Nanodisc-based kinetic assays reveal distinct effects of phospholipid headgroups on the phosphoenzyme transition of sarcoplasmic reticulum Ca2+-ATPase

Sarco(endo)plasmic reticulum Ca2+-ATPase catalyzes ATP-driven Ca2+ transport from the cytoplasm to the lumen and is critical for a range of cell functions, including muscle relaxation. Here, we investigated the effects of the headgroups of the 1-palmitoyl-2-oleoyl glycerophospholipids phosphatidylcholine (PC), phosphatidylethanolamine (PE), phosphatidylserine (PS), and phosphatidylglycerol (PG) on sarcoplasmic reticulum (SR) Ca2+-ATPase embedded into a nanodisc, a lipid-bilayer construct harboring the specific lipid. We found that Ca2+-ATPase activity in a PC bilayer is comparable with that of SR vesicles and is suppressed in the other phospholipids, especially in PS. Ca2+ affinity at the high-affinity transport sites in PC was similar to that of SR vesicles, but 2–3-fold reduced in PE and PS. Ca2+ on- and off-rates in the non-phosphorylated ATPase were markedly reduced in PS. Rate-limiting phosphoenzyme (EP) conformational transition in 0.1 m KCl was as rapid in PC as in SR vesicles, but slowed in other phospholipids, especially in PS. Using kinetic plots of the logarithm of rate versus the square of mean activity coefficient of solutes in 0.1–1 m KCl, we noted that PC is optimal for the EP transition, but PG and especially PS had markedly unfavorable electrostatic effects, and PE exhibited a strong non-electrostatic restriction. Thus, the major SR membrane lipid PC is optimal for all steps and, unlike the other headgroups, contributes favorable electrostatics and non-electrostatic elements during the EP transition. Our analyses further revealed that the surface charge of the lipid bilayer directly modulates the transition rate.

lyzes ATP-driven Ca 2ϩ transport from the cytoplasm to the lumen against a Ca 2ϩ concentration gradient of several thousand times (1)(2)(3). The pump consists of three cytoplasmic domains (nucleotide binding (N), phosphorylation (P), and actuator (A)) and 10 transmembrane helices (M1-M10) (4) (Fig. 1). The ATPase is first activated by binding of two Ca 2ϩ ions from the cytoplasmic side to the transport sites composed of M4, M5, M6, and M8 (E2 ϩ 2Ca 2ϩ 3 E1Ca 2 ; Fig. 1) and forms an autophosphorylated intermediate (EP) with ATP at the catalytic residue Asp-351 on the P domain and thereby occludes the two Ca 2ϩ at the transport sites (E1PCa 2 ). This EP is ADP-sensitive (E1P) and rapidly regenerates ATP in the reverse reaction upon the addition of ADP. Subsequently, E1P proceeds through an isomeric conformational transition to ADP-insensitive E2PCa 2 (E1PCa 2 3 E2PCa 2 ) followed by rapid Ca 2ϩ release (E2PCa 2 3 E2P ϩ 2Ca 2ϩ ). Finally, the aspartylphosphate in E2P is hydrolyzed, and the ATPase returns to the Ca 2ϩ -unbound ground E2 state. The isomeric EP transition is rate-limiting in the ATPase cycle and involves large rotation of the A domain, inclination of the P domain with M4/M5 toward the A domain, and consequent tight A-P domain association, thereby releasing Ca 2ϩ and readying the catalytic site for hydrolysis (5)(6)(7)(8)(9). In the Ca 2ϩ transport cycle, ϳ2-3H ϩ are transported to the cytoplasmic side under physiological conditions in exchange for the 2Ca 2ϩ , making the Ca 2ϩ pump electrogenic (10 -13).
As Ca 2ϩ -ATPase is a membrane-embedded protein, the surrounding lipid bilayer is crucial for function. The length of the fatty acid acyl chains strongly affects activity. Short (ϽC14) or long (ϾC24) acyl chains are inhibitory, and a chain length around C18 -C16 is optimal; thus, the thickness of the lipid bilayer is critical (14,15). Membrane perturbation by nonionic detergent C 12 E 8 at a non-solubilizing concentration stabilizes a transient E2P intermediate state involved in the Ca 2ϩ -release process, which, otherwise, cannot easily be discerned (16). Phospholipid content of SR membrane indicates mostly PC (68%) and smaller amounts of phosphatidylethanolamine (PE) (16%) and phosphatidylserine (PS) (11%), altogether comprising ϳ95%, and they are asymmetrically located in the outer and inner leaflets of the SR membrane (17). A high content of PE in artificial liposomes causes a marked reduction of ATPase activity (18,19), although the reason is unknown.
In a recent study (20), electron density maps by solvent contrast modulation on SERCA1a crystals made with a 1,2-dio-leoyl-sn-glycero-3-phosphocholine bilayer revealed the locations of phospholipids surrounding the crystalized SERCA1a molecules. Many first-layer phospholipids are hydrogen-bonded with amino acid residues, and some phospholipids are anchored by Arg/Lys-phosphate salt bridges and can follow the movements of transmembrane helices, causing local distortions and changes in thickness of the lipid bilayer. Interestingly, exchanging PC with PE does not alter the Ca 2ϩ -ATPase crystal structure (21). In Na ϩ ,K ϩ -ATPase, particular lipids are directly embedded within specific sites of the protein and have a stabilizing and/or activating role (22,23). Such specific sites are not known in SERCA1a.
Such previous biochemical and very recent structural interactions underscore the necessity to evaluate the contribution of phospholipid headgroups to ATPase function. The advent of nanodisc, which is composed of a phospholipid bilayer defined in size and constituents and held in place by two membrane scaffold proteins (24) and used to study membrane proteins, including P-type H ϩ -ATPase (25), offers an opportunity to study purified SR Ca 2ϩ -ATPase protein in pure 1-palmitoyl-2oleoyl (16:0, 18:1) glycerophospholipids: phosphatidylcholine (PC), PE, PS, and phosphatidylglycerol (PG) or a mixture. The nanodisc made with PC without an introduced membrane protein is well characterized and contains two MSP1D1 (membrane scaffold protein 1D1) bundle molecules and ϳ120 PC molecules (26,27). The nanodisc construct allows evaluation of the phospholipid bilayer contribution to structure and function without detergent and without a confounding membrane potential or Ca 2ϩ gradient. We are able to show distinct effects of the headgroups of phospholipids on pump cycle kinetic properties and dissect out electrostatic and non-electric contributions to the rate-limiting EP transition.

