Mutations of either or both Cys876 and Cys888 residues of sarcoplasmic reticulum Ca2+-ATPase result in a complete loss of Ca2+ transport activity without a loss of Ca2+-dependent ATPase activity. Role of the CYS876-CYS888 disulfide bond.

Disulfide-containing peptides in pepsin digest of sarcoplasmic reticulum vesicles were identified by using a fluorogenic thiol-specific reagent 4-fluoro-7-sulfamoylbenzofurazan and a reductant tributylphosphine. Sequencing of the purified peptides revealed the presence of a Cys(876)-Cys(888) disulfide bond on the luminal loop connecting the 7th and 8th transmembrane helices (loop 7-8) of the Ca(2+)-ATPase (SERCA1a). We substituted either or both of these cysteine residues with alanine and made three mutants (C876A, C888A, C876A/C888A), in which the disulfide bond is disrupted. The mutants and the wild type were expressed in COS-1 cells, and functional analysis was performed with the microsomes isolated from the cells. Electrophoresis performed under reducing and non-reducing conditions confirmed the presence of Cys(876)-Cys(888) disulfide bond in the expressed wild type. All the three mutants possessed high Ca(2+)-ATPase activity. In contrast, no Ca(2+) transport activity was detected with these mutants. These mutants formed almost the same amount of phosphoenzyme intermediate as the wild type from ATP and from P(i). Detailed kinetic analysis showed that the three mutants hydrolyze ATP in the mechanism well accepted for the Ca(2+)-ATPase; activation of the catalytic site upon high affinity Ca(2+) binding, formation of ADP-sensitive phosphoenzyme, subsequent rate-limiting transition to ADP-insensitive phosphoenzyme, and hydrolysis of the latter phosphoenzyme. It is likely that the pathway for delivery of Ca(2+) from the binding sites into the lumen of vesicles is disrupted by disruption of the Cys(876)-Cys(888) disulfide bond, and therefore that the loop 7-8 having the disulfide bond is important for formation of the proper structure of the Ca(2+) pathway.

bound protein (1, 2) that catalyzes Ca 2ϩ transport coupled to ATP hydrolysis (3,4). In the catalytic cycle, the enzyme is activated by binding of two Ca 2ϩ ions to the transport sites from the cytoplasmic side, and then ␥-phosphoryl group of ATP is transferred to Asp 351 (5)(6)(7) to form ADP-sensitive EP, which can react with ADP to form ATP (8 -10). Upon formation of this EP, the two Ca 2ϩ ions are occluded. A subsequent rate-limiting transition of ADP-sensitive EP to ADP-insensitive EP, which cannot react with ADP, results in release of the Ca 2ϩ ions into the lumen. Finally, ADP-insensitive EP is hydrolyzed to form P i and the dephosphoenzyme. This EP can also be formed from P i in the absence of Ca 2ϩ by reversal of its hydrolysis (11,12).
The Ca 2ϩ -ATPase contains ten transmembrane-helices (M1 to M10), and the bound two Ca 2ϩ ions are shown to be located side by side near the center of four helices, M4, M5, M6, and M8 in the crystal structure (13). The ATP binding site and phosphorylation site are located on the large cytoplasmic loop between M4 and M5 (13,14). Luminal loops are short except for the one connecting M7 and M8, the loop 7-8 (approximately Ala 853 -Glu 892 ), which protrudes into the luminal space in the crystal structure (13). Possible roles of this loop have not yet been well understood. This loop contains two cysteine residues, Cys 876 and Cys 888 . These residues have been predicted to participate in disulfide bonds (15), although the exact disulfide structure has yet to be identified. Because mutations of these residues were reported to cause partial loss of function (14), the possible disulfide bonds formed with Cys 876 and with Cys 888 likely to be important for structure and function of the enzyme.
