Functional Properties of Sarcoplasmic Reticulum Ca2+-ATPase after Proteolytic Cleavage at Leu119-Lys120, Close to the A-domain

By measuring the phosphorylation levels of individual proteolytic fragments of SERCA1a separated by electrophoresis after their phosphorylation, we were able to study the catalytic properties of a p95C-p14N complex arising from SERCA1a cleavage by proteinase K between Leu119 and Lys120, in the loop linking the A-domain with the second transmembrane segment. ATP hydrolysis by the complex was very strongly inhibited, although ATP-dependent phosphorylation and the conversion of the ADP-sensitive E1P form to E2P still occurred at appreciable rates. However, the rate of subsequent dephosphorylation of E2P was inhibited to a dramatic extent, and this was also the case for the rate of “backdoor” formation of E2P from E2 and Pi. E2P formation from E2 at equilibrium nevertheless indicated little change in the apparent affinity for Pi or Mg2+, while binding of orthovanadate was weaker. The p95C-p14N complex also had a slightly reduced affinity for Ca2+ and exhibited a reduced rate for its Ca2+-dependent transition from E2 to Ca2E1. Thus, disruption of the N-terminal link of the A-domain with the transmembrane region seems to shift the conformational equilibria of Ca2+-ATPase from the E1/E1P toward the E2/E2P states and to increase the activation energy for dephosphorylation of Ca2+-ATPase, reviving the old idea of the A-domain being a phosphatase domain as part of the transduction machinery.

Large scale relative movements of the three main sub-domains in the cytosolic head of P-type ATPases, as well as associated movements of the transmembrane segments to which these domains are connected, probably play a major role in the active transport mediated by these membrane proteins (1,2). We have therefore become interested in the functional effects of proteolytic disruption of the integrity of the loops connecting these various ATPase domains. For rabbit SERCA1a, 1 a representative of P-type ATPases involved in Ca 2ϩ transport into intracellular compartments, we have previously studied how Ca 2ϩ transport and ATP hydrolysis are affected by proteolytic cleavage at a site located in the segment linking the A-domain with the third transmembrane segment, M3 (3). In that previous work, the proteolysis conditions had allowed us to obtain relatively large (milligram) amounts of fairly pure "p83C-p28N" peptide complexes, derived from ATPase by excision of a short 239 MAATE 243 sequence from this segment, and procedures that previously had been developed to study the catalytic cycle of intact ATPase were used unaltered to study the properties of the proteolytic product. The results showed that the cleaved ATPase was deficient in Ca 2ϩ transport but remained phosphorylatable by ATP in a manner indistinguishable from that of intact ATPase. A detailed study of the partial reactions revealed that a major defect resulting from proteolysis was the difficulty for the cleaved ATPase to enter what is known as the "E2P" state, i.e. the ADP-insensitive form of phosphoenzyme. In the present work, we have aimed at studying whether SERCA1a function is affected by proteolytic cleavage in another part of the link between the A-domain and the membrane. To do so, we again resorted to proteolytic attack by proteinase K, but instead of performing the proteolysis in the presence of Ca 2ϩ and AMP-PCP, we now treated the ATPase in the absence of Ca 2ϩ and nucleotide: this is known to allow cleavage at Leu 119 -Lys 120 , in the segment linking the A-domain with the second transmembrane segment, resulting in the formation of complementary "p14N" and "p95C" peptides (4). In this case, however, proteolysis also simultaneously occurs at other sites, so that a mixture of peptides is formed, among which p95C is only present in relatively small amounts compared with p83C or residual intact ATPase. To study the catalytic properties of the "p95C-p14N" complex, we therefore had to resort to new procedures. We combined two types of established experiments: first, after phosphorylation of the various peptides, we separated them by electrophoresis under conditions previously found to allow adequate separation between fragments of large size and yet minimize dephosphorylation (5); second, we made use of various phosphorylation and dephosphorylation protocols previously used to character-ize kinetically the individual steps in the catalytic cycle of heterologously expressed ATPase mutants. This allowed us to simultaneously study to what extent the p95C and the p83C peptides were phosphorylated or dephosphorylated under various situations and to compare their properties with those of residual intact ATPase. Results concerning the p83C-p28N complex were consistent with those that we had previously obtained with conventional methods (3). The present report therefore focuses on the catalytic properties of the p95C-p14N complex resulting from cleavage of the Lys 119 -Leu 120 bond: we find that following this cleavage Ca 2ϩ -ATPase turnover is severely perturbed but in a manner quite different from that obtained by excision of the 239 MAATE 243 sequence in the peptide chain linking the A-domain with M3. The results make it possible to get a better understanding of the role of the A-domain for the intermediary processes involved in Ca 2ϩ transport.

MATERIALS AND METHODS
Proteolysis by Proteinase K-For controlled proteolysis of sarcoplasmic reticulum (SR) Ca 2ϩ -ATPase, SR membranes (6) were suspended at a concentration of 2 mg protein/ml in a 100 mM Mops-NaOH medium (pH 6.5) containing 5 mM Mg 2ϩ and 0.5 mM EGTA, and treated with PK at 0.03 mg/ml, generally for 15 min at 20°C unless otherwise indicated. Proteolysis was stopped by adding 0.5 or 1 mM phenylmethylsulfonyl fluoride (as well as 0.6 mM Ca 2ϩ in some cases) and storing the samples on ice for 10 min. For some of the phosphorylation experiments, namely those involving [␥-32 P]ATP, the preparation with inactivated PK was used without further treatment as detailed below. For other phosphorylation experiments, namely those involving [ 32 P]P i that were best performed with a higher concentration of protein than those involving [␥-32 P]ATP, a large batch of PK-treated concentrated membranes was first prepared (see Supplemental Material).