Properties of nanodisc formed with Ca 2؉ -ATPase
SR Ca 2ϩ -ATPase (SERCA1a) was purified by red-agarose column chromatography and reconstituted into nanodisc with various phospholipid components, and then the SR Ca 2ϩ -ATPase-containing nanodisc (CND) was purified by size-exclusion chromatography (Fig. 2). The EP formation activity of SERCA1a eluted at around 11.5-12.0 min (Stokes diameter 10.8 -11.8 nm) and corresponded to a major 280-nm protein peak followed by a smaller peak attributable to empty nanodisc (also see below). The CND fraction collected at 11.5-12.0 min shows almost a single band in native PAGE (Fig. 3A) without major impurities, and the empty nanodisc fraction collected at 12.2-13.3 min (corresponding to a Stokes diameter of 8.4 -10.0 nm) also shows one major band.
The homogeneity and dimensions of empty nanodiscs and CNDs constructed of PC were further examined by transmission electron microscopy (Fig. 4). The images show that the nanodiscs exist as single particles (Fig. 4, A and B) with Feret diameters for the CND of 14.9 Ϯ 1.5 nm (n ϭ 100) and for the empty nanodisc of 11.9 Ϯ 1.2 nm (n ϭ 100) (Fig. 4C), both values consistent with the homogeneous behavior and Stokes diameters of all nanodiscs in size-exclusion chromatography.
The molar ratio of MSP1D1 to SERCA1a in the CND fraction was determined by SDS-PAGE and Coomassie Brilliant Blue R-250 staining. All of the CNDs with different phospholipids contained only SERCA1a and MSP1D1 (two bands by SDS-PAGE) without any other proteins (Fig. 3B), and the empty nanodisc had only MSP1D1 protein. The ratio MSP1D1/ SERCA1a is 2.0 in CND formed with 1-palmitoyl-2-oleoyl-snglycero-3-phospho-L-serine (POPS), an ideal value for a nanodisc composed of two MSP1D1 molecules and one SERCA1a molecule (as shown in Fig. 4D), and somewhat larger in CNDs with other lipids. Then as a test sample, the fraction of CND constructed with PC (MSP1D1/SERCA1a molar ratio 2.5) was  (4)). The cytoplasmic domains A, P, and N are colored yellow, cyan, and pink, respectively, and transmembrane helices are shown in gray with the two bound Ca 2ϩ ions in red. The approximate position of the membrane region is shown by horizontal lines.