Previous studies demonstrated (15)(16)(17) that totally 3 or 4 disulfide bonds are present in the Ca 2ϩ -ATPase. We showed (16) that all of the disulfide bonds in the enzyme of the SR vesicles were not readily reduced with dithiothreitol even at high concentrations but can be readily reduced if both Ca 2ϩ and a purine nucleotide are present. We further revealed that the reduction time courses of the disulfide bonds were not well separated from each other and thus that selective reduction of the disulfide bonds in the enzyme is impossible (16). The functional roles of the disulfide bonds can be explored, therefore, only by the site-directed mutagenesis of the cysteine residues identified as to form the disulfide bond of interest.
In the present study, we have explored first the exact disul-fide structure of Cys 876 and Cys 888 on the loop 7-8, and then to identify the possible role of the disulfide bond. Sequencing of the disulfide-containing peptides purified from the pepsin digest of SR vesicles showed that a disulfide bond is formed between Cys 876 and Cys 888 of the Ca 2ϩ -ATPase. We have then substituted either or both of these residues with alanine and made three SERCA1a mutants (C876A, C888A, and C876A/ C888A), in which the disulfide bond is disrupted. All the three mutants hydrolyzed ATP at high rates in the mechanism well accepted for the Ca 2ϩ -ATPase. In contrast, none of the mutants could transport Ca 2ϩ at a detectable rate. Results indicate that the loop 7-8 having the disulfide bond is important for formation of the proper structure of the Ca 2ϩ -release pathway.

EXPERIMENTAL PROCEDURES
Preparation of SR Vesicles-SR vesicles were prepared from rabbit skeletal muscle as described previously (18). The content of phosphorylation site determined with [␥-32 P]ATP according to Barrabin et al. (19) was 4.1 Ϯ 0.1 nmol/mg of SR protein (n ϭ 6). The number of disulfide bonds in the Ca 2ϩ -ATPase of SR vesicles estimated with ABD-F and TBP according to Kirley (20) was 2.4 Ϯ 0.4 disulfide bonds (n ϭ 4), and agreed closely with the number determined by Thorley-Lawson and Green (three disulfide bonds) (17).
Identification of Disulfide Bonds-SR vesicles (1 mg/ml) in 12 mM HCl (pH 2.0) were degassed and flushed with nitrogen, and then digested with pepsin (0.1 mg/ml) under nitrogen at 37°C for 1 h under acidic conditions that were previously shown to prevent possible disulfide interchanges (21) and oxidation of cysteine residues (22, 23). After centrifugation, the supernatant was subjected to reversed phase HPLC performed as previously described (24). All the elution buffers used were acidic (pH below 4) and degassed with an on-line degasser to prevent the possible oxidation and disulfide interchanges (21)(22)(23). Aliquots of each fraction were assayed for disulfides by simultaneous reduction and labeling of disulfides with a fluorogenic thiol-specific reagent ABD-F and a reductant TBP. ABD-F reacts with neither disulfides (25) nor TBP (26). Samples were labeled maximally with 0.2 mM ABD-F for 30 min at 60°C in the presence and absence of 0.5 mM TBP in 125 mM Na 2 B 4 O 7 (pH 8.0), 2 mM EDTA, and 2% SDS. Fluorescence intensity of bound ABD (excitation, 382 nm; emission, 504 nm) was measured in 50 mM ammonium acetate (pH 4.0) and 50% (v/v) ethanol. Sequencing of purified peptides was performed with an Applied Biosystems 477/120A sequencer.
Mutagenesis and Expression-Overlap extension PCR (27) was utilized for the substitution of Cys 888 with alanine in the rabbit SERCA1a cDNA. The ApaLI-SalI restriction fragments were excised from the PCR products and ligated back into the corresponding region in the full-length SERCA1a cDNA in the pMT2 expression vector (28). The alanine substitution of Cys 876 in the SERCA1a cDNA was generated by the oligodeoxyribonucleotide-directed dual amber method (29). The KpnI-SalI fragments from the SERCA1a cDNA were cloned into pKF19c vector (29) and hybridized with an oligonucleotide containing the desired mutation. The resulting mutant KpnI-SalI fragments were excised and ligated back into the SERCA1a cDNA in the pMT2 vector. DNA sequence was confirmed by the dideoxy method. The pMT2 DNA was transfected into COS-1 cells by the liposome-mediated transfection method. Microsomes were prepared from the cells in the presence of 3 mM 2-mercaptoethanol as described (30). The "control microsomes" were prepared from COS-1 cells transfected with the pMT2 vector containing no SERCA1a cDNA. The amount of the expressed SERCA1a was quantified by a sandwich enzyme-linked immunosorbent assay as described (31). The expression levels of the mutants C876A, C888A, and C876A/C888A in the microsomes were 11 Ϯ 1.7, 18 Ϯ 1.5, and 17 Ϯ 1.9% (n ϭ 4) of that of the wild type, respectively. The reduced expression of SERCA1a due to substitutions of Cys 876 and of Cys 888 was also previously observed in HEK-293 cells (32).
