Intramolecular Interactions of the Regulatory Region with the Catalytic Core in the Plasma Membrane Calcium Pump*

The access of three proteases to their sites of cleavage was used as a measure of regulatory interactions in the plasma membrane Ca 2 (cid:1) pump isoform 4b (PMCA4b). When the proteases could not cut at their sites in the C-terminal regulatory region, the interaction was judged to be tight. This was the case in the absence of Ca 2 (cid:1) , when chymotrypsin and caspase cut PMCA only very slowly. Ca 2 (cid:1) accelerated the fragmentation, but the digestion remained incomplete. In the presence of Ca 2 (cid:1) plus calmodulin, the digestion became nearly complete in all cases, indicating a more flexible conformation of the carboxyl terminus in the fully activated state. The acceleration of proteolysis by Ca 2 (cid:1) or Ca 2 (cid:1) plus calmodulin occurred equally at the caspase site upstream of the calmodulin-binding domain and the chymotrypsin and calpain sites downstream of that domain. Replacing Trp 1093 (a key residue within the calmodulin-binding domain) with alanine had a much more specific effect, because it exposed only proteolytic sites within the cal-modulin-binding domain that had previously been shielded in the native protein. At these sites,

The plasma membrane Ca 2ϩ ATPase or pump (ATP2B1-4, referred to here as PMCA) 1 is a P-type ATPase that is an essential element of cellular Ca 2ϩ homeostasis. Its role is to remove excess Ca 2ϩ from the cell in order to maintain the large Ca 2ϩ concentration gradient existing between the cytosol and the extracellular space. This pump is encoded by four different genes, and further diversity is produced by alternative splicing of the primary transcripts at two different sites (1). Isoforms PMCA1 and 4 are ubiquitous, whereas the expression of PMCA2 and 3 are more cell-and tissue-specific.
The activity of the plasma membrane Ca 2ϩ pump is regulated by calmodulin. It has a unique carboxyl terminus that is responsible for the regulation. A high affinity, ϳ28-residue calmodulin-binding sequence is located within this region, which is also a built-in inhibitor of the enzyme (2, 3) (Fig. 1). Removing the carboxyl terminus either by proteolysis or truncation by mutagenesis causes full activation of the pump similar to that caused by calmodulin (4). Synthetic peptides representing the calmodulin-binding sequence were able to restore the inhibition when added to the fully active forms (2,4). Cross-linking experiments using synthetic peptides have revealed that the N-terminal half of this sequence interacts with the large cytoplasmic loop (N domain), whereas the C-terminal part interacts with the small cytoplasmic loop (A domain) of the catalytic core (5)(6)(7). These interactions must interfere with the movements of the A and N domains during the reaction cycle so that, in the absence of calmodulin, the pump has low affinity for its substrate Ca 2ϩ and a low maximal turnover rate. Binding of Ca 2ϩ -calmodulin to the pump removes the inhibitor from this interaction, allowing high affinity substrate binding and the reaction cycle to proceed with full speed.
The PMCA isoforms show great diversity in their regulation with calmodulin. They have different basal activities, different affinities for calmodulin, and different rates of activation and inactivation in response to calmodulin binding and dissociation (8 -10). Among the isoforms, PMCA4b has the lowest basal activity and the greatest stimulation by calmodulin. An important feature of this pump is that binding of calmodulin followed by activation is very slow, with a half-time of ϳ20 -60 s depending on the Ca 2ϩ concentration (11). The inactivation by calmodulin removal is even slower, with a half-time of about 20 min. These features of PMCA4b suggest that it is actively involved in the developing, shaping, and duration of the Ca 2ϩ signal; it allows the first Ca 2ϩ spike to develop for several seconds before becoming activated and remains activated for several minutes after the spike is dissipated so that it will respond to the next spike much faster.
