Localization of a K+ -binding site involved in dephosphorylation of the sarcoplasmic reticulum Ca2+ -ATPase.

K+ plays an important role for the function of the sarco(endo)plasmic reticulum Ca2+ -ATPase (SERCA), but its binding site within the molecule has remained unidentified. We have located the binding site for a K+ ion in the P-domain by means of x-ray crystallography using crystals prepared in the presence of the K+ congener Rb+. Backbone carbonyls from the loop containing residues 711-715 together with the side chain of Glu732 define the K+/Rb+ site in the Ca2+ -ATPase conformation with bound Ca2+, ADP, and AlF4-. Functional analysis of Ca2+ -ATPase mutants with alterations to Glu732 shows that this site is indeed important for the stimulatory effect of K+ on the dephosphorylation rate. Comparison with the Ca2+ -ATPase in a dephosphorylated E2 conformation suggests that the K+ site is involved in the correct movement and positioning of the A-domain during translocation and dephosphorylation.

(SERCA) 1 is a P-type ATPase (1) that actively transports Ca 2ϩ against a large concentration gradient, fuelled by energy derived from ATP. In addition to the primary substrates, Mg:ATP and Ca 2ϩ , monovalent alkali metal ions like K ϩ and Na ϩ (but with the exception of Li ϩ ) also influence the enzymatic cycle of the Ca 2ϩ -ATPase (2). Of particular functional importance is the K ϩ -induced acceleration of dephosphorylation of the ADP-insensitive E2P phosphoenzyme intermediate (2)(3)(4). Thus, addition of K ϩ at the concentration found in the cytoplasm was found to increase the rate of dephosphorylation more than 10-fold relative to the situation without added monovalent cation (3). Furthermore, there is evidence that K ϩ modulates step(s) involved in Ca 2ϩ binding to the dephosphoenzyme (5)(6)(7). Despite the fact that the dependence of Ca 2ϩ -ATPase function on alkali metal ions was discovered more than 25 years ago, the underlying molecular mechanism(s) and the binding site(s) involved have not been determined, but as far as dephosphorylation is concerned kinetic evidence suggests that K ϩ is bound with high affinity in a 1:1 complex and from the cytosolic side (8,9).
Atomic structures of four different states of the Ca 2ϩ -ATPase, obtained by x-ray crystallography, are now available (10,11). The Ca 2ϩ -ATPase consists of three distinct cytoplasmic domains named A ("actuator"), P ("phosphorylation"), and N ("nucleotide binding"), as well as a Ca 2ϩ -binding transmembrane sector comprised by ten mostly helical segments, M1 through M10. With the aim of identifying binding sites for K ϩ we have taken advantage of the fact that this cation can be exchanged by the congener Rb ϩ , thereby allowing for the identification of K ϩ /Rb ϩ -binding sites by difference Fourier techniques. This approach was successfully used before to elucidate the potassium ion-binding sites of for example the potassium channel (12) and the 50 S ribosomal subunit (13). Here, we describe a K ϩ /Rb ϩ -binding site present at Glu 732 in the crystal structure of the Ca 2ϩ -ATPase with bound Ca 2ϩ , ADP, and AlF 4 Ϫ (11), and we provide evidence by analysis of mutant enzyme that this site is functionally important in connection with K ϩ modulation of the dephosphorylation rate.

EXPERIMENTAL PROCEDURES
For crystallization, Ca 2ϩ -ATPase isolated from rabbit skeletal muscle was purified by extraction with a low concentration of deoxycholate according to established procedures (4). The protein was solubilized by 30 mM C 12 E 8 in 100 mM MOPS (pH 6.8), 20% glycerol, 80 mM KCl or RbCl, 10 mM CaCl 2 , 3 mM MgCl 2 , and 1 mM ADP plus 0.33 mM AlCl 3 , and 5 mM NaF, at 12 mg/ml final protein concentration (11). The supernatant was used directly at 19°C in hanging drops by mixing equal volumes of protein and crystallization buffer (8% polyethylene glycol-6000, 4% tert-butanol, 15% glycerol, 5 mM ␤-mercaptoethanol in 200 mM NaOAc). Large, single crystals grew over a few days and were mounted from mother liquor in nylon loops and flash-frozen in liquid nitrogen. Diffraction data were collected at the Protein Structure Factory beamline BL14.1 of Free University Berlin at BESSY and Deutsches Elektronen-Synchrotron (DESY/EMBL) beamline X11 and were processed and merged using the HKL package (14). Oligonucleotidedirected mutagenesis of cDNA encoding rabbit SERCA1a, expression in COS-1 cells, and functional analysis of the overall and partial reactions of expressed enzyme were carried out as described previously (15,16

RESULTS AND DISCUSSION
Structural Analysis-Crystals of the Ca 2ϩ -ATPase in the Ca 2 E1-ADP:AlF 4 Ϫ form grown in the presence of either 80 mM KCl or 80 mM RbCl belong to the same crystal form. The structure of the Ca 2 E1-AMPPCP form refined at 2.6-Å resolution was used as the starting model for structure determination and refinement of the Ca 2 E1-ADP:AlF 4 Ϫ (KCl) form at 2.9-Å resolution as described previously (11). Crystals grown in the presence of Rb ϩ diffracted to 3.3-Å resolution, and data were collected using synchrotron radiation with a wavelength of 0.81 Å where a significant anomalous signal of rubidium is present (fЉ ϭ 3.5 e) (Table I). CNS was used to generate an anomalous difference Fourier map (17) based on data from 40-to 3.3-Å resolution and phases previously derived from the refined model of the KCl form. A single positive peak at 7.1 (background maximum of 4.5 ) was identified at a site displaying the structure of a typical cation-binding site as defined by interactions with several carbonyl and carboxylate oxygens (Fig. 1A). Similarly, a F o (RbCl) Ϫ F o (KCl) difference Fourier map showed a positive peak at 7.2 in the same site (Fig. 1A). The site was therefore assigned as a K ϩ site in the native enzyme.
