Functional Consequences of Proline Mutations in the Cytoplasmic and Transmembrane Sectors of the Ca2'-ATPase of Sarcoplasmic Reticulum*

Site-specific mutagenesis was used to investigate whether Pro'", Pro1", Pro3'', Pro312, Proso3, and Pro8" play essential roles in the function of the sarco- plasmic reticulum Ca2+-ATPase. All six prolines were substituted with alanine; and in addition, Pro3'' was replaced by glycine and by glycine as well as by leucine. Mutant cDNAs were expressed in COS-1 cells, and mutant Ca2+-ATPases located in the isolated microsomal fraction were examined with respect to ca'+ uptake activity, Ca2+ dependence of phosphorylation from ATP, and the kinetic properties of the phosphoenzyme intermediates formed from both ATP and Pi. The enzymatic cycle was little affected by substitution of ProlB0, and Pro8", which are located in the cytoplasmic domain; but replacement of Pro3OS, Pro312, and Proso3, in the putative transmembrane helices, had a profound impact on the function of the enzyme. All mutations of Pro3'' and ProSo3 led to ATPases which were characterized by a reduced affinity for Ca". These prolines may therefore be involved in the structure of the high affinity Ca2+-binding sites in the en- zyme. Substitution of Pro312 with alanine or glycine gave rise to mutants unable to transport ported per enzymatic cycle (Inesi, 198.5) and a site concentration as determined hv the steady-state level ofphosphorylation from ATP. Analxsis of Phosphoenzyme Intermediotcs-Phosphorylation from either ATP or inorganic phosphate to ohtain BIP or E2P, respectively, was performed as descrihed (Clarke et al., 1989h). Studies of the intermediate reaction steps were carried out as descrihed in the accompanying paper (Andersen et al., 1989).


Functional Consequences of Proline Mutations in the Cytoplasmic and Transmembrane Sectors of the Ca2'-ATPase of Sarcoplasmic
Site-specific mutagenesis was used to investigate whether Pro'", Pro1", Pro3'', Pro312, Proso3, and Pro8" play essential roles in the function of the sarcoplasmic reticulum Ca2+-ATPase. All six prolines were substituted with alanine; and in addition, Pro3'' was replaced by glycine and by glycine as well as by leucine. Mutant cDNAs were expressed in COS-1 cells, and mutant Ca2+-ATPases located in the isolated microsomal fraction were examined with respect to ca'+ uptake activity, Ca2+ dependence of phosphorylation from ATP, and the kinetic properties of the phosphoenzyme intermediates formed from both ATP and Pi. The enzymatic cycle was little affected by substitution of ProlB0, and Pro8", which are located in the cytoplasmic domain; but replacement of Pro3OS, Pro312, and Proso3, in the putative transmembrane helices, had a profound impact on the function of the enzyme. All mutations of Pro3'' and ProSo3 led to ATPases which were characterized by a reduced affinity for Ca". These prolines may therefore be involved in the structure of the high affinity Ca2+-binding sites in the enzyme. Substitution of Pro312 with alanine or glycine gave rise to mutants unable to transport Ca2+ even though their apparent affinities for Ca2+ in the phosphorylation reaction with ATP were increased.
In these enzymes, the ADP-sensitive phosphoenzyme intermediate was stable for at least 5 min at 0 "C, whereas the ADP-insensitive phosphoenzyme intermediate decayed at a rate similar to that of the wild type. Thus, the inability to transport Ca2+ could be accounted for by a block of ADP-sensitive to ADPinsensitive phosphoenzyme intermediate conformational transition. In contrast, substitution of Pro312 with leucine gave rise to a mutant enzyme that retained about 7% of the normal Ca2+ transport rate. Phos-phoenzyme turnover in this mutant also occurred at a low but significant rate, suggesting that the leucine side chain can substitute to some extent for proline.