The low basal activity of the PMCA4b isoform indicates that the interaction between the carboxyl terminus and the catalytic core is very strong. We have shown that Trp 1093 within the calmodulin-binding sequence is an essential anchor for the auto-inhibitory interaction as well as for the interaction with calmodulin. Changing this residue to an alanine increased the basal activity in the absence of calmodulin, the rate of calmod-ulin activation, and, more dramatically, the rate of inactivation by calmodulin removal (12). In addition to the calmodulinbinding sequence, regions both upstream and downstream provide additional anchor points for the auto-inhibitory interaction. Replacing Asp 1080 five residues upstream of the calmodulin-binding sequence (13) or removing the downstream inhibitory region (3) caused substantial activation of the pump. We have to emphasize that any change made within the carboxyl terminus that increased the basal activity also increased the rate of activation by calmodulin.
Controlled proteolysis of the purified and membrane-bound erythrocyte Ca 2ϩ pump (which is mainly PMCA4b) has revealed several sites at the carboxyl terminus that are cut by proteases such as trypsin (14), chymotrypsin (2), calpain (15)(16), V8 protease (17), and, more recently, caspase-3 (18). Among these proteases, chymotrypsin, calpain, and caspase-3 cut the protein preferentially at the carboxyl terminus. The N-terminal fragments were all fully or partially active depending on what portion of the carboxyl terminus was removed. Irreversible activation of the erythrocyte Ca 2ϩ ATPase by calpain has been reported (15). In the absence of calmodulin, the pump was proteolyzed to a group of fragments of 124 -125 kDa, whereas in the presence of calmodulin a 127-kDa fragment was identified. A later study using a purified erythrocyte pump or synthetic peptides determined the cleavage sites and found that all were located within the calmodulin-binding domain (16). In more recent experiments, we have demonstrated that PMCA4b is also cut by caspase-3 during the early phase of apoptosis (18). This cleavage occurs immediately downstream of Asp 1080 , removing the whole regulatory region and producing a fully active 120-kDa fragment.
In this study three different proteases, i.e. calpain, chymotrypsin, and caspase-3, were used to digest PMCA4b in membrane preparations from COS cells over-expressing PMCA4b. Because all three proteases cut the pump at the carboxyl terminus, conformational changes of this region induced by Ca 2ϩ and Ca 2ϩ -calmodulin could be monitored by limited proteolysis. We found that the accessibility of the proteolytic sites depended strongly on the conformational state of this protein.
The sites within the calmodulin-binding domain were all shielded either by the autoinhibitory interactions with the catalytic core or the interaction with calmodulin. Sites within this sequence became accessible only when a key residue, Trp 1093 , was mutated to an alanine. The proteolytic sites upstream and downstream of the calmodulin-binding domain were somewhat more exposed in the auto-inhibited state; however, they became fully accessible only after calmodulin binding.

MATERIALS AND METHODS
Chemicals-Calmodulin and ␣-chymotrypsin were obtained from Sigma. Recombinant caspase-3 was from Upstate Biotechnology, and -calpain was obtained from Calbiochem. All other chemicals used for this study were of reagent grade.
Construction of Trp 1093 3 Ala-The construction of the Trp 1093 3 Ala mutant is described in our previous paper (12). The expression vector used was pMM2, which was derived from pMT2PC (19) and uses the SV40 promoter.
Cell Culture and Transfection-COS cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, 100 units/ml penicillin, and 2 mM L-glutamine. All cells were kept at 37°C, 5% CO 2 in a humidified atmosphere. COS cells were transfected as described previously (9) using the LipofectAMINE reagent (Invitrogen) based on the protocol as described by the manufacturer. Transfection using 175-cm 2 flasks was initiated when the cells were 70 -80% confluent. The cells were incubated at 37°C with a DNA-Lipo-fectAMINE complex (formed by incubating 8 g DNA and 100 l Lipo-fectAMINE in 3.6 ml of serum-free Opti-MEM medium for 30 min) in 14.5 ml of serum-free Opti-MEM medium. After 5 h of incubation, the cells were supplemented with serum, and incubation was continued for a total of 24 h. The medium containing the DNA-LipofectAMINE com-plex was then replaced with fresh tissue culture medium with 10% serum, and the cells were cultured for an additional 24 h.
