely Gln 244 , Thr 242 , and Met 239 (as indicated by dashed pink lines in Fig.  7). This rotation of the A domain is analogous, although not identical, to the one previously proposed for the E2⅐vanadate form of ATPase (an E2P-like form) in two-dimensional crystals (1,55).
On this basis, deleterious effects of excision of the MAATE sequence from the loop connecting the A domain to M3 can be easily predicted. First, the absence of the excised sequence may lead to destabilization (because of loss of hydrogen bonds) of the interaction of the rotated A domain with the P domain, prohibiting the ATPase from forming an E2P state (either from inorganic phosphate or after phosphorylation with ATP) or from forming a native-like "E2" state (especially in the presence of TG). Second, by ruining mechanical cross-talk between the phosphorylation site and the third membrane helix, the excision of this sequence may prevent opening of the luminal gate and transport of Ca 2ϩ through the membrane. These outlines of the transduction process, derived from proteolytic analysis, may serve as a starting point for future thinking about the subject, in connection with future elucidation of the actual structures of all phosphorylated and non-phosphorylated forms of Ca 2ϩ -free intact ATPase. Also, comparative proteolytic experiments and comparisons of the sequences of the corresponding extended loops in the various ion-transporting ATPases will be of interest.
From the above discussion, we can propose the following overall view of the Ca 2ϩ pump cycle, in both functional and mechanistic terms. The Ca 2ϩ -ATPase cycle may be considered to compose a cyclic wave of conformational changes, starting with binding of two calcium ions by the M4, M5, M6, and M8 transmembrane helices. Initial changes probably are transmitted to the N and P domain via the L6 -7 loop and S4/S5 (1,66), and in the presence of ATP they result in ATPase phosphorylation; it is remarkable that these events are intact after derangement of the function of the N-terminal part of Ca 2ϩ -ATPase by proteolytic cleavage. In intact ATPase, the conformational energy gained by phosphorylation with ATP then results in a further conformational change (1,67) in which new interactions of the now rotated A domain with the bent forward P domain are formed, resulting in a change in the chemical properties of the phosphorylation site and in transfer of information to the membrane region via S3/M3 (with concomitant opening of the luminal gate and release of Ca 2ϩ ). This is followed by E2P dephosphorylation and conversion of the ATPase to a more relaxed E2 state. Our results show that this re-orientation of the P and A domains is catalytically important, as normal cycling of the enzyme is blocked when stabilization of the re-oriented domains is made impossible by PK-induced disruption of the chain at MAATE 243 . Presumably, this is because Glu 183 in the conserved TGES segment of the A domain can only be positioned correctly close to the phosphorylation site after this reorientation. When the connection of the A domain to the membrane is disrupted by proteolytic cleavage, cross-talk between the P domain and the membranous region can no longer occur in the same manner, FIG. 7. Relative location of A and P domains in nonphosphorylated, Ca 2؉ -free, and TG-ligated ATPase, compared with Ca 2؉bound ATPase, MAATE 243 /P6 proximity. The crystal structure of the TG-ligated ATPase form, "E2.TG " (bottom frame), has recently been elucidated (70). The present figure emphasizes the location in this new structure of selected residues, e.g. Glu 183 and Arg 198 in the A domain, as well as the close proximity between P6 helix in the P domain and the MAATE 243 region (highlighted in red in the bottom frame). Hydrogen bonds (in pink) form between Q244Ne and K712O, between T242OH and K712Nz, and between M239O and K712Nz. For comparison, the Ca 2ϩ -bound structure, "E1Ca 2 ," is shown in the top smaller frame. the ADP-sensitive phosphoenzyme is stabilized, P i release is very slow (much like in the soluble MJ0968 P-type ATPase from Methanococcus jannaschii, ref. 21), and the luminal gate remains closed, thus blocking lumenal Ca 2ϩ release. Hopefully, the stabilized ADP-sensitive phosphoenzyme formed from our p83C-28N complex will prove useful for studying structural characteristics of an E1P-like conformation.