Substrate-induced Conformational Fit and Headpiece Closure in the Ca 2 (cid:1) ATPase (SERCA)*

Protection of the Ca 2 (cid:1) ATPase (SERCA) from proteinase K digestion has been observed following the addition of Ca 2 (cid:1) , Mg 2 (cid:1) , and nucleotide and interpreted as a substrate-dependent conformational change (1). The protected digestion site is located on the loop connecting the A domain and the M3 transmembrane helix. We studied by mutational analysis the protective effect of AMP-PCP, an ATP analog that is not utilized for enzyme phosphorylation. We found that the nucleotide protective effect is interfered with by single mutations of Arg-560 and Glu-439 in the N domain and Lys-352, Lys-684, Thr-353, Asp-703, and Asp-707 in the P domain. This is consistent with a transition from the open to the compact configuration of the ATPase headpiece and approximation of the N and P domains by interactions with the nucleotide adenosine and phosphate moieties, respec-tively. The A domain-M3 loop is consequently involved. Protection by nucleotide substrate increased following the mutations of Asp-351 (the residue undergoing phosphorylation by ATP) and neighboring Asn-706 to Ala, underlying the importance of side chain specificity in positioning the nucleotide terminal phosphate and limiting the stability of the substrate-enzyme complex. Protection is not observed when AMP-PCP is added in the absence iments based on ATPase estimates by Western blotting and compensa- tion with empty microsomes. The reaction was quenched at serial times by the addition of trichloroacetic acid to yield a final 2.5% concentration. The quenched protein was then solubilized by adding sodium dodecyl sulfate (1%), Tris (0.312 M ), pH 6.8, sucrose (3.75%), (cid:1) -mercaptoethanol (1.25 m M ), and bromphenol blue (0.025%). The samples were then subjected to electrophoretic analysis (10) on 12.5% gels followed by staining with Coomassie Blue or Western blotting. For this purpose, the monoclonal antibody mAb CaF3–5C3 to the chicken SERCA-1 protein and goat anti-mouse IgG-horseradish peroxidase-conjugated secondary antibodies were used followed by densitometry of the bands visualized with an enhanced chemiluminescence-linked detection system (Amer-sham Biosciences). Amino acid sequencing of peptide fragments eluted from electrophoretic gels was performed at Johns Hopkins University.

The sarcoplasmic reticulum (SR) 1 ATPase is a 110-kDa enzyme that utilizes 1 mol of ATP for active transport of two calcium ions in exchange for 2 mol of protons. The catalytic cycle begins with cooperative binding of two calcium ions from the cytosolic medium followed by ATP utilization and formation of a phosphorylated enzyme intermediate. The bound calcium is then released onto the luminal side of the SR, and the cycle is completed by hydrolytic cleavage of acylphosphate.
The ATPase molecular structure includes a transmembrane region made of 10 clustered helical segments including the Ca 2ϩ binding domain and a cytosolic headpiece including the nucleotide binding (N), phosphorylation (P), and actuator (A) domains. Structural studies have shown that if crystallization is performed in the presence of Ca 2ϩ the three headpiece domains are distinctly separate in an open configuration (2). On the other hand, in the absence of Ca 2ϩ , those domains gather to form a compact headpiece (3,4). This finding suggests that in solution the three domains undergo fluctuations, yielding different enzyme conformations depending on the presence of specific ligands. Taking advantage of a selective ATPase cleavage (5), it was then shown that the addition of nucleotides in the presence of Mg 2ϩ and Ca 2ϩ protects the ATPase from digestion by proteinase K (1). Protection was attributed to a change in the positions of the headpiece domains with respect to the loop connecting the A domain to the transmembrane helices. These movements are directly related to gate opening and closing in the Ca 2ϩ pathway. This is a very important finding, suggesting that the open headpiece of the Ca 2ϩ -activated enzyme acquires a compact conformation upon substrate binding.
We describe here a series of experiments on the interference of various mutations with the protective effect of nucleotide. We demonstrate that occurrence of the nucleotide-dependent conformational change requires participation of amino acid residues in both N and P domains and that nucleotide-dependent approximation of the headpiece domains occurs only when both Ca 2ϩ sites are filled.