Isolation of Microsomes from COS Cells-Crude microsomal membranes from COS cells were prepared as described previously (4) with the following modifications. Cells were washed with ice-cold phosphatebuffered saline solution and then harvested in the same medium containing 0.1 mM phenylmethylsulfonyl fluoride, 6 g/ml aprotinin, 2.2 g/ml leupeptin, and 1 mM EGTA, pH 7.4. After centrifugation, cells were resuspended in an ice-cold hypotonic solution containing 10 mM Tris-HCl, pH 7.4, 1 mM MgCl 2 , 0.5 mM EGTA, 4 g/ml aprotinin, 2 g/ml leupeptin, and 4 mM dithiothreitol. After lysis, homogenization, and centrifugation, the final pellet was resuspended in a solution of 0.25 M sucrose, 0.15 M KCl, 10 mM Tris-HCl, pH 7.4, 2 mM dithiothreitol, 20 M CaCl 2 , and the suspension was stored in liquid N 2 .
In Vitro Digestion with Recombinant Caspase-3-10 g of microsomal membrane vesicles isolated from COS cells transfected with the appropriate construct were preincubated for 3 min at 37°C in a 100-l medium containing 100 mM KCl, 25 mM TES-triethanolamine, pH 7.2, 0.09 mM EGTA, 8.5% sucrose, 5 mM dithiothreitol, 20 g/ml aprotinin, 20 g/ml leupeptin, 0.1% CHAPS, 0.1 mM CaCl 2 (10 M free Ca 2ϩ ; when added), and 485 nM calmodulin (when added). The proteolysis was started by the addition of 0.25 g of recombinant caspase-3 and stopped by the addition of ice-cold trichloroacetic acid (6% final concentration). The precipitate was supplemented with 100 g of bovine serum albumin, washed once with distilled water, and then dissolved in electrophoresis sample buffer.
In Vitro Digestion with -Calpain or Chymotrypsin-10 g of microsomal membrane vesicles isolated from COS cells transfected with the appropriate construct were preincubated for 3 min at 37°C in a 100-l medium containing 100 mM KCl, 25 mM TES-triethanolamine, pH 7.2, 0.2 mM CaCl 2 (100 M free Ca 2ϩ ; when added), 0.1 mM EGTA, 5 mM dithiothreitol, 1 mM MgCl 2 , 0.05% Triton X-100, and 485 nM calmodulin (when added). The proteolysis was initiated by the addition of 5 g/ml -calpain or 0.05 g/ml ␣-chymotrypsin and stopped by the addition of ice-cold trichloroacetic acid (6% final concentration). The precipitate was supplemented with 100 g of bovine serum albumin, washed once with distilled water, and then dissolved in electrophoresis sample buffer.
Gel Electrophoresis, Electrotransfer, and Immunostaining-The samples were electrophoresed on 7.5% acrylamide gel following Laemmli's procedure except that the sample buffer contained 62.5 mM Tris-HCl, pH 6.8, 2% SDS, 10% glycerol, 5 mM EDTA, 100 mM dithiothreitol, and 125 mg/ml urea. The samples were subsequently electroblotted, and the blots were immunostained by the antibody 5F10. The constructs and their fragments were quantified using the Bio-Rad Molecular Imager system.
Structural Model of PMCA4b-The sequence of PMCA4b was aligned with that of SERCA1a using a combination of computer methods and accumulated knowledge about the properties of different parts of the pump sequences. This alignment was then processed with the program MODELLER 6.2 (20), using the SERCA1a structure files 1EUL (SERCA E1) or 1IWO (SERCA E2), taken from the Protein Data Bank (21,22).

