Systematic Comparison of Molecular Conformations of H+,K+-ATPase Reveals an Important Contribution of the A-M2 Linker for the Luminal Gating

Background: The gastric H+,K+-ATPase proton pump achieves gastric acid secretion. Results: A newly determined (SCH)E2·MgF structure represents a hybrid conformation of (SCH)E2·BeF and luminal-closed (Rb+)E2·AlF states. Conclusion: Comparison of E2P-related structures revealed that the A-M2 linker importantly contributes to conformational changes of the enzyme. Significance: A-M2 linker-mediated conformational change is conserved between H+,K+-ATPase and Ca2+-ATPase and perhaps other P-type ATPases.

Gastric H ؉ ,K ؉ -ATPase, an ATP-driven proton pump responsible for gastric acidification, is a molecular target for anti-ulcer drugs. Here we show its cryo-electron microscopy (EM) structure in an E2P analog state, bound to magnesium fluoride (MgF), and its K ؉ -competitive antagonist SCH28080, determined at 7 Å resolution by electron crystallography of twodimensional crystals. Systematic comparison with other E2P-related cryo-EM structures revealed that the molecular conformation in the (SCH)E2⅐MgF state is remarkably distinguishable. Although the azimuthal position of the A domain of the (SCH)E2⅐MgF state is similar to that in the E2⅐AlF (aluminum fluoride) state, in which the transmembrane luminal gate is closed, the arrangement of transmembrane helices in the (SCH)E2⅐MgF state shows a luminal-open conformation imposed on by bound SCH28080 at its luminal cavity, based on observations of the structure in the SCH28080-bound E2⅐BeF (beryllium fluoride) state. The molecular conformation of the (SCH)E2⅐MgF state thus represents a mixed overall structure in which its cytoplasmic and luminal half appear to be independently modulated by a phosphate analog and an antagonist bound to the respective parts of the enzyme. Comparison of the molecular conformations revealed that the linker region connecting the A domain and the transmembrane helix 2 (A-M2 linker) mediates the regulation of luminal gating. The mechanistic rationale underlying luminal gating observed in H ؉ ,K ؉ -

ATPase is consistent with that observed in sarcoplasmic reticulum Ca 2؉ -ATPase and other P-type ATPases and is most likely conserved for the P-type ATPase family in general.
Gastric H ϩ ,K ϩ -ATPase is an ATP-driven proton pump responsible for the gastric acid secretion (1,2). This enzyme catalyzes the energetically uphill, electro-neutral exchange of H ϩ /K ϩ coupled with ATP hydrolysis, which generates a million-fold H ϩ -gradient across the parietal cell membrane (i.e. pH 1 in the stomach versus pH 7 in the cytosol; see Refs. 3 and 4). Like other P-type ATPases, the enzyme undergoes cyclical conformational changes between two principal reaction states (E1, E2) and their auto-phosphorylated forms (E1P, E2P) during its transport cycle (see Fig. 1 and Refs. [5][6][7]. H ϩ ,K ϩ -ATPase comprises both an ␣and an ␤-subunit. Like other P2-type ATPases (8), including the housekeeping sodium pump (Na ϩ ,K ϩ -ATPase; Ref. 9) and sarcoplasmic reticulum Ca 2ϩ -ATPase 1a (SERCA 2 ; Ref. 10), the catalytic ␣-subunit contains 10 transmembrane (TM) helices in which cation binding sites are located, and three cytoplasmic domains (A, P, and N domains) where ATP hydrolysis and the accompanying autophosphorylation occur (11). In addition to the ␣-subunit, H ϩ ,K ϩ -ATPase and Na ϩ ,K ϩ -ATPase require an accessory ␤-subunit that is indispensable for functional expression and trafficking of the ␣␤-complex to the cell surface (12). Furthermore, recent studies of H ϩ ,K ϩ -ATPase revealed important contributions of the ␤-subunit for regulating the E1P/E2P equilibrium, which is an important characteristic for generating the large H ϩ gradient between gastric membranes (13,14).