Tryptic Digestion and Immunoblotting-SR vesicles or microsomes from COS-1 cells (0.6 mg/ml) were digested with trypsin (Sigma) (0.9 g/ml) at 21°C for 10 min in 5 mM CaCl 2 , 0.5 M sucrose, and 50 mM MOPS/NaOH (pH 7.0). The samples were treated with a modified Laemmli sample buffer with or without 0.7 M 2-mercaptoethanol and separated by SDS-PAGE according to Laemmli (33). Proteins on the gels were blotted onto a polyvinylidene fluoride membrane and incubated with VE121G9 or IIH11 monoclonal antibody to the rabbit SERCA1a (Affinity Bioreagents). After incubation with secondary antibody (goat anti-mouse IgG-horseradish peroxidase-conjugated), the bound proteins were probed using an enhanced chemiluminescence-linked detection system (Amersham Pharmacia Biotech).
Ca 2ϩ Transport Activity-Ca 2ϩ transport activity was assayed as described previously (31) at 25°C in the presence and absence of 0.5 M thapsigargin in a mixture containing 10 g/ml microsomal protein, 1 mM ATP, 7 mM MgCl 2 , 0.1 M KCl, 20 mM MOPS/NaOH (pH 7.0), 5 mM potassium oxalate, 0.5 mM EGTA, and various concentrations of 45 CaCl 2 . The specific transport rate/mg of SERCA1a protein was calculated from the amount of the expressed SERCA1a and the Ca 2ϩ transport activity of the expressed SERCA1a, which was obtained by subtracting the thapsigargin-sensitive transport activity of the control microsomes from that of the microsomes expressing SERCA1a.
ATPase Activity-The rate of ATP hydrolysis was determined at 25°C in the presence and absence of 0.5 M thapsigargin in a mixture containing 10 g/ml microsomal protein, 1 M A23187, 0.1 mM [␥-32 P]ATP, 7 mM MgCl 2 , 0.1 M KCl, 50 mM MOPS/NaOH (pH 7.0), 0.5 mM EGTA, and various concentrations of CaCl 2 . The specific ATPase activity/mg of expressed SERCA1a protein was calculated from the amount of the expressed SERCA1a and the ATPase activity of the expressed SERCA1a, which was obtained by subtracting the thapsigargin-sensitive ATPase activity of the control microsomes from that of the microsomes expressing SERCA1a.
Formation and Hydrolysis of EP-Phosphorylation of SERCA1a in microsomes with [␥-32 P]ATP or 32 P i , and dephosphorylation of 32 Plabeled SERCA1a were performed under conditions described in the figure legends. The reactions were quenched with ice-cold trichloroacetic acid containing P i . The precipitated proteins were separated at pH 6.0 by 5% SDS-PAGE according to Weber and Osborn (34). The radioactivity associated with the separated Ca 2ϩ -ATPase was quantitated by digital autoradiography as described (35). The amount of EP formed with the expressed SERCA1a was obtained by subtracting the background radioactivity with the control microsomes. This background was less than 5% of the radioactivity of EP formed with the expressed wild type SERCA1a. The amount of EP/mg of SERCA1a protein was calculated from the amount of EP thus obtained and the amount of the expressed SERCA1a.
Miscellaneous-ABD-F was obtained from Dojindo (Kumamoto, Japan). TBP was from Sigma. Protein concentrations were determined by the method of Lowry et al. (36) with bovine serum albumin as a standard. The concentration of free Ca 2ϩ was calculated as described (37). Data were analyzed by nonlinear regression using the program Origin (Microcal Software, Inc., Northampton, MA).

