Glycine 105 as Pivot for a Critical Knee-like Joint between Cytoplasmic and Transmembrane Segments of the Second Transmembrane Helix in Ca2+-ATPase*

The cytoplasmic actuator domain of the sarco(endo)plasmic reticulum Ca2+-ATPase undergoes large rotational movements that influence the distant transmembrane transport sites, and a long second transmembrane helix (M2) connected with this domain plays critical roles in transmitting motions between the cytoplasmic catalytic domains and transport sites. Here we explore possible structural roles of Gly105 between the cytoplasmic (M2c) and transmembrane (M2m) segments of M2 by introducing mutations that limit/increase conformational freedom. Alanine substitution G105A markedly retards isomerization of the phosphoenzyme intermediate (E1PCa2 → E2PCa2 → E2P + 2Ca2+), and disrupts Ca2+ occlusion in E1PCa2 and E2PCa2 at the transport sites uncoupling ATP hydrolysis and Ca2+ transport. In contrast, this substitution accelerates the ATPase activation (E2 → E1Ca2). Introducing a glycine by substituting another residue on M2 in the G105A mutant (i.e. “G-shift substitution”) identifies the glycine positions required for proper Ca2+ handling and kinetics in each step. All wild-type kinetic properties, including coupled transport, are fully restored in the G-shift substitution at position 112 (G105A/A112G) located on the same side of the M2c helix as Gly105 facing M4/phosphorylation domain. Results demonstrate that Gly105 functions as a flexible knee-like joint during the Ca2+ transport cycle, so that cytoplasmic domain motions can bend and strain M2 in the correct direction or straighten the helix for proper gating and coupling of Ca2+ transport and ATP hydrolysis.

Sarco(endo)plasmic reticulum Ca 2ϩ -ATPase (SERCA1a), 2 a representative member of P-type ion-transporting ATPases, catalyzes Ca 2ϩ transport coupled with ATP hydrolysis (Fig. 1A) (for recent reviews, see Refs. [1][2][3]. The enzyme is activated by the binding of two Ca 2ϩ ions to high affinity transport sites facing the cytoplasmic side (E2 to E1Ca 2 in Fig. 1) and autophos-phorylated at Asp 351 with MgATP to form an ADP-sensitive phosphoenzyme (E1P), which reacts with ADP to regenerate ATP in the reverse reaction. Upon E1P formation, the two bound Ca 2ϩ are occluded in the transport sites (E1PCa 2 ). The subsequent isomeric transition to the ADP-insensitive E2P form results in rearrangements of the Ca 2ϩ binding sites to deocclude Ca 2ϩ , open the release path, and reduce the affinity, thus releasing Ca 2ϩ into the lumen. Finally, the Asp 351 -acylphosphate in E2P is hydrolyzed to form a Ca 2ϩ -free inactive E2 state.
In the transport cycle, the three cytoplasmic domains N, P, and A undergo large movements and change their organizational state, a repositioning that is coupled to rearrangements in transmembrane helices and thereby changes in the transport sites (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15). Most remarkable is the motion of the A domain, which functions in cytoplasmic and luminal gating to regulate Ca 2ϩ binding and release, as well as E2P hydrolysis. The long helix M2 connects directly with the A domain at their junction (A/M2-junction) and moves largely together with the A domain and also changes its secondary structure, unwinding/rewinding with consequent length changes, during the Ca 2ϩ transport cycle (Fig. 1A). The long helix structure of M2 and the A domain motions are common features in P-type ion-transporting ATPases (16 -20).
We have demonstrated by extensive mutations throughout M2, disrupting and elongating the helix with glycine insertions (21), that its transmembrane part (M2m), cytoplasmic part (M2c), and junctional region with the A domain (M2top) (Fig.  1B) play different roles in the various catalytic steps, a separation of function apparently needed to control gating at the transport sites and thereby coupling Ca 2ϩ transport with ATP hydrolysis. M2 has a role distinct from that of the A/M1Ј-linker loop (22)(23)(24), although both are needed to coordinate coupling.
There is a glycine residue, Gly 105 , in the region connecting M2c and M2m. A glycine residue in an ␣-helix is known to break the helical structure and provide conformational freedom (25), and bending at Gly 105 may possibly translate M2c and M2m motions between membrane helices and the cytoplasmic domains, expediting the structural functions of M2. Actually, according to crystal structures, M2 moves, bends, unwinds, and possibly registers tension during the A domain motions in the transport cycle (Fig. 1). In this study, we explore possible structural effects and conformational freedom provided by the glycine residue at the M2c and M2m junction by introducing mutations, which are intended to 1) reduce conformational freedom (G105A) and distort helix structure (G105P, V104P, and V106P), 2) increase flexibility (single G-substitution of each residue and GG-substitution of successive residue pairs), and 3) evaluate the positional importance of the glycine (single G-shift substitutions in the G105A mutant). The results demonstrate that Gly 105 is crucial for Ca 2ϩ occlusion in EP and rapid EP isomerization and, therefore, for rapid coupled ATP hydrolysis. Surprisingly, a G-shift substitution at the 7th residue from Gly 105 located on the same face of the helix (G105A/A112G) fully restores wild-type kinetic properties and function. The glycine evidently functions as a flexible joint or hinge for M2-mediated coupling between catalytic events and gating during Ca 2ϩ transport, allowing directionally oriented bending and absorbing strain when needed. Its important role in pumping is underscored by the fact that a glycine is conserved in this region in P-type ATPases. Fig. 2, we first determined the total amount of EP (EP total , sum of E1P and E2P) and fraction of E2P at steady state for wild type and all mutants in 0.1 M K ϩ , which strongly accelerates E2P hydrolysis and therefore suppresses its accumulation at steady state in the wild type (26). The EP total is Ͼ50 pmol/mg of microsomal protein in the wild type and all of the mutants and thus is sufficient to perform Ϫ ⅐ADP as the E1ϳP⅐ADP⅐Ca 2 ‡ analog (PDB entry 1T5T (10)), E2⅐BeF 3 Ϫ as the E2P ground state analog (7) (PDB 2ZBE (11)), E2⅐AlF 4 Ϫ (TG) as the transition state (E2ϳP ‡ ) analog for E2P hydrolysis (7) (PDB 2ZBG (11)), E2(TG) as the E2 state fixed with thapsigargin (PDB 1IWO (6)), E1Mg 2ϩ (PDB 3W5A (14)), and E1Ca 2 (PDB 1SU4 (4)). The structures are aligned with the static M8 -M10 helices. The nucleotide binding (N), phosphorylation (P), and actuator (A) domains and M1-M6 helices are colored as indicated. The approximate position of membrane is shown by light green lines. The binding sites for two Ca 2ϩ (purple spheres) consist of residues on M4, M5, M6, and M8. The yellow, pink, and red arrows indicate the approximate motions of the A and P domains and M2 (depicted in red), respectively, to the next structural state during the EP processing (isomerization and hydrolysis) and the E2 3 E1 transition. The light blue arrow indicates the unwound M2 part. B, the ␣-carbon of residues on M2 mutated in this study is indicated in the crystal structure E1Ca 2 ⅐AlF 4 Ϫ ⅐ADP. Ca 2ϩ ligand Glu 309 , which is occluding Ca 2ϩ as a closed cytoplasmic gate, and Leu 65 on M1, which is fixing the Glu 309 side chain configuration for the Ca 2ϩ occlusion, are also depicted. The position of Gly 105 is indicated by a pink triangle. functional analyses. In wild type and all mutants, E1P accumulates almost exclusively.