Because the first atomic model of SERCA was reported in 2000 (15), studies of its structure and function have revealed its detailed molecular mechanisms, with crystal structures cover-ing almost the entire reaction cycle (16). Several cryo-EM structures of H ϩ ,K ϩ -ATPase have been determined by electron crystallography of two-dimensional crystals, allowing for evaluation of its important physiologic roles (13,(17)(18)(19). The conformational changes of H ϩ ,K ϩ -ATPase that accompany its transport cycle, however, are not yet fully elucidated.
To clarify the transport mechanism of H ϩ ,K ϩ -ATPase, a better understanding of the E2P state structure as a key reaction intermediate is essential to reveal the conformational changes that occur for the H ϩ /K ϩ exchange against a luminal solution with opening and closing of the transport pathway and gate. Extensive studies of SERCA with fluorinated phosphate analogs (XFs) led to the characterization of its structural and functional properties in various states ( Fig. 1 and Refs. 20 -22). Tetrahedral BeF 3 (BeF) shows the closest coordination of aspartyl phosphate (E2⅐BeF, E2P ground state), a trigonal bipyramidal complex of AlF 4 and water (AlF), which is assigned as a transition analog from aspartyl phosphate to its hydrolysate (E2⅐AlF, E2⅐P transition state), and a tetrahedral MgF 4 (MgF), which mimics the inorganic phosphate produced by E2P hydrolysis (E2⅐MgF, E2⅐P i product state). Because each bound XF is tightly coordinated by a TGES loop located at the A domain, subtle differences in their environment generate a different type of cytoplasmic arrangement, leading to distinct conformational changes of the whole enzyme structure (16,21). Therefore, even at a limited resolution of ϳ8 Å determined by electron crystallography, molecular events occurring at the phosphorylation site and the luminal gate can be observed based on the rearrangement of the cytoplasmic domains and TM helices, respectively, in response to each bound XF (13,(17)(18)(19).
Here we determined new cryo-EM structures of H ϩ ,K ϩ -ATPase using a phosphate analog (either AlF or MgF) and a specific antagonist, SCH28080. Although these structures represent an identical conformation, their cytoplasmic and luminal halves appeared to be modulated independently by the phosphate analog and the bound antagonist. Systematic comparison of the cryo-EM structures in the E2P-related states allows us to propose a luminal gating mechanism of H ϩ ,K ϩ -ATPase.

EXPERIMENTAL PROCEDURES
Isolation of H ϩ ,K ϩ -ATPase-enriched Membrane Fractions-Purification of the H ϩ ,K ϩ -ATPase-enriched membrane fractions (G 1 and G 2 ) was performed as described previously (23) with some modifications, and the fractions were further purified with SDS (24). Purified membrane fractions were frozen in liquid nitrogen and stored at Ϫ80°C until use. For two-dimensional crystallization and other biochemical analyses, an SDSpurified G 1 fraction was used as described previously (13).
Samples were negatively stained with 2% (w/v) uranyl acetate to screen for crystallization conditions. Specimens for cryo-EM were prepared in a cold room using the carbon sandwich method (25,26). Images were recorded with a JEM-3000SFF electron microscope (JEOL) equipped with a super fluid helium stage ( Lower panel, E2P and its dephosphorylation steps mimicked by XFs are shown (see "Cytoplasmic Domains" for details). The atomic models represent the coordination chemistry of XFs (sticks for an aspartate residue and spheres for BeF 3 (left), AlF 4 and a water (middle), and MgF 4 (right), respectively). Schematics of the molecular conformations presented in this study are shown (see Fig.  9 for details).