Phosphorylation and Dephosphorylation-Phosphorylation from [␥-32 P]ATP of ATPase fragments was generally measured on ice, by adding 5 M [␥-32 P]ATP (at 0.5 mCi/mol) to PK-treated membranes suspended at 0.2 mg/ml in buffer A (100 mM KCl, 5 mM Mg 2ϩ , and 50 mM Mops-Tris, adjusted to pH 7.0 at room temperature), supplemented with various concentrations of Ca 2ϩ (or a Ca 2ϩ /EGTA buffer) (typically 600 l per sample). When desired, ATPase dephosphorylation was subsequently triggered by adding EGTA and the appropriate additive, with only small dilution (see below for details). In some cases, phosphorylation was also measured after a few milliseconds reaction at 20°C, with a Biologic QFM5 quenched-flow apparatus (48).
Phosphorylation from [ 32 P]P i of ATPase fragments was measured by adding [ 32 P]P i (at 0.5 mCi/mol), generally at 200 M, to concentrated PK-treated membranes suspended at 1-2 mg/ml in buffer B (10 mM Mg 2ϩ , 1 mM EGTA, and 50 mM Mops-Tris, adjusted to pH 7.0, at room temperature) supplemented with 25% Me 2 SO (v/v) (typically 60 l per sample). When desired, ATPase dephosphorylation was then triggered either by adding excess non-radioactive P i (resulting in only little dilution; experiments of this kind were performed at 20°C), or by 10-fold dilution of the phosphorylated sample into an Me 2 SO-free, KCl-containing, and ATP-containing buffer (experiments of this kind were performed on ice, see below).
In all cases the reaction was stopped by acid quenching with trichloroacetic acid (or in some cases PCA) and phosphoric acid, at final concentrations of 500 mM and 30 mM, respectively (typically in a total volume of 900 l). Generally, the quenched sample, previously left on ice for 20 min, was pelleted by centrifugation (15,000 rpm for 25 min, 4°C). The supernatant was sucked off and replaced by an equivalent volume of acid solution (diluted 10-fold compared with the final quenching solution), and the sample was centrifuged as before. After removal of the second supernatant, the pellet was resuspended in 250 l of a sample buffer containing 2% SDS, 10 mM EDTA, 150 mM Tris-HCl (at pH 6.8), 16% (v/v) glycerol (or 8 M urea in some cases), 0.8 M ␤-mercaptoethanol, and 0.04% bromphenol blue, by vortexing for 1 min (thorough vortexing was found to be critical for reproducible resuspension of the acid-treated pellet); 20-l aliquots were then loaded onto gel for electrophoretic separation as described below.
Electrophoretic Separation and 32 P Counting-This was performed according to protocols previously described (5, 7-9; see also Ref. 10 for a related approach), but with some modifications. The stacking gel contained 4% acrylamide (with an acrylamide to bisacrylamide ratio of 29 to 1), 65 mM Tris-H 3 PO 4 , pH 5.5, 0.1% SDS, 2% ammonium persul-fate, and 0.1% TEMED. The separating gel was a continuous 7% gel, containing 65 mM Tris-H 3 PO 4 , pH 6.5, 0.1% SDS, 0.4% ammonium persulfate, and 0.05% TEMED. Unless otherwise indicated, gels were run in a cold room. The pre-cooled running buffer contained 170 mM Mops-Tris, pH 6.0 and 0.1% SDS; it was kept under stirring during electrophoresis to prevent heating of the gels. Electrophoresis lasted about 3 h, with a current intensity set constant to 10 mA/gel (70 -80 V). Gels were then stained and fixed for 10 min in 40% methanol, 10% acetic acid, and 0.1% Coomassie Blue R-250. Destaining was carried out in 10% acetic acid, 10% methanol, and 1% glycerol within less than 1 h, and the gels were then dried overnight between two sheets of cellophane paper, prior to exposure to a phosphor screen the next morning. Radioactivity was revealed with a STORM 860 PhosphorImager (Amersham Biosciences) and quantified by comparison with known amounts of [␥-32 P]ATP or [ 32 P]P i .
In some cases, samples were also run in a Laemmli-type system (e.g. see Fig. A in Supplemental Material). This was especially the case for simple immuno-decoration experiments, performed as previously described (3). Tricine gels were also sometimes used for peptide separation (4).
Solubilization and Elution of the p95C-p14N Complex in the Presence of Detergent-After proteolysis, solubilization with C 12 E 8 or DM was sometimes performed and followed by chromatography on a detergentequilibrated size exclusion column, essentially as previously described (e.g. Refs. 3 and 11). Selected fractions were used for SDS-PAGE and immuno-reactivity analysis (see Supplemental Material), whereas other aliquots were used for MALDI-TOF analysis (see below) of p14N, either without further separation, or after additional trypsinolysis (12) (see Supplemental Material).
Mass Spectrometry Analysis-After sample preparation (see Supplemental Material), spectra were acquired on a MALDI-TOF spectrometer (Perseptive Biosystems, Voyager STR-DE) equipped with a nitrogen laser (337 nm), in linear mode or in reflector mode (see Supplemental Material).
Edman Degradation Analysis and Other Methods-After SDS-PAGE separation of PK-treated fragments on an 8% Laemmli gel (with 1 mM thioglycolate in the running buffer), the N terminus of the blotted p95C fragments was sequenced using a Procise (Applied Biosystems) protein sequencer.
Low concentrations of free Ca 2ϩ were obtained by buffering with EGTA, using the following apparent dissociation constants for Ca 2ϩ and Mg 2ϩ at pH 7: K dCa⅐EGTA ϳ 0.4 M and K dMg⅐EGTA ϳ 25 mM. Structures were visualized using the SwissPDB Viewer (Ref. 13; available at www.expasy.org/spdbv/). Additional information has been deposited as web-available Supplemental Material, together with several figures and more detailed discussion.