The Ca2+-ATPase of the sarcoplasmic reticulum catalyzes the uphill transport of Ca2+ from the sarcoplasm to the membrane lumen at the expense of free energy derived from the hydrolysis of ATP (Hasselbach, 1964;Martonosi and Beeler, 1983). A structural model for the Ca2+-ATPase has been proposed on the basis of the amino acid sequence deduced from the cDNA sequence (MacLennan et al., 1985;Brand1 et al., 1986). In this model, two globular cytoplasmic domains are separated from a transmembrane channel region consisting of 10 a-helices by a stalk sector made up of five ahelices. The formation of an ADP-sensitive phosphoenzyme intermediate (EIP)' by reaction with ATP is central to the mechanism of energy transduction in the enzyme. This form is then converted to an ADP-insensitive phophoenzyme intermediate ( E 2 P ) during the course of Ca2+ transport (de Meis and Vianna, 1979). These reactions are linked with conformational changes in the enzyme which apparently bring about a reorientation of the two Ca2+-binding sites, leading to a loss in their affinity for Ca2+. This eventually leads to discharge of two Ca2+ ions into the luminal space. In EIP, the bound Ca2+ ions are in an occluded state characterized by a low rate of isotopic exchange with free Ca2+ on either side of the membrane (Dupont, 1980;Vilsen and Andersen, 1987). It is unclear whether the occluded Ca2+ ions are bound at the primary Ca2+-binding sites in the enzyme or whether one or both of the ions have been transferred to low affinity sites in a release channel leading to the luminal surface (Green et al., 1986;Petithory and Jencks, 1988).
One way to improve our understanding of the molecular details of the Ca2+ transport mechanism is to identify the residues involved in the specific partial reactions of the pump cycle through site-directed mutagenesis. This approach has recently proven valuable in the assignment of residues making up the ATP-binding site (Maruyama and MacLennan, 1988;Maruyama et al., 1989) and the high affinity Ca2+-binding sites (Clarke et al., 1989a(Clarke et al., , 1989b. Thus, substitution of more than 30 negatively charged and polar residues in the protein pinpointed 6 residues in the putative transmembrane helices as possible candidates for Ca2+ ligands. Of great interest for our understanding of the mechanism  of Ca2+ transport are proline residues because of their unique structural and functional properties. The structural destabilization induced by prolines located in the middle of a-helices and the possibility of cis-trans isomerization of peptide bonds between prolines and their preceding residues make prolines important candidates for specific functions related to conformational changes in the protein (Schulz and Schirmer, 1979;Rarlow andThornton, 1988 Brandl andDeber, 1986). Furthermore, increased basicity of the carbonyl oxygen atom in peptide bonds involving proline may be of importance in the creation of cation-binding sites. Indeed, it has been found that the putative transmembrane segments of several membrane transport proteins contain a significantly higher number of proline residues than do the transmembrane portions of nontransport membrane proteins, suggesting some functional role of membrane-buried prolines in the transport reaction (Brandl and Deber, 1986).
In this study, we have mutated 6 proline residues in the Ca"-ATPase. Three are located in putative transmembrane helices, 2 are in the predicted @-strand sector in the globular cytoplasmic projection, and 1 is in a relatively short loop connecting putative transmembrane helices M 6 and M 7 (Fig.  1). All of the prolines are substituted initially with alanine. In addition, Pro:'"* was changed to glycine and Proal2 to glycine as well as to leucine to test for the importance of a-helix destabilization. Information about Ca'+ binding in the mutants was obtained by studying the Ca' ' dependence of Ca2+ transport and of phosphorylation of the Ca"-ATPase from ATP. The EIP to E2P interconversion in the enzyme was analyzed by measurement of ADP sensitivity of the phosphorylated intermediate and of the rate of dephosphorylation, starting from either EIP or E'P.
We have found that all 3 prolines located in the transmembrane sector are important for function. Pro'"R and seem to be involved in Cat+ binding, whereas Pro"" is essential to the transition between EIP and E,P. On the other hand, the enzyme cycle seems to be little affected by the mutations of prolines in nontransmembrane portions of the protein.