images were processed with MRC image processing programs (29), and their initial contrast transfer function parameters were determined (30) for correction. The data tilted to 60°were merged using LATLINE (31) and used to calculate a three-dimensional density map at 8 Å and 7 Å resolution for the (SCH)E2⅐AlF and (SCH)E2⅐MgF states, respectively ( Table 1). The EM density maps of (SCH)E2⅐AlF and (SCH)E2⅐MgF were deposited in the EMDataBank (accession code EMD-2759 and EMD-2760, respectively). The coordinates of the homology models were also deposited in the PDB (accession code 4UX1 and 4UX2, respectively). Homology Models and Structural Comparison of E2P-related States of H ϩ ,K ϩ -ATPase-Homology models of (SCH)E2⅐AlF and (SCH)E2⅐MgF were built with MODELLER v9.12 (32) using the atomic model of Na ϩ ,K ϩ -ATPase with K ϩ and ouabain (Ref. 33, PDB code 3A3Y) as a starting template. The initial manual fitting of the homology model into the EM density map was achieved using COOT (34). The adjustment for each individual domain and TM helix with the EM density map was performed using SITUS (35). After the positional search, further fine-fitting and connecting of the split loop region were performed manually using COOT with regularization refinement. For systematic comparison with other cryo-EM structures, the same procedures were applied to build homology models for the (SCH)E2⅐BeF state and (Rb ϩ )E2⅐AlF states using atomic models of Na ϩ K ϩ -ATPase in the high affinity ouabain binding state (Ref. 36, PBD code 4HYT) and K ϩ -bound E2⅐MgF state (Ref. 37, PDB code 2ZXE), respectively, although the resulting models are almost identical to those reported previously (18,19).
To compare the relative orientation of the cytoplasmic domains, the EM density map and homology models were aligned based on the superpositioning of the molecules on their P domain (see Fig. 3). To compare other structural parts, including the A-M2 linker and TM helices, the molecules were aligned by superpositioning of the TM segments M7-M10 and ␤M, which are the least variable of the reported SERCA structures.
The change in FITC fluorescence at the steady state was determined at 37°C in a 3-ml solution containing 50 mM HEPES/Tris, pH 7.0, 1 mM MgCl 2 , 25 mM sucrose, and 20 g of the FITC-labeled H ϩ ,K ϩ -ATPase membrane fraction. Reactions were initiated by adding 5 mM MgCl 2 and 5 mM P i for "Mg 2ϩ ϩ P i (E2P)," 1 mM BeSO 4 and 4 mM NaF for "BeF (E2⅐BeF)," 1 mM AlCl 3 and 4 mM NaF for "AlF (E2⅐AlF)," or 5 mM MgCl 2 and 10 mM NaF for "MgF (E2⅐MgF)" conditions followed by 10 M SCH28080. The addition of the reagents had a negligible effect on the fluorescence intensity of the FITC dye itself. All FITC fluorescence measurements were performed using a Shimadzu F5500 spectrophotometer fitted with a magnetic device to stir the cuvette contents and control the water temperature. The FITC fluorescence intensity was optimized by measurement at the wavelength pair (excitation 495 nm /emission 515 nm ) with the slit width set to 5 nm for either excitation or emission.
Enzyme Activity-H ϩ ,K ϩ -ATPase activity and K ϩ -p-nitrophenyl phosphatase activity of membrane-bound enzyme were determined as described previously (17,39). For the measurement of the dose-dependent inhibition by XFs, FITC-modified membrane preparations were incubated in the indicated concentrations of BeSO 4 or AlCl 3 in the presence of 1 mM NaF, 0.1 mM MgCl 2 , 250 mM sucrose, and 40 mM HEPES/Tris, pH 7.0. For MgF, an equal amount of MgCl 2 and NaF was used. After incubation for 1 h at 37°C, residual K ϩ -p-nitrophenyl phosphatase activity was measured colorimetrically (20 mM p-nitrophenyl phosphate/Tris, 20 mM MgCl 2 , 16 mM KCl, 250 mM sucrose, and 40 mM imidazole/HCl, pH 7.8).