Multiple Fragmentation of Ca 2ϩ -ATPase Treated with PK:
Formation of a p95C-p14N Complex in the Absence of Ca 2ϩ and Assay of Phosphorylatable Proteolytic Peptides with Sarkadi Gels-The typical pattern of ATPase proteolysis by PK in the presence or absence of Ca 2ϩ , as deduced from electrophoresis on Laemmli gels, is shown in the top panel of Fig. 1, lanes 2-5 in this panel show that, as previously documented, treatment of SR vesicles with PK in the presence of Ca 2ϩ gives rise to a number of well characterized initial products, comprising membranous peptides derived from either the N-terminal (p28N) or the C-terminal part (p83C and p19C) of the ATPase, together with water-soluble p29/30 fragments derived from the central nucleotide binding domain (4,14). In the absence of Ca 2ϩ , fragments with the same apparent masses as the p83C and p28N bands again show up (lanes 6 -9 in Fig. 1A), but in this case a new band with a higher molecular mass, the "p95" peptide is also transiently formed in the first 15 min, together with a new p14 band; both p95C and p14N bands are degraded after 60 min (compare lanes 6 -9 with lanes 2-5, and see similar results in Refs. 4 and 15). This p95 band represents a Cterminal peptide that previously was found to arise from ATPase cleavage at Leu 119 -Lys 120 (4), and by Edman degradation analysis we confirmed herein this assignment (data not shown). Thus, the absence of Ca 2ϩ appears to result in an increased susceptibility of the ATPase chain to PK cleavage at For functional studies, the question whether the complementary p14N and p95C fragments arising from the abovedescribed cleavage remain associated in the membrane is an important issue. To address this question, after treatment with PK we solubilized the treated SR membranes with a mild detergent (C 12 E 8 or DM) and chromatographed the solubilized sample on an high-performance liquid chromatography size exclusion column equilibrated in the presence of the same detergent. By both Western blotting and Coomassie Blue staining, we found that the small p14N peptide indeed elutes at the position where monomeric intact ATPase normally elutes, and where, after PK treatment, the p83C-p28N complex resulting for cleavage at Thr 242 -Glu 243 also elutes together with residual intact ATPase (3): i.e. p14N presumably elutes in association with the larger p95C peptide (Fig. B in Supplemental Material). Furthermore, judging from Coomassie Blue staining, p95C and p14N are present in these fractions (corresponding to the "monomeric" region of the elution pattern) in relative amounts consistent with the formation of stoichiometric complexes ( Fig. C in Supplemental Material), suggesting that the two ATPase fragments resulting from cleavage at Leu 119 -Lys 120 remain closely associated despite the presence of detergent, as was previously demonstrated for the p83C and p28N fragments (3). We therefore consider it likely that the former fragments also remain associated in the membrane after cleavage, hence our designation of these fragments in the remainder FIG. 1. Proteinase K (PK) treatment in the absence of Ca 2؉ results in cleavage at Leu 119 -Lys 120 and formation of a p95C-p14N complex (together with other peptides). Sarcoplasmic reticulum (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, 5 mM Mg 2ϩ , pH 6.5), in the presence of 0.3 mM Ca 2ϩ (lanes [2][3][4][5] or the presence of 0.5 mM EGTA (lanes 6 -9). Here, after proteolysis arrest, aliquots of the various samples were added to a urea-containing denaturation buffer (essentially as described in Ref. 46), boiled, and loaded (about 4 g of protein per lane) onto a 12% SDS-PAGE Laemmli gel prepared in the presence of 1 mM Ca 2ϩ . Lanes 1 and 10 contain molecular mass markers ("LMW" Amersham Biosciences kit). The bottom part of the figure schematizes the identified points of primary cleavage and the various proteolytic products. Note that because Fig. B in Supplemental Material shows that p95C and p14N fragments remain associated after membrane solubilization by detergent, these fragments can be considered to remain associated in the PK-treated membrane itself. of this paper as a p95C-p14N complex.
We checked that after cleavage in the absence of Ca 2ϩ , the N-terminal peptide formed, p14N, does have its C-terminal end at Leu 119 , i.e. that proteinase K has not cleaved off a few more residues on the C-terminal side of this peptide as we previously found to be the case for the p83C-p28N complex. Mass spectrometry analysis showed that the 1-119 peptide was indeed not subject to further degradation ( Fig. D in Supplemental Material) (and simultaneously confirmed that, as previously established (16), the N terminus of Ca 2ϩ -ATPase is acetylated). The identity of the p14N fragment is thus fully established. Note that this p14N fragment, formed after cleavage in the absence of Ca 2ϩ , was previously designated as p14a in Ref. 4 (and, in the legend to Table I of that report, it was erroneously indicated that this peptide corresponded to residues 1-19, instead of 1-119). After treatment with PK for 15 min in the absence of Ca 2ϩ , the amount of the p95C-p14N complex formed is somewhat smaller than that of undegraded ATPase and definitely smaller than the amount of p83C-p28N complex; after longer proteolysis periods, secondary cleavage sites evidently result in further degradation of the primary products (see lane 9 in Fig To study phosphorylation of the Ca 2ϩ -ATPase fragments, we have adopted a moderately acidic gel electrophoresis system that previously has been described not only to retain the ability to resolve protein fragments of relatively similar molecular masses as efficiently as the Laemmli system, but also to allow phosphorylated P-type ATPases or fragments to remain phosphorylated during electrophoretic migration (5, 7). We have been able to fully confirm the value of this SDS-PAGE system. The proteolytic peptides in the PK-treated SR membranes, submitted to phosphorylation according to various protocols, followed by acid quenching and resuspension in SDS, could be clearly separated without