Identification of Disulfide Bonds in SR Ca 2ϩ -ATPase-To
identify disulfide bonds present in the native Ca 2ϩ -ATPase, proteolysis of SR vesicles with pepsin and chromatographic isolation of peptides were performed under acidic conditions (pH below 4), which prevent possible disulfide interchanges or oxidation of cysteine residues (21)(22)(23), and in solutions degassed or flushed with nitrogen prior to and during use (see "Experimental Procedures"). When SR vesicles were digested extensively with pepsin, 78% of disulfides in the whole digest was recovered in the supernatant. The supernatant was subjected to the first reversed phase HPLC, and aliquots of each fraction were assayed for disulfides ( Fig. 1, A-C). Peptides containing disulfides were present in a major sharp peak, fraction I, and two minor broad peaks, fraction II and fraction III (Fig. 1C).
In the second reversed phase HPLC, fraction I gave a single disulfide-containing peak (peak Ia, Fig. 1, D-F). Peptides in this peak were further purified in the third reversed phase HPLC, and then labeled with ABD-F in the presence of TBP. In the fourth reversed phase HPLC, the labeled sample gave one major (peptide IaЈ) and one minor (peptide IaЉ) fluorescent peaks of bound ABD, which gave the sequences Met 874 -Glu 889 and Met 874 -Cys 888 , respectively, of SERCA1a (Table I). These peptides contained only two cysteine residues, Cys 876 and Cys 888 , and both of these residues had been labeled with ABD-F only in the presence of TBP to generate ABD-cysteine (Fig. 1E). The results show that a disulfide bond(s) is formed with these cysteine residues. Furthermore, we found in SDS-PAGE performed under non-reducing conditions that the observed molecular mass of the Ca 2ϩ -ATPase corresponded to its monomeric form but not to a possible dimeric or higher oligo-meric form (see also Fig. 2), indicating that the enzyme possesses only intramolecular disulfide bonds. Collectively, the results demonstrate that an intramolecular disulfide bond is formed between Cys 876 and Cys 888 in the SR Ca 2ϩ -ATPase.
In the second reversed phase HPLC, fraction II and fraction III gave two (peak IIa and peak IIb) (Fig. 1, G-I) and four (peak IIIa, peak IIIb, peak IIIc, and peak IIId) (Fig. 1, J-L) disulfidecontaining peaks, respectively. The disulfide-containing peptides in these peaks were further purified and labeled with ABD-F in the presence of TBP as above. The peptide from peak IIIc gave a sequence Met 874 -Phe 891 of SERCA1a, in which Cys 876 and Cys 888 had been labeled with ABD only in the presence of TBP (peptide IIIc, Table I): results again show that a disulfide bond is formed between Cys 876 and Cys 888 . The peptide from peak IIa gave a sequence Leu 832 -Phe 843 of a 160-kDa glycoprotein (and its splice variant 53-kDa glycoprotein) of SR (38) (Table I), in which a disulfide bond is formed between Cys 836 and Cys 842 . Peak IIb gave no distinct sequence. Peptides from peak IIIa and peak IIIb gave sequences in the 160-kDa glycoprotein, and a peptide from peak IIId gave a sequence in pepsin (data not shown).
Effects of 2-Mercaptoethanol on Mobilities of SERCA1a and Its Tryptic Fragments in SDS-PAGE-To explore possible roles of the Cys 876 -Cys 888 disulfide bond, we made three SERCA1a substitution mutants, C876A, C888A, and C876A/C888A, in which the disulfide bond of interest is disrupted, and expressed in COS-1 cells. The microsomes were prepared from the cells in the presence of 3 mM 2-mercaptoethanol to prevent possible formation of non-native disulfide bonds in the wild type and mutants. We first confirmed the presence of the intramolecular Cys 876 -Cys 888 disulfide bond in the expressed wild type SERCA1a as in the SR Ca 2ϩ -ATPase. In Fig. 2, the microsomes, SR vesicles, and their tryptic digests were either unreduced or reduced in SDS with 2-mercaptoethanol at a high concentration (0.7 M), and then subjected to SDS-PAGE and immunoblotting.