Measurement of [ 3 H]SCH28080
Binding-In equilibrium binding experiments, to determine the K d value (41,42), the purified membrane fractions (5-100 g/ml) were suspended in a buffer comprising 40 mM MES (pH 6.5, adjusted with Tris), 2% glycerol, and the addition of 5 mM MgCl 2 and 5 mM P i for "MgP i ," 1 mM MgCl 2 , 1 mM BeSO 4 , and 4 mM NaF for "BeF,"; 1 mM MgCl 2 , 1 mM AlCl 3 , and 4 mM NaF for "AlF," or 5 mM MgCl 2 and 10 mM NaF for "MgF" followed by a 30-min incubation at room temperature. Each reaction tube was then incubated at 0 -37°C, and 5 nM--1 M [ 3 H]SCH28080 (synthesized by PerkinElmer Life Sciences) was added. The level of nonspecific binding was determined in the presence of a 100-fold excess of unlabeled SCH28080 over the concentration range of [ 3 H]SCH28080 used. The enzyme suspension (0.1-2 ml) was incubated for 1 h at the indicated temperature and rapidly filtered through a nitrocellulose membrane filter (HAWP Millipore filter, 0.45 m) pre-wetted with washing buffer comprising 10 mM MES (pH 6.5, adjusted with Tris), 1 mM MgCl 2 , and 10% PEG 3500, which was placed on top of a glass fiber filter. The membrane was washed 3 times with 5 ml of washing buffer to remove unbound inhibitor. The membrane was placed in a 20-ml scintillation vial; 1 ml of 2% SDS was added to dissolve the H ϩ ,K ϩ -ATPase and bound SCH28080 from the filter membrane, and 10 ml of scintillation solvent was added and the contents counted. Binding of [ 3 H]SCH28080 was assessed by subtracting the level of nonspecific binding of [ 3 H]SCH28080, obtained in the presence of a 100-fold excess of nonradioactive SCH28080, from the amount of [ 3 H]SCH28080 bound to the membrane in the absence of the cold inhibitor. The dose dependence of SCH28080 binding was fit to the hyperbolic binding curve, and the binding maximum (B max ) and dissociation constant (K d ) were determined at the indicated temperatures. The B max values were 2.0 -2.4 nmol/mg regardless of added XFs and reaction temperature, which was close to the maximum amount of E 32 P formed from [␥-32 P]ATP (ϳ2 nmol/mg for the purified membrane fractions; Ref. 43) and thus consistent with previous results (42).
Overall changes in the enthalpy (⌬H) and entropy (⌬S) of SCH28080 were obtained from the van't Hoff plot using the relationship (41), where R is the gas constant. Free energy terms were calculated using the equation,

Two-dimensional Crystallization of (SCH)E2⅐AlF and (SCH)E2⅐
MgF States-Previously, we reported several cryo-EM structures of H ϩ ,K ϩ -ATPase bound to different XFs in the presence or absence of ions and substrates for the TM domain (such as transported cation K ϩ , its congener Rb ϩ , and a specific antagonist SCH28080), which include the reaction state analog of E2⅐BeF (17), E2⅐AlF (13), (SCH)E2⅐BeF (18), and (Rb ϩ )E2⅐AlF (19). For systematic comparison of the molecular conformations induced by XFs mimicking the reaction substeps of E2P dephosphorylation, we obtained two-dimensional crystals in the presence of SCH28080 in combination with either AlF or MgF and determined their three-dimensional structures at 8 Å and 7 Å resolution, respectively ( Fig. 2A, Table 1). It is notable that no two-dimensional crystals were produced in the presence of either MgF alone or MgF with Rb ϩ (not shown). Because inhibition of H ϩ ,K ϩ -ATPase activity by MgF is reversible and weaker than that of other XFs (17), stabilization of the molecular conformation by SCH28080 is required for the twodimensional crystal formation when MgF is used as the phosphate analog. The two newly determined structures, however, were almost identical in their overall conformation, including the arrangement of the TM helices, the azimuthal position of the A domain, and the connecting linker between them (Fig. 2, A and B). In fact, x-ray structures of SERCA in the E2⅐AlF and E2⅐MgF states also indicated identical molecular conformations (21). Therefore, we concluded that these two structures of H ϩ ,K ϩ -ATPase adopt an indistinguishable conformation and thus used the (SCH)E2⅐MgF structure as a representative with better crystallographic characteristics for subsequent comparison of the molecular conformation with other cryo-EM structures.