EP Formation at Steady State-In
ATP Hydrolysis, Ca 2ϩ Transport, and Coupling-In Fig. 3A, the Ca 2ϩ -ATPase activity and oxalate-dependent Ca 2ϩ transport activity were determined at a saturating 10 M Ca 2ϩ during the initial linear part of P i liberation and Ca 2ϩ transport (inset) with the inclusion of 5 mM oxalate (to trap transported Ca 2ϩ in the lumen), and mutants were compared with wildtype activities. The activities were markedly reduced by alanine or proline substitutions of Gly 105 (G105A or G105P) and also proline substitutions of Val 104 and Val 106 (V104P and V106P) at the M2c-M2m connecting region. The activities were also largely reduced by glycine (G) or two-glycine (GG) substitution at the Ile 97 -Leu 98 region on M2m and the Gln 108 -Ala 115 region on M2c but considerably accelerated by substitutions at the Asn 101 -Val 104 region immediately adjacent to Gly 105 .
In Fig. 3B, the ratio of Ca 2ϩ transport activity to ATPase activity (Ca 2ϩ /ATP) of each mutant relative to the wild type is shown. The substitutions G105A, G105P, V104P (but not V104G), V106G, V106P, and W107G around Gly 105 markedly reduced Ca 2ϩ /ATP, indicating severe uncoupling (i.e. almost no Ca 2ϩ transport despite fair ATP hydrolysis). The G-or GGsubstitutions at the Leu 96 -Val 104 and Gln 108 -Ala 115 regions did not cause such uncoupling. Interestingly, the single G-substitutions V106G and W107G significantly reduced the transport activity and thus the Ca 2ϩ /ATP ratio, but the GG-substitution of these residues (V106G/W107G, resulting in three successive glycines GGG 107 and accompanying conformational freedom to the helical wheel at the M2c-M2m connecting region (Fig. 3D)) somehow restored the wild-type activities and coupled Ca 2ϩ transport.
We then made "G-shift substitutions," in which the G105A substitution was combined with a glycine substitution of another residue on the M2 helix, to examine whether shifting the position of the glycine residue can restore function. Surprisingly, only one G-shift substitution, G105A/A112G, within the Leu 96 -Ala 115 wide region restored wild-type ATPase activity and coupled Ca 2ϩ transport, both of which were disrupted by the G105A substitution. Residues 105 and 112 are positioned on the same side of the M2 helical wheel in E1Ca 2 ϳPϳADP ‡ , E2P ground state, and E2ϳP ‡ , facing M4C, the cytoplasmic part of M4 (Figs. 3 (C and D), 5, and 8). Results demonstrate the critical importance of the glycine position on the M2 helix, both in terms of sidedness and distance from Gly 105 , for structural communication between the cytoplasmic catalytic domains and transmembrane transport sites.
E1P 3 E2P Isomerization-We then analyzed each of the steps in the Ca 2ϩ transport cycle. In Fig. 4, the rate of EP isomerization E1P 3 E2P (i.e. the loss of the ADP sensitivity at the catalytic site) was determined in the presence of K ϩ under conditions essentially the same as those for EP formation in Fig.  2. The isomerization rate is strongly reduced in the G105A and G105P substitutions, but it is comparable with or even higher than that of the wild type in the G-or GG-substitution mutants of other residues, including Ala 112 , and in the V104P and V106P mutants. Results show the critical importance of Gly 105 for rapid EP isomerization, which involves a large A-domain rotation and association with the P domain. Interestingly, the GGsubstitution at Ile 103 -Val 104 (I103G/I104G) accelerates EP isomerization 10-fold; thus, the introduction of three successive glycines and the increase in flexibility of the helix here G-shift substitutions to mutant G105A restored (G105A/ I97G, G105A/N101G, G105A/Q108G, G105A/A112G, and G105A/A115G) or even accelerated (G105A/V104G) the suppressed EP isomerization rate (Fig. 4A). It is notable that all of the glycines introduced in these restoring G-shift substitutions are situated on the same side of the M2 helical wheel as Gly 105 (Fig. 4, B and C, green), suggesting a directional structural change in the helix, such as in a knee-bending motion.