When reduced, all the SR Ca 2ϩ -ATPase, wild type, and three mutants migrated at 110 kDa ( Fig. 2A, dotted lines), which agrees with the molecular mass (1,2). When unreduced, the SR Ca 2ϩ -ATPase migrated at a substantially faster rate (95 kDa) ( Fig. 2A, arrows), but the three mutants migrated only very slightly faster than the reduced 110-kDa ATPase chains. The unreduced wild type SERCA1a migrated at exactly the same rate (95 kDa) as the unreduced SR Ca 2ϩ -ATPase. These results together with the above sequencing results show that the expressed wild type SERCA1a as well as the SR Ca 2ϩ -ATPase possesses the intramolecular Cys 876 -Cys 888 disulfide bond. This was further confirmed with the tryptic B fragment (Ala 506 to C terminus) on which both of the cysteine residues are located. The unreduced B fragment of the SR Ca 2ϩ -ATPase and wild type SERCA1a migrated substantially faster (45 kDa) (Fig. 2B, arrows) than the reduced B fragment (55 kDa) (Fig.  2B, dotted lines), whereas unreduced B fragment of three mutants migrated only very slightly faster than the reduced B fragment. We infer that this slightly faster rate is due to the presence of some other disulfide bonds (16,17), of which reduction would not affect the migration rate substantially. The tryptic A fragment (N terminus to Arg 505 ) of all the samples migrated at ϳ55 kDa regardless of whether untreated or treated with the reductant (Fig. 2C, dotted lines), the results being consistent with the observation (17) that disulfide bonds of the enzyme are confined to its B fragment.
Effects of Substitutions of Cys 876 and Cys 888 on Ca 2ϩ Transport and ATP Hydrolysis- Fig. 3A shows typical examples of the thapsigargin-sensitive Ca 2ϩ transport catalyzed by microsomes from COS-1 cells at 10 M Ca 2ϩ . Microsomes expressing the mutant C876A or C888A showed no transport activity over the background level determined with the control microsomes, which were prepared from the COS-1 cells transfected with the vector containing no SERCA1a cDNA. This background level was as low as 4% of the activity of microsomes expressing the wild type SECRA1a. The specific transport rate/mg of the wild type SERCA1a increased with increasing Ca 2ϩ concentration, and reached the maximal level at 10 M Ca 2ϩ with the halfmaximal rate at 0.4 M Ca 2ϩ (Fig. 3B). In contrast, none of the mutants C876A, C888A, and C876A/C888A was able to transport Ca 2ϩ at a detectable rate at all the Ca 2ϩ concentrations examined.
ATP hydrolysis was assayed with 0.1 mM [␥-32 P]ATP in the presence of Ca 2ϩ ionophore A23187 under conditions otherwise similar to those for the Ca 2ϩ transport assay. In sharp contrast to non-detectable Ca 2ϩ transport activity, fairly high thapsigargin-sensitive ATPase activities were observed with the microsomes expressing the mutants over the background level of the control microsomes at 10 M Ca 2ϩ (Fig. 4A). This background level was 8% of the activity of microsomes expressing the wild type SERCA1a. The specific ATPase activity/mg of SERCA1a protein increased with increasing Ca 2ϩ concentration, and reached the maximal level at 10 M Ca 2ϩ with the half-maximal activity at 1.0 -1.2 M Ca 2ϩ for the three mutants and 0.4 M Ca 2ϩ for the wild type (Fig. 4B). The maximal specific activities of the three mutants agreed closely with each other, and ϳ1.5 times higher than that of the wild type.
Ca 2ϩ Concentration Dependence of Phosphorylation with ATP and with P i -EP was formed from 10 M ATP at 0°C under conditions otherwise similar to those for the ATPase assay (Fig. 5A). The amounts of EP formed with the mutant and wild type SERCA1a at steady state increased with increasing Ca 2ϩ concentration and reached the maximal level at 10 M Ca 2ϩ . The half-maximal EP formation was obtained at 0.8 -1.0 M Ca 2ϩ for the three mutants and the wild type. The maximal EP levels of the three mutants were almost the same as that of the wild type.