RESULTS
PMCA4b was expressed in COS cells, and limited proteolysis was performed in native COS cell membranes at 37°C. We used caspase-3, chymotrypsin, and calpain digestion of PMCA4b to study conformational changes induced by Ca 2ϩ and Ca 2ϩ -calmodulin. These proteases cut the C-terminal regulatory region of PMCA4b within a short segment (about 40 -50 residues) downstream of Asp 1080 . Following proteolysis, the samples were run on SDS-gels and immunoblotted. The immunoblots shown in the figures were stained with the antibody 5F10. 5F10 recognizes a linear epitope from residues 719 -738 that is common to all known human PMCA gene products and splicing variants. To verify that the pump was cut C-terminally, Western blot analysis was also performed with two PMCA4-specific antibodies (data not shown). The monoclonal antibody JA9 reacts specifically with PMCA isoform 4 near the N terminus within residues 51-75, whereas monoclonal antibody JA3 binds near the C terminus from residues 1156 -1180 (23). All of the proteolytic fragments detected by 5F10 were also detected by JA9; however, JA3 did not recognize any of the fragments. The antibody results indicate that all three proteases cut PMCA4b at the C terminus upstream of the JA3 recognition sequence.
The sizes of the fragments were determined by comparing their electrophoretic mobilities to those of C-terminally truncated mutants ct120, ct92, and ct71 (Fig. 1). These PMCA4b mutants are missing 120, 92, or 71 residues from the C terminus (24). Fig. 2A shows the fragmentation of PMCA4b by the different proteases, and Fig. 2B shows the relative electrophoretic mobility of the same fragments as compared with that of the truncated mutants. The calpain digest produced two N-terminal fragments. The larger fragment migrated close to the position of ct71, whereas the smaller fragment migrated close to that of ct92. Based on relative mobility, the calculated molecular masses of the fragments were 125 and 124 kDa, respectively. These results indicated that one calpain cut site is ϳ20 residues downstream of the calmodulin-binding domain, and a second site is located close to the C-terminal end of the calmodulin-binding domain (Fig. 1). The chymotrypsin digest produced one major N-terminal fragment of 124.5 kDa, which is smaller than ct71 but larger than ct92. Thus, the chymotryptic cut site must also be located downstream of the calmodulin-binding domain between Ser 1113 and Pro 1134 . Based on the estimated molecular weight of the fragment, the most likely position of the chymotryptic cut site is after Tyr 1122 as indicated in Fig. 1. A more precise determination of the cut sites by N-terminal sequencing of the small fragments was not possible, because Ͻ1% of the total membrane protein is accounted for by the expressed PMCA. We have to emphasize that mild conditions were used to produce these fragments. In the case of chymotrypsin especially, increasing concentrations of the protease or longer digestions induced further fragmentation of the protein. As also shown in Fig. 2, the caspase-3 digest produced a single N-terminal fragment of 120 kDa even after prolonged incubations, because PMCA4b contains only one caspase consensus site, Asp 1077 -Glu-Ile-Asp 1080 , which is located five residues upstream of the calmodulin-binding domain (Fig. 1). In a previous paper, we have demonstrated that caspase-3 cuts immediately after Asp 1080 (18).