MATERIALS AND METHODS
SR vesicles were obtained with the microsomal fraction of rabbit leg muscle homogenate as described by Eletr and Inesi (6). Recombinant ATPase was obtained by exogenous gene expression in COS-1 cells infected with adenovirus vectors carrying chicken WT or mutant SERCA1 cDNA (7). Adenovirus vector construction, site-directed mutations, COS-1 cells culture methods, and preparation of microsomal fractions were previously described in detail (8). Protein concentration was measured using bicinchoninic acid with the biuret reaction (Pierce). ATPase hydrolytic activity was determined following P i production by a colorimetric method (9).
Experiments on limited digestion of ATPase with proteinase K and protection by nucleotides or Ca 2ϩ were performed in media containing 50 mM MOPS, pH 7.0, and 50 mM NaCl in the presence or in the absence (CDTA or EGTA present) of 5 mM MgCl 2 and/or 0.1 mM CaCl 2 . In experiments with native SR vesicles, the protein concentrations were 0.3 mg of SR and 0.01 mg of proteinase K/ml. In experiments with recombinant ATPase, the protein concentrations were 1.2 mg of microsomal protein and 0.04 mg of proteinase K/ml. The concentration of recombinant ATPase was adjusted to the same level in all of the exper-* This work was supported by National Institutes of Health Program Project HL27867 and the Human Frontier Science Program. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Nucleotide and Divalent Cation Protection of ATPase from
Digestion by Proteinase K-Electrophoretic analysis (Fig. 1A) shows that limited digestion of ATPase with proteinase K yields a different pattern depending on the absence (EGTA present) or the presence of Ca 2ϩ . In fact, whereas 95-and 83-kDa proteolytic products are obtained in the absence of Ca 2ϩ , the 95-kDa band is not observed in the presence of Ca 2ϩ . Amino acid sequencing demonstrates that these fragments are produced by cleavage at Leu-119 and Thr-242, respectively. Therefore, the pattern shown in Fig. 1 indicates that Ca 2ϩ binding protects the Leu-119 site while leaving the Thr-242 site accessible to proteinase K. On the other hand, in agreement with Danko et al. (1), we found that ATP or AMP-PCP protects the enzyme from proteinase K digestion at the Thr-242 site (Fig. 1). Protection by nucleotide occurs in the presence of Ca 2ϩ but not in the absence of Ca 2ϩ , and the overall protection of the ATPase band by nucleotide and Ca 2ϩ is much more effective than that of Ca 2ϩ alone ( Fig 1B). It is of interest that while ADP has still a protective effect, AMPCP (i.e. ADP analog) does not (does not shown), suggesting a role of the oxygen atom between the ␣and ␤-phosphate for substrate stabilization at the catalytic site.
As a preliminary to mutational analysis, we confirmed that identical results are obtained with recombinant ATPase. Because the microsomal preparations derived from COS-1 cells contain several additional proteins, the digestion products of recombinant ATPase were evidenced by Western blotting, which is favored by their reactivity to the same antibody. In fact, since unrelated protein bands are not detected by Western blotting, the results of proteinase K digestion and the protection by AMP-PCP can be demonstrated quite clearly (Fig. 2).
Effects of Site-directed Mutations-Considering that protection from proteinase K digestion is produced by functionally relevant ligands such as Ca 2ϩ and nucleotide, we studied the effects of site-directed mutations that may interfere with binding and/or utilization of such ligands for catalytic reactions. A list of mutants produced for these experiments is shown in Table I. It is also shown in Table I that most mutants were expressed at levels nearly as high as those obtained with WT protein. Few mutants such as D703A and D707A were expressed at lower levels, perhaps because of defective folding of the nascent peptide. Generally, the digestion of mutants by proteinase K proceeded somewhat faster than digestion of WT ATPase as shown by the half-time of disappearance of the main ATPase band (Table I). This was more pronounced as a consequence of D707A and the K684A mutations and may be due to some folding destabilization by the mutations.

