The role of the M6-M7 loop (L67) in stabilization of the phosphorylation and Ca(2+) binding domains of the sarcoplasmic reticulum Ca(2+)-ATPase (SERCA).

The amino acid sequence (L67) intervening between the M6 and M7 transmembrane segments of the Ca(2+) transport ATPase was subjected to mutational analysis. Mutation of Pro(820) to Ala interferes with protein expression even though transcription occurs at normal levels. Single mutations of Lys(819) or Arg(822) to Ala, Phe, or Glu allow good expression, but produce strong inhibition of ATPase activity. The main defect produced by these mutations is strong interference with enzyme phosphorylation by ATP in the presence of Ca(2+), and also by P(i) in the absence of Ca(2+). The Lys(819) and Arg(822) mutants undergo slight and moderate reduction of Ca(2+) binding affinity, respectively. Reduction of overall steady state ATPase velocity is then due to inhibition of phosphorylated intermediate formation. On the other hand, a cluster of conservative mutations of Asp(813), Asp(815), and Asp(818) to Asn interferes strongly with enzyme activation by Ca(2+) binding and formation of phosphorylated enzyme intermediate by utilization of ATP. Enzyme phosphorylation by Pi in the absence of Ca(2+) undergoes slight or no inhibition by the triple aspartate mutation. Therefore, the triple mutation interferes mainly with the calcium-dependent activation of the ATPase. The effect of the triple mutation can be to a large extent reproduced by single mutation of Asp(813) (but not of Asp(815) or Asp(818)) to Asn. Functional and structural analysis of the experimental data demonstrates that the L67 loop plays an important role in protein folding and function. This role is sustained by linking the cytosolic catalytic domain and the transmembrane Ca(2+) binding domain through a network of hydrogen bonds.

The sarcoplasmic reticulum (SR) 1 ATPase is a membranebound enzyme that plays a crucial role in sequestration of cytosolic Ca 2ϩ in muscle fibers (1)(2)(3). The catalytic and trans-port cycle begins with high affinity binding of two Ca 2ϩ per ATPase, whereby the enzyme is activated and proceeds to utilization of ATP. Enzyme phosphorylation by ATP induces vectorial translocation of bound Ca 2ϩ , and the cycle is then completed by hydrolytic cleavage of the phosphoenzyme. The SR ATPase contains 994 amino acids (4), folded to form 10 transmembrane segments (M1-M10) and a large extramembranous (cytosolic) region (5,6). Mutational (7) and structural studies (8) have shown that the Ca 2ϩ binding domain resides within the membrane-bound region of the ATPase, at a 50-Å distance from the phosphorylation domain in the cytosolic region of the enzyme. Functional interdependence of Ca 2ϩ and phosphorylation domains occurs through long range intramolecular linkage (9).
The cytosolic region of the ATPase includes a short N-terminal segment (Met 1 -Glu 58 ) and the loop (Trp 107 -Ser 261 ) between the M2 and M3 transmembrane segments, folded in the A domain. Furthermore, the cytosolic region includes the large loop between the M4 and M5 transmembrane segments, folded in separate P (phosphorylation) and N (nucleotide binding) domains (8). Although the loops between the remaining transmembrane segments are relatively small, attention has been brought to L67 (Phe 809 -Gly 831 ) by Falson et al. (10) and Menguy et al. (11), who found that cluster mutations of aspartic residues to alanine raise the Ca 2ϩ concentration required for ATPase activation. Our interest on L67 was heightened by its close relationship with the P domain (8), and the major role played by the contiguous M6 segment in Ca 2ϩ binding (12). Therefore, we performed a detailed mutational analysis of L67.

MATERIALS AND METHODS
DNA Constructs and Vectors-The chicken fast muscle SR ATPase (SERCA-1) cDNA (13) was inserted into the pUC19 plasmid for amplification, and then subcloned into the pSELECT-1 vector for site directed mutagenesis. Mutations were carried out by the Altered Sites in vitro mutagenesis system made available by Promega (Madison, WI), or by overlap extension using the polymerase chain reaction.