Effects of lipid headgroups on Ca 2؉ -ATPase
subjected to an additional size-exclusion chromatography step (Fig. 3C). The early part of the protein peak shows the ideal molar ratio of 2.0 for MSP1D1/SERCA1a, and the ratio increases in the latter part of the peak, indicating that the higher ratio is due to empty nanodisc without SERCA1a. Such repeat chromatography led to a large loss of CND sample; therefore, we used the 11.5-12.0 min fraction in the first chromatography for functional analysis of Ca 2ϩ -ATPase in nanodisc. Note again that all of the collected CND samples contained only SERCA1a and MSP1D1 proteins without any other major proteins (Fig.  3B) and are thus suitable for following functional analyses of Ca 2ϩ -ATPase embedded in nanodisc.

Native PAGE reveals the difference in surface charge of nanodisc with different lipids
Advantageously, nanodisc (with or without membrane-embedded protein) can be handled as soluble material and can be analyzed by native PAGE without detergent (Fig. 3A). Migration distance depends solely on charge, size, and shape of the nanodisc. Then where the size and shape of the complex are nearly the same with different phospholipids, the migration distance depends solely on charge. The migration distances of CNDs with acidic phospholipids (PG and PS) are much larger than those of CNDs with neutral phospholipids (PE and PC), showing the effect of the negatively charged headgroups of the lipids. The CNDs with PG and with PS migrate more slowly than empty nanodiscs with the respective phospholipid, probably due to the smaller number of phospholipid molecules when Ca 2ϩ -ATPase is present as well as the larger size. However, PC CND and the empty PC nanodisc migrate similarly, so it seems that the increase in size may be compensated for by a negatively charged protein. Empty PE nanodisc migrates farther than empty PC nanodisc, as is the case with PE CND and PC CND (although the difference is very small in the CNDs); this finding indicates that PE in the nanodisc is forced to possess negative charge under the nanodisc intrinsic atmosphere, as also noted previously (28). CND made with lipids obtained from SR vesicles (SRL) exhibits an intermediate migration distance between CND with PC (or PE) and CND with PS, as also seen with empty nanodiscs, consistent with the composition of the lipids of the SR membrane of PC (68%), PS (11%), and PE (16%) (17).

Effects of lipid headgroups on Ca 2؉ -ATPase Ca 2؉ -ATPase activity
The Ca 2ϩ -ATPase activity was determined in 0.1 M KCl under physiological conditions in the presence and absence of Ca 2ϩ ionophore A23187, and the turnover rate is shown (Fig.  5A). The activity is highly sensitive to the phospholipid component in CND. In A23187, the activity of PC CND is comparable with that of SR vesicles, and activities with other lipids (PG, PS, and PE) are much lower, especially with PS (Ͻ4% of that with PC). SRL CND has high activity, but still almost half of that with PC, consistent with the content of PC at 68% and not 100% (17).
In the absence of A23187, the Ca 2ϩ -ATPase activity of SR vesicles is inhibited due to back-inhibition by luminally accumulated Ca 2ϩ . By contrast, the activities of all CNDs are hardly affected by A23187, in agreement with the structural feature of the nanodisc that both the cytoplasmic and luminal sides of SERCA1a embedded in a nanodisc are exposed to the same solution, and thus there can be no Ca 2ϩ uptake and no Ca 2ϩ gradient formation. This is very different from the situation of SERCA1a embedded unidirectionally in liposomes.
The slight effect of A23187 on activity in CNDs is consistent with A23187 directly influencing the activity of deoxycholatepurified SR Ca 2ϩ -ATPase (29). It may be possible that such a direct A23187 effect is observed here lipid-dependently with the nanodisc system.
In Fig. 5B, the Ca 2ϩ -ATPase activity was determined after solubilization with 2 mM C 12 E 8 , a non-ionic detergent. In this assay, 1 mM POPC was included in the assay solution to mask the effects of lipids originating from CND; the amount of added POPC exceeds that of CND lipids over 100 times. Once the CND sample is solubilized with C 12 E 8 in POPC, the activities of all of the CND samples become nearly equal, showing that the reduced activity of SERCA1a in certain CNDs is reversible, even in PS CND with the lowest activity. The Ca 2ϩ -ATPase activity of solubilized SR vesicles is lower than that of the CND samples, possibly accounted for by an inhibitory factor in native vesicles that is absent from purified Ca 2ϩ -ATPase.