Next, we studied the effects of Ca 2ϩ -calmodulin and free Ca 2ϩ on the fragmentation of the pump. In the absence of Ca 2ϩ -calmodulin, the pump is in an inhibited (closed) state and has very low activity and low affinity for its substrate, Ca 2ϩ . In the presence of Ca 2ϩ -calmodulin, the pump is in an activated (open) state with high activity and high Ca 2ϩ affinity. In addition to the open and closed states of the pump, an equilibrium exists between the E2 and E1 catalytic conformations. Low concentrations of Ca 2ϩ shift the E2-E1 equilibrium more toward E1 (25). Figs. 3 and 4 show the fragmentation of PMCA4b with caspase-3 and chymotrypsin. These proteases do not require Ca 2ϩ for activity; therefore, we could study the fragmentation of PMCA4b in the absence of Ca 2ϩ (in the presence of EGTA). These figures show that, in the absence of Ca 2ϩ , both caspase-3 and chymotryptic cleavage sites are protected from proteolysis because very little fragmentation was seen under FIG. 2. Electrophoretic migration of the fragments produced by caspase-3, calpain, and chymotrypsin is compared with that of the truncated mutants. 10 g of microsomes isolated from COS cells transfected with PMCA4b were digested with calpain, chymotrypsin, or caspase-3 as described under "Experimental Procedures." A, the proteolyzed samples and C-terminally truncated PMCA4b mutants ct92, ct71, and ct120 were immunoblotted with the anti-PMCA antibody 5F10 as indicated. B, the migration of PMCA4b and its fragments was analyzed using the Bio-Rad Molecular Imager system. The intensity of chemiluminescence is plotted as a function of mobility. Solid lines represent the migration of the fragments produced by calpain, chymotrypsin, or caspase-3. Dashed lines represent the migration of the C-terminally truncated mutants on the same gel. Results are representative of three separate experiments.
FIG. 1. Amino acid sequence of the C-terminal regulatory region of PMCA4b. The putative cleavage sites for caspase-3, calpain, and chymotrypsin are marked with black arrows. The gray arrows show the calpain sites of the purified erythrocyte Ca 2ϩ pump as determined by James et al. (16). The region marked CaM binding corresponds to the high affinity calmodulin-binding domain of PMCA4b. ct120, ct92, ct71, and ct57 are C-terminally truncated mutants missing 120, 92, 71, and 57 residues from the C terminus, respectively (an arrow indicates the Cterminal end of each mutant). these conditions. The addition of Ca 2ϩ to the incubation media substantially increased the fragmentation. However, only ϳ30 -40% of the protein could be digested even after prolonged incubation, indicating that 60 -70% of the protein remained protected from cleavage. Binding of Ca 2ϩ -calmodulin, on the other hand, exposed the cleavage sites, and fragmentation became complete with both caspase-3 and chymotrypsin. Fig. 3A shows some intact protein remaining at 3 h of digestion with caspase-3; this intact material disappeared completely after longer exposure to the protease (not shown). Fig. 3B shows the progressive formation of the 120-kDa fragment in the presence of calmodulin by caspase-3. Fig. 4B also shows a more progressive time-dependent accumulation of the chymotryptic 124.5-kDa fragment in the presence of calmodulin; however, only a maximum of 60% of the total protein was converted to this fragment because of further degradation (smaller fragments are not shown). Fig. 5 shows the degradation of PMCA4b by -calpain. Calpain requires Ca 2ϩ for activity; therefore, only the effect of Ca 2ϩ -calmodulin on digestion could be tested. In the absence of Ca 2ϩ -calmodulin, calpain produced two fragments (125 and 124 kDa); however, the digestion was incomplete. In the pres-ence of Ca 2ϩ -calmodulin, the 124-kDa fragment was not seen, although the formation of the125-kDa fragment became more pronounced, resulting in a nearly complete disappearance of the intact protein by the end of the digestion. These data indicate that Ca 2ϩ -calmodulin binding protects the calmodulin-binding site from calpain cleavage but exposes the cleavage site downstream.
Mutating Trp 1093 to an alanine has been shown to increase the basal activity of PMCA4b by 3-fold (12). In Fig. 6 we show that chymotrypsin and calpain cut the Trp 1093 mutant more readily in the absence of calmodulin and that its cleavage pattern is rather different from that of the wild type. With calpain, the 124-kDa fragment became predominant in the absence of calmodulin, indicating that the cleavage site within the calmodulin-binding domain is more exposed in the mutant. With chymotrypsin the change in the pattern was even more dramatic. Presumably, the mutation of Trp 1093 exposed a site that was otherwise inaccessible in the wild type pump. The new chymotryptic fragment is slightly smaller than ct92, with an estimated molecular mass of 123 kDa. The most likely cleavage site is after Phe 1094 within the calmodulin-binding domain that is protected from proteolysis when Ca 2ϩ -calmodulin is bound to the mutant.