Wild type and mutated cDNA was subcloned into the shuttle plasmid p⌬E1sp1A (Microbix BioSystems). In the final constructs, the cDNA was preceded by the SV40 or the cytomegalovirus promoter, and followed by the SV40 polyadenylation signal. The shuttle plasmids were either used directly for transfection of COS-1 cells by the DEAE-dextran method, or for cotransfection of HEK293 cells in conjunction with the replication defective adenovirus plasmid pJM17 (Microbix BioSystems) to obtain recombinant adenovirus vectors (14). Alternatively, cDNA constructs were subcloned into pAd-lox shuttle vector and cotransfected with purified ⌿5 adenovirus genome in CRE8 cells derived from the HEK293 line. CRE8 cells constitutively express the Cre recombinase, which catalyzes efficient recombination between loxP sites in the ⌿5 genome and in the pAd-lox to yield recombinant adenovirus (15). The recombinant products were plaque-purified and cesiumbanded, yielding concentrations on the order of 10 10 plaque-forming units/ml.
Transfections of COS-1 cells with wild type or mutated SERCA-1 cDNA, subcloned into the shuttle vector p⌬E1sp1A or pAd-lox, were conducted by the DEAE-dextran method (16).
Recombinant adenovirus vectors were used for infection of COS-1 cells as described by Zhang et al. (12).
Northern blots were performed with total RNA obtained by aspirating the medium from cell culture dishes, adding 0.3-0.4 ml of Trizol-LS (Life Technologies, Inc.)/10-cm 2 cell lawn, and following the Life Technologies procedure. The final mRNA pellet was dissolved in diethyl pyrocarbonate-water (Quality Biological) to yield ϳ1 mg of RNA from one 150-mm dish.
Electrophoresis and blotting were performed by the use of a Northern Max-Gly blotting kit from Ambion, and mRNA detection was obtained with DIG High Prime labeling and detection kit from Roche Molecular Biochemicals. Alternatively, a 32 P-labeled probe was used. A SERCA-1 cDNA cassette was used as a template for the probe.
Microsome Preparation and Immunodetection of Expressed Protein-The procedure for microsome preparation from infected COS-1 cells was as described by Autry and Jones (17), and the final product was stored in small aliquots at Ϫ70°C. The total microsomal protein was determined using bicinchoninic acid with the Biuret reaction (Pierce). The expressed SERCA-1 ATPase was detected by Western blotting (12). Quantitation of immunoreactivity was obtained by densitometry, and standardized with samples of wild type ATPase to be used as controls for the functional studies.
Functional Studies-ATPase hydrolytic activity was assessed by measuring P i production (18). The reaction mixture contained 20 mM MOPS, pH 7.0, 80 mM KCl, 10 mM MgCl 2 , 5 mM sodium azide, 1.0 mM EGTA, and CaCl 2 to yield various free Ca 2ϩ concentrations, 10 -30 g of microsomal protein, and 3 M A23187 Ca 2ϩ ionophore. The reaction was started by addition of 3.0 mM ATP and run at 25°C temperature. Serial samples were taken every 2 min for 12 min. Due to the presence of the Ca 2ϩ ionophore, the ATPase reaction proceeded at constant velocity, yielding linear plots of P i production.