much loss of the covalently bound 32 P (less than half of the radioactivity was lost from the phosphorylated fragments; see Fig. A in Supplemental Material) and with higher resolution than by the Weber and Osborn procedure commonly used for revealing phosphorylated P-type ATPases (data not shown). In the following, we have therefore combined the use of such Sarkadi gels with the use of various phosphorylation and dephosphorylation protocols, previously developed to study individual steps in the catalytic cycle of either SR intact Ca 2ϩ -ATPase or expressed wild-type or mutant ATPase (e.g. Refs. [17][18][19]. The p95C-p14N Complex Can Be Phosphorylated from [␥- 32 Fig. 2A show PhosphorImager analysis of an experiment in which PKtreated SR vesicles, equilibrated in the presence of Ca 2ϩ , were reacted with [␥-32 P]ATP for 5, 15, or 30 s on ice before acid quenching and electrophoresis in the presence of SDS. Besides residual intact ATPase, 32 P was retained to a significant extent only by the large p95C and p83C proteolytic fragments, as expected on the basis of previous experiments performed at steady state (3,4). The final phosphorylation levels were similar, as shown by dividing the number of picomoles of 32 P associated with each band (shown in panel A) by the extent of Coomassie Blue staining of the same band (shown in panel C), a procedure that also allowed us to take into account any small difference in the total amount of protein loaded onto each lane that could be due to a slightly variable recovery of the acidprecipitated proteins. On a time scale of a few seconds, phosphorylation of the p95C fragment on ice was fast and indistinguishable from that of either intact ATPase or p83C-p28N complex. When the exact time course of phosphorylation was resolved, using a quenched flow equipment with millisecond time resolution at 20°C, we found that the rate of phosphorylation of the p95C peptide was actually somewhat slower (3fold) than that of intact ATPase (Fig. F in Supplemental Material). In separate experiments performed on ice, steady-state phosphorylation levels apparently revealed a fairly similar dependence on free Ca 2ϩ for intact ATPase and the proteolytic peptides ( Fig. G in Supplemental Material; see, however, below experiments designed to reveal the true affinity of the peptides for Ca 2ϩ in the absence of steady-state phosphorylation). The reaction of phosphorylation from ATP in the p95C-p14N complex (the "Ca 2 E1 to Ca 2 E1P" transition) therefore appears not to be very different from that of intact ATPase or the p83C-p28N complex.

P]ATP at a Relatively High Rate, but Dephosphorylates Very Slowly, Even in the Presence of ADP-Lanes 2-4 in
In contrast, after steady state was reached, EGTA-induced dephosphorylation of the p95C-p14N complex was very slow, even slower than that of the p83C-p28N complex, whose dephosphorylation is itself much slower than that of intact ATPase (as previously demonstrated in Ref. 3). This was the case both when dephosphorylation was triggered by addition of EGTA and excess non-radioactive MgATP in the presence of an ATP regenerating system (to induce dephosphorylation in the normal direction of the cycle while reducing the ADP content as much as possible, see lanes 5-7 in Fig. 2), and when dephosphorylation was triggered by addition of EGTA and excess ADP (to assay the amount of ADP-sensitive phosphoenzyme, see lanes 8 -10 in Fig. 2). In the latter case, too, most of the p95 fragment remained phosphorylated in an unusually stable state, although most of intact ATPase and p83C had dephosphorylated within 5 s.
If interpreted within the framework of the four-step catalytic cycle generally used for intact Ca 2ϩ -ATPase (Ca 2 E1 3 Ca 2 E1P 3 E2P 3 E2; see Ref. 20 and inset in Fig. 2), this fact suggests that for p95C the processing of phosphoenzyme from its initial ADP-sensitive form to a subsequent ADP-insensitive form remains possible (the "Ca 2 E1P to E2P" transition), but hydrolysis of the ADP-insensitive form is dramatically slowed down (the "E2P to E2" step), so that at steady-state a significant proportion of the latter phosphoenzyme form accumulates. Using previously established conditions (alkaline pH and no potassium) to directly measure the rate of the transition at which phosphoenzyme becomes insensitive to ADP, i.e. is converted from Ca 2 E1P to E2P (e.g. Ref. 21), we were indeed able to confirm that the latter transition is reasonably fast in p95C peptide, as it is in intact ATPase (see Fig. H, Supplemental Material).
The p95C-p14N Complex Can Also Be Phosphorylated from [ 32 P]P i , with Unaltered Apparent Affinity for P i at Equilibrium, but with Kinetics Very Significantly Slowed Down for Both Phosphorylation and Dephosphorylation-To further study the properties of the ADP-insensitive form of phosphoenzyme in the p95C-p14N complex, we attempted to form it from P i via the "backdoor" reaction, in the absence of Ca 2ϩ and presence of Me 2 SO (22). After equilibrium was reached, at 20°C, the p95C peptide was found to be phosphorylated from P i to the same extent as intact ATPase (or even to a slightly larger extent) and with essentially the same apparent affinity for P i (lanes 2-5 in Fig. 3A and the left portion of Fig. 3B), whereas phosphorylation was much less favorable for the p83C peptide, in accordance with our previous results (3). Phosphorylation from P i was inhibited by Ca 2ϩ , as expected, although slightly less efficiently for the p95C peptide than for the other peptides (lanes 5-9 in Fig. 3A and the right portion of Fig. 3B; a detailed discussion of this result will be given below). As P i -derived phosphorylation of ATPase is known to be Mg 2ϩ -dependent (possibly because Mg⅐P i is the real substrate, see Ref. 6), we also checked the effect of increasing concentrations of Mg 2ϩ (see Fig. I in Supplemental Material) and again found no obvious difference in the apparent affinity of p95C-p14N and intact ATPase for Mg 2ϩ concerning P i -derived phosphorylation at equilibrium (despite a reduced apparent affinity for vanadate; this will be further discussed below).