Description of the Identified Rb ϩ /K ϩ Site- Fig. 1B shows the overall structure of the Ca 2ϩ -ATPase in the E1-ADP:AlF 4 Ϫ conformation with the K ϩ site located between the two C-terminal helices of the P-domain, over the M3 and M5 segments. This localization in the cytoplasmic domain is in agreement with biochemical data demonstrating that K ϩ exerts its effect on the function of the Ca 2ϩ -ATPase when present on the cytoplasmic side (8).
In the electron density maps of the Ca 2ϩ -ATPase with bound AMPPCP or ADP:AlF 4 Ϫ we had previously identified additional density at this position that could not originate from the atoms of the enzyme. Also, the density could not be accounted for by a water molecule since this led to unrealistic low B-factors compared with the B-factors of surrounding protein atoms and an impossible hydrogen bonding network. On the other hand, placing a potassium ion at the site provided a sensible chemical model which was substantiated by further refinement at 2.6-Å resolution of the Ca 2 E1-AMPPCP form, in agreement with the finding with rubidium. This leads to the conclusion that K ϩ is coordinated with carbonyl oxygens from residues 711, 712, and 714 together with one side chain oxygen of Glu 732 , thereby providing a total of four of the six ligands required for an octahedral coordination of the potassium ion. Moreover, the observed site is surface-located, resulting in the potassium ion being exposed (Fig. 1B), thereby allowing solvent molecules to fulfil the coordination.
A Na ϩ to (Rb ϩ or K ϩ ) ratio of 2.5:1 was present at the given experimental conditions and a partial occupancy by Na ϩ in the crystals cannot be excluded, since Na ϩ also stimulates dephosphorylation, although with a somewhat lower affinity than K ϩ (2, 3). Indeed, Toyoshima and co-workers (10, 18) place a Na ϩ ion at the same site in their refined structures of the Ca 2 E1 and Ca 2 E1-AMPPCP forms obtained in the absence of K ϩ . However, at typical physiological conditions with a very high K ϩ to Na ϩ ratio the site is most likely occupied by K ϩ . We have also identified density corresponding to this K ϩ site in preliminary maps of the Ca 2ϩ -ATPase in the E2-thapsigar-  Ϫ form and suggests that this is an integral potassium-binding site of SERCA that remains occupied throughout the enzyme cycle.
Mutagenesis Analysis of the Importance of Glu 732 in K ϩ Modulation of Function-To test the functional importance of the K ϩ site described above, Glu 732 was replaced by either alanine or glutamine, and the mutant and wild-type enzymes, expressed in COS-1 cell endoplasmic reticulum membranes, were compared with respect to K ϩ modulation of function. Table II, left part, shows the steady-state turnover rates for ATP hydrolysis determined at 37°C in the presence of optimal concentrations of Ca 2ϩ and ATP, with and without K ϩ added. It is seen that the addition of 100 mM K ϩ induced a ϳ2-fold increase of the turnover rate for wild-type enzyme, whereas mutant Glu 732 3 Ala showed close to maximum activity already in the absence of K ϩ with little further activation occurring upon the addition of K ϩ . Mutant Glu 732 3 Gln, on the other hand, behaved much like wild type, with K ϩ inducing a ϳ2-fold activation.