Steady state levels of phosphorylated intermediate by utilization of ATP was obtained in the presence of 20 mM MOPS, pH 7.0, 80 mM KCl, 10 mM MgCl 2 , 30 g of microsomal protein, 5 M Ca 2ϩ , and 5 M [␥-32 P]ATP, in a total volume of 50 l. The reagents were pre-cooled in ice. The reaction was started by the addition of ATP, run for 10 s at 2-3°C temperature, and quenched by the addition of 1.0 ml of cold 1 M PCA. Rapid mixing upon addition of ATP and PCA was obtained by vortexing. The quenched samples were transferred into a 1.7-ml Eppendorf tube containing 100 g of bovine serum albumin as carrier protein, and placed in ice. The samples were then spun in a refrigerated clinical centrifuge at 5000 rpm for 5 min, and the sediments ware washed three times with 1.0 ml of cold 0.125 MPCA and once with cold water. The final pellets were dissolved in 40 l of denaturing buffer (50 mg of lithium dodecyl sulfate, 0.01 ml of 2-mercaptoethanol, and 0.05 ml of Weber-Osborn buffer per ml) and 10 l of tracking dye solution (1 mg of bromphenol blue and 0.3 g of sucrose per ml). The entire samples were then run on 6.5% acrylamide gels at pH 6.2, with a current limit of 100 milliamperes, at 15°C temperature. The radioactive phosphoenzyme was detected both by autoradiography and phosphoimaging.
The time course of phosphoenzyme formation following addition of ATP and Ca 2ϩ , or ATP to enzyme preincubated with Ca 2ϩ , was determined by preincubating 30 g of WT protein, or corresponding aliquots of mutant ATPase (as indicated by Western blot analysis) at 3°C. The reaction medium contained 20 mM MOPS, pH 7.0, 80 mM KCl, 10 mM MgCl 2 , and 1 mM EGTA in the absence or in the presence of 1 mM CaCl 2 . The reaction was started by the addition of 1 M [␥-32 P]ATP and 1 mM CaCl 2 , or 1 M [␥-32 P]ATP. Acid quenching at serial times, washing, and detection of phosphorylated intermediate was obtained by electrophoresis, autoradiography, and phosphoimaging. The time of autoradiographic exposure was the same for all samples, so as to reveal differences in phosphoenzyme levels, if any.
The time course of radioactive phosphoenzyme decay was determined by adding a chase pulse of 0.5 ml of 1.0 mM non-radioactive ATP, 10 s following the initial addition of radioactive ATP. Several samples were obtained by quenching the reaction at serial time with PCA. Radioactive phosphoenzyme was then determined by electrophoresis, autoradiography, and phosphoimaging.
Reverse enzyme phosphorylation by P i was obtained by adding 30 g of microsomal protein to 0.2 ml of reaction mixture containing 50.0 mM MesTris, pH 6.2, 10 mM MgCl 2 , 20.0% Me 2 SO, and 2.0 mM EGTA. Alternatively, the EGTA was omitted and 10, 50, or 100 M CaCl 2 was added. Following a 10-min incubation at 25°C, 1.0 ml of 1.0 M ice-cold PCA was added, the samples were transferred into a 1.7-ml Eppendorf tube containing 100 g of bovine serum albumin as a carrier protein, and placed in ice. Centrifugation, washing, electrophoresis, and detection of phosphoenzyme were then conducted as described above for enzyme phosphorylation with ATP.
Measurements of Ca 2ϩ binding to recombinant ATPase in the absence of ATP were performed in microsomes obtained from COS-1 cells infected with adenovirus vectors (as opposed to simple transfections), taking advantage of the higher concentration of recombinant ATPase in these samples. The measurements were performed exactly as described by Zhang et al. (12), in the presence of 3 M free Ca 2ϩ , as determined by an EGTA-Ca buffer. The measured Ca 2ϩ binding levels were adjusted to compensate for variations of recombinant ATPase expression in various preparations, with reference to a wild type preparation as indicated by Western blots. The difference between samples incubated in the absence and in the presence of thapsigargin was considered to be specific Ca 2ϩ binding.
Computations-Calculation of free Ca 2ϩ in various reaction mixtures were based on the concentrations of total calcium and EGTA as originally described by Fabiato and Fabiato (19).
Simulations of steady state kinetics, yielding overall ATPase velocity and levels of intermediate species, were based on user entered rate constants and concentrations of substrate, ligands, and products, as described by Inesi et al. (20). Computations were performed on a PC microcomputer, using a 14-digit precision MEGABASIC with BCD coding (American Planning Corp., Alexandria, VA).