We also tested whether mutation at Trp 1093 affected the accessibility of the cleavage sites upstream (caspase-3) or downstream (calpain and chymotrypsin) of the calmodulinbinding domain. The digestion of the mutant with caspase-3 was very similar to that of the wild type (not shown), as the fragmentation of the mutant was highly dependent on the presence of Ca 2ϩ and calmodulin. This indicated that mutation at Trp 1093 did not affect the accessibility of the caspase-3 cleavage site. To test the downstream cut sites, we studied the digestion of the proteins by calpain in the absence of calmodulin. Fig. 7 shows a time course of the formation of the 125-and 124-kDa calpain fragments. Fragmentation of the wild type protein resulted in a small amount of the 124 kDa fragment (ϳ10% of the total enzyme) and somewhat more of the 125 kDa fragment (ϳ20%). Mutation of Trp 1093 to an alanine increased the formation of the 124-kDa calpain fragment ϳ3-fold but did not affect or even decreased the formation of the 125-kDa fragment. Similar results were obtained with chymotrypsin (data not shown). The results with calpain and chymotrypsin indicated that mutation at Trp 1093 did not affect the accessibility of the downstream cut sites but increased the accessibility of the sites within the calmodulin-binding sequence. DISCUSSION It is widely known that the plasma membrane Ca 2ϩ pump is regulated by calmodulin. At its carboxyl terminus it has a high-affinity calmodulin-binding sequence that also serves as an auto-inhibitor. Previous studies have predicted that in the absence of calmodulin, i.e. in the inhibited or closed conformation, the calmodulin-binding inhibitory sequence interacts with the catalytic core. By using site-directed mutagenesis it has been demonstrated that regions both upstream and downstream of this sequence are also involved in the auto-inhibition. The limited proteolysis experiments described above were aimed at identifying the conformational changes accompanying Ca 2ϩ -calmodulin binding to PMCA4b in order to further understand the role of the carboxyl terminus in the regulation of the activity of this pump. Conformational changes induced by Ca 2ϩ -calmodulin binding to the calmodulin-binding site, predicted on the basis of these experiments, are illustrated in Fig. 8.
Three proteases with different substrate specificities were used to digest PMCA4b in its natural membrane environment. It has been demonstrated that all three proteases cut the pump near the carboxyl terminus (2, 14, 18). Previously, we showed that caspase-3, a key protease of apoptosis, cuts the protein five residues upstream of the calmodulin-binding sequence after Asp 1080 . This cut removes the whole inhibitory region and produces a single 120-kDa N-terminal fragment that is fully active without calmodulin (18). We also showed that Asp 1080 is essential in auto-inhibition, because mutation of this residue increased the basal activity substantially (13).
Here we demonstrated that digestion of PMCA4b at this site is highly conformation-dependent. In the presence of EGTA, when the E2-E1 equilibrium is shifted more toward the E2 conformation, this site is completely protected. In the presence of Ca 2ϩ in the E1 conformation ,this site seems more exposed, and the pump becomes partially degraded by the protease. Binding of Ca 2ϩ -calmodulin to the calmodulin-binding sequence induces a larger conformational change that fully exposes Asp 1080 to the protease so that the digestion is complete. This suggests that the carboxyl terminus of PMCA4b should be protected from caspase-3 like proteases in non-stimulated cells at low intracellular Ca 2ϩ concentrations and that an increase in cytosolic Ca 2ϩ followed by calmodulin binding is needed to cleave PMCA4b by caspase-3 during apoptosis.