Ca 2؉ binding
Ca 2ϩ affinity at the high-affinity transport sites of Ca 2ϩ -ATPase (E2 ϩ 2Ca 2ϩ 7 E1Ca 2 ) was estimated by the Ca 2ϩ concentration dependence of EP formation (Fig. 6A). The affinities in CNDs formed with PC and with SRL are comparable with that of SR vesicles (K d ϭ 0.26 Ϯ 0.01 M with Hill coefficient 2.1 Ϯ 0.1 (Table 1)). The affinity in PE CND is somewhat reduced (K d ϭ 0.59 Ϯ 0.03 M) and rather lower in PS CND (K d ϭ 0.92 Ϯ 0.09 M). On the other hand, the affinity in PG CND (K d ϭ 0.17 Ϯ 0.02 M) is slightly higher than that in SR vesicles.
In Fig. 6B, the time course of Ca 2ϩ release from the Ca 2ϩbound active E1Ca 2 state was assessed by the loss of EP forma-   Table 1. The amount of EP was normalized to the maximum level of EP. B, samples were incubated with 0.2 mM CaCl 2 , otherwise as described above, and mixed with an equal volume of 2 mM EGTA in the same buffer without added CaCl 2 . At the indicated times after the EGTA addition, the reaction mixture was further mixed with the same volume of a solution containing 20 M [␥-32 P]ATP and 2 mM EGTA without added CaCl 2 , otherwise as above. Then the reaction was terminated by acid at 2 s after the ATP addition. The solid lines show the least-squares fit to a single exponential decay. The rate thus obtained represents Ca 2ϩ off-rate, and is listed in Table 1 together with the Ca 2ϩ on-rate calculated with the off-rate and K d .

Effects of lipid headgroups on Ca 2؉ -ATPase
tion ability upon the addition of excess EGTA, and the Ca 2ϩ off-rate was calculated (Table 1). Ca 2ϩ release in PS CND is extremely slow (see Fig. 6B, inset), and that of the other CNDs with SRL, PG, PC, and PE is only slightly slower than that of SR vesicles. The Ca 2ϩ on-rate (E2 ϩ 2Ca 2ϩ 3 E1Ca 2 ), estimated from the Ca 2ϩ affinity and Ca 2ϩ off-rate, revealed that binding is extremely slow with PS CND, only 1% of that of SR vesicles, whereas it is somewhat slowed in the other CNDs, from 70 to 20% of that of SR vesicles (Table 1).

EP transition
The time course of the rate-limiting isomeric conformational transition of EP (E1PCa 2 3 E2PCa 2 ) followed by a rapid Ca 2ϩ release (E2PCa 2 3 E2P) was examined in 0.1 M KCl (Fig. 7, A  and B). In the experiments, EP is first formed with ATP for 10 s to reach a steady state, and then EP decay is followed by chasing Ca 2ϩ with EGTA. In the steady state under these conditions, EP is mostly ADP-sensitive E1P (Fig. 7C, open bar); therefore, EP decay represents the rate-limiting EP transition process (E1PCa 2 3 E2PCa 2 , which is followed by rapid Ca 2ϩ release and E2P hydrolysis (Fig. 1)). Decay rates of SRL and PC CNDs are comparable with those of SR vesicles. PG and PE CNDs are slower, and PS CNDs markedly are retarded to 2% of the rate of SR vesicles. The effects of phospholipids on the rate-limiting EP transition and on Ca 2ϩ -ATPase turnover are consistent (cf. Fig. 5A).
We previously found (30) that the plot of logarithm of rate versus the square of mean activity coefficient of solutes (␥ Ϯ 2 ) for a reaction of interest gives a linear relationship, thereby dividing the activation energy into two components: electric and non-electric forces. The slope reflects the amplitude of the contribution of the electrostatic energy, and the intercept at ␥ Ϯ 2 ϭ 0 reflects the non-electrostatic contributions (or steric effects), because in this hypothetical state, electrostatic interactions on the protein surface are completely shielded. We apply this approach to the EP transition in 0.1-1 M KCl, whereby electrostatic contributions are gradually suppressed.
In Fig. 8, the log(rate) versus ␥ Ϯ 2 plot indeed shows a linear relationship for all CND samples and SR vesicles. The slopes for neutral lipid CNDs (PC and PE) and with SRL are slightly positive ( Fig. 8 and Table 2) as with SR vesicles, indicating that electrostatic interactions between SERCA1a protein and the lipid heads, even with neutral lipids, facilitate the isomeric conformational EP transition. It is also possible that electrostatic interactions within SERCA1a protein needed for rapid EP transition (30) are favored by these lipids surrounding the protein.
On the other hand, the slopes for PG and PS CNDs are negative, especially with PS, indicating that their negative charge hampers the EP transition, reflected in the extremely low Ca 2ϩ -ATPase turnover rate (cf. Fig. 5A).
The intercept at ␥ Ϯ 2 ϭ 0 for PE CND is markedly reduced, indicating that non-electrostatic restrictions come into play with this lipid during the conformational transition ( Fig. 8 and Table 2). Thus, the low Ca 2ϩ -ATPase turnover rate with PE, seen above, is due to direct non-electrostatic inhibition of the EP transition.