MATERIALS AND METHODS
Site-specific Mutagenesis-Oligonucleotide-directed site-specific mutagenesis was carried out according to Kunkel (1985) as described previously (Maruyama and MacLennan, 1988). Mutations were introduced into short restriction fragments that had been excised from the full-length rabbit fast-twitch muscle Ca"-ATPase cDNA clone and inserted into the Bluescript vector (Stratagene, La Jolla, CA).
The fragments used were the following: SmaI (position 442)-KpnI (position 663) for mutation of Pro"' and Pro'"", RarnHI (position R65)-Smnl (position 1600) for Pro"' and Pro"?, and AuaI (position 2:153)-RstElI (position 2716) for Pro":' and Pro"". Mutant clones were selected hy hybridization screening with radiolabeled mut.ant oligonucleotides according to standard procedures (Maniatis et al. 1982). In order to verify that mutations were appropriate and to ensure that no unwanted mutations or deletions had occurred, se-quencina was perfnrmed using the dideoxynucleotide chain termination method (Sanger et a/., 1977) with either Sequenase (IBI Inc.) or T7-DNA polymerase I (Pharmacia LKR Biotechnology Inc.). T o solve compressions sometimes occurring in (G + C)-rich regions, the dITP or 7-deaza-dGTP analogues of dGTP were used (Barnes et al. 1983: Mizusawa et al., 1986. For two mutations, it was possible to design the mutant oligonucleotides so that a new restriction site (PouI site) was generated, allowing additional confirmation of these mutations (Fig. 2).
The mutated fragments were excised from the Bluescript vector and religated hack into their original positions in the full-length clone contained in the pBSF4 vector (Maruyama and MacLennan, 1988). It is important that the vector fragment used in the religation he uncontaminated by a single digested vector which would readily circularize to reform the wild t-we. Therefore, the vector fragment was purified hv agarose gel electrophoresis followed by extraction with Geneclean (Hio/Can Scientific Inc.) before use in the religation reaction. For expression in COS-1 cells (Gluzman, 1981), the entire CaY+-ATPase cDNA containing the mutation was cloned into the EcoRI site of vector p91023(B) (Wnng et al., 1985).
Expression of Mutant DNA-Transfection of COS-1 cells by the DEAE-dextran chloroquine shock method (Sompayrac and Danna, 1981;Gorman, 1985) and isolation of the microsomal fraction containing the expressed Ca"-ATPase were carried out as described previously (Maruyama and MacLennan 1988;Clarke et al., 1989b). Expression levels were examined and quantitated by immunoblotting following SDS gel electrophoresis and by sandwich enzyme-linked immunosorbent assay (Clarke et al., 1989b) using the highly specific monoclonal antibody A52 (Zubrzycka-Gaarn et al., 1984). In these assays, the standard enzyme used was the deoxycholate-purified fasttwitch skeletal muscle enzyme (MacLennan, 1970).
Ca" Transport Activity-For measurement of ATP-driven uptake of Ca2+ in the isolated microsomes, the reaction mixture contained 0.05-0.2 pg of Cay'-ATPase protein/ml, 20 mM MOPS, pH 6.8, 80 mM KCI, 5 mM MgCI?, 5 mM ATP, 0.5 mM EGTA, 5 mM potassium oxalate, 2 pCi/ml "Ca'+, and various concentrations of 4"Ca" to, obtain the desired concentrations of free Ca", calculated according to published stability constants for CaEGTA, MgEGTA, CaATP, and MgATP (Vianna, 1975;Dupont, 1982). Incubation was performed at 27 "C for 5, 10, and 25 min and terminated by transferring 150-p1 aliquots to 3 ml of quench solution containing 0.15 M KC1 and 1 mM LaCL The quenced samples were filtered through 0.3-pm PHWP or 0.45-pm HAWP Millipore filters. The filters retaining the calciumloaded microsomes were washed with 15 ml of quench solution, and the radioactivity on the filters was measured by liquid scintillation counting. The Cay+ uptake referable to the expressed fast-twitch Cay'-ATPase or mutants was calculated after subtraction of the background (usually less than 5%) corresponding to microsomes derived from nontransfected COS-1 cells. The linear part of the Ca'+ uptake curve (the first 5 min) was used in this calculation. The specific rate of Cay+ uptake was expressed as the molecular turnover number calculated assuming a stoichiometry of two Ca" ions trans- ported per enzymatic cycle (Inesi,198.5) and a site concentration as determined hv the steady-state level ofphosphorylation from ATP.