Calpain and chymotrypsin were used to map the region downstream of the calmodulin-binding sequence of PMCA4b. Within this region, between residues Ser 1113 and Pro 1134 we identified one major cut site for each protease. Digestion of the protein at these sites also depended on the apparent conformational state similar to that seen with caspase-3. In the presence of EGTA, no digestion with chymotrypsin could be seen, whereas in the presence of Ca 2ϩ the cut sites were more exposed, and the pump became partially degraded by either calpain or chymotrypsin. Full digestion, however, was observed only in the presence of Ca 2ϩ -calmodulin. These data suggest that, in the auto-inhibited or closed conformation, the sequences both upstream and downstream of the calmodulinbinding domain interact with the catalytic core that makes the proteolytic sites fully or partially resistant to proteolysis, depending on the equilibrium between the E1 and E2 configurations. In addition to the major cut site, digestion with calpain has revealed another site closer to the calmodulin-binding sequence. This cut site, however, is only weakly exposed in the absence of calmodulin, whereas it is completely shielded by calmodulin binding. Based on the size of this fragment we concluded that the cut site must be located near the end of the calmodulin-binding sequence.
Carboxyl-terminal sequencing of the calpain fragments derived from the erythrocyte Ca 2ϩ pump (mainly PMCA4b) has revealed two additional sites within the calmodulin-binding sequence (16). In those experiments, the digestion was carried out in the absence of calmodulin on ice and for long incubation times. When we digested the erythrocyte pump in its natural environment in the membrane under the same conditions used in our experiments, the sizes of the fragments were identical to those of the fragments of the over-expressed protein in COS cells (not shown in detail). Thus, the over-expressed protein showed a similar proteolytic pattern as the pump native to erythrocytes. To determine why we did not see the additional fragments seen by others, we also digested these pumps on ice for 30 min (not shown in detail) and found that more fragments smaller than ct92 but bigger than ct120 were formed when the protease was calpain. When chymotrypsin was used, one fragment migrating faster than ct92 could be identified in addition to that seen at 37°C. These results indicated that, at low temperatures, the sites within the calmodulin-binding sequence are more exposed to proteolysis than at 37°C. It has to be emphasized that in the presence of calmodulin, these smaller fragments could not be detected even after digestion at 0°C, and only the typical fragments that were seen at 37°C were detected using either calpain or chymotrypsin. In previous experiments the calpain cut site in the presence of calmodulin was identified using synthetic peptides representing the calmodulin-binding sequence (16). In that case it is evident that calpain cut the enzyme within this sequence, because the native cut site was missing from the peptides. Our data show that the whole calmodulin-binding sequence is protected against proteolysis in the native enzyme at 37°C either in the absence or presence of calmodulin. In the absence of calmodulin this sequence must be shielded by intra-molecular interactions with the catalytic core. Binding of calmodulin induces a conformational change, allowing activation of the pump and exposing sites both upstream and downstream of the calmodulinbinding sequence, although it remains protected by the interaction with calmodulin.
We reported previously that replacing Trp 1093 , a key residue within the calmodulin-binding sequence, increased the basal activity of the pump while lowering its affinity for calmodulin (12). The higher basal activity observed in the mutant indicates a less effective interaction between the calmodulin-binding sequence and the catalytic core. Here we demonstrated that the calmodulin-binding sequence became more flexible and that sites otherwise inaccessible became more exposed to proteolysis in the mutant. With calpain the smaller fragment became predominant, whereas with chymotrypsin a new fragment was formed when proteolysis was performed in the absence of calmodulin. Based on the relative mobility of these fragments, the most probable chymotryptic cut site is after Phe 1094 , whereas the calpain cut site is located near the end of the calmodulinbinding sequence. Binding of calmodulin to the mutant protected these sites from the proteases, which is also consistent with the notion that these sites must be located within the calmodulin-binding sequence. Digestion at the sites upstream and downstream of the calmodulin-binding sequence was not affected by the mutation, indicating that the effect of the mutant was very local and specific. Even though it loosened the binding to the catalytic core of the immediate surroundings of the mutant, its effect did not extend to the regions upstream and downstream of the calmodulin-binding domain. This was in contrast to the more global effects of calmodulin and Ca 2ϩ . These experiments confirm that Trp 1093 has an essential role in the auto-inhibitory interaction of the calmodulin-binding sequence with the catalytic core.