Relationship between effects of electric forces on EP transition and surface charge of membrane
To further explore the nature of the electric component of the lipid-protein interaction and possible effects of surface

Effects of lipid headgroups on Ca 2؉ -ATPase
charge, a mixture of phospholipids was introduced into the CNDs. As seen with representative examples in Fig. 9A, slope decreases with increasing acidic lipid content. The increase in acidic lipid content is confirmed by native PAGE (Fig. 9, B and C). For PG/PC mixtures, the slope decreases almost linearly with the increase in R f value (Fig. 9D), indicating that surface charge directly affects the EP transition rate (rather than a specific PG binding effect on SERCA1a). For PS/PC CNDs, the slope values again decrease almost linearly with increases in the R f value up to ϳ0.3 (PS content ϳ0.5 (cf. Fig. 9D)) and then decrease sharply. The latter phenomenon, at high PS proportions, suggests a PS-specific effect on SERCA1a, which impedes the transition, in addition to the unfavorable electrostatic interactions described above.

Single Ca 2؉ -ATPase molecule in one CND
The properties of nanodisc constructed with PC and two belts of MSP1D1 proteins have been well characterized (26,27), and the 4.6-nm thickness with acyl chain length of C16 -C18 is optimal for SERCA1a activity (31). The inner diameter and bilayer area of the MSP1D1 bundle are about 7.6 nm and 44 nm 2 , respectively (27), which should be sufficient for harboring the transmembrane section of one SERCA1a molecule (ϳ4 -5 nm in diameter and 9.5-11 nm 2 in area) (20), but probably not two. Indeed, we found the MSP1D1/SERCA1a ratio of all our samples to be ϳ2 or higher (due to the presence of empty nanodisc) and far from the value 1 (expected for two ATPases per nanodisc). The Stokes diameters of the CNDs estimated from the retention time in the size-exclusion chromatography are between 10.8 and 11.8 nm (Fig. 2), and the Feret diameter determined with PC CND (representative of the four phospholipids) in transmission electron microscopy is 14.9 Ϯ 1.5 nm, and this size range is close to the SERCA1a's long diameter of ϳ12 nm. Thus, a single SERCA1a molecule is embedded in one nanodisc. The system is eminently suitable for exploring lipid headgroup effects on function.

PC is the major lipid in native SR membrane and optimal for function
CNDs with PC exhibit rapid ATPase activity, high-Ca 2ϩ affinity binding, rapid ATPase activation upon Ca 2ϩ binding, and a relatively fast rate-limiting EP transition. PS, PG, and PE are all inhibitory, except for a small effect of PG on increasing Ca 2ϩ -binding affinity. In fact, PC performs as well as the SR lipids themselves, which begs the question as to why small amounts of PS and PE are present at all, obviously not for pump function. The answer may lie in the need to introduce curvature to the reticular membrane (32,33) and also for PS to possibly bind Ca 2ϩ in the lumen with its carboxyl group (34), as, in fact, a large part (84%) of PS is located in the inner leaflet of SR membrane (17).
It has recently emerged that phospholipids play an intimate role in the dynamics of Ca 2ϩ -ATPase pumping, acting both as deformable entities and anchors during the substantial conformational changes among the transmembrane helices (20). Deformability is intrinsic to the acyl chain region through the hydrophobic effect (35). The stabilizing role is through hydrogen bonding and salt linkages, some temporary or shifting, between phosphoryl entities and basic amino acid residues. PC, with its large head of three hydrophobic methyl groups, contributes to deformability, spread, and anchoring, compared with PE, which has a tendency to self-associate and presumably form a tighter mat (36). In the electron density maps (20), there are several salt linkages between basic amino acid residues and phosphoryl groups of first-layer phospholipids, but Arg-324 on the cytoplasmic P domain at its junction with M4 is prominent in linking to one phospholipid in E1PCa 2 (E1Ca 2 ⅐AlF 4 Ϫ ⅐ADP crystal), only to release it and join with two others in E2ϳP (E2⅐AlF 4 Ϫ crystal). Actually, we have found that the alanine and glutamate substitutions of this critical Arg-324 markedly retard the EP transition (37). The large methyl groups of choline, projecting away from the protein, may allow a more favorable presentation of the phosphoryl groups to the arginine.