Analxsis of Phosphoenzyme Intermediotcs-Phosphorylation from either ATP or inorganic phosphate to ohtain BIP or E2P, respectively, was performed as descrihed (Clarke et al., 1989h). Studies of the intermediate reaction steps were carried out as descrihed in the accompanying paper (Andersen et al., 1989). Fig. 3 shows examples of immunoblots of the wild-type fast-twitch Ca"-ATPase and of the proline mutants expressed in COS-1 cell microsomes. As shown, the mutants are expressed to approximately the same level as that of the wild type. As described previously, the antibody used reacts specifically with epitopes on the fasttwitch Ca'+-ATPase (Clarke et al., 1989b). There is no crossreactivity with the endogenous slow-twitch endoplasmic reticulum Ca"-ATPase of the COS-1 (monkey kidney) cells. For each of the mutants, the average expression level in four to six microsomal preparations was estimated to be between 80 and 100% of the average wild-type level, suggesting that no major structural perturbation was introduced by substitution of the proline residues.

Expression of Mutants-
Ca'+ Uptake- Table I presents the molecular turnover rates calculated from measurements of the rates of Ca'+ transport for the wild-type and mutant Ca-ATPases at pCa 5.0. To extend the time period in which the Ca" uptake showed a linear time dependence beyond seconds, oxalate was included Samples of the various microsomal preparations (12.5 pg of total microsomal protein) were separated on a 10% SDS-polyacrylamide gel followed hy transfer to nitrocellulose. The highly specific monoclonal antihody A52 was used to detect the Ca"-ATPase protein. An alkaline phosphatase-conjugated goat anti-mouse secondary antihody (Promega Riotec) was hound on top of A52, and the blot was visualized using 5-hromo-4-chloro-3-indolyl phosphate and p-nitro blue tetrazolium chloride.  " Determined from the calcium dependence of phosphorylation from ATP. *Calculated on the hasis of the initial Ca" uptake rate (nanomoles of Cay+ transported/s/mg total microsomal protein) measured a t pCa 5.0 and the active site concentration ohtained by measurement of phosphorylation (nanomoles of phosphoprotein/mg of total microsomal protein). Since two Ca" ions are transported per molecule of ATP hydrolyzed, the Ca'+ uptake rate was divided by 2 to ohtain the turnover rate. in the reaction mixture to precipitate transported calcium ions inside the microsomes (Maruyama and MacLennan, 1988). For the wild-type enzyme in the COS-1 cell microsomes, the observed turnover rate of 12.5 s-' agrees well with the value measured under the same conditions for the Ca'+-ATPase in sarcoplasmic reticulum vesicles. The Pro''o -+ Ala, Pro'!''' + Ala, Pro'''* + Ala, Prox0:' + Ala, and ProR" + Ala mutants all transported Ca'+ a t maximum rates close to or identical to that of the wild type. In contrast to the ProInR -+ Ala mutant, a decrease of the transport activity to 12% of the wild type a t pCa 5.0 was observed with the Pro'"R + Gly mutant. The Pro:"' + Ala and Pro:"' + Gly mutants were completely unable to transport Car+, and very little transport activity was detected with the Pro"' + Leu mutant (7% of the wild type).