Mutations of Residues Involved in Ca 2ϩ
Binding-The specific Ca 2ϩ requirement for protection by AMP-PCP was most convincingly demonstrated by the use of the E771Q and N796A mutants. Consistent with inhibition of specific Ca 2ϩ binding (11), we found that these mutations interfere also with AMP-PCP protection of ATPase digestion by proteinase K (Fig. 3 and Table I). It is of interest that protection by AMP-PCP is totally abolished by a mutation that interferes with the binding of only one Ca 2ϩ (i.e. N796A) as well as by a mutation that interferes with the binding of both Ca 2ϩ (i.e. E771Q). Because it is known that enzyme activation requires the binding of both Ca 2ϩ , it is apparent that protection by nucleotide requires specific enzyme activation by Ca 2ϩ . Mutations of Residues in the N Domain-We recently found that mutation of Arg-560 to Ala produces strong inhibition of ATP utilization but much less inhibition of reverse enzyme phosphorylation with P i (12). We find now that the R560A mutation interferes completely with AMP-PCP protection of the enzyme from digestion with proteinase K (Fig. 4 and Table  I). These combined observations indicate that the Arg-560 side chain is specifically involved in the stabilization of nucleotide in the N domain.
Because of their proximity to the nucleotide binding site, we also characterized the effects of Glu-439 and Arg-489 mutations. We found partial catalytic inhibition as a consequence of their mutations, but interference with the AMP-PCP-protective effect was observed only in the E439A mutant (Table I).
Considering that the bound nucleotide must reach the P domain to be utilized as a catalytic substrate and that nucleotide protection form proteinase K may be related to N and P domain approximation, we extended our mutational analysis to the P domain.
Mutations of Residues in the P Domain-Ca 2ϩ -dependent utilization of ATP results in phosphorylation of Asp-351 to form a phosphorylated enzyme intermediate. Therefore, mutation of this residue results in complete enzyme inactivation. On the other hand, AMP-PCP protection from proteinase K digestion is retained following mutation of Asp-351 to Asn (Table I) and is actually much improved by mutation of Asp-351 to Ala (Fig.  5) or Asn-706 to Ala (Table I).
The behavior of the Asp-351 mutant in conjunction with the effect of the AMP-PCP pseudo substrate demonstrates unambiguously that phosphoryl transfer is not required for the protective effect, but simple nucleotide binding is effective. In addition, the greater protection observed with the D351A and N706A mutants indicates that the native side chains of these amino acids play a role in positioning the nucleotide ␥-phosphate and limiting the stability of the substrate-enzyme complex. In fact, a rather weak substrate-enzyme complex, relative to the subsequent transition state, is required for optimal enzyme kinetics.
It was previously reported that Lys-684, a P domain residue, reacts with adenosine triphosphate pyridoxal in the presence of Ca 2ϩ , and this reaction is blocked by ATP (13). Accordingly, we found that mutation of neighboring Lys-684 to Ala produces catalytic inactivation as well as interference with the protective effect of AMP-PCP (Table I). It is then apparent that the  Lys-684 side chain plays an important and direct role in stabilization of the nucleotide substrate terminal phosphate.
We also found that mutations of Lys-352, Thr-353, Asp-703, and Asp-707 produce total or partial catalytic inactivation and interfere with the protective effect of AMP-PCP ( Table I). All of these residues reside in close proximity of the phosphorylation site (i.e. Asp-351). Analogous mutational analysis of the Na ϩ K ϩ -ATPase (14) indicates that electrostatic interactions around the phosphorylation site may play an important role in substrate positioning and utilization. We considered that their mutation may alter direct interactions with the nucleotide terminal phosphate or ligation of Mg 2ϩ in conjunction with oxygen atoms of the ATP-terminal phosphate. On the other hand, the K352E mutation is not as effective as the K352A mutation. Furthermore, Mg 2ϩ is not required for the nucleotide-protective effect (see below). Fig. 6A that the protective effect of AMP-PCP is obtained to the same extent, even when Mg 2ϩ is omitted (CDTA present), and Ca 2ϩ is present at concentrations (20 -100 M) that are much lower than the effective nucleotide concentrations (1 mM).