Ca 2ϩ -independent phosphorylation from P i in the presence of Me 2 SO was also measured on ice. In this case, phosphorylation of p95C again proved possible, up to quite high a final phosphorylation level (whereas phosphorylation of p83C was very faint). But the rate of p95C phosphorylation from P i was much slower than for intact ATPase (lanes 1-4 in Fig. 4). After half  2-4). After phosphorylation for 30 s, dephosphorylation of the various fragments was either triggered by addition of 2 mM EGTA, 2 mM MgATP, 1 mM phosphoenol pyruvate, and 0.02 mg/ml pyruvate kinase (to remove ADP as much as possible) and monitored after 5, 30, or 60 s (lanes 5-7), or triggered by addition of EGTA together with 1 mM ADP and again monitored after 5, 30, or 10 s (by mistake) (lanes 8 -10). lane 1 contained prestained molecular weight markers (PSMW). Panel A shows the PhosphorImager scan of the gel. From this scan and from the Coomassie Blue staining ability of the various bands (panel C), we computed the relative ability of the various bands to retain 32 P (panel B), taking the phosphorylation level of intact ATPase at 30 s as 100%.
an hour of phosphorylation under these conditions, dephosphorylation was triggered by a 10-fold dilution of the phosphorylated membranes into a Me 2 SO-free and KCl-and ATP-containing medium similar to that previously used to measure dephosphorylation after ATP-dependent phosphorylation. Although intact ATPase dephosphorylated very rapidly (as expected), this was not the case for p95C, which remained unusually stable. This fully confirms that the slow dephosphorylation rate and reduced ADP sensitivity inferred from the ATP phosphorylation experiment in Fig. 2 above are mainly due to a severe slowing down of the "E2P to E2" transition in the p95C-p14N complex. At 20°C, we also observed both a slowing down of the phosphorylation rate (from P i ) and a slowing down of the dephosphorylation rate of p95C (here, after addition of excess non-radioactive P i in the continued presence of Me 2 SO), compared with intact ATPase (see Fig. J in Supplemental Material). The reduction of both phosphorylation and dephosphorylation rates is consistent with the similar apparent affinities found at equilibrium for p95C-p14N and for intact ATPase (lanes 2-5 in Fig. 3 above). Thus, after ATPase cleavage at Leu 119 -Lys 120 , the P i -derived phosphorylated form of the p95C-p14N complex is neither stabilized nor destabilized relative to the unphosphorylated form, but the kinetics of the phosphorylation/dephosphorylation reactions are slowed down, i.e. the energy barrier for the reaction is higher.
The Unphosphorylated p95C-p14N Complex Has a Poorer Affinity for Ca 2ϩ and a Reduced Rate of Reaction with Ca 2ϩ , Compared with Intact ATPase-The unaltered affinity with which the p95C-p14N complexes can be phosphorylated from P i makes it possible to interpret in more detail the apparent competition, at equilibrium, between Ca 2ϩ binding and phosphorylation from P i illustrated in lanes 5-9 of the preceding Fig. 3. It is well known that, in such experiments, the efficiency of Ca 2ϩ for preventing phosphorylation from P i depends on both the true affinity of the ATPase for Ca 2ϩ and the equilibrium prevailing between phosphorylated and non-phosphorylated forms of Ca 2ϩ -free ATPase. Thus, the fact that the EC 50 of Ca 2ϩ for inhibiting phosphorylation from P i of intact ATPase was about 6 M, i.e. somewhat higher than the Ca1 ⁄2 for equilibrium binding of Ca 2ϩ at pH 7 in the absence or presence of Me 2 SO (e.g. Ref. 6), is consistent with the fact that, in the presence of 200 M P i and Me 2 SO, phosphorylation from P i efficiently reduces the amount of Ca 2ϩ -free ATPase available for reaction with Ca 2ϩ . Similarly, the fact that for the p83C peptide the EC 50 for inhibition by Ca 2ϩ was now less than 1 M is consistent with the above-mentioned unfavorable phosphorylation of this peptide from P i and the unaltered true affinity of this peptide for Ca 2ϩ (3). On these grounds, because intact ATPase and the p95C-p14N complex are phosphorylated with roughly the same apparent affinity for Pi, the fact that phosphorylation of the p95C peptide was reduced by half only at about 20 M Ca 2ϩ (pCa 4.7) suggests that this peptide has experienced a true loss in affinity for Ca 2ϩ .
To investigate the kinetics of the Ca 2ϩ -dependent transition from E2 to Ca 2 E1, we then turned back to Ca 2ϩ -dependent phosphorylation from [␥-32 P]ATP. We designed an experiment, on ice, similar to the one illustrated in Fig. 2 except that before being added to the phosphorylation medium (which contained [␥-32 P]ATP together with an excess of Ca 2ϩ ), PK-treated membranes were now left in the presence of EGTA, to allow Ca 2ϩfree fragments to react with Ca 2ϩ and ATP at the same time (Fig. 5). Phosphorylation of p95C was definitely slower under these conditions (compare with the results in lanes 2-4 in Fig.  2), and much slower than phosphorylation of p83C or intact ATPase. This indicates that in the p95C-p14N complex the calcium-induced transition from the Ca 2ϩ -free form to the Ca 2ϩ -bound and phosphorylatable form is relatively slow. The same fact was observable when experiments were repeated at 20°C with a rapid mixing and quenching machine (see Fig. K in Supplemental Material).