Phosphorylation with P i was performed in the presence of 35% (v/v) Me 2 SO, which greatly favors EP formation (39). The three mutants and the wild type formed almost the same maximal amounts of EP at pCa 9 -7 (Fig. 5B). The EP formation was almost completely inhibited at 100 M Ca 2ϩ with the half-maximal inhibition at 2.3-2.9 M Ca 2ϩ .

Formation of ADP-sensitive EP and ADP-insensitive EP from
ATP-EP was formed from ATP in the presence and absence of added K ϩ at 0°C and 50 M Ca 2ϩ under conditions otherwise similar to those for the ATPase assay, and the total amount of EP and amount of ADP-insensitive EP were determined at steady state. In the presence of K ϩ , which strongly accelerates hydrolysis of ADP-insensitive EP and thus suppresses its accumulation (40), only a very small fraction of EP was ADPinsensitive in the three mutants as well as in the wild type (Fig.  6A). On the other hand, in the absence of K ϩ , EP formed was largely ADP-insensitive in the three mutants as well as in the wild type (Fig. 6B).
Dephosphorylation of EP-Dephosphorylation of ADP-sensitive EP in the presence of K ϩ was examined by first phosphorylating the Ca 2ϩ -ATPase with [␥-32 P]ATP at 50 M Ca 2ϩ under the conditions in Fig. 6A, in which EP formed at steady state is almost completely ADP-sensitive and then terminating phosphorylation by adding excess EGTA to allow dephosphorylation of 32 P-labeled EP (Fig. 7A). The dephosphorylation of the ADP-sensitive EP in the three mutants and in the wild type proceeded with first-order kinetics. The results are consistent with the view that the transition of EP from ADP-sensitive from to ADP-insensitive form is rate-limiting for the dephosphorylation in the presence of K ϩ . Dephosphorylation rates of the three mutants agreed well with each other, and ϳ1.7 times higher than that of the wild type.
Hydrolysis of ADP-insensitive EP was examined in the absence of K ϩ by first phosphorylating the Ca 2ϩ -ATPase with 32 P i and then chasing with an excess of non-radioactive P i (Fig. 7B). Hydrolysis of the 32 P-labeled ADP-insensitive EP in the three mutants and in the wild type proceeded with first-order kinetics. The hydrolysis rates of the three mutants were almost the same as that of the wild type. DISCUSSION The present study shows that an intramolecular disulfide bond is formed between Cys 876 and Cys 888 on the luminal loop 7-8 in SR Ca 2ϩ -ATPase and expressed wild type SERCA1a and that the disruption of this disulfide bond by the mutations results in a complete loss of the Ca 2ϩ transport activity without a loss of the Ca 2ϩ -dependent ATP hydrolysis activity. It is likely that the pathway for delivery of Ca 2ϩ from the binding sites into the lumen of vesicles is disrupted by disruption of the Cys 876 -Cys 888 disulfide bond, and therefore that the loop 7-8 having the disulfide bond is important for formation of the proper structure of the Ca 2ϩ pathway.
It is unlikely that this disulfide bond is a non-native one formed by oxidation or disulfide interchanges during the proteolysis and chromatographic isolation procedures, because all the procedures were carried out under the conditions that prevent oxidation and disulfide interchanges (21)(22)(23). It should also be noted that the disulfide bond is present in the expressed wild type SERCA1a prepared in the presence of 2-mercaptoethanol (Fig. 2). This result is consistent with our previous finding that the disulfide bonds in the enzyme are not readily reducible (see Ref. 16 and the introduction). The observed identical behavior of the wild type SERCA1a and the SR Ca 2ϩ -ATPase in SDS-PAGE performed under reducing and nonreducing conditions (Fig. 2), therefore, also indicates that the Cys 876 -Cys 888 disulfide bond is present in the native enzyme.