E2P Hydrolysis-In Fig. 5, we determined the E2P hydrolysis rate by first phosphorylating the enzyme with 32 P i in the absence of Ca 2ϩ and K ϩ and the presence of 30% (v/v) Me 2 SO, which strongly favors E2P formation in the reverse reaction (27), followed by diluting the phosphorylated protein with a large volume of nonradioactive P i and K ϩ without Ca 2ϩ . The hydrolysis rate was not significantly inhibited by the alanine or proline substitutions of Gly 105 , Val 104 , or Val 106 ; nor was it inhibited by G-or GG-substitutions of all of the residues in the entire Leu 96 -Ala 115 region. Rather, the hydrolysis was accelerated with substitutions L98G on M2m and Q108G, A112G, E113G, and A115G on M2c. Consistently, in the structural change that takes place during E2⅐BeF 3 G-shift substitutions at Val 104 and Val 106 next to Gly 105 (i.e. G105A/V104G and G105A/V106G (but not the G-substitutions V104G and V106G possessing Gly 105 )) largely retarded the hydrolysis, suggesting that the required flexing is strictly in a particular direction.
The E2P ground state possesses luminally open low affinity Ca 2ϩ transport sites with K d of ϳ1 mM (23, 28 -34), and in E2P 3 E2ϳP ‡ during E2P hydrolysis, the luminal gate is tightly closed, which prevents luminal Ca 2ϩ access and possible Ca 2ϩ leakage (7). Actually, E2P hydrolysis is inhibited by a high concentration of luminal Ca 2ϩ due to Ca 2ϩ binding at the luminally open transport sites in E2P. We determined the E2P hydrolysis rate of the G105A mutant in the presence of 3 and 20 mM luminal Ca 2ϩ otherwise, as in Fig. 5A, and found that the hydrolysis is markedly retarded by the luminal Ca 2ϩ in the G105A mutant as in the wild type. Actually, the hydrolysis rates (s Ϫ1 ) in the presence of 0, 3, and 20 mM Ca 2ϩ are 0.587 Ϯ 0.078 (n ϭ 6), 0.006 Ϯ 0.005 (n ϭ 3), and 0.003 Ϯ 0.003 (n ϭ 3), respectively in the wild type and 0.501 Ϯ 0.112 (n ϭ 3), 0.018 Ϯ 0.004 (n ϭ 3), and 0.007 Ϯ 0.001 (n ϭ 3), respectively in the G105A mutant. Thus, luminal Ca 2ϩ access to the transport sites in the E2P ground state in G105A is as in the wild type, and evidently gating here is not impaired.
E2 3 E1 Transition-For ATPase activation E2 3 E1 3 E1Ca 2 , the enzyme is isomerized first to a transient E1 state to gain an open cytoplasmic gate and high Ca 2ϩ affinity at the FIGURE 3. Ca 2؉ -ATPase activity and oxalate-dependent Ca 2؉ transport activity. A, the activities of the expressed SERCA1a mutants were determined as described under "Experimental Procedures" and shown as the values relative to the respective wild-type activities (ATP hydrolysis, 47.8 Ϯ 1.8 nmol P i /min/mg microsomal protein (n ϭ 5); oxalate-dependent Ca 2ϩ transport, 12.7 Ϯ 0.5 nmol Ca 2ϩ /min/mg microsomal protein (n ϭ 5)). Typical time courses of P i liberation and Ca 2ϩ accumulation in the wild type and mutant G105A are shown in the inset. B, the ratio, Ca 2ϩ transport activity per Ca 2ϩ -ATPase activity (Ca 2ϩ /ATP) is shown as the percentage of the wild-type ratio. The wild type (light gray), G105A and G-substitution mutant (open), G-shift substitution mutant (closed), and GG-substitution mutant (lattice) are shown as indicated. C and D, M2 is viewed from a direction parallel to the membrane plane (C) or from the position indicated by the red arrow in C (D) in the crystal structure as indicated (E1Ca 2 ⅐AlF 4 Ϫ ⅐ADP). The effect of G-shift substitution of residues on Ca 2ϩ /ATP in B is visualized with ␣-carbon coloring as follows. Green, wild type (Gly 105 )-like coupled transport; red, Ͻ36% of wild type (severely uncoupled). Error bars, S.D. transport sites and then binds two Ca 2ϩ ions. The E2 3 E1 transition rate can be assessed by measuring the rate of E1PCa 2 formation from the Ca 2ϩ -deprived E2 state following the addition of Ca 2ϩ plus ATP. The assay takes advantage of the fact that subsequent steps (Ca 2ϩ binding, ATP binding, and phosphorylation) are relatively fast (21). In Fig. 6A, the rates of E1PCa 2 formation from the Ca 2ϩ -deprived E2 state and from the Ca 2ϩ -bound activated E1Ca 2 state were determined, and the ratio of the two rates is shown in Fig. 6B.
The rate of E1PCa 2 formation from the E1Ca 2 state was hardly affected by the mutations, but that from the E2 state was mostly increased or reduced, depending on the mutations, as compared with the wild type. In the alanine substitution G105A, the E2 3 E1 transition was faster, and the ratio increased 3-fold (Fig. 6B). In contrast, in the proline substitution G105P as well as V104P and V106P, the transition was markedly retarded, and the ratio was accordingly reduced.
The G-substitution of Ile 97 and Leu 98 on M2m and of Gln 108 , Ala 112 , and Ala 115 on M2c markedly reduced the E2 3 E1 transition rate and ratio (Fig. 6, C and D, asterisks). The G-shift substitutions of these residues, G105A/L98G, G105A/Q108G, G105A/A112G, and G105A/A115G, restored the wild-type rate and ratio by reducing the increased rate and ratio of the G105A mutant (except for G105A/I97G) (Fig. 6, C and D, green). These residues are again on one side of the M2 helical wheel facing M4C/M6, suggesting the importance of a directional motion of M2 and consequent formation of a more rigid integral helix structure in the E2 3 E1 structural transition.
Ca 2ϩ Occlusion in E1PCa 2 -For coupling Ca 2ϩ transport with ATP hydrolysis, the Ca 2ϩ occlusion in E1PCa 2 and subsequent deocclusion with luminal gate opening during E1PCa 2 3 E2PCa 2 3 E2P ϩ 2Ca 2ϩ are key processes. To examine a possible cause of the uncoupling of mutants, V104P, G105A, V106G, V106P, W107G, and the G-shift mutants having the G105A substitution (except for the coupled G105A/E112G) (cf. Fig. 3), the Ca 2ϩ occlusion in E1PCa 2 was assessed in Fig. 7 under the conditions in Fig. 2, in which all of the mutants as well as the wild type accumulate mostly E1P at steady state. In the experiments, E1P was formed by ATP with the Ca 2ϩ -activated enzyme in the presence of 45 Ca 2ϩ , and then unbound and unoccluded 45 Ca 2ϩ were washed out with membrane filtration. By this method, two Ca 2ϩ ions were determined to be occluded in E1PCa 2 in wild type.