This slow transition allowed us to design a protocol for estimating by a phosphorylation assay the "true" affinity of Ca 2ϩ binding to p95C. In this experiment, PK-treated fragments were pre-equilibrated at various pCa values, and we measured then the amount of phosphorylated p95C formed after only 5-s incubation with [␥-32 P]ATP, anticipating from the results in Fig. 5 that fragments of p95C phosphorylated after this reaction period would correspond to those ATPase fragments that already had Ca 2ϩ bound to them at time zero (because Ca 2ϩfree p95C fragments are unable to react with ATP within such a short period). Fig. L in the Supplemental Material shows that the Ca1 ⁄2 for EP formation by the p95C-p14N complex after only 5s was between 2 and 20 M, i.e. higher than what is known from literature for intact SR or p83C under the same conditions (e.g. Ref. 3) and in fact also found in this experiment. This finding therefore confirms the above conclusion that the true affinity for Ca 2ϩ of the p95C-p14N complex is slightly reduced, relative to that of intact ATPase. Note that when EP measurements were performed after phosphorylation for 30 s instead of only 5 s, the difference between Ca1 ⁄2 values for the p95C-p14N complex and intact ATPase was not as pronounced, as previously mentioned (see Fig. G in Supplemental Material), presumably because in a steady-state situation the strong inhibition of dephosphorylation in the p95C-p14N complex compensates for the loss in true calcium affinity (see, for instance, a related discussion about the effects of steady-state cycling in Ref. 23).
The "True" Affinity with Which Ca 2ϩ -free Unphosphorylated p95C-p14N Complex Binds Orthovanadate Tightly Is Poorer Than That of Intact ATPase-The small shift of the equilibrium affinity for Ca 2ϩ suggested by both Figs. 3 and Fig. L, combined with the much more pronounced slowing down of the Ca 2ϩ -induced transition shown in Fig. 5 (and Fig. K), might suggest that the (reverse) events accompanying Ca 2ϩ dissociation from the p95C-p14N complex are also slightly slowed down, although not to the same extent as the Ca 2ϩ -induced transition. We did not check this directly. Something we did check, however, is the above suggestion, deduced from the kinetic and equilibrium characteristics of the P i -derived phosphorylation of the p95C-p14N complex, that the energy barrier for phosphorylation of this complex from P i is higher than for intact ATPase. Such a relative destabilization of the transition state for the phosphorylation/dephosphorylation reactions was established in experiments in which we measured the true affinity with which this complex can bind orthovanadate, FIG. 4. In the p95C-p14N complex, dephosphorylation of E2P, formed from [ 32 P]P i , is indeed slowed down compared with intact ATPase, and the (reverse) rate of phosphorylation is also slowed down. PK-treated and resuspended membranes were reacted with 200 M [ 32 P]P i in buffer B supplemented with 25% Me 2 SO and 1 mM EGTA, on ice, and the kinetics of phosphorylation of the various bands was first measured (lanes 1-4). After 30-min phosphorylation, dephosphorylation was triggered by 10-fold dilution into Me 2 SO-free buffer A, supplemented with 2 mM MgATP and 1 mM EGTA (similar results were obtained if excess non-radioactive P i was also present, data not shown), and monitored for a few more minutes (lanes 4 -7). Panel A shows the PhosphorImager scan. From this scan and from the Coomassie Blue staining ability of the various bands (not shown), we computed the relative ability of the various bands to retain 32 P (panel B), taking the phosphorylation level of intact ATPase after 30 min at 20°C as 100%. whose binding is generally considered to provide an analog of the transition state formed during phosphorylation from P i . The rationale of such experiments (24 -26) is to incubate the membranes with vanadate for an extended period of time (to attain equilibrium) and, subsequently, to phosphorylate the sample from ATP for a short period of time and thereby measure the amount of ATPase that has not reacted with vanadate. Care was taken, on the one hand, to keep the phosphorylation reaction period short enough to minimize dissociation of previously formed ATPase-vanadate complexes, and, on the other, to make it long enough to ensure full phosphorylation of all vanadate-free ATPase fragments. Such experiments, performed with PK-treated SR membranes, showed that after 1-h incubation at 20°C, vanadate bound to intact ATPase with the expected sub-micromolar affinity, whereas the apparent affinity of the p95C-p14N complex for vanadate was shifted to higher concentrations by more than one order of magnitude (Fig. 6). If, after preincubation with vanadate, the phosphorylation period was increased, the amount of phosphorylated p95C became larger on a relative basis (i.e. compared with phosphorylated intact ATPase), suggesting for vanadate bound to p95C-p14N a faster rate of dissociation which fits with poorer binding. Using this assay, the p83C-p28N complex was found completely unable to bind orthovanadate tightly, as previously found (3) and attributed to a deficient "E2" conformation.

DISCUSSION
Methodology-The present data demonstrate the usefulness of combining phosphorylation protocols with an appropriate gel electrophoretic separation technique, to study the partial reactions of various Ca 2ϩ -ATPase proteolytic fragments, even when only small relative amounts of these fragments are formed along with other cleavage products. Such a combination might also prove useful for the study of other P-type ATPases. For one of the Ca 2ϩ -ATPase complexes resulting from treatment with proteinase K, p83C-p28N, formed after cleavage at Glu 243 , the results we obtained with the present combination of techniques are in agreement with those that we had obtained earlier by other, more standardized techniques (Refs. 3 and 27; see additional discussion about the p83C peptide in the Supplemental Material). This gives credit to the results obtained here concerning the p95C-p14N complex, resulting from cleavage at Lys 120 , which will be discussed more specifically.
The Main Effects of Cleavage at Lys 120 -We found that disruption of the link between the A-domain and the second transmembrane span by cleavage of the Leu 119 -Lys 120 peptide bond gives rise to a pronounced decrease in the ATPase activity of SERCA1a, yet with retention of the capacity for almost full phosphorylation from ATP and P i . The detailed analysis of the properties of the resulting p95C-p14N complex indicates that the reaction rates associated with E2P dephosphorylation or formation of E2P from E2 and P i are both reduced by at least two orders of magnitude, without appreciably affecting the equilibrium constant of the reaction. In comparison, the rate for the E2 3 Ca 2 E1 transition is reduced less strongly, whereas phosphorylation of Ca 2 E1 from ATP and the subsequent Ca 2 E1P3 E2P transition are affected to an even lesser extent. This pattern differs significantly from the pattern found for the previously studied N-terminal proteinase K cleavage product, the p83C-p28N complex, in which the link between the A-domain and M3 (the third membrane span) has been disrupted by excision of the short 239 MAATE 243 sequence (3): in that case proteinase K cleavage primarily results in severe reduction of the Ca 2ϩ translocating Ca 2 E1P3 E2P transition and an impaired capability to form E2P from P i in the absence of Ca 2ϩ . To facilitate discussion, Fig. 7 summarizes the effects on the various transition rates that we have observed in the present work to result from ATPase cleavage at either Lys 120 or Glu 243 (this work and Ref. 3) (the numbers in Fig. 7 should nevertheless be regarded as a crude summary only, because data obtained under different conditions have been pooled together).