Previous studies by Green and co-workers (15, 17) as well as ours (16) have demonstrated that totally three or four disulfide bonds are present in the enzyme molecule. The Cys 876 -Cys 888 bond has been considered as a most likely candidate for one of the disulfide bonds (15,32). Close location of Cys 876 to Cys 888 (at 7.7 Å between their ␣-carbons) in the crystal structure (13) suggests formation of the Cys 876 -Cys 888 disulfide bond being potentially possible. However, no disulfide bonds are seen in the crystal structure (13). It may be possible that disulfide bonds of the enzyme in the crystal had been reduced by dithiothreitol during the purification and dialysis for crystallization of the enzyme. We think this is quite possible, because all the disulfide bonds in the enzyme can not be readily reduced by dithiothreitol but can be readily reduced if both Ca 2ϩ and ADP are present (16) and because, in fact, Ca 2ϩ , ADP, and dithiothreitol were all present during the purification with a Redagarose column chromatography and dialysis (in its initial period) employed for crystallization of the enzyme (13).
The requirements for the disulfide bond would be demonstrated unambiguously if it were possible to obtain functional alternations by treating the native enzyme with reductants and thus by selective reduction of the disulfide bond of interest. However, our previous finding showed that such selective reduction is impossible to achieve (see Ref. 16  tion): Treatment with dithiothreitol in the presence of Ca 2ϩ and purine nucleotides results in non-selective reduction of all the disulfide bonds with only a very slightly faster reduction rate for one over that of the others, the time courses thus not being separated (16). Actually the Cys 876 -Cys 888 disulfide bond was found even after reduction of this slightly fast reacting disulfide bond (data not shown). Thus, only the site-directed mutagenesis is available to explore functional roles of the Cys 876 -Cys 888 disulfide bond.
All the three mutants C876A, C888A, and C876A/C888A, in which the Cys 876 -Cys 888 disulfide bond is disrupted, hydrolyzed ATP at high rates in a Ca 2ϩ -dependent reaction, in contrast to the complete loss of the Ca 2ϩ transport activity (Figs. 3 and 4). The characteristics of the mutants in the ATP hydrolysis reaction are in essential agreement with those of the wild type. The activation of the catalytic site by the high affinity Ca 2ϩ binding at the transport sites through the long-range interaction (41,42) was not inhibited in the mutants (Figs. 4 and 5). The observed predominant accumulation of ADP-sensitive EP in the presence of K ϩ is in harmony with the presence of the rate-limiting transition of EP from ADP-sensitive form to ADPinsensitive form in the ATPase cycle (Fig. 6). In fact, the decay rate of the accumulated ADP-sensitive EP and the specific ATPase activity were changed to almost the same extent (1.5-1.7 times increase) by the mutations (Figs. 4B and 7A). Hydrolysis of the ADP-insensitive EP in the mutants was directly demonstrated to take place by the EP formation from P i in the reversal of its hydrolysis (Fig. 5B) and by dephosphorylation of this EP occurring with almost the same rate as the wild type (Fig. 7B). Collectively, the results show that the Ca 2ϩ -activated mutants hydrolyze ATP through formation of ADP-sensitive EP, subsequent rate-limiting transition of this EP to ADPinsensitive EP, and hydrolysis of the latter EP, in the mechanism well accepted for the Ca 2ϩ -ATPase. It should be noted that the high affinity ATP binding at the catalytic site is also not impaired in the mutants, because the level of EP formed from ATP as low as 10 M was almost the same as that of the wild type (Fig. 5A).
Because the substitution of either or both of the two residues resulted in the complete loss of the Ca 2ϩ transport activity (Fig.  3), the loss is most likely a consequence of the disruption of the Cys 876 -Cys 888 disulfide bond. The observation that the three mutants are identical to each other also in the altered kinetic properties in the ATP hydrolysis is consistent with the view that each of the two cysteine residues has a specific function in the formation of the disulfide bond. However, it cannot be excluded that the effect of mutations may be due to other effects such as steric hindrance in the site caused by the different side chain.