Divalent Cation Specificity and Nucleotide Concentration Dependence-It is shown in
We also found that irrespective of the presence or the absence of Mg 2ϩ the protective effect occurs at 0.1-1.0 mM nucleotide concentrations. This range is higher than the 10 micromolar K m observed with ATP as a substrate for ATPase activity. It is of interest that the concentration dependence of the protective effect of AMP-PCP shifts to a lower range when the D351A mutant is used (Fig. 6B). A high affinity of the D351A mutant for the nucleotide substrate was previously noted by McIntosh et al. (15). DISCUSSION Definition of the SR ATPase crystal structure (2, 3) has been a major step in the understanding of this enzyme. With regard to nucleotide binding to the ATPase, studies of crystals soaked with TNP-AMP (2) revealed that the adenosine moiety is located within a flap of the N domain near the Phe-487, Lys-492, and Lys-515 and Arg-560 residues whose derivatization or mutation interferes with nucleotide binding (12, 16 -20). A recent model based on ATP-Fe 2ϩ (replacing Mg 2ϩ ) catalyzed oxidation and on the crystal structure of the ATPase in the E2(TG) state places the adenosine moiety in a pocket delimited by Leu-492 and Lys-515 with stabilization provided by Arg-560 (21). The ATP-Mg complex is in a folded configuration with Mg 2ϩ stabilized by ␣and ␤-phosphate oxygen atoms and by the Thr-441 side chain (N domain), whereas the ␤-phosphate approximates Thr-353 (P domain) and the ␥-phosphate approximates Asn-359 and Asp-601 (hinge region). However, although this arrangement may result from early collision of the ATP-Mg complex with the ATPase open conformation, the geometry of ATP must then be altered to place its terminal phosphate near Asp-351 (P domain) so that phosphoryl transfer to this residue can take place.
Useful information on this subject was provided by studies on nucleotide-induced protection of the ATPase from digestion with proteinase K. The digestion site is at Thr-242 located on the loop connecting the A domain to the M3 helix. Thus, the susceptibility is expected to be affected by the position of the A domain. In fact, this is an isolated loop in E1 Ca 2ϩ (open configuration) but is attached to the P domain in E2(TG) (compact configuration), reflecting the different configuration of the cytoplasmic domains. Steric hindrance by the P domain ap- pears to be the origin of the protection of this site in the E2(TG) state. However, different degrees of protection from proteolysis are observed in the E2(TG), CaE1ATP, and E2P states (even though the headpiece resides in a compact configuration in all of these states), reflecting a graded response of the A domain to these ligand-induced transitions (1).
Our mutational analysis indicates that Arg-560 and Glu-439 (N domain) as well as Asp-351, Lys-352, Thr-353, Lys-684, Asp-703, Asn-706, and Asp-707 (P domain) are involved in the nucleotide effect. Participation of N and P residues demonstrates that approximation of the two domains does indeed occur as required to accommodate the nucleotide by means of adenosine moiety interaction with the N domain and phosphate interaction with the P domain. The A domain must also reposition as shown by the protection from proteinase K. The Ca 2ϩ requirement for nucleotide protection indicates that even though a compact arrangement of the headpiece is favored by nucleotide binding, the transmembrane domain retains bound Ca 2ϩ . The compact headpiece conformation obtained under these conditions is not identical to that observed in the absence of Ca 2ϩ (1) but evidently represents an additional specific state produced by nucleotide binding on the Ca 2ϩ -activated enzyme. An apparently similar approximation of nucleotide binding ("fingers") and catalytic ("palm") domains is known to occur in DNA polymerases (22).
It is of interest that the nucleotide concentration dependence of the protective effect is in the 0.1-1.0 mM range. This range is higher than the 10 M K m observed with ATP as a substrate for ATPase activity, suggesting that the kinetics of catalytic ATP utilization have a significant influence on the overall ATP concentration dependence of enzyme activity. It should be pointed out that a rather weak substrate-enzyme complex, relative to the stability of the subsequent transition state, is required for effective enzyme kinetics.