The A-domain, a Phosphatase Domain?-A few years ago, the mutation of some of the transmembrane residues critical Phosphorylation from [␥-32 P]ATP of the p95C-p14N complex is slow when the complex initially is in a Ca 2؉ -deprived state, suggesting a slow E2 to Ca 2 E1 transition. Here, in contrast with the experiments illustrated in Fig. 2, PK-treated membranes were left in EGTA (0.5 mM) after proteolysis arrest, and phosphorylation was triggered by 10fold dilution of these membranes into a [␥-32 P]ATP-containing phosphorylation medium that had been supplemented with 100 M total Ca 2ϩ . Phosphorylation was measured after various periods ( lanes  2-4). Dephosphorylation after 30-s phosphorylation was also measured in the same experiment ( lanes 5-7), as a repetition of the dephosphorylation experiment previously illustrated in lanes 5-7 of for Ca 2ϩ binding was also found to result in reduced rates of E2P dephosphorylation (reviewed in Ref. 28). However, these mutants generally had apparent affinities for P i higher than normal, and therefore defective proton binding to the countertransport sites was suspected to be responsible for their reduced rates of dephosphorylation. Subsequently, mutation of Arg 198 in the A-domain was found to moderately reduce the rate of dephosphorylation (29), and more recently, a dramatically reduced rate of E2P dephosphorylation was observed for many mutants of Val 200 near the T2 site in the A-domain of Ca 2ϩ -ATPase (30). This is also the case for a mutant of Glu 183 in the conserved TGES sequence of the A-domain, 2 and a closely related situation has been described to occur after mutation of Thr 214 in the conserved TGES sequence of the A-domain of Na ϩ ,K ϩ -ATPase, where a 5-fold reduction of the dephosphorylation rate was found for the Thr 214 3 Ala mutant (25). Independently, it has been suggested for Na ϩ ,K ϩ -ATPase (32,33) and H ϩ ,K ϩ -ATPase (34) that the glutamate residue of the TGES motif in the A-domain may be interacting with the phosphorylation site through the Mg 2ϩ residue bound to the phosphorylated Asp 351 residue in the catalytic site (this conclusion was derived from cleavage experiments performed in the presence of hydrogen peroxide and ferrous ion, with Fe⅐ATP presumably acting as a substitute for Mg⅐ATP in the catalytic site). Thus, it is likely that the A-domain in general, and the TGES motif in particular, play a specific role in acid/base catalyzed cleavage of the aspartylphosphate bond at Asp 351 in E2P (an effect that may be enhanced if the cavity formed by the regions of the P-domain and A-domain which surround the catalytic site in the E2P state is fairly hydrophobic, see Ref. 22). Our present demonstration of a dramatic inhibition, after cleavage at Lys 120 of the link between M2 and the A-domain, fits well into this picture of the role of the A-domain. All these results revive a concept, previously proposed for H ϩ -ATPases, of the A-domain being a "phosphatase" domain, acting at a late step of the cycle (35)(36)(37), whereas the N-domain would act as a "kinase" domain in an early step of the ATPase cycle. As discussed by Toustrup-Jensen and Vilsen (25), this would also fit with the demonstrated existence, in non-ATPase members of the HAD family, of catalytic residues outside the Rossman fold (corresponding to the ATPase P-domain), to assist the nucleophilic attack of water during dephosphorylation.
Structural Implications of the Effect of Proteolysis-From a structural point of view, it seems clear that the movements of the A-domain are key events for active transport (31). For instance, extensive rotation parallel to the membrane accompanies the ATPase transition from its Ca 2ϩ -bound E1 conformation to its Ca 2ϩ -free conformations, either stabilized by TG or, for two-dimensional membrane crystals, in the additional presence of decavanadate (2, 15, 38; PDB accession numbers 1EUL, 1IWO, and, for the modeled structures, 1FQU and 1KJU); this rotation is probably an essential element for the 2 J. P. Andersen and J. D. Clausen, personal communication.
FIG. 6. The true affinity with which orthovanadate binds to Ca 2؉ -free p95C-p14N complex and inhibits its phosphorylation from [␥-32 P]ATP is clearly poorer than for intact ATPase. PK-treated membranes were diluted to 0.1 mg/ml protein in buffer A (100 mM KCl, 5 mM Mg 2ϩ , and 50 mM Mops-Tris at pH 7.0) supplemented with 0.2 mM EGTA, and incubated for 65 min at 20°C with various concentrations of orthovanadate. Samples were subsequently cooled on ice for 10 min, and phosphorylation was triggered by the simultaneous addition of 5 M [␥-32 P]ATP and 300 M Ca 2ϩ . Acid quenching was performed after 30-s phosphorylation (120 s in the sample indicated by an asterisk, lane 10). Panel A shows the PhosphorImager scan. From this scan and from the Coomassie Blue staining ability of the various bands (not shown), we computed the relative ability of the various bands to retain 32 P (as in Fig. 2B), taking the maximal phosphorylation level of the intact ATPase band (in the absence of vanadate, lane 1) as 100% (panel B). Note that if the phosphorylation period was increased from 30 to 120 s (lane 10), the amount of phosphorylated p95C became larger, consistent with a relatively fast rate of dissociation for vanadate bound to p95C-p14N.