The complete loss of Ca 2ϩ transport without a loss of normal Ca 2ϩ -dependent ATP hydrolysis is likely caused by disruption of the pathway for delivery of Ca 2ϩ from the binding sites into the lumen of vesicles, or possibly by increased rate of passive efflux of transported Ca 2ϩ through the mutants. Cys 876 and Cys 888 are located on the long luminal loop 7-8, which is directly connected to M7 and M8. In the crystal structure (13), M7 is in close contact with M5 near the luminal surface and the likely Ca 2ϩ outlet to lumen is predicted to be in the area surrounded by M5, M4, and M3. M8 is directly involved in the formation of the Ca 2ϩ channel (13). It is possible that the tertiary structure of loop 7-8 stabilized by the Cys 876 -Cys 888 disulfide bond is essential for proper orientation of M7 and of M8 relative to other helices, conferring appropriate packing of the helices required for the delivery of Ca 2ϩ into the lumen and prevention of passive Ca 2ϩ efflux. Interestingly, mutations of the residues on M5, Val 772 , Cys 774 , and Ile 775 , which are located at the position in close contact with M7 in the crystal structure (13), were reported to cause inhibition of Ca 2ϩ uptake with little or no inhibition of ATPase activity, i.e. the uncoupling effect (43). In agreement with the view that local interference with proper packing of the helices results in the uncoupling, several other mutations within M4, M5, M6, or M8 were also reported to cause the uncoupling (43,44). A possible role in sealing the lumen and preventing passive Ca 2ϩ efflux was predicted (43) for Lys 297 located at the luminal end of M4.
Alternatively, the loop 7-8 may directly be involved in the Ca 2ϩ release process. Kinetic studies on the effect of luminal Ca 2ϩ on EP formation from P i by Myung and Jencks (45) predicted a second set of Ca 2ϩ binding sites on the luminal surface, possibly on the loop 7-8 (46), through which Ca 2ϩ is released into the lumen. If such sites exist on this loop, it is possible that disruption of the Cys 876 -Cys 888 disulfide bond FIG. 7. Dephosphorylation of ADP-sensitive EP formed from [␥-32 P]ATP and that of ADP-insensitive EP formed from 32 P i . A, microsomes were phosphorylated with [␥-32 P]ATP at 0°C for 15 s in 100 l of the mixture containing 15 g of microsomal protein, otherwise as described in Fig. 6A. Phosphorylation was terminated by addition of EGTA to give 5 mM, and dephosphorylation was quenched with trichloroacetic acid at different times after the addition of EGTA. B, the microsomes were phosphorylated with 32 P i at 25°C for 10 min in 50 l of a mixture containing 15 g of microsomal protein, 0.1 mM 32 P i , 10 mM MgCl 2 , 35% (v/v) Me 2 SO, and 100 mM MOPS/Tris (pH 7.0), and 2 mM EGTA. The mixture was then cooled and diluted at 0°C by addition of 950 l of a mixture containing 2 mM non-radioactive P i , 10 mM MgCl 2 , 100 mM MOPS/Tris (pH 7.0), and 2 mM EGTA. At different times after the dilution, dephosphorylation was quenched with trichloroacetic acid. could result in the loss of the Ca 2ϩ transport activity.
The loop 7-8 regions of the Ca 2ϩ -ATPases in the SERCA family have high homology to each other, and both Cys 876 and Cys 888 are conserved in this family (1,2,(47)(48)(49). On the other hand, other members of P-type ATPases, plasma membrane Ca 2ϩ -ATPase (50), Na ϩ ,K ϩ -ATPase (51,52), and H ϩ ,K ϩ -ATPase (53) contain no cysteine residues (or only one in human gastric H ϩ ,K ϩ -ATPase (54)) on the loop 7-8 region. It is possible that the role of the disulfide bond on the loop 7-8 found in this paper is specific to the SERCA family.
Recently, Darier disease, a human autosomal dominant skin disorder, was shown to be caused by mutations in the SERCA2b gene (55), and a single missense mutation at Cys 875 (C875G) was found in one of the Darier-disease pedigrees (56). Because Cys 875 in SERCA2b corresponds to Cys 876 in SERCA1a, our present observations suggest that the mutation of Cys 875 (i.e. disruption of the Cys 875 -Cys 887 disulfide bond) in SERCA2b results in the loss of the Ca 2ϩ transport and therefore causes perturbation in calcium homeostasis and the disease.