Comparing the crystal structure of the SERCA pump in the E1 versus E2 conformation (21,22), it is evident that the pump undergoes large-scale movements involving both transmembrane and cytoplasmic domains. Based on the structural homology between SERCA and PMCA, we may hypothesize how binding of the inhibitory calmodulin-binding sequence to the catalytic core inhibits the activity of PMCA. Cross-linking experiments using synthetic peptides representing the sequence of the calmodulin-binding region of PMCA have identified two sites for the interaction; one is located within the nucleotide binding or N domain downstream of the phosphoenzyme forming aspartic acid (between residues 537-544), and the other is between residues 206 -271 within the actuator or A domain (5,6). These two binding sites are in regions that are homologous between SERCA and PMCA, so it is possible to draw valid conclusions by consideration of a model of PMCA based on the structure of SERCA. When such a model is constructed, it is evident that the two binding sites move relative to each other during the enzyme cycle. The binding site in the N domain is in a region whose maximum dimension is about 1.8 nm. Because the exact location of the binding site in the A domain was not well defined by the cross-linking study, the binding site could be anywhere within a region defined approximately by a sphere of 3.2 nm diameter. In the E2 state, this large sphere of possible binding locations in the A domain comes within 1.2 nm of the binding site in the N domain. In the E1 state the binding sites move apart, and the minimum distance between them is 2.5 nm. Although our present uncertainty about the location of the binding site in the A domain does not allow us to define more accurately the distances between the sites, it is clear that they move relative to one another during the enzyme cycle. Binding of the inhibitory sequence to the N domain and the A domain would interfere with the movements of these domains and slow down the reaction cycle, thus inhibiting the activity of PMCA. The possible role of the regions upstream and downstream of the calmodulin-binding sequence is to stabilize the auto-inhibited conformation. This model is also supported by previous observations that synthetic peptides similar to those used in the cross-linking experiments were able to inhibit efficiently the activity of the SERCA pump (26).
The proteolytic patterns obtained with the three proteases verify the conformational changes associated with E1/E2 conformation. We found that, in the presence of EGTA (which favors the E2 conformation), the carboxyl terminus of PMCA was completely resistant to proteolysis, whereas in the presence of Ca 2ϩ the cut sites were somewhat more exposed. Full digestion was observed only in the presence of Ca 2ϩ -calmodulin. Our data agree very well with the idea that the A and N domains move relative to one another as the enzyme goes between the E1 and E2 conformation. In the E2 conformation, the inhibitory region is tightly bound to the catalytic core so that it forms a compact structure together with the A and N domains, preventing access of these proteases to their cut sites.
In the E1 configuration, where the A and N domains are more separated, this binding could be weaker, allowing some flexibility of the C-terminal region and thus exposing more of the proteolytic sites. Removal of the contact points that anchor the C terminus to the catalytic core by site-directed mutagenesis (changing Asp 1080 or Trp 1093 to alanine or truncating the downstream inhibitory sequence) may perturb this interaction, resulting in a higher basal activity as has been observed (3,12,13). The effect of these mutations appeared to be very local and specific, because mutation of Asp 1080 did not affect the proteolytic sites downstream, and mutation of Trp 1093 exposed cut sites only in its immediate surroundings. Binding of Ca 2ϩcalmodulin to the calmodulin-binding sequence has a more global effect on the structure; it displaces the whole inhibitory region, making the carboxyl terminus more flexible. As a result, in the presence of Ca 2ϩ -calmodulin the proteolytic sites are more sensitive to proteases, and the pump is fully active.