interaction of the A-domain with the P-domain and concomitant positioning of the TGES sequence in the vicinity of the catalytic site. The S2 and S3 ATPase segments linking the cytosolic A-domain to the transmembrane region and containing the proteinase K cleavage sites (Lys 120 and Glu 243 ) could conceivably play a strategic role in this rotation, because the 1IWO Ca 2ϩ -free structure shows that after rotation these segments interact with helices P6 and P7 of the P-domain (Fig. 8). Nevertheless, the fact that the p95C-p14N complex remains capable of being phosphorylated from P i (at a slow rate, but with an unaltered apparent affinity for P i ) implies that E2 and E2P conformations with a rotated A-domain can still be formed after cleavage of S2. In fact, some of our results with the p95C-p14N complex (the decreased affinity for Ca 2ϩ , the slower E2 to Ca 2 E1 transition, and the accumulation of E2P during turnover) might even be interpreted as suggesting that both the E2 and E2P conformations are slightly stabilized in the proteolyzed complex (relative to the Ca 2ϩ -bound conformation), as if cleavage at Lys 120 does not prevent, but even favors stable reorientation of the A-domain (in contrast with cleavage at Glu 243 , see Ref. 3).
Thus, in continuation of the above considerations we still need to address the most remarkable effect of proteolytic cleavage of S2, i.e. the dramatic slowing down of the rates of phosphorylation from P i and of dephosphorylation, an effect likely to reflect perturbation of the pathway through which A-and P-domains can change their mutual interactions during the enzymatic cycle. Of course, this perturbation may stem in part from entropic effects, because new conformations made avail-able (at least locally) by the chain disruption may reduce the probability for proceeding along the normal reaction pathway. But in addition, as previously discussed in connection with similar effects observed for some mutants (25,39), the fact that in the p95C-p14N complex both forward and reverse rates of backdoor phosphorylation from P i are apparently reduced to the same extent may imply relative destabilization of the transition state for the phosphorylation reaction. In support of this view we indeed observed a reduction in the affinity with which the p95C-p14N complex binds orthovanadate, a ligand that is considered as an analog of the pentacoordinated transition state of the phosphoryl group (40).
It is difficult at present to discuss the structural basis for this observation, because we do not have any high resolution structure of this transition state at our disposal. However, by fitting the E1 high resolution three-dimensional structure into low density maps of two-dimensional membrane crystals, two models have been proposed for decavanadate-reacted Ca 2ϩ -ATPase (Protein Data Bank accession codes 1FQU (1) and 1KJU (38)), an ATPase form that probably also contains orthovanadate and may therefore provide the best presently available approximation of the transition state of the protein during dephosphorylation. Judging from these structures, it is not impossible that part of the explanation for the effect of cleavage at Lys 120 may indeed be that cleavage prevents the formation of bonds essential for attaining the transition state (see additional discussion in the Supplemental Material). In this respect, it is remarkable that results very recently obtained by directed mutagenesis confirmed the prominent role for the E2P to E2 transition (and the more modest role for the E2 to Ca 2 E1 transition) of residue Tyr 122 (and also, to a lesser extent, of residue Glu 123 ), immediately above the Leu 119 -Lys 120 bond disrupted by our PK treatment (41).
As indicated in Fig. 7, cleavage of Ca 2ϩ -ATPase at Glu 243 in the segment linking the A-domain and S3 gives rise to a different phenotype for the resulting p83C-p28N complex, where the Ca 2 E1P to E2P transition and the E2P dephosphorylation are both strongly inhibited. In the case of this complex, we could measure inhibition of Ca 2ϩ transport, too (3), whereas in the case of the p95C-p14N complex, we have no evidence at present that the disrupted S2 link has any specific effect on translocation (beyond the expected slowing down of activity). However, it remains a tempting idea that in addition to locally changing the A/P interactions, disruption of the various loops between the A-domain and the transmembrane region prevents mechanical forces to be transmitted from the catalytic domains to the transport sites and vice versa. Together with the links through S2 and S3, the link through S1 will obviously also play a major role in these events, because it has been beautifully demonstrated recently that single deletion of any of the residues in the Glu 40 -Ser 48 loop between M1 and the A-domain slows down both the Ca 2 E1P to E2P transition and the E2P to E2 transition (Ref. 51; see also Supplemental Material). If movement of the A-domain (parallel and perpendicular to the membrane plane) leads to both phosphoenzyme hydrolysis and ion occlusion or deocclusion, this could perhaps justify revival of the old name of "transduction domain" that was once given to the A-domain, although on the basis of controversial experiments (42)(43)(44). The segments that have been proteolyzed in the present work could be among the controlling elements that coordinate events at the cytosolic catalytic site with events at the membranous transport sites, as requested by the "coupling rules" of active transport (45).
Acknowledgments-We thank Pr. Robin Post for suggesting us to try and measure the effect of cleaving the ATPase at Leu 119 -Lys 120 , as well as for helpful comments on our draft. We also thank J. P. Andersen and FIG. 7. Summary of the functional effects of ATPase cleavage at either Lys 120 or Glu 243 . Numbers given here are only very tentative figures, obtained by pooling for each transition the numbers obtained under different experimental conditions (temperature, Me 2 SO, etc.). As concerns the p83C-p28N complex, the reduced rates indicated for phosphorylation from P i and dephosphorylation of E2P are also somewhat tentative; they are based on the results in Fig. 2  J. D. Clausen for communicating to us the results of their unpublished experiments with ATPase mutants and for extensive discussion, as well as for tips on how to make phosphoenzyme measurements more reproducible, Bitten Holm for her assistance in some of these experiments, Paulette Decottignies-Le Maréchal (Université Paris XI, Orsay) for her help with N-terminal sequencing, and Chikashi Toyoshima for discussion.