The hydrolysis of the ADP-insensitive phosphoenzyme intermediate, E2P, is one of the partial reaction steps contributing to rate limitation of the enzyme cycle under the experimental conditions used for measurement of ATPase activity described above, particularly in the absence of K ϩ (19). This step was examined separately by determining the dephosphorylation rate following phosphorylation of the enzyme with 32 P i (Fig. 2 and Table II, right part). The addition of 100 mM K ϩ increased the dephosphorylation rate by a factor of 14 in wild type and by a factor of 11 in mutant Glu 732 3 Gln, whereas for Glu 732 3 Ala, the dephosphorylation rate was activated only 2.6fold, thus indicating a very significant reduction of the sensitivity to K ϩ . In the absence of K ϩ , the dephosphorylation rate of Glu 732 3 Ala was 1.5-fold higher than that corresponding to wild type, in line with the finding of higher ATPase activity, relative to wild type, in the absence of K ϩ . From these data it is clear that Glu 732 is important for the modulation of E2P dephosphorylation by K ϩ and that this function is disrupted by removal of both side chain oxygen atoms of Glu 732 (in Glu 732 3 Ala) but not by the selective removal of only one of them (in Glu 732 3 Gln), thus indicating that Glu 732 donates one of its side chain oxygen atoms, but not necessarily both, to K ϩ binding. This notion correlates well with the crystal structure, as shown in Fig. 1A. It is also clear that the mutation to alanine relieves an inhibition existing when the side chain oxygen is not coordinated by K ϩ . The Glu 732 3 Ala mutation will destabilize the K ϩ site substantially, yet a residual stimulatory effect of K ϩ was still observed in this mutant, which may be ascribed to low affinity binding in the identified site.
Proposed Mechanism of K ϩ Modulation of Dephosphorylation-Evidence has accumulated that dephosphorylation re-quires that domain A is rotated relative to its position in E1 and E1P forms. Upon dissociation of ADP the highly conserved TGES motif of domain A is inserted into the catalytic site of domain P, with the glutamate of TGES participating in the hydrolysis of the aspartyl phosphoryl bond as catalytic residue in the E2P form (10,20,21). The dephosphorylation of E2P has previously been found to be sensitive to amino acid substitutions or proteolytic cleavage in the linker segments connecting either domain A or domain P with the transmembrane segments M2, M3, and M5 (15,16,22,23). Hence, the linker segments appear to hold domain A and domain P in the optimal positions for catalysis of dephosphorylation. We believe that the importance of the identified K ϩ site should be seen in this context. Interestingly, the crystal structure of the thapsigargin-bound E2 dephosphoenzyme form (24) shows that the A-M3 linker is within a 5-Å distance of the K ϩ -binding site identified here, and the models for the structure of the vanadate-bound FIG. 2. Dephosphorylation of phosphoenzyme formed with 32 P i . Phosphorylation was performed at 25°C for 10 min in 100 mM MES/Tris (pH 6.0), 2 mM EGTA, 10 mM MgCl 2 , 30% (v/v) dimethyl sulfoxide, and 0.5 mM 32 P i . The samples were then chilled on 0°C ice water, and dephosphorylation was studied at 0°C by a 19-fold dilution of the phosphorylated microsomes into 50 mM MOPS/Tris (pH 7.0), 2 mM EGTA, 2 mM MgCl 2 , 5 mM non-radioactive P i , without (open symbols) or with (solid symbols) 100 mM KCl, followed by acid quenching at the indicated time intervals. The lines show the best fits of a monoexponential decay function, giving the rate constants listed in Table II. E2 form suggested on the basis of analysis of two-dimensional membrane crystals (25,26) also indicate that these sites come close together. Hence they might interact directly in the E2P state (23), and the K ϩ ion could, thus, play a role as a "crosslinker" that stabilizes the interaction between the lower part of domain P and the A-M3 linker. Indeed, it should be possible for the K ϩ ion with free, solvent-exposed ligand-binding sites to coordinate incoming oxygen atoms from residues of the A-M3 linker. Which advantages would the K ϩ ion offer as such an interdomain cross-linker, compared with, for example, a salt bridge? First of all, the interaction must conform to the coordination of the K ϩ ion, and it will therefore be sensitive to conformational changes. This may serve as a regulatory switch in the E2P to E2 transition, where a close yet transient interaction between the P-domain and the A-M3 linker may stimulate the dephosphorylation reaction. In support of this view, Na ϩ and Rb ϩ display coordination chemistries similar to K ϩ and are also observed to have a stimulatory effect on the dephophorylation rate, whereas the chemically far more distinct Li ϩ does not.
Is the K ϩ Site Conserved in the P-type ATPase Family?-The alignment in Fig. 1c shows that the Glu 732 residue is highly conserved as either Glu or Asp. The residues making up the 710 -715 loop are only partly conserved, but since only the backbone carbonyls are contributing to the binding site this is not surprising. More importantly Pro 709 is highly conserved and is likely to be essential for the formation of the loop by discontinuing the preceding helix. Taken together, this suggests that the site may be a general feature of P-type ATPases. A stimulatory effect of K ϩ has also been found in the plant H ϩ -ATPase (27), indicating the possible functional role of the site in this pump, as well. For Na ϩ ,K ϩ -and H ϩ ,K ϩ -ATPases, K ϩ is a direct substrate for intramembraneous cation-binding sites, which then either activates or inhibits the pump (when binding from the extracellular or cytoplasmic side, respectively). An additional, modulatory effect of the cytoplasmic K ϩ site described here may therefore be obscured in most func-tional studies of Na ϩ ,K ϩ -and H ϩ ,K ϩ -ATPases, and the importance of a possible, modulatory effect of K ϩ on these pumps may need to be addressed by renewed interpretation and acquisition of experimental data similar to those described here for the Ca 2ϩ -ATPase.