Electric forces for rapid EP transition
In our analysis of the electrostatic and non-electrostatic effects on the crucial rate-limiting EP transition, we found favorable electrostatic contributions to the transition with PC and PE, but inhibitory effects with the negatively charged PS and PG. Thus, not only hydrogen bonding and salt phosphorylbasic amino acid residue linkages are important for the transition, but also the positive charges of the amino and choline entities of PE and PC, respectively. The EP transition is a com-  Table 2.

Table 2
Fitting parameters in the analysis in Fig. 8 Values shown are the mean Ϯ S.E.

Effects of lipid headgroups on Ca 2؉ -ATPase
plicated step involving a large movement of M4 to incline the P-domain and a 90º rotation of the A-domain, concomitant with an opening of the transport sites to the lumen and release of Ca 2ϩ . The Ca 2ϩ binding sites are located close to the membrane center and include a cluster of acidic amino acid residues. The two Ca 2ϩ released to the lumen are exchanged for about two H ϩ , which means that destabilizing negative charges in the transmembrane binding sites appear during the EP transition. An intriguing possibility is that charges are stabilized by the positive electrostatic charges of the choline moiety operating through the low dielectric of the membrane interior. The advantageous effect of the positive charge of PC on the EP transition was confirmed by the introduction of negatively charged PS and PG into POPC CNDs. Increasing the negative charge content of the nanodisc proportionately decreased the slope of the log(rate) versus ␥ Ϯ 2 plot, showing that the stimulating electrostatic effect of PC is exerted through the overall bilayer surface charge. Surface charge then is what is helping to drive the EP transition, possibly through a stabilizing effect on the developing negative charges at the emptying Ca 2ϩ sites.

Non-electrostatic effects on EP transition
Whereas PC and PE positively influence the EP transition through electrostatic forces, we also found that PE exerts a direct non-electrostatic, probably steric, inhibitory effect on the EP transition when electrostatic influences are quelled. Inhibition of Ca 2ϩ -ATPase activity by PE has also been observed in reconstituted liposomes with a PE content of Ͼ80% (19). The small size of the PE head may allow it entry into a protein crevasse that normally only opens transiently during the transition, thereby inhibiting the step. In fact, in many crystal structures of SERCA1a, PE is bound, although no PE was added during purification and crystallization (7, 38 -42). There may be a negative charge component to this steric inhibition, as empty PE nanodiscs migrate further than empty PC nanodiscs in native PAGE (Fig. 3A), signifying more negative charge, which is detrimental according to the effects of PS and PG. It is also possible that the hydrogen bonds between residues of the ATPase protein and lipid heads are disrupted with PE due to its small headgroup.

PS-specific unfavorable electrostatic effects
PS has a carboxyl group at the head, is more negatively charged and distinct from the other lipids, and has the strongest unfavorable electrostatic effects on the steps that involve large conformational changes: the ATPase activation by Ca 2ϩ binding (E2 ϩ 2Ca 2ϩ 3 E1Ca 2 ) and the EP transition. Actually, the rates at 0.1 M KCl under physiological conditions are only 1-2% of the respective ones of SR vesicles. PG, lacking the carboxyl, is much less inhibitory.

Effects of lipid headgroups on Ca 2؉ -ATPase
Although it is not possible to pinpoint the sites responsible for such unfavorable interactions, there is a strong possibility that such sites may be located at the lipid-protein interface at the cytoplasmic region of the protein. The cytoplasmic Ca 2ϩ entry path in E2 ϩ 2Ca 2ϩ 3 E1Ca 2 is located at the space created around M1 and M2 (42), and the carboxyl group of the PS head could well retard Ca 2ϩ movements in and out of the binding sites. Also, one of the most crucial and dramatic changes at the protein-lipid interface occurs during the EP transition when M1Ј is formed by an M1 kink and embedded at the surface in the membrane (modeled with E1Ca 2 ⅐AlF 4 Ϫ ⅐ADP 3 E2⅐BeF 3 Ϫ ), a structural change that is largely reversed during the E2 3 E1Ca 2 structural change (see Fig. 2 in Ref. 43). Such motions of M1Ј may be severely impeded by the PS headgroup. In fact, the M1Ј helix has negatively charged and polar residues (Glu-51, Glu-55, Gln-56, and Glu-58) on the upper side facing away from the membrane and non-polar residues (Leu-49, Trp-50, Val-53, Ile-54, and Phe-57) on the other side facing the membrane core, stabilizing M1Ј on the membrane. It is not surprising then that negatively charged phospholipids disrupt its movements and these two steps in the catalytic cycle. As another possibility, arginine/lysine residues (including the above critical Arg-324) form salt bridges with the first-layer phospholipids (20), and PS may enhance such interaction and disrupt proper orientation and conformation of the Ca 2ϩ -ATPase molecule.
Ca 2ϩ , at high concentrations, binds to PS and displaces water molecules, causing dehydration of the serine headgroup, which could potentially affect activity. However, we performed ATPase activity measurement and Ca 2ϩ -affinity studies at free Ca 2ϩ concentration 0.01-100 M, too low to see such effects; only partial dehydration is observed at 3 mM Ca 2ϩ (28). Kinetic analyses of both EP transition and Ca 2ϩ -off time course are followed upon the removal of Ca 2ϩ by a large excess of EGTA, avoiding an effect due to dehydration of the serine headgroup.
Finally, whereas the N, P, and A domains are somewhat above the membrane, it is not unreasonable that membrane electrostatic effects could extend to influencing electrostaticsensitive movements here. We have found (30) that long-range electrostatic interactions between specific polar residues on the N domain and P domain play a critical role in the EP transition, probably by helping to open the N and P domain junction to allow docking of the A domain with the P domain. Actually, perturbation of these long-range electrostatic N-P domain interactions by site-specific mutations (reducing the slope in the log(rate) versus ␥ Ϯ 2 plot) retards the EP transition.