Copies of the computational programs can be obtained from M. Kurzmack (La Porte, CO).

Expression of Mutants-L67
includes the Gly 808 -Gly 832 segment of the ATPase sequence, in which we produced several site directed mutations (Table I). As a comparison, we also mutated residues in the small M8/M9 loop (Ser 917 -Glu 920 ). Expression of all mutants in COS-1 cells was similar to, or somewhat (10 -30%) lower than, that of WT ATPase. An exception was the Pro 820 3 Ala mutant, whose protein recovery was negligible, even though its mRNA level was comparable to that produced by WT cDNA (Fig. 1). It is of interest that Menguy et al. (11) obtained expression of a totally inactive Pro 820 3 Ala mutant. This suggests that mutation of Pro 820 interferes profoundly with protein folding, resulting in degradation of product by the COS-1 cells used for expression in our experiments, and production of inactive protein by the yeast cells used by Menguy et al. (11). Functional Characterization of Aspartate Mutants-Falson et al. (10) and Menguy et al. (11) reported that cluster mutations of Asp 813 , Asp 815 , and Asp 818 to Ala reduce the affinity of the ATPase for Ca 2ϩ , an effect that was confirmed by direct measurements of Ca 2ϩ binding by equilibrium experimentation (12). We now find that even conservative mutation of the Asp 813 , Asp 815 , and Asp 818 cluster to Asn produces strong inhibition of ATPase steady state velocity, and maximal activity cannot be reached even by raising the Ca 2ϩ concentration to the 0.1 mM level ( Fig. 2A). The effect of the triple mutant can be to a large extent reproduced (data not shown) by single mutation of Asp 813 (but not Asp 815 or Asp 818 ) to Asn.
The triple aspartate mutation strongly reduces the steady state levels of phosphorylated intermediate formed with ATP in the presence of Ca 2ϩ ( Fig. 3 and Table I). However, it affects much less prominently enzyme phosphorylation by P i in the absence of Ca 2ϩ (Fig. 4). In agreement with previous findings (12), direct measurements with radioactive tracer show that the cluster mutation of Asp 813 , Asp 815 , and Asp 818 to Asn displays very low levels of high affinity Ca 2ϩ binding in the presence of 3 M Ca 2ϩ (Table II). Direct measurements of Ca 2ϩ binding cannot be performed at higher Ca 2ϩ concentrations due to unfavorable signal to noise ratio. Nevertheless, we found that phosphorylation of the cluster mutant with P i is only moderately inhibited by 10 -100 M Ca 2ϩ (Fig. 5), indicating that Ca 2ϩ binding occurs at lower levels than in the WT enzyme.
Time resolution of enzyme phosphorylation with ATP at low temperature show again significant inhibition, especially when the reaction is started by addition of ATP and Ca 2ϩ to enzyme deprived of Ca 2ϩ , as compared with addition of ATP to enzyme already activated by Ca 2ϩ (Fig. 6). On the other hand, pulsechase experiments indicate that decay of (already formed) phosphoenzyme is not significant delayed by the triple mutation (Fig. 7).
Single mutations of Asp 815 to Asn or Ala do not produce significant ATPase inhibition, although slight displacements of the Ca 2ϩ concentration dependence of ATPase activation (  (Table II). Mutation of Leu 814 to Ala does not produce significantly effects.