For those mutants which were able to transport Ca'+ at measurable rates, it was possible to examine the Ca'+ affinity by titration of the Ca2+ dependence of Ca'+ transport (Fig. 4). As seen in Fig. 4, the Pro:'('" + Ala, Pro:'(1R + Gly, and ProHn3 + Ala mutants all displayed a lower Ca'+ affinity than that of the wild type, with the pCa a t which half-saturation occurs being shifted about 0.5 unit. A less significant deviation from the wild-type Ca'+ affinity was seen for the Pro"' -+ Ala mutant. For the Pro'95 + Ala mutant, Ca" dependence was identical to that of the wild type. This was also the case for the Pro'6o + Ala mutant (data not shown).
Phosphorylation of Mutants from ATP-The finding of a reduced Ca2+ affinity for the Pro:"'" and Pro"":' mutants and a complete loss of Ca'+ transport activity for the Pro:"' + Ala and Pro:"' + Gly mutants led us to examine the partial reactions of Ca'+ transport. The initial transfer of the terminal phosphate group of ATP to the enzyme, forming the phosphorylated intermediate, is a Ca'+-dependent reaction which requires the binding of two Ca'+ ions at high affinity sites in the wild-type enzyme (de Meis and Vianna, 1979;Petithory and Jencks, 1988). The proline mutants examined in this study all showed the same level of phosphorylation from ATP as the wild-type enzyme when the reaction was carried out at a free Ca2+ concentration of approximately 100 Ca" uptake activity (calculated per milligram of protein) is shown relative to that measured for the wild t.ge at saturating Ca" concentration.

Ca2+-ATPase Proline Mutants
PM. When the free Ca'+ concentrations were varied below this level, however, striking differences from the wild-type enzyme were observed for the Pro""', Pro"' , and ProfflI3 mutants.
Examples of this are seen in Fig. 5, which presents autoradiographs of phosphorylation data obtained with the Pro"' + Ala and Pro"' "-f Ala mutants. Whereas the wild-type enzyme was saturated with Ca2+ at pCa 6, the Pro"' + Ala mutant required a pCa close to 5.5 for saturation, confirming the reduced affinity observed by measurement of Ca2+ uptake. On the other hand, the Pro"' + Ala mutant showed a higher apparent affinity for Ca2+ relative to that of the wild type, with saturation occurring at pCa 6.2-6.5. Fig. 6 shows quantitation of similar data obtained when prO"l' and Pro"2 were substituted with glycine instead of B n-a- alanine. Again, an increase of apparent affinity was observed after substitution of Pro"".
We also replaced Pro"I2 with leucine, the residue which is present at the homologous position in the Na+, K+-ATPase (Shull et al., 1985). This substitution induced an increase in affinity similar to that observed with the other two substitutions of Pro3".
The Ca2+ titration data for all of our proline mutants are presented in Table I as the Caz+ concentration at which halfmaximum phosphorylation occurred (Koa). The relative affinities of mutants and the wild type measured in the phosphorylation experiments are consistent with those observed in the Ca'+ uptake measurements (Fig. 4). A comparison of the results presented in Table I with Fig. 4, however, shows that for both the wild type and mutants, the activation occurred a t higher Ca'+ concentrations in the phosphorylation experiments than in the Ca'+ uptake experiments. This can be accounted for by the higher concentration of free Mg2' present in the phosphorylation experiments (ATP concentration of only 2 PM) since Mg" competes with Ca'+ at the high affinity Ca2+-binding sites (Vilsen and Andersen, 1987).