In the mutants G105A, G105P, V104P, and V106P, the Ca 2ϩ occlusion in E1P was largely reduced and thus defective, showing that the Ca 2ϩ in E1PCa 2 was not occluded. Because Ca 2ϩ that escaped or transported into the lumen would be trapped by oxalate present in the transport assay system, the observed uncoupling in these mutants (Fig. 2) indicates that Ca 2ϩ escape is not toward the lumen but to the cytoplasmic side. All of the G-or GG-substitution mutants occluded two Ca 2ϩ ions in E1PCa 2 as in the wild type, in agreement with their coupled FIGURE 4. EP isomerization. A, microsomes expressing wild type or mutant were first phosphorylated with [␥-32 P]ATP at 0°C for 30 s as in Fig. 2. The phosphorylation reaction was terminated at zero time by Ca 2ϩ removal by the addition of an equal volume of a buffer containing 10 mM EGTA, 0.1 M KCl, 7 mM MgCl 2 , and 50 mM MOPS/Tris (pH 7.0) at 0°C, and the amounts of EP were determined at the indicated times. Note that the wild type and all of the mutants accumulate mostly E1P at steady state (Fig. 2); therefore, the rate-limiting E1P to E2P isomerization rate was determined by EP decay kinetics. The inset shows typical examples; solid lines show the least squares fit to a single exponential, and the E1P decay rates thus obtained are shown in the main panel. Bars are colored as in Fig. 3. B and C, the effects of G-shift substitution of residues in A are visualized with ␣-carbon coloring on M2. Green, rate Ͼ0.060 s Ϫ1 , which is comparable with or higher than the wild-type (Gly 105 ) rate, 0.078 Ϯ 0.015 (n ϭ 4) s Ϫ1 ); red, rate Ͻ0.025 s Ϫ1 (i.e. marked retardation). M2 is viewed as in Fig. 3, C and D. Error bars, S.D.
Ca 2ϩ transport, but an exception was found with the mutants V106G and W107G, which showed uncoupling despite Ca 2ϩ occlusion in E1PCa 2 (for more analysis, see below).
Approximately two Ca 2ϩ ions were occluded in the wild type and control 4Gi-46/47 mutant as E1PCa 2 and E2PCa 2 , respectively, as found previously (23,24). In the A/M1Ј-linker elongated V106G and W105G mutants and all of the G-shift mutants except for G105A/A112Gϩ4Gi-46/47, Ca 2ϩ occlusion in E2PCa 2 was markedly reduced and thus defective, in contrast to their Ca 2ϩ occlusion in E1PCa 2 . This finding agrees with the view that in these mutants, the occluded Ca 2ϩ in E1PCa 2 escapes to the cytoplasmic side during EP isomerization E1PCa 2 3 E2PCa 2 , causing uncoupling.
Structural State of Trapped E2PCa 2 State Revealed by Proteolytic Analysis-In the EP isomerization and Ca 2ϩ release E1PCa 2 3 E2PCa 2 3 E2P ϩ 2Ca 2ϩ , the A domain largely rotates and docks on the P domain (E1PCa 2 3 E2PCa 2 ), causing loss of ADP sensitivity, and then the associated A and P domains are pulled and inclined by the A/M1Ј-linker, thereby causing deocclusion and release of bound Ca 2ϩ into the lumen (E2PCa 2 3 E2P ϩ 2Ca 2ϩ ) (23,24). These structural changes can be monitored by changes in the availability of specific cleavage sites for trypsin and proteinase K (prtK) (5,7,23,35).
In the two top panels (trypsin) of Fig. 9, trypsin proteolysis was performed, and the ATPase chain and its fragments were probed with a monoclonal antibody that recognizes Ala 199 -Arg 505 (the tryptic fragment A1) of SERCA1a. In the wild type and all of the mutants, the T1 site (Arg 505 ) on the outermost loop of the N domain is very rapidly cleaved to produce the fragment A (Met 1 -Arg 505 , probed by the antibody) and the fragment B (Ala 506 to the C terminus Gly 994 , not probed).
In E1PCa 2 , which accumulates exclusively in the wild type, the T2 site Arg 198 on the outermost Val 200 loop (Asp 196 -Asp 203 ) of the A domain is rapidly cleaved to produce the fragments A1 (probed) and A2 (Met 1 -Arg 198 , not probed). By contrast, in E2PCa 2 , exclusively accumulated with the elongated A/M1Ј-linker control mutant (4Gi-46/47) and all of the ϩ4Gi-46/47 mutants on M2, the A1 fragment band is very faint and extremely slow to make its appearance. Thus, the control mutant 4Gi-46/47 and all of the ϩ4Gi-46/47 mutants possess the characteristic property of the ADP-insensitive EP (E2PCa 2 as well as E2P); namely, the A domain has largely rotated from its position in E1PCa 2 and associated with the P domain at the Val 200 loop, including Arg 198 , thereby causing the loss of the ADP sensitivity and blocking sterically the tryptic attack (5,7,23,35).
In the two bottom panels (prtK) of Fig. 9, the same set of experiments was performed with prtK. As demonstrated previously (23,24), in the E2PCa 2 state trapped by the A/M1Ј-linker elongation, prtK cleavage occurs at Leu 119 on M2, producing the fragment p95 (see the control mutant 4Gi-46/47), in contrast to its nearly complete resistance in E1PCa 2 (see the wild type E1PCa 2 ) as well as in the Ca 2ϩ -released E2P, as demonstrated previously (23).