Cytoplasmic Domains-Like inorganic phosphate, XFs bind to the Asp residue in the invariant sequence of DKTG located at the P domain, and each bound XF is covered by the outermost TGES loop on the A domain. The different coordination chemistries of the XFs (Fig. 1) change the relative orientation of the XF-bound aspartate residue and TGES-loop, and thus the P and A domains assume different relative positions in respective XFbound structures (21). When the E2P-related cryo-EM structures of H ϩ ,K ϩ -ATPase were aligned according to their P domain, the location of the A domain was significantly different (Fig. 3, A and B, and Ref. 18). The A domain in (SCH)E2⅐MgF was rotated ϳ20°around an axis approximately perpendicular to the membrane plane compared with that in the E2P ground state analog (SCH)E2⅐BeF (18) (AlF and MgF). Furthermore, the A domain in (Rb ϩ )E2⅐AlF was more inclined (ϳ5°) toward the membrane surface than that in (SCH)E2⅐MgF (Fig. 3F). This inclination of the A domain induced the TGES loop to segregate even further from the phosphorylation site, which may allow for water penetration at this site, facilitating the hydrolysis of the aspartyl phosphate. Hence, the observed inclination of the A domain induced by the Rb ϩ binding (19) in the H ϩ ,K ϩ -ATPase (Rb ϩ )E2⅐AlF state suggests that this conformation is a transient point that leads to the formation of the dephosphorylated (K ϩ )E2 state. Observed differences in the azimuthal positions of the A domain in E2P analog states of H ϩ ,K ϩ -ATPase were consistent with reported SERCA structures (21,22); thus a similar mechanism can be applied for the relationship between the phosphorylation states and the arrangement of the cytoplasmic domains in both ATPases.
Monitoring the Enzyme Conformation by the FITC Fluorescence Change-The fluorescence probe FITC preferentially forms a covalent bond with the ⑀-amino group of the Lys-518 residue, which is embedded in the conserved Lys-518 in the ATP binding site of the N domain (38). This chemical modification of the Lys residue impairs H ϩ ,K ϩ -ATPase activity (1.7% of residual H ϩ ,K ϩ -ATPase activity compared with that of mock-treated enzyme) due to a loss of ATP-binding ability, suggesting that the FITC probe is located at the nucleotide binding position. The FITC-modified H ϩ ,K ϩ -ATPase, however, can hydrolyze substrates less bulky than ATP, such as acetyl phosphate or p-nitrophenyl phosphate (39), showing 76% residual K ϩ -p-nitrophenyl phosphatase activity compared with that of mock-treated enzyme. The FITC-modified enzyme also has affinities for XFs comparable with those of the unmodified enzyme (Fig. 4A). Therefore, the FITC-modified enzyme remains active and undergoes a conformational change in response to substrate binding, allowing us to monitor the conformational changes, especially those occurring at the nucleotide binding site in the N domain, based on the fluorescence intensity (40).