Dephosphorylation Experiments-The inability of the fully phosphorylated Pro"' mutants to pump Ca2+ suggests that a partial reaction following phosphorylation is critically dependent on this proline residue. Therefore, we set up assay conditions with the purpose of testing the later steps in the reaction cycle. Fig. 7 shows the decay of phosphoenzyme in the wild type and mutants observed starting from either EIP or EzP (cf. Scheme 1). In order to form EIP (Fig. 7, left   panels), the enzyme was phosphorylated for 15 s in the presence of 2 PM ATP, 0.1 mM Ca2+, 80 mM K' a t 0 "c and pH 6.8. As seen in the first lanes of Fig. 7 and illustrated by the leftward reaction pathway in Scheme 1, all of the phosphoenzyme disappeared within 5 s when 1 mM ADP was added, indicating that no E2P was accumulated under these conditions. When EGTA was added to EIP in the absence of ADP, the Ca"-dependent phosphorylation from ATP ceased, and the dephosphorylation occurring through conversion to EzP and its subsequent hydrolysis could be observed (see Scheme 1, rightward reaction pathway). As seen in Fig. 7, EGTAinduced dephosphorylation was almost completed within 20 s for the wild-type enzyme. By contrast, dephosphorylation in the absence of ADP was very slow in the Pro:"2 mutants. Quantitation of the radioactivity associated with the protein in the gel showed that for the Pro"' -+ Ala mutant, more than 80% of the phosphoenzyme remained 5 min after the addition of EGTA. For the Pro"" + Leu mutant, less than 30% of the phosphoprotein was left after 30 s. Analogous experiments conducted with the Pro"" + Gly mutant demonstrated a very low dephosphorylation rate, similar to that observed with the Pro"" +Ala mutant (data not shown). The differences between mutants and the wild type cannot be ascribed to Ca'+ accumulated in the wild-type microsomes since all experiments were conducted in the presence of ionophore A23187.
In order to examine separately the dephosphorylation of the "low energy" phosphoenzyme intermediate (E'P), phosphorylation was performed with inorganic phosphate as substrate (Fig. 7, right panels). The Pro"' mutants (as well as the other proline mutants examined in this study) reacted with Pi to the same extent as the wild-type enzyme. When the phosphorylation by Pi was terminated a t 0 "C by dilution, the subsequent decay of the phosphoenzyme in the mutants occurred a t a rate similar to that of the wild-type enzyme. This rate was much higher than the rate of dephosphorylation observed after the EGTA quench of EIP. This shows that the EIP to EzP interconversion, rather than EzP hydrolysis, is Dephosphorylation was then initiated by the addition of 1 mM EGTA (lost thrw lanes), and acid quenching was performed at the times indicated after addition of EGTA. In the lirst lanes, 1 mM ADP was added with EGTA to demonstrate the ADP sensitivity, and acid quenching was performed 5 s later. Right panels, wild-type and Ca"-ATPase mutants were phosphorylated by 500 p~ ''2P, a t room temperature lor 1 0 min in an incubation mixture containing 50 mM MES, pH 6.2, 10 mM MgCI,, 2 mM EGTA, 20% (v/v) dimethyl sulfoxide. Following cooling of the sample to 0 "C, dephosphorylation was initiated hy 20-fold dilution of an aliquot into an ice-cold medium containing 60 mM MOPS, pH 6.8, 80 mM KCI, 5 mM MgCI,, 1 mM nonradioactive Pi, and 200 p~ CaCI,, and acid quenching was performed at serial time intervals as indicated.

ADP 2 ca"
ATP cL E, PCaZA 5 P + 9 SCHEME 1 rate-limiting for phosphoenzyme turnover and that the inhibition of dephosphorylation from EIP in the Pro:"' mutants is caused mainly by an effect on the EIP to E2P transition.