In all of the ϩ4Gi-46/47 mutants on M2 with the elongated A/M1Ј-linker, the accumulated E2P species is cleaved at Leu 119 , producing the p95 fragment, as in the control elongation mutant 4Gi-46/47 (i.e. there is no indication of a large change by these substitutions on M2 in the overall structure of the for both Ca 2ϩ -bound and Ca 2ϩ -unbound states in the above preincubation), otherwise as above, was added at 0°C, and the EP formation time course was followed. Typical examples are shown with the wild type and the mutant G105A in the inset. Solid lines, least squares fit to a single exponential; the rates thus determined are shown in the main panel. The Ca 2ϩ -unbound state was denoted as "E2" for simplicity, and the ratio of the two rates is shown in B. Bars are colored as in Fig. 3. C and D, residues of which G-substitution severely retarded the E2 3 E1 transition rate are highlighted (*). The effects of G-shift substitution of residues on the ratio (i.e. on the rate-limiting E2 3 E1 transition in B) are visualized with ␣-carbon coloring on M2 in the E1Mg 2ϩ structure. Green, ratio 28 -37%, comparable with the wild type 31.4 Ϯ 2.4% (n ϭ 4); red, 66 -93% (i.e. marked acceleration); blue, Ͻ3% (i.e. marked retardation). Error bars, S.D.
trapped E2PCa 2 species). Thus, all of the ϩ4Gi-46/47 mutants accumulate E2PCa 2 with its characteristic structure found in the control elongation mutant (4Gi-46/47). The defect of Ca 2ϩ occlusion in E2PCa 2 caused by the mutations on M2 is not due to a large structural effect on the E2P species but can be ascribed to some specific effects on the cytoplasmic Ca 2ϩ gate, Glu 309 .

Discussion
Gly 105 on M2 Functions as a Flexible Joint-Glycine is a typical helix breaker and as part of an ␣-helix gives conformational freedom to the structure, such that bending and loosening, with or without unwinding and elongation, become possible. A typical example exists in the Ca 2ϩ -ATPase at Gly 770 on M5 at the transport sites, where there is a pivoting point for a tilting motion essential for controlling the high affinity Ca 2ϩ binding sites in E1Ca 2 7 E2 ϩ 2Ca 2ϩ (6). Our results point to the well conserved Gly 105 on M2 playing a similar role, although none of the crystal structures of the catalytic intermediates or analogs thereof shows destabilization here, except for a slight bend in Ca 2ϩ -occluded E1Ca 2 ⅐AlF 4 Ϫ ⅐ADP (Fig. 1). M2 exists as a long straight helix in E1Mg 2ϩ and E1Ca 2 , becomes transiently unwound at M2top in E2P, and becomes unwound at Asn 111 -Ala 115 in the middle of M2c above Gly 105 in E2ϳP ‡ and E2 (Fig. 1).
It comes as a bit of a surprise then that mutation G105A has a profound effect on the occlusion of Ca 2ϩ on phosphorylation of the pump by ATP, causing uncoupling of Ca 2ϩ transport from ATP hydrolysis. The defect occurs in both E1PCa 2 and the more transient E2PCa 2 . Evidently, the cytoplasmic gate rendered by the Ca 2ϩ -coordinating residue Glu 309 and normally fixed by residues on M1 (particularly Leu 65 ) has been loosened to allow Ca 2ϩ to escape to the cytoplasmic side. Interestingly, the defect is corrected by a second mutation two turns up on the M2 helix at Ala 112 . Thus, pump G105A/A112G has wild typelike activity and coupling.
A second significant finding is that the E2 3 E1 structural transition is accelerated by mutation G105A, in contrast to the marked retardation of the EP isomerization, and both wild-type rates are recovered by the G-shift mutation at Ala 112 . The inhibition of ATPase activity can be ascribed to the retardation of either EP isomerization (e.g. G105A and G105P) or E2 3 E1 transition (e.g. G105P, but not G105A) or both. We will argue through analysis of our mutations in this region and the crystal structures that Gly 105 is actually a hinge point, where conformational flexibility in a particular direction (a knee-like bending, possibly with elongation) is paramount for occlusion but where excessive flexing is largely inhibitory and less freedom beneficial for straightening of M2 to gain the E1 state.
E2 3 E1 Transition-Examining the E2 3 E1 transition first, where mutation G105A facilitates a rapid E2 3 E1 structural transition, whereas forcing helix disruption by G105P, V104P, and V106P and increased flexibility by G-substitutions of resi-  ϭ 10 -20)). Bars are colored as in Fig. 3. B and C, the effect of G-shift substitution of residues on the Ca 2ϩ occluded/EP total in A are visualized with ␣-carbon coloring on M2 in the E1Ca 2 ⅐AlF 4 Ϫ ⅐ADP structure. Green, 1.7-1.9 (i.e. occlusion as in the wild type (Gly 105 )); red, Ͻ0.9 (i.e. severe disruption of the occlusion). Error bars, S.D.
dues located on the same side of M2 as Gly 105 facing M4C/M6 are inhibitory, and where these same G-shift substitutions restore the wild type rate from the accelerated one in G105A (Fig. 6), it is apparent that the intermediate properties of the single wild-type glycine control the structure here such that the helix is neither too bent and flexible nor too rigid.
In the crystal structure E2(TG), the M2 helix is unwound at Asn 111 -Ala 115 and in the transition to E1Mg 2ϩ moves toward M4C/M6 and straightens, a rigidity stabilized by interactions of M2c/M2m with M4C/M6 (cf. Fig. 1). The marked retardation of E2 3 E1 transition by the GG-substitutions at the Gln 108 -Ala 115 region on M2c is consistent with these changes of M2. How then is Gly 105 influencing this step; does the residue bear on the rewinding higher up or perhaps on the interactions with M4 and M6? We speculate that M2 is actually bent at Gly 105 in the E2 structure of wild type (not seen in the crystal structure of E2(TG) but possible in the physiological "flexible" E2 assisted by unoccupied Ca 2ϩ transport sites and no interaction with M4/M6), which would fit with the unwinding higher up being due to a pulling strain imposed by A domain movements, and that mutation G105A helps reverse this tug and facilitate helix straightening, thus favoring the rapid E2 3 E1 transition.