An additional finding is the lack of Mg 2ϩ requirement for nucleotide protection of ATPase digestion by proteinase K (Fig.  6A). It is of interest that in structural snapshots (23) of the phosphoserine phosphatase (PSPase) (its catalytic domain is analogous to the P domain of Ca 2ϩ ATPase) reaction cycle, the initial substrate-enzyme complex does not include Mg 2ϩ , whereas interaction with Mg 2ϩ occurs in concomitance with the phosphoryl transfer and hydrolytic reactions. With regard to the Ca 2ϩ ATPase, it is probable that a complex with Mg 2ϩ is formed initially (21) when ATP is the substrate due to the cation binding property of the nucleotide. On the other hand, our present experiments suggest that nucleotide binding (and its conformational effect) can be obtained either in the presence or in the absence of Mg 2ϩ , although the subsequent phosphoryl transfer and hydrolytic reactions require Mg 2ϩ in analogy to PSPase. This suggestion is consistent with the random mechanism proposed by Reinstein and Jencks (24) for ATP and Mg 2ϩ binding to the Ca 2ϩ ATPase. In analogy to the PSPase, it is likely that catalytically required Mg 2ϩ binding occurs in concomitance with the phosphorylation transition state, including coordination by Asp-703, Thr-353, and ␥-phosphate.
Mutational analysis demonstrates direct roles of Arg-560 (N domain) and Lys-684 in nucleotide stabilization in the N and P domains, respectively. Arg-560 may interact with the nucleotide ␣-phosphate. Lys-684, on the other hand, is likely to interact with ␥-phosphate in analogy to stabilization of phosphorylserine by Lys-144 in the PSPase (23). The roles of other residues are more complex. For instance, the absence of nucleotide protection in the Asp-703 mutant could be attributed to a role of this residue in Mg 2ϩ stabilization. On the other hand, nucleotide protection is obtained even in the absence of Mg 2ϩ . Therefore, Asp-703 must have an additional and important role in structural stabilization by hydrogen bonding with neighboring residues. Similar considerations can be made regarding Asp-707, Lys-352, and Thr-353, whose side chains are likely to establish hydrogen bonding with neighboring residues and/or water molecules (23). Note that the K352A mutation is much more effective than the K352E mutation, indicating that the ability of K352 to establish stabilizing interactions in the open or compact headpiece conformation is more important than a possible interaction with ␥-phosphate oxygen atoms.
Another interesting finding is related to the higher protective effect of nucleotide in the D351A and N706A mutants as compared with WT Asp-351 and Asp-706 (as well as with the D351N mutant). Asp-351 is in fact the residue undergoing phosphorylation, and Asn-706 (Asn-170 in the PSPase) is a close neighbor that contributes its side chain nitrogen (3,23) for coordination of the same ␥-phosphate oxygen as Lys-684 (Lys-144 in the PSPase). It is then apparent that in the WT enzyme, a number of steric and electrostatic constraints guide  (23). A, a view along the plane of ribose of ATP. B, a view corresponding to that presented previously (21), approximately orthogonal to the view in A. Side chains of important residues are shown. Their positions match closely to those of the corresponding residues in PSPase with the substrate phosphoserine-bound (23). Green sphere, Mg 2ϩ bound to ATP; purple sphere, Mg 2ϩ found in the MgF x complex. The arrow in broken line in A shows the direction of approximation of the N domain to the P domain. the ␥-phosphate to an optimal position for covalent interaction. On the other hand, even tighter binding may be obtained by removing some of these constraints.
To understand our experimental results, an atomic model was built based on the atomic model of the PSPase with its substrate bound (23). This was possible because of the close similarity between the atomic structures of the catalytic domain of the PSPase and the P domain of the MgF x complex of the Ca 2ϩ ATPase (considered to be an E-P analog). 2 The model for the Ca 2ϩ ATPase was built combining that for the N domain taken from the E2(TG) form (Protein Data Bank code 1IWO) (3) and that for the P domain from the MgF x complex. ATP in an extended form was placed so that the ␥-phosphate comes exactly to the same position as that of the phosphate in phosphoserine (Protein Data Bank code 1L7P).