To determine whether the observed rearrangement of the cytoplasmic domains in the H ϩ ,K ϩ -ATPase structure in fact occurred in the solution or was an artifact due to the crystal condition and/or packing, the conformational change in the enzyme in response to XF and SCH28080 binding was evaluated using FITC fluorescence as a probe (Fig. 4). Because the rotational motion of the A domain between (SCH)E2⅐BeF and (SCH)E2⅐MgF (Fig. 3, A and B) also changes the distance between the A and N domains in each structure, which may affect bound ADP at the nucleotide binding site in the N domain (Fig. 4, B and C). Bound ADP is likely to be coordinated to both the N domain and the outermost A domain, as seen in the (SCH)E2⅐BeF state in which the distance between the A and N domains is shortest among all the other E2P-related structures (Fig. 4B). Therefore, changes in the distance between the A and N domains are related to the absence of the EM density responsible for the bound ADP in the (SCH)E2⅐MgF (A domain ϳ20°rotated, as shown in Fig. 4C) as well as in (Rb ϩ )E2⅐AlF states (A domain ϳ20°rotated and ϳ5°inclined, as shown in Fig. 3, E and F).
We found that the addition of BeF increased the FITC fluorescence signal (6.9 Ϯ 1.2% increase), whereas that of MgF decreased it (8.3 Ϯ 0.6% decrease). In every experiment the addition of SCH28080 to the respective condition slightly decreased the fluorescence signal (Fig. 4D). The fluorescence intensity rapidly increased in response to the addition of BeF (Fig. 4D, 0.5 Ͻ 1.3 s), consistent with its high affinity binding (17) and the fast inhibition of the ATPase activity of the FITCfree enzyme in the presence of 1 mM BeF (not shown). The addition of Mg 2ϩ and P i (MgP i ) induced a genuine E2P formation, which also increased the fluorescence signal (7.7 Ϯ 1.0%), suggesting that BeF acts as a close mimetic of genuine acylphosphate as reported for SERCA (20). The rate of E2 32 P formation from radioactive 32 P i ( 0.5 ϭ ϳ15 s at 20°C, from Ref. 44, or 23 s at 0°C 3 ) was close to that of the fluorescence increase occurring after the addition of MgP i (Fig. 4D, 0.5 ϭ 19 s at 25°C), supporting the notion that the observed change in fluorescence intensity was due to E2P formation. Although the addition of MgF decreased the signal intensity, the slow response of the fluorescence signal (Fig. 4D, 0.5 ϭ 71 s) was consistent with its slow inhibition of ATPase activity ( 0.5 ϭ 120 s, from Ref. 17), thus reflecting its binding to the enzyme. Unfortunately, FITC 3 K. Abe, unpublished data. Cryo-EM Structure of Antagonist-bound E2P Gastric H ؉ ,K ؉ -ATPase fluorescence measurements are not applicable to the AlFbound form, because the addition of AlF induces a large amount of nonspecific quenching of the FITC fluorescence intensity, making it unlikely to reflect the conformational change of the enzyme (Fig. 4D, gray trace), probably due to a direct reaction between AlF and the FITC dye itself (45).
Because FITC fluorescence intensity generally increases in a hydrophobic environment, bound FITC located at the nucleotide binding site might be surrounded by the outermost amino acids of the A domain, just like bound ADP in the (SCH)E2⅐BeF condition (Fig. 4B). In contrast, as indicated by the decreased fluorescence intensity in the (SCH)E2⅐MgF condition, bound FITC might be exposed to the bulk solution due to the separated N and A domains in the (SCH)E2⅐MgF state in which bound ADP is disordered (Fig. 4C). Bound ADP was also visible in the SCH28080-free E2⅐BeF state (17), the condition in which FITC fluorescence intensity was increased (Fig. 4, D and E). Therefore, although fluorescence measurements cannot be used for the AlF-bound form, the result is consistent with the appearance of the ADP density in other available XF-bound cryo-EM structures of H ϩ ,K ϩ -ATPase, and thus the observed rearrangements in the cytoplasmic domains are likely occur in the native condition.