Functional Characterization of Lysine and Arginine Mutants-Single mutations of Lys 819 to Ala or Glu, and of Arg 822 to Phe, Glu, or Ala produce very strong inhibition of the steady state ATPase velocity even at relatively high Ca 2ϩ concentrations (Fig. 2 B ). WT enzyme and mutants reach their own maximal activity within the same ATP concentration range (data not shown), indicating that the steady state ATPase depend- ATPase velocities of WT enzyme and mutants were obtained by serial determination of P i production, as described under "Materials and Methods." The ATP concentration was 3 mM, and the free Ca 2ϩ concentration was obtained with EGTA-Ca buffers. The ATPase velocities were obtained from linear plots of several time points, and then corrected to reflect the concentrations of mutant and WT enzyme in different microsomal preparations. Average values obtained from three different measurements are listed in Table I. ence on the ATP concentration is not significantly changed by the Lys and Arg mutations. Direct measurements of Ca 2ϩ binding in the absence of ATP show slight and moderate reduction in the Lys 819 and Arg 822 mutants, respectively (Table  II). This finding is in agreement with the interference with the inhibitory effect of Ca 2ϩ on the P i reaction (Fig. 5).
A marked effect of the Lys 819 and Arg 822 mutations is strong reduction of the steady state levels of phosphorylated intermediate formed by utilization of ATP. This reduction is only slightly compensated for by raising the Ca 2ϩ concentration in the medium (Fig. 3). Low levels of phosphoenzyme are already apparent in the initial phase of its formation upon simultaneous addition of ATP and Ca 2ϩ to enzyme pre-incubated with EGTA, or even when ATP alone is added to enzyme preincubated with Ca 2ϩ . In the former case the phosphoenzyme level rises slowly due to slow activation by Ca 2ϩ (Fig. 6A), while in the latter case the steady state level of phosphoenzyme is reached faster since the enzyme is already activated by Ca 2ϩ (Fig. 6B). It is clear that, in either case, net formation of phosphoenzyme is much lower in the mutants than in the WT samples.
We also conducted pulse-chase experiments to test whether hydrolytic cleavage of the phosphorylated intermediate (once formed) is influenced by the Lys 819 and Arg 822 mutations (Fig.  7). We found that the time course of phosphoenzyme cleavage is not significantly affected by the Lys and Arg mutations (Fig. 7).
It should be pointed out that the experiments on phosphoenzyme formation and cleavage were conducted at low temperature and non-saturating ATP concentrations in order to obtain kinetic resolution. The relevance of these findings to the steady state behavior of the enzyme at 25°C is considered under "Discussion." It is of interest that the equilibrium levels of phosphoenzyme formed with P i in the absence of Ca 2ϩ and ATP, are also very much reduced (Fig. 4). These experiments were conducted at 25°C and, consistent with the ATP experiments conducted at low temperature, indicate that Lys 819 and Arg 822 play an important role in determining the functional integrity of the phosphorylation domain. It is then noteworthy that the phosphorylation defect resulting from the Lys 819 and Arg 822 mutations can be observed in the presence and in the absence of Ca 2ϩ , and at low and high temperature.
Functional Characterization of Proline Mutants-We found no inhibition of ATPase activity in the Pro 811 3 Ala and Pro 812 3 Ala mutants, and moderate inhibition in the Pro 821 3 Ala, Pro 824 3 Ala, and Pro 827 3 Ala mutants (Fig. 2). As noted above, however, the Pro 820 mutation resulted in negligible protein recovery, despite normal mRNA levels. All tested proline mutants exhibited a Ca 2ϩ concentration dependence nearly identical to that of the WT ATPase, independent of whether the ATPase velocity was reduced or not (Fig. 2). The proline mutants yielded phosphoenzyme levels similar to those obtained with WT ATPase, by utilization of either ATP in the presence of Ca 2ϩ ( Table I. FIG. 4. Equilibrium levels of phosphoenzyme obtained by incubation with P i in the absence of Ca 2؉ . Phosphoenzyme was obtained by incubating COS-1 cell microsomes with [ 32 P]P i at 25°C as described under "Materials and Methods." The concentration of microsomal protein was adjusted to yield the same concentration of WT or mutant ATPase, as indicated by Western blot analysis. The time of radiographic exposure was the same for all samples, in order to reveal differences in the equilibrium levels of phosphoenzyme, if any. The greater width of some bands (e.g. P811A) is due to spread of larger amounts of total microsomal protein in the sample, as required to yield the same amount of recombinant ATPase. Average values obtained from three different experiments are listed in Table I. 3) or P i in the absence of Ca 2ϩ (Fig. 4). In agreement with previous reports (11), we found that the proline mutations did not interfere significantly with Ca 2ϩ inhibition of enzyme phosphorylation with P i (Fig. 5).