Materials
The pMSP1D1 plasmid (Addgene plasmid 20061) was purchased from Addgene. Non-ionic detergent octaethylene glycol monododecyl ether (C 12 E 8 ) was purchased from Tokyo Chemical Industry Co. LTD (Tokyo, Japan). POPS was purchased from Avanti Polar Lipids, and other phospholipids were from NOF Corp. (Tokyo, Japan). Lipids were dissolved in buffer containing 10 mM Tris/HCl (pH 7.5) and 100 mM (54 mg/ml) C 12 E 8 and stored at Ϫ80°C.

Expression and purification of nanodisc scaffold protein MSP1D1
An expression vector containing the MSP1D1 gene construct was transformed into E. coli BL21-Gold(DE3) cells (26). E. coli cell cultures were induced with 1 mM isopropyl ␤-D-1thiogalactopyranoside and cultured for 3 h at 37°C. The cells were harvested by centrifugation and stored at Ϫ20°C. The E. coli cell pellets were resuspended in a lysis buffer (20 mM sodium phosphate (pH 7.4), 1% Triton X-100, and 1 mM phenylmethylsulfonyl fluoride) and sonicated. The cell lysate was cleared by centrifugation and was loaded onto His60 Ni Superflow Resin column (Clontech). The column was washed sequentially with the following buffers:

Preparation of SR vesicles
SR vesicles were prepared from rabbit skeletal muscle as described (44,45); all of the methods were carried out in accordance with institutional laws and regulations of the Asahikawa Medical University, and the experimental protocols were approved by the Animal Experimentation Ethics Committee of the Asahikawa Medical University (license 16006).

SERCA1a purification
SR vesicles (2 mg of protein/ml) were solubilized with 10 mg/ml C 12 E 8 in 10 mM Tris/HCl (pH 7.5) and 10 mM CaCl 2 . Subsequently, SERCA1a was allowed to bind to Reactive Red 120 resin column (Sigma-Aldrich). The flow-through fraction of this step, which contains lipids from SR vesicles, was kept and used as SRL solution for reconstitution. The column was washed by 2 column volumes of buffer (10 mg/ml C 12 E 8 , 20 mM Tris/HCl (pH 7.5), and 10 mM CaCl 2 ). The SERCA1a protein was eluted in 2 column volumes of an elution buffer containing 10 mg/ml C 12 E 8 , 20 mM Tris/HCl (pH 7.5), 10 mM CaCl 2 , and 0.5 M NaCl. The SERCA1a sample thus purified was desalted and concentrated by a centrifugation with Vivaspin (10,000 molecular weight cutoff; GE Healthcare).