To substantiate this conclusion, phosphorylation experiments with ATP were performed at pH 8.35 in the presence of 10 mM Mi'+ and in the absence of alkali metal ions. This medium composition promotes the EIP to E2P interconversion and slows dephosphorylation of E2P (Shigekawa et al., 1983;Andersen et al., 1985), leading to steady-state accumulation of K,P in the wild-type enzyme (Fig. 8, left panel). No E,P accumulated with the Pro:"' + Gly mutant, however, as demonstrated by the full dephosphorylation obtained in the presence of ADP, even 5 min after addition of EGTA (Fig. 8,   a b c d e f g h FIG. 8. Effect of Pro"" 4 Gly mutation on E I P to E2P interconversion. Phosphorylation was performed at 0 "C in the presence of 2 pM [y-:"P]ATP, 10 mM MgCI,, and 50 p M Cay' at pH 8.35. The concentration of alkali metal ions in the reaction mixture was kept below 5 mM to promote accumulation of E,P. Lanes a and e , acid quenching was performed after a 20-s incubation; lanes b and f, 1 mM ADP with 1 mM EGTA was added after a 20-s incubation, and the samples were acid-quenched 6 s later; lane c, 1 mM EGTA was added after a 20-s incubation, and the samples were acidquenched 1 min later; lane d, 1 mM EGTA was added after a 20-s incubation followed by 1 mM ADP 1 min later and by acid quenching 5 s after addition of ADP; lane g, 1 mM EGTA was added after a 20s incubation, and the samples were then acid-quenched 5 min later; lane h, 1 mM EGTA was added after a 20-s incubation followed by 1 mM ADP 5 min later and by acid quenching 5 s after addition of ADP. The phosphoenzyme remaining after a 5-s reaction with ADP represents I;,P, whereas the phosphoenzyrne observed in the absence of ADP represents the sum of E,P and E,P.

right panel).
Similar results were obtained with the Pro:"2 + Ala mutant, but the Pro"' + Leu mutant accumulated E2P to a level intermediate between the wild type and the Pro:"' + Gly mutant (data not shown).

DISCUSSION
Of the proline mutants examined in this study, none of those located in the predicted cytoplasmic region of the Ca2+-ATPase behaved strikingly differently from the wild-type enzyme. Surprisingly, Pro""', which is highly conserved among the cation-transporting ATPases (Serrano, 1988;, could be replaced by alanine without any major consequences for the function of the enzyme. Pro'"5 has been suggested to be part of a Ca2+-binding torus (Gangola and Shamoo, 1986), but such a role seems to be excluded by the normal behavior of the Pro19s + Ala mutant as well as by the earlier finding that the acidic residues in the proposed torus could be mutated without loss of Ca2+ transport function (Clarke et al., 1989b).
By contrast, mutations of those prolines located in the transmembrane region of the molecule had profound effects on the function of the Ca"-ATPase. The transmembrane mutants fell into two functional groups. The Pro:"" and Proflo' mutants were characterized by a reduced affinity for Ca2+ as judged from the Ca2+ dependence of Ca2+ transport and of phosphorylation, whereas the Pro"I2 mutants were unable to transport Ca2+, but displayed a higher Ca2+ affinity than that of the wild-type enzyme in the phosphorylation assay. Prolines 308 and 803 are located in the middle of predicted transmembrane helices M4 and M6, respectively, close to carboxylic groups which, in our previous mutagenesis study (Clarke et al., 1989a), were assigned as Ca" ligands. On the basis of the reduced Ca2+ affinity that we observed after replacement of Pro:"'* and Pro"':', it seems likely that these prolines are also involved in formation of one or both high affinity Ca"-binding sites, either by donating the electrons associated with the carbonyl group in the preceding peptide bond (Brand1 and Deber, 1986) or by kinking the helix in such a way that the nearby carboxylic acid residues are brought into the optimal position for interaction with Ca". The standard free energies (AG") of Ca'+ binding, calculated from the measured values (Table I), are -8.0, -7.4, and -7.5 kcal/mol for the wild type and the Pro3'' "-* Ala and Prom3 + Ala mutants, respectively. Thus, the contribution of each of these proline residues to the total binding energy of the Ca'+ sites would lie in the range of 6-7%.