Functional Characterization of Mutants in the L89 -A set of single mutations were produced in the loop intervening between the M8 and M9 transmembrane segments (L89), to obtain a comparative evaluation with the effects obtained FIG. 5. Ca 2؉ inhibition of enzyme phosphorylation with P i . Equilibrium levels of phosphoenzyme were obtained by incubation of WT or mutant enzyme with [ 32 P]P i at 25°C, as described under "Materials and Methods," in the absence of Ca 2ϩ (EGTA present) or in the presence of Ca 2ϩ as indicated. The acid-quenched samples were washed and subjected to electrophoresis, and the radioactive phosphoenzyme was detected by autoradiography. As the phosphoenzyme levels were different in WT and mutant proteins (see Fig. 7), autoradiographic exposure was adjusted to yield signals of similar intensity at 0 Ca 2ϩ , in order to render possible a comparison of the effects of Ca 2ϩ . Average values (relative to EP levels in the absence of Ca 2ϩ ) obtained from three different experiments are plotted in the right panels. Standard deviations ranged from 1 to 13. with mutations in L67. We found that the Ser 917 3 Ala and Gln 920 3 Ala mutants sustained ATPase activity at rates equal to the WT enzyme, while the Glu 918 3 Ala and Asn 919 3 Ala exhibited slightly lower rates. The Ca 2ϩ concentration dependence of the M8/M9 loop mutants was nearly identical to that of the WT ATPase (Fig. 2).

DISCUSSION
Functional Analysis-Our experiments demonstrate that single or cluster mutations within L67 interfere profoundly with the ATPase function. It is remarkable that L67 can affect the molecular events occurring at both phosphorylation site (Asp 351 ) and Ca 2ϩ binding sites, which are separated by a more than 50-Å distance. L67 plays a crucial role even in protein folding. In fact, mutation of Pro 820 (unique among six L67 prolines) interferes with the appearance of significant levels of expressed protein. By comparison, mutations within L89 produce little or no interference. We want to clarify here how the twenty five residues of the L67 loop can interfere with such various aspects of the enzyme structure and function.
Particularly interesting are single mutations of Arg 822 and Lys 819 . Mutations of these residues inhibit strongly ATPase activity, due to interference with formation of phosphorylated enzyme intermediate both by utilization of ATP in the presence of Ca 2ϩ , and of P i in the absence of Ca 2ϩ . While these mutations affect events that occur at the phosphorylation site, they interfere only to a lower degree with Ca 2ϩ binding. Yet another type of functional defect is produced by mutations of L67 aspartate residues, as originally pointed out by Falson et al. (10) and Menguy et al. (11) who mutated these residues to Ala. We find that removal of negative charge by conservative mutations of aspartate residues interferes with high affinity Ca 2ϩ binding and Ca 2ϩ dependent enzyme phosphorylation by ATP. Inhibition of steady state velocity is observed even at saturating Ca 2ϩ . No significant inhibition of enzyme phosphorylation with P i in the absence of Ca 2ϩ is produced by the aspartate mutations.