The ATP orientation in the model (Fig. 7) suggests that Mg 2ϩ coordinated by the ␤and ␥-phosphates would not come to a position suitable for coordination by Asp-703. In fact, in the D11N mutant of the PSPase (corresponding to the D351N mutant of the Ca 2ϩ ATPase and used for the substrate bound state), no Mg 2ϩ is found as opposed to the subsequent states of the catalytic cycle in which Asp-167 (Asp-703 in the Ca 2ϩ ATPase) participates. This clearly indicates that the initial step of substrate binding does not require Mg 2ϩ , consistent with the observation that Mg 2ϩ is not required for the nucleotide-induced protection from proteinase K digestion of the Ca 2ϩ ATPase.
In the D11N mutant of the PSPase, the conformation of Asn-11 is different from other states. The conformation of Asn-11 is one of the standard rotamers and places the side chain nitrogen atom close to the phosphate. This conformation is stabilized by hydrogen bonding between oxygen atom of Asn-11 and main chain amide of Asp-167 (Asp-703 in Ca 2ϩ ATPase). Thus, this is likely to be the conformation in the native enzyme when substrates bind initially. Here phosphates are coordinated by atoms corresponding to Lys-352 (main chain NH), Thr-353 (main chain NH), Thr-625 (O␥1), Gly-626 (main chain NH), Lys-684 (N), and Asn-706 (N␦2) in the Ca 2ϩ ATPase. Thus Lys-684 is the only positively charged residue likely to be a key attractor for the ␥-phosphate. Hence the position of the side chain nitrogen atom must be very important. In fact, in phosphoglucomutase (25,26), the C␣ position of the corresponding residue is different by one residue but the position of the nitrogen atom is exactly the same. This position appears to be controlled by Asp-351 and Asp-707 in the Ca 2ϩ ATPase when substrates are absent. Thus, mutations of these residues may strongly affect the affinity for ATP and consequently its protective effect. This may explain the higher protection observed with the D351A as compared with the D351N mutant.
The effect of the N706A mutant is particularly interesting. N706 in the MgF x complex of the Ca 2ϩ ATPase as well as the corresponding Asn-170 in the PSPase provides a nitrogen atom for coordination of the same phosphate oxygen as Lys-684. Thus, the strong nucleotide protection observed in the N706A mutant was contrary to our expectation. Although this Asn residue is not well conserved in the haloacid dehologenase superfamily, the corresponding residue (Asn-170) of the PSPase is involved in a large conformation change and appears to regulate the flap (or lid) of the substrate binding site by changing the interaction with Phe-49 in the subdomain. Asn-706 may bear a similar role and might adjust the position of the N or A domain. At least in the MgF x complex, this residue does interact with the key loop in the A domain ( 181 TGES 184 ) or it may have an influence on the position of Lys-684. If mutated to Ala, Lys-684 might take a position that allows stronger interaction with ATP. This may be a likely reason for the strong nucleotide protection observed in the Asn-706 mutant.
Finally, it should be pointed out that in the model (Fig. 7) Arg-560 takes a very strategic position. It stabilizes ATP by interacting with the ␣-phosphate and also stabilizes the closed configuration of the N and P domains by making a salt bridge with Asp-627, one of the critical residues in the P domain. In this sense, Arg-560 may have a similar role to Arg-56 in the PSPase. Why Asp-627 is important has never been explained. This conformation of Asp-627 is realized by an interaction with Lys-352, which appears to serve also in positioning Thr-353, another critical residue. This may be the reason for the profound effect of the K352A mutation.
In conclusion, our experiments indicate that nucleotide binding occurs by collision with the N domain in the Ca 2ϩ -dependent open conformation of the enzyme headpiece. A substrateinduced conformational fit then takes place relative to stabilization of the headpiece domains in a compact configuration. This allows approximation of the ATP ␥-phosphate to Asp-351 in the P domain, whereas the membrane-bound domain still binds two Ca 2ϩ . The roles of several amino acid residues are demonstrated, in some cases related to direct substrate binding and in other cases related to short and long range interactions of protein structure.