Reconstitution of purified SERCA1a in lipid bilayer nanodisc
Purified SERCA1a protein (0.5 mg/ml) was incubated with 20 M MSP1D1 in 1.1 mM phospholipid (POPC, POPE, POPG, or POPS or their mixture), 10 mM CaCl 2 , 20 mM Tris/HCl (pH 7.5), and 10 mg/ml C 12 E 8 on ice for 30 min. For reconstitution of SERCA1a in nanodisc with lipids from SR membrane, SERCA1a was incubated with MSP1D1 in SRL solution obtained as above. Then to remove C 12 E 8 , the mixture was incubated with 0.4 mg of BioBeads SM2 (Bio-Rad)/ml at 4°C for 4 h with gentle agitation. The beads and aggregated proteins were removed by centrifugation (10,000 ϫ g, 4 min, at 4°C), and the supernatant was passed through a filter (0.45 m, cellulose acetate). The nanodisc thus constructed as a protein/lipid mixture was subjected to size-exclusion column chromatography

Effects of lipid headgroups on Ca 2؉ -ATPase
(ENrich SEC 650 10 ϫ 300 mm; Bio-Rad) in 10 mM MOPS/Tris (pH 7.0), 0.1 M KCl, and 0.1 mM CaCl 2 as an elution buffer with a flow rate of 1 ml/min. The position of nanodisc containing SERCA1a was detected by phosphoenzyme formation activity. For constructing empty nanodiscs, MSP1D1 and lipids were incubated in 1.3 mM phospholipids, 20 mM Tris/HCl (pH 7.5), 10 mg/ml C 12 E 8 , and 20 M MSP1D1 and treated with Biobeads and purified otherwise as described above. The SERCA1a-containing nanodiscs and empty nanodiscs thus obtained were subjected to centrifugation with Vivaspin (10,000 molecular weight cutoff; GE Healthcare) to change the solution to 0.3 M sucrose, 0.1 M KCl, 0.1 mM CaCl 2 , and MOPS/Tris (pH 7.0) and to concentrate. The samples were stored at Ϫ80°C after flashfreezing in liquid nitrogen. The molar ratio of the SERCA1a to MSP1D1 in the samples was analyzed by quantitative SDS-PAGE using 12.5% polyacrylamide gels and ImageJ software. The content of phosphorylation sites was determined essentially according to Barrabin et al. (45).

Polyacrylamide gel electrophoresis
Native PAGE was performed at 100 V with 5% polyacrylamide gels in 50 mM BisTris/HCl (pH 7.0), 0.1 M 6-aminocaproic acid, and 1 mM EGTA with running buffer 50 mM BisTris/ HCl (pH 7.0) and 1 mM EGTA. The sample (3-5 g of protein) in sample buffer 4% glycerol, 0.1 M 6-aminocaproic acid, 10 mM BisTris/HCl (pH 7.0), and 0.01% bromphenol blue was applied. For quantification of SERCA1a and MSP1D1 protein contents, Laemmli SDS-PAGE (46) was performed with 12.5% polyacrylamide gels, and the SERCA1a protein purified by a deoxycholate treatment of SR vesicles (47) and the above purified MSP1D1 protein were applied on the same gel as standards. The gels were stained by Coomassie Brilliant Blue R-250, and the densitometric analysis and determination of R f value were performed by ImageJ software.

Ca 2؉ -ATPase activity
The rate of ATP hydrolysis was determined with A23187 or C 12 E 8 at 25°C in a mixture containing 10 g/ml protein, 1 mM [␥-32 P]ATP, 0.1 M KCl, 7 mM MgCl 2 , 50 mM MOPS/Tris (pH 7.0), and 0.1 mM CaCl 2 , or 2 mM EGTA without added CaCl 2 otherwise as described in the figure legends. The reaction was terminated with 0.1 M HCl, and the amount of 32 P i released from [␥-32 P]ATP was quantified with digital autoradiography. The Ca 2ϩ -ATPase activity was obtained by subtracting the Ca 2ϩ -independent ATPase activity determined in 2 mM EGTA without added CaCl 2 , otherwise as above. The turnover rate was calculated with the content of the phosphorylation site in each sample.

EP formation
SERCA1a phosphorylation was performed with 10 M [␥-32 P]ATP for 10 s at 0°C in MOPS/Tris (pH 7.0), 7 mM MgCl 2 , 10 M CaCl 2 , and 0.1 M KCl, otherwise as described in the figure legends. The total amount of EP and the fraction of E2P were determined with digital autoradiography after separation by 5% SDS-PAGE at pH 6.0 according to Weber and Osborn (48) as described previously (49).

Transmission electron microscopy
Samples were negatively stained using 2% (w/v) uranyl acetate on the carbon-coated copper grid (400 mesh). Images were acquired on a JEM-1010 electron microscope (JEOL) operated at 100 kV with a cMOS camera (TemCamF416, TVIPS) and a nominal magnification of ϫ40,000. The Feret diameter was determined with ImageJ software.

Miscellaneous methods
Protein concentrations were determined by the method of Lowry et al. (50) or absorbance at 280 nm. Data were analyzed by nonlinear regression using the program Origin (Microcal Software, Inc.).