Transmembrane Helices-Like in the (SCH)E2⅐BeF state, bound SCH28080 imposed a luminal-open conformation of TM helices in the (SCH)E2⅐MgF state (Fig. 5, A and B). Comparison between the luminal-open conformation of the SCH28080-bound forms and the luminal-closed conformation of E2⅐AlF and (Rb ϩ )E2⅐AlF revealed that opening and closing of the luminal helices was accomplished by rearranging the M1-M4 helices (Fig. 5, C and D). During the transition from an open to closed form of the luminal helices, the luminal half of the M3M4 helices was shifted laterally to close the SCH28080binding site, which was coupled with the movement of the M1M2 helices (see Fig. 5, C and D, green and blue arrows).
Affinity and Binding Mode of SCH28080 with Different XFs-Despite the identical luminal-open arrangement of TM helices in the (SCH)E2⅐BeF and (SCH)E2⅐MgF states, however, most of the SCH28080 moiety was invisible in the density map of (SCH)E2⅐MgF (Fig. 6B), which is in marked contrast to that of (SCH)E2⅐BeF (Fig. 6A). Because SCH28080 was required for the two-dimensional crystallization in the (SCH)E2⅐MgF condition as well as the different arrangement of the TM helices in (SCH)E2⅐AlF from its absent E2⅐AlF state (Fig. 5C), the observed poor density at the SCH28080-binding site was likely due to the disorder of this antagonist at the luminal cavity. Such a difference around the presumed SCH28080 density in the (SCH)E2⅐BeF and the (SCH)E2⅐MgF states prompted us to investigate the SCH28080 binding affinity to H ϩ ,K ϩ -ATPase bound to the various XFs (Fig. 6C). Using 3 H-labeled SCH28080, the effects of XFs on the SCH28080 binding affinity were examined (41). BeF-bound H ϩ ,K ϩ -ATPase had a high affinity for the SCH28080 binding (K d ϭ 14.8 Ϯ 1.1 nM), which was as high as that for the genuine E2P formed from MgP i (K d ϭ 13.4 Ϯ 1.4 nM) at 25°C. In contrast, AlF-or MgF-bound H ϩ ,K ϩ -ATPase had 3ϳ4 times lower affinity for SCH28080 (K d ϭ 33.6 Ϯ 2.2 nM, K d ϭ 54 Ϯ 4.5 nM) than BeF-bound H ϩ ,K ϩ -ATPase, consistent with the observed poor density of SCH28080 at the luminal cavity in the (SCH)E2⅐MgF (Fig. 6B) and (SCH)E2⅐AlF (not shown) states.
A-M2 Linker That Connects Cytoplasmic Domains and TM Helices-Rotation of the A domain was in turn transmitted to the TM region, which was mediated by the connecting linker between them (Fig. 7). The middle of the A-M2 linker, indicated as dark-colored tubes in Fig. 7, A-D, assumed an unwound loop structure in the (SCH)E2⅐MgF state (Fig. 7B) almost identical to that in the E2⅐AlF and (Rb ϩ )E2⅐AlF states (Fig. 7, C and D, respectively), but significantly different from that in the (SCH)E2⅐BeF state (Fig. 7A). The conformation of the juxta-membranous portion of the A-M2 linker and M2 helix in (SCH)E2⅐MgF, however, was largely different from those in E2⅐AlF and (Rb ϩ )E2⅐AlF, but is similar to that in (SCH)E2⅐BeF (Fig. 7A, Fig. 8). These differences were related to the conformational rearrangement of the A domain and TM  In the (SCH)E2⅐MgF state, the cytoplasmic portion assumed a typical "luminal-closed" form (ϳ20°rotated A domain and unwound loop structure of the A-M2 linker), which might subsequently induce the luminal gate closure in the absence of SCH28080 as seen in the E2⅐AlF structure (Fig. 5C). The M1M2 helices, however, could not move due to the luminal-open conformation of the TM helices fixed by bound SCH28080 at the luminal cavity. As a result, the (