Considering the heterogeneity of effects produced by mutations within the L67 loop, we sought to verify by kinetic analysis whether the observed functional behavior of various mutants could be accounted quantitatively for by perturbations of the phosphorylation reaction, or rather by interference with Ca 2ϩ binding and enzyme activation. For this purpose we utilized a complete reaction sequence (Fig. 8) to simulate the steady state ATPase behavior. The rate constants listed in Fig.  8, obtained at 25°C as explained previously (20), generate a simulated pattern of ATPase Ca 2ϩ activation that is identical to that obtained experimentally with the WT enzyme (Figs. 2  and 9). Any of the rate constants can then be changed to explore the effects of specific perturbations of partial reactions on the overall steady state behavior of the ATPase. A change of the forward rate constant for any particular reaction must be accompanied by a proportional change of the reverse rate constant of the same or of another appropriate reaction so that the overall equilibrium constant (corresponding of that ATP hydrolysis to ADP and P i ) remains unchanged.
To simulate the triple mutant (Asp cluster) behavior we considered that the mechanism of Ca 2ϩ binding and ATPase activation includes three steps: fast binding of the first Ca 2ϩ , followed by a relatively slow isomeric transition, and cooperative binding of the second Ca 2ϩ (21). Enzyme activation occurs only after binding of the second Ca 2ϩ , when involvement of FIG. 8. ATPase reaction scheme and bidirectional rate constants for the partial reactions. The prime (Ј) refers to the occurrence of the Ca 2ϩ -induced conformational transition involved in cooperative binding and enzyme activation. The asterisk (*) refers to enzyme acquisition of the low affinity state for Ca 2ϩ . The bidirectional rate constants (25°C) for the partial reactions were obtained by kinetic and equilibrium experimentation and related analysis as described previously (20). transmembrane helices M4, M5, M6 and M8 in Ca 2ϩ complexation is complete (12). We then found in our simulation that a simple inhibition of the enzyme affinity for Ca 2ϩ through reduction of fast steps (reactions 1 forward and 8 reverse in Fig.  8) yields a displacement of the Ca 2ϩ curve, with full activation at higher Ca 2ϩ (Fig. 9A, squares). This type of curve was not observed experimentally ( Fig. 2A). Combined inhibition of the slow Ca 2ϩ induced transition and of enzyme phosphorylation with ATP (reactions 2 and 5 in Fig. 8) is required to generate a pattern (Fig. 9A, diamonds) matching precisely the pattern observed experimentally with the cluster Asp 813 , Asp 815 , and Asp 818 to Asn mutant ( Fig. 2A, diamonds). Thus the simulation suggests that the aspartate negative charges are not involved in direct coordination of Ca 2ϩ , but play an important role in maintaining the structural integrity of the L67 loop, thereby permitting Ca 2ϩ binding and transmission of the activation signal to the phosphorylation domain.
Simulating the behavior of the Lys and Arg mutant is more straightforward, and is obtained just by lowering the rate constants of ATP phosphorylation by ATP and P i (reactions 5 FIG. 9. Simulated steady state velocities of WT and mutant ATPase at various Ca 2؉ concentrations. A and B, q, rate constants as shown in Fig. 8. A, f, as in Fig. 8 but k 1f ϭ 4 ϫ 10 6 M Ϫ1 s Ϫ1 , and k 8r ϭ 5 ϫ 10 4 M Ϫ1 s Ϫ1 ; ࡗ, as in Fig. 8 but k 2f ϭ 24 s Ϫ1 , k 2r ϭ 5 s Ϫ1 , k 5f ϭ 6 s Ϫ1 , and k 5r ϭ 10.5 s Ϫ1 . Note that f does not, but ࡗ does, correspond to the experimental data obtained with the aspartate cluster mutant (Fig. 2A). The analysis is consistent with mutational perturbation of a rate-limiting step in the Ca 2ϩ binding and activation mechanism. B, q, rate constants as in Fig. 8; f, as in Fig. 8 but k 5f ϭ 10 s Ϫ1 and k 5r ϭ 3 s Ϫ1 ; ࡗ, k 5f ϭ 5 s Ϫ1 and k 5r ϭ 1.5 s Ϫ1 . Note the correspondence to the experimental patterns obtained with the Lys and Arg mutants (Fig. 2C). The analysis is consistent with mutational perturbation of the phosphorylation domain, affecting enzyme phosphorylation with ATP in the presence of Ca 2ϩ , and with P i in the absence of Ca 2ϩ . Note that any manipulation of rate constants was balanced in the forward and reverse directions so as to maintain an identical overall equilibrium constant for the entire reaction cycle. forward and 11 reverse in Fig. 8). The simulations reproduce satisfactorily the curves observed experimentally, including the low steady state velocity and an apparent saturation by Ca 2ϩ at lower concentrations than observed with the WT enzyme (Compare Figs. 2B and 9B). This indicates that the observed reduction of overall steady state ATPase velocity at 25°C can be attributed to mutational perturbation of the phosphorylation domain and consequent inhibition of phosphoenzyme formation (Fig. 6).