It is noteworthy that Pro3'' is highly conserved among cation-transporting ATPases (Serrano, 1988;Green et ~1 . ' ) . Proso3, on the other hand, is conserved only in the Na+,K+-ATPase. In the plasma membrane Ca2+-ATPases, the residue located in the homologous position is an alanine (Shull and Greeb, 1988;Verma et al., 1988), the residue used to replace proline in this study. Since both Pro3'' and Prom3 can be substituted with alanine without a major reduction of the expression level or of the maximum turnover rate, it seems unlikely that these prolines are involved in formation of intramembranous 6-turns or other loop structures in the transmembrane sector (Lodish, 1988). It also seems unlikely that cis-trans isomerization of the peptide bonds at Pro3" and Proso3 contributes to conformational changes associated with Ca2+ transport.
In contrast to the mutants discussed above, the Pro312 -Ala and Pro312 "-* Gly mutants were unable to transport Ca2+, even at Ca2+ concentrations permitting full phosphorylation from ATP. We have shown that these mutants are defective in the E 1P to E2P transition, which accounts for their inability to translocate Ca2+. Our evidence was based on kinetic experiments carried out with the ADP-sensitive and -insensitive phosphoenzyme intermediates formed by phosphorylation from ATP and Pi, respectively. As a consequence of the block of the EIP to E2P step in the reaction cycle, the ADP-sensitive phosphoenzyme intermediate accumulated in the Pro312 + Ala and Pro3" --* Gly mutants. This kinetic effect can explain the fact that the mutants display a higher apparent affinity for Ca2+ in the phosphorylation reaction than does the wild type.
It is remarkable that the Pro312 + Ala and + Gly mutants remained almost fully phosphorylated and ADPsensitive even 5 min after dephosphorylation was initiated by addition of EGTA in the presence of ionophore (Figs. 7 and 8). As the ADP sensitivity has been rationalized in terms of a mutual destabilization of bound Ca2+ and phosphate (Inesi, 1985;Pickart and Jencks, 1984), our findings raise the question whether the Caz+ ions might still be bound to the mutant enzyme 5 min after addition of EGTA. If so, the Ca2+ ions would be resting in a completely occluded state in the protein, with no access to either cytoplasmic or luminal spaces.
Replacement of Pro312 by leucine (the residue present at the homologous position in Na+,K+-ATPase) also produced an enzyme with a defective EIP to E2P transition. In this case, however, the inhibition was less severe than with alanine or glycine (7% transport activity left and dephosphorylation to less than 30% within 30 s). This seems to exclude a critical role of cis-trans isomerization of the peptide bond at Proal2 in the EIP to EzP conformational transition. As leucine is a good helix former and as the "helix breaker" glycine was found to be a poor substitute for Pro"', there is no firm basis for assigning the functionality of Pro3" to helix termination. It should, however, be kept in mind that proline and glycine may appear at the ends of a-helices for widely different reasons (Richardson and Richardson, 1988). It is also possible that the bulky side chain of leucine can introduce a defect in helix packing in the membrane, to some extent mimicking the effect of a kink dictated by the presence of proline (Barlow and Thornton, 1988).
Although the Pro3" residue is, by itself, conserved only in other Ca"'-ATPases and in the yeast H+-ATPase (Serrano, 1988), it is located in a highly conserved region which forms a physical link between the cation-binding domain and the phosphorylated aspartic acid residue This may be of importance from the point of view of a general mechanism for energy transduction in transport ATPases. We have proposed a model in which the EIP to EzP conformational change comprises a rotation or tilting of one or more of the transmembrane helices, disrupting the Ca2'-binding domain and thereby making the previously occluded Ca2+ ions accessible to a release channel Clarke et ul., 1989a).
On the basis of our observation that there is a crucial dependence of the EIP to EzP transition on the Pro312 residue, we now suggest that movement of helix Mq, in which this proline is located, forms the central element in the conformational change. By introducing a defect in helix packing in the transmembrane sequences, the residue may be decisive in signal transduction leading to channel opening.