The marked reduction of the E2 3 E1 transition rate by G-substitutions L98G, Q108G, A112G, and A115G facing M4C/M6 (Fig. 6, B and C, asterisk) is in contrast to the acceleration of E2P hydrolysis by these substitutions (Fig. 5; see below) but fits very well with the reverse structural change that occurs during the hydrolysis; namely, M2 unwinds at M2c, bends, and moves away from M4C/M6, the opposite, as we have discussed, of what occurs during E2 3 E1.
EP Isomerization E1PCa 2 3 E2PCa 2 -Phosphorylation of E1Ca 2 is hardly affected by any of the mutations. In contrast, the next step, the EP isomerization, is markedly retarded by substitutions G105A and G105P, an inhibition reversed by G-shift substitutions of residues facing M4C (Fig. 5). The G-or GG-substitutions of all other residues as well as V104P and V106P, all possessing Gly 105 , actually accelerate the isomerization. Thus, flexibility of M2 in one direction (and possibly a  Fig. 2. B, microsomes expressing wild type or mutants were first phosphorylated with [␥-32 P]ATP at 0°C for 30 s as in A. Then the EP decay rate was determined and shown as in Fig. 4. Note that the A/M1Ј-linker elongation causes almost exclusive E2P accumulation (A) and almost completely blocks its decay (B). C, the amount of Ca 2ϩ occluded in EP total was determined and shown as the mean Ϯ S.D. (n ϭ 10 -20), otherwise as in Fig. 7. D and E, the effect of G-shift substitution of residues on the Ca 2ϩ -occluded/EP total in C are visualized with ␣-carbon coloring on M2 in the E2⅐BeF 3 Ϫ structure. Green, 1.6 -1.8 (i.e. occlusion as in the wild type (Gly 105 )); red, Ͻ0.5 (i.e. severe disruption of the occlusion). Error bars, S.D.
loosening and elongation; see schematic model in Fig. 10) is crucial for rapid EP isomerization, which involves a large A-domain rotation swinging away from the P domain and docking on the P domain that inclines toward the A domain, and consequent rearrangement of the M2top/M2c interaction with the M4C/P domains and strain imposed on M2 in E2PCa 2 state (as indicated by the prtK cleavage at Leu 119 on M2top) (23,24).
Ca 2ϩ Occlusion in E1PCa 2 -We determined that the uncoupling of Ca 2ϩ transport from ATP hydrolysis in mutants G105A, G105P, V104P, and V106P is due to defective Ca 2ϩ occlusion in E1PCa 2 , and evidently flexibility without distortion in M2 around the glycine is needed to prevent this in E1PCa 2 . The Ca 2ϩ occlusion is restored in G-shift substitution of the residues facing M4C and some facing other sides (Fig. 7), indicating the importance of bending and loosening (elongation). To occlude Ca 2ϩ at the transport sites, the side-chain configuration of the Ca 2ϩ ligand and cytoplasmic gate Glu 309 needs to be fixed by residues on M1 (particularly Leu 65 (36)), which forms the rigid V-shaped structure with M2m upon the kinking of M1 during the E1PCa 2 formation (see FIGURE 9. Structural analysis of E2PCa 2 state by limited proteolysis. Microsomes expressing wild type or the mutants shown in Fig. 8 were phosphorylated at 25°C for 10 s in 6 l of a mixture containing 0.12 mg/ml microsomal protein, 0.5 mM ATP, 0.1 M KCl, 7 mM MgCl 2 , 5 mM CaCl 2 , 1 M A23187, and 50 mM MOPS/Tris (pH 7.0), and then 0.72 mg/ml trypsin (top panels) or prtK (bottom panels) was added in a small volume and incubated for the indicated time periods. In the wild type, EP accumulated was exclusively E1PCa 2 , and its decay was extremely slowed during the proteolysis periods due to the feedback inhibition by the high concentration of Ca 2ϩ . In the mutant, EP accumulated was exclusively E2P, and its decay was extremely slow (see Fig. 8, A and B). The proteolysis was terminated by 2.5% (v/v) trichloroacetic acid, and the digests were subjected to Laemmli SDS-PAGE. The ATPase chain and its fragments separated on the gel were blotted onto a polyvinylidene difluoride membrane and visualized by immunodetection with a monoclonal antibody that recognizes the Ala 199 -Arg 505 peptide (tryptic fragment A1) of SERCA1a, as described under "Experimental Procedures." The tryptic fragments were as follows; A, Met 1 -Arg 505 ; A1, Ala 199 -Arg 505 . The fragments formed by prtK were p95 (Lys 120 -Gly 994 ), p81 (Met 1 -Met 733 ), and p83 (Glu 243 -Gly 994 ) (48,49). The positions of the Ca 2ϩ -ATPase chain and its fragments and those of the molecular mass markers are indicated on the left and right, respectively. NOVEMBER 18, 2016 • VOLUME 291 • NUMBER 47

JOURNAL OF BIOLOGICAL CHEMISTRY 24697
Ϫ ⅐ADP in Fig. 1) (9). This gating closes the Ca 2ϩ access/exit channel to the transport sites, which passes alongside M2 (14). As mentioned above, the straight, long helix of the E1PCa 2 intermediate is bent and probably loosened at Gly 105 in E1PCa 2 , and the bending and loosening must optimize "V" interactions and positioning of Leu 65 relative to Glu 309 . The M2top/M2c region is fixed by interactions with M4C and M1Ј, and the other membrane end is fixed by interactions at the "V" bottom. Straightening M2 with these constraints at the ends, as is likely in G105A, could force the mid-helix away from M4 to open the Ca 2ϩ channel and allow Ca 2ϩ release. Whatever the path, what is clear is that the Ca 2ϩ does escape to the cytoplasmic side in E1PCa 2 with the mutation.