Structural Considerations-L67 is an extended loop, consisting of ϳ25 residues. Its pathway seems to be determined by a hydrogen bond network with other protein segments; that is, with the L89 loop near its N-terminal end, with helices in the P domain and M5 in the middle part, and with M3 and M5 near its C-terminal end. Two pairs of Pro may also play an important part in determining the L67 path, by restricting the main chain torsion angles (Fig. 10).
The effect of L67 aspartate mutations on Ca 2ϩ binding are likely to be indirect, since their distance from the transmembrane domain (8) precludes their direct participation in the binding of the two activating calcium ions. As illustrated in Fig.  10, Asp 813 , Asp 815 , and Asp 818 are involved in a hydrogen bonding network with Arg 751 and Asn 755 (cytosolic extension of M5), Ser 917 and Gln 920 (L89), and Ala 617 (P2). Our experiments suggest that involvement of Asp 813 in these bonds is most important. Since aspartate mutations impair Ca 2ϩ binding, the hydrogen bonding network is likely to maintain M6 (with its three Ca 2ϩ binding residues) and neighboring segments in optimal position for Ca 2ϩ binding and Ca 2ϩ dependent enzyme activation.
It is of interest that the L67 apex sustains a crucial role in protein folding, as the Pro 820 3 Ala mutant fails to yield significant levels of expressed protein even though transcription occurs normally. Mutation of the neighboring Arg 751 (in the cytosolic extension of the long M5 transmembrane helix) also interferes with expression and functional competence of recombinant ATPase (22). It is shown in Fig. 10 that Pro 820 is very close to Arg 751 , and hydrogen-bonded at the carbonyl oxygen atom of the preceding residue (Lys 819 ). Thus this hydrogen bond appears critical for proper folding of the protein and Pro 820 is likely to be important for the correct positioning of this carbonyl group (presumably by restricting the main chain torsion angle).
Mutations of Lys 819 and Arg 822 , also within the L67 apex, produce strong functional inhibition and specific interference with formation of the phosphorylated enzyme intermediate. Lys 819 and Arg 822 are in close proximity and hydrogen-bonded to residues of the P2 (Asp 616 ) and P1 (Glu 340 ) helices of the phosphorylation domain (Fig. 10). Mutation of Ser 346 , which is located in the loop connecting ␤1 and P1, and is hydrogenbonded to Glu 696 (Fig. 10), also produces catalytic interference (23). It is likely that, in the process of enzyme activation, M4 and M5 undergo large conformational changes (8), and M6 is repositioned by engagement of three of its residues (Asn 796 , Thr 799 , and Asp 800 ) in Ca 2ϩ complexation (12). Thereby, the L67 loop and the P1 and P2 helices are affected as well. This would in turn affect the neighboring ␤-strand (strand 1 in Fig.  10) that includes the residue undergoing phosphorylation (Asp 351 ). This strand is connected to transmembrane segment M4, which is also involved in Ca 2ϩ complexation through the intervention of Glu 309 . The entire region appears to be stabilized by the M5 helix that extends from the lumenal surface of the membrane up to the end of the P domain (8). Thus, precise conservation of structural interactions in this region is required for competence of the phosphorylation and the Ca 2ϩ binding domains, and their long range functional linkage.