Notably also, our previous biochemical analyses on the structural change E1Ca 2 ⅐AlF 4 Ϫ ⅐ADP 3 E1Ca 2 ⅐BeF 3 Ϫ (the genuine E1PCa 2 analog not crystalized yet) demonstrated (13) that upon the formation of E1PCa 2 from the transition state, the A domain rotates to some extent and comes close to the P domain, and such domain arrangement is between E1Ca 2 ⅐ AlF 4 Ϫ ⅐ADP and E2Ca 2 ⅐BeF 3 Ϫ (E2PCa 2 ) (13, 23, 24). Therefore, it is possible that M2 is more bent during the phosphoryl transfer to form E1PCa 2 . Ca 2ϩ Occlusion in E2PCa 2 -The structural requirement for maintaining occlusion in E2PCa 2 is stricter than in E1PCa 2 , and only wild type and G-shift substitution G105A/A112G are able to restore Ca 2ϩ occlusion in E2PCa 2 damaged by the G105A substitution (Fig. 8). In the crystal structure E1Ca 2 ⅐AlF 4 Ϫ ⅐ADP (as an available E1PCa 2 model), a part of M2c/M2top including the prtK site at Leu 119 is associated with M4C but detaches during EP isomerization and could become strained and unwound in the transient E2PCa 2 state, as seen in E2P, due to the A-domain motion and positioning above the P domain ( Fig.  10) (23,24). This destabilization at the M2top region evidently is transmitted to the V-shaped structure of M2m and M1 in most of the mutations to disrupt the cytoplasmic gate. Only strict directional flexing of M2 at the glycine in wild type and in G105A/A112G maintains occluded Ca 2ϩ .
In the subsequent Ca 2ϩ release to the lumen, E2PCa 2 3 E2P ϩ 2Ca 2ϩ , the associated A and P domains and connected transmembrane helices incline due to the strain of the A/M1Ј-linker The yellow arrow on E1Ca 2 and E1PCa 2 and the red arrow on E2PCa 2 indicate the motions (tilting and rotation) of the A domain and the tilting of associated A-P domains, respectively, for the subsequent step. In E1Ca 2 3 E1PCa 2 , the A domain slightly tilts due to the P domain's conformational change upon the Mg 2ϩ ligation and phosphorylation, thereby pulling up M1 and M2 (broken purple arrows), and the M1/M2m is produced and fixes the Glu 309 cytoplasmic gate; thus, the Ca 2ϩ occlusion is accomplished. In E1PCa 2 3 E2PCa 2 , the A domain largely rotates and docks on the P domain; thereby, M2 connected with the A domain is pulled and moved, M2c is detached from the cytoplasmic part of M4 (M4C) and strained, and the top part of M2c (M2top) is unwound (as found with Leu 119 exposed to prtK in E2PCa 2 (see Fig. 9)). In the latter, E2PCa 2 3 E2P ϩ 2Ca 2ϩ , the associated A-P domains are inclined due to the strain imposed on the A/M1Ј-linker in E2PCa 2 (23); thereby, the M2m/M1 rigid V-shaped body pushes M4L to open the luminal path (gate) to release Ca 2ϩ (9). B, as a consequence, in E2P (depicted with its model E2⅐BeF 3 Ϫ (PDB 2ZBE (11)), M2 straightens a steric collision of the Gly 105 region with M4C that inclines toward Gly 105 -M2c by the P domain inclination, and the structure is stabilized by interactions in the Tyr 122 -hydrophobic cluster (Leu 119 /Tyr 122 on M2c/M2top with the A and P domains), the M1Ј-M2c (Val 106 -Arg 110 ) interaction, and the M2m/M1 V-shaped body (M2m). Gly 105 (or Gly 112 of G-shift G105A/A112G mutant) with its conformational freedom is critical for rapid processing of these large motions while keeping the cytoplasmic gate closed. C, the protein sequence of rabbit SERCA1a Ca 2ϩ -ATPase (UniProt P04191) is aligned using ClustalW version 2.1 with pig H ϩ ,K ϩ -ATPase (UniProt P19156), pig Na ϩ ,K ϩ -ATPase (UniProt P05024), and human flippase ATP8A2 (UniProt Q9NTI2). Letter colors denote fully conserved (green) or highly conserved residues (blue). The color bars above the sequence alignment indicate the M2m, M2c, and M2top regions of Ca 2ϩ -ATPase. The glycine residue at the M2m-M2c connecting region is shown in red. (23,24), and the M1/M2m rigid V-shaped body leans and pushes M4L to open the luminal gate while keeping the cytoplasmic gate closed, as seen in the E2⅐BeF 3 Ϫ crystal structure (Fig. 8, D and E). M2 straightens significantly from the structure seen in E1Ca 2 ⅐AlF 4 Ϫ ⅐ADP (E1PCa 2 ), and flexibility at the glycine would obviously be needed for this (see Figs. 8 (D and E) and 10). The straightening may be due to a steric collision of the Gly 105 region with M4C that inclines toward Gly 105 -M2c by the P domain inclination (Fig. 10). The E2P ground state structure with luminally open gate, under the influence of M2, is stabilized by the interaction networks at the Tyr 122 -hydrophobic cluster (formed with Leu 119 /Tyr 122 on M2top and hydrophobic residues on the A and P domains (28,37,38)), at M1Ј-M2c (Val 106 -Arg 110 ) interaction, and at the M1/M2m V-shaped body.
E2P Hydrolysis-Finally, during the subsequent E2P hydrolysis process from the ground state to the transient state E2P 3 E2ϳP ‡ , the luminal gate becomes tightly closed (7), due to the unwinding at M2top and overall downward motion of M2 upon the slight (25°) rotation of the A domain by the water attack (according to the crystal structural model (11)). The G105A substitution does not affect the E2P hydrolysis kinetics with coupled luminal gate closure; thus, it appears that there is no bending movement, at least around the glycine, during the hydrolysis reaction. It fits with the need for most of the helix to remain straight for the downward movement to close the luminal gate (see Fig. 5B as a model for E2ϳP ‡ ).
An interesting question is why the unwinding of M2 helix (due to distortion imposed on M2 upon the A domain motion) occurs in the Asn 111 -Ala 115 region and not at Gly 105 , as seen in the crystal structural change E2⅐BeF 3 Ϫ 3 E2⅐AlF 4 Ϫ (TG) (Figs. 5 and 8). It needs to be borne in mind that unwinding and elongation at any position in the Ile 103 -Ala 115 region forced by a five-successive glycine insertion exhibits very rapid E2P hydrolysis with coupled tight luminal gate closure (21). The unwinding in the wild type seen in the E2ϳP ‡ model probably occurs because 1) the Gly 105 region is fixed by the interaction with M1Ј, and 2) the M2top region (Leu 119 /Tyr 122 ) is fixed by the Tyr 122 -hydrophobic cluster with the A and P domain, whereas 3) the Asn 111 -Ala 115 region is detached from the P domain upon the A domain's tilting motion in E2P 3 E2ϳP ‡ (E2⅐BeF 4 Ϫ 3 E2⅐AlF 4 Ϫ ) and therefore unsupported and susceptible to unwinding. Retardation of hydrolysis by the G-shift substitution at Val 104 and Val 106 (Fig. 5) may possibly be due to a disruption of the interaction with M1Ј, resulting in M2 bending in a direction unfavorable to proper A domain motion relative to the P domain on water attack of the phosphoryl group.
Gly 105 Conservation-Gly 105 of SERCA1a on the M2c-M2m connecting region is conserved in the P-type ATPase family; the position of the glycine is shifted three residues earlier in the sequence in H ϩ ,K ϩ -ATPase, Na ϩ ,K ϩ -ATPase, and flippase (Fig. 10C) and presumably serves the same function. A kneelike bending movement of M2 may be needed in these P-type ATPases.

Experimental Procedures
Mutagenesis and Expression-The pMT2 expression vector (39) carrying rabbit SERCA1a cDNA with a desired mutation was constructed as described previously (23). Transfection of pMT2 DNA into COS-1 cells and preparation of microsomes from the cells were performed as described (40).
Ca 2ϩ -ATPase Activity and Ca 2ϩ Transport Activity-Activities of expressed SERCA1a were obtained essentially as described previously (41). The rate of ATP hydrolysis was determined at 25°C in a mixture containing 1-5 g of microsomal protein, 0.1 mM [␥-32 P]ATP, 0.1 M KCl, 7 mM MgCl 2 , 10 M CaCl 2 , 5 mM potassium oxalate, and 50 mM MOPS/Tris (pH 7.0). The Ca 2ϩ -ATPase activity of expressed SERCA1a was obtained by subtracting the ATPase activity determined in the presence of 1 M thapsigargin (TG), a highly specific and subnanomolar affinity SERCA inhibitor (42) with conditions otherwise as above. The rate of Ca 2ϩ transport was determined with 45 Ca 2ϩ and nonradioactive ATP and otherwise as above. The Ca 2ϩ transport activity of expressed SERCA1a was obtained by subtracting the activity determined in the presence of 1 M TG, with conditions otherwise as above.
Formation and Hydrolysis of EP-Phosphorylation of SERCA1a in microsomes with [␥-32 P]ATP or 32 P i and dephosphorylation of 32 P-labeled SERCA1a were performed under conditions described in the figure legends. The reaction was quenched with ice-cold trichloroacetic acid containing P i . Precipitated proteins were separated by 5% SDS-PAGE at pH 6.0 according to Weber and Osborn (43). The radioactivity associated with the separated Ca 2ϩ -ATPase was quantified by digital autoradiography as described (44). The amount of EP in expressed SERCA1a was obtained by subtracting the background radioactivity determined in the presence of 1 M TG, with conditions otherwise as above. We confirmed that 1 M TG reduces the EP value in the wild type and all mutants to a background radioactivity level (i.e. Ͻ1% of the maximum EP level, which is actually the same as that obtained in the absence of Ca 2ϩ without TG).
Ca 2ϩ Occlusion in EP-Microsomes were phosphorylated for 1 min at 0°C in a mixture containing 1-5 g of microsomal protein, 10 M ATP, 0.1 M KCl, 7 mM MgCl 2 , 10 M 45 CaCl 2 , 1 M A23187, and 50 mM MOPS/Tris (pH 7.0), and immediately filtered through a 0.45-m nitrocellulose membrane filter (Millipore). The filter was washed extensively with a washing solution (1 mM EGTA, 0.1 M KCl, 7 mM MgCl 2 , and 20 mM MOPS/ Tris (pH 7.0)), and 45 Ca 2ϩ remaining on the filter was quantified as described (23). The amount of Ca 2ϩ occluded at the transport sites of EP in the expressed SERCA1a was obtained by subtracting the amount of nonspecific Ca 2ϩ binding determined in the presence of 1 M TG, with conditions otherwise as above. The amount of EP formed was determined with nonradioactive Ca 2ϩ and [␥-32 P]ATP under otherwise the same conditions as above by membrane filtration, and the radioactivity remaining on the filter was quantified.
Limited Proteolysis and Western Blotting Analysis-Ca 2ϩ -ATPase was phosphorylated and subjected to structural analysis by limited proteolysis with trypsin and prtK as described in the legend to Fig. 9. The digests were separated by 10.5 or 7.5% SDS-PAGE according to Laemmli (45) and blotted onto a polyvinylidene fluoride membrane and then incubated with IIH11 monoclonal antibody to rabbit SERCA1a (Affinity Bioreagents), which recognizes an epitope between Ala 199 and Arg 505 . After incubation with secondary antibody (goat antimouse IgG, horseradish peroxidase-conjugated), the signal was detected with Pierce Western blotting substrate (Thermo Fisher Scientific).
Miscellaneous-Protein concentration was determined by the method of Lowry et al. (46) with bovine serum albumin as a standard. Data were analyzed by nonlinear regression using the program Origin (Microcal Software, Inc., Northampton, MA). Three-dimensional models of the enzyme were reproduced by the program VMD (47). The data represent the mean Ϯ S.D. for 3-6 independent experiments (or 10 -20 experiments in Figs. 7 and 8C).
Author Contributions-T. D. conceived, designed, performed, and analyzed the experiments. T. D. and H. S. coordinated the study and wrote the paper. K. Y. and S. D. provided critical discussion and technical advice. All authors reviewed the results and approved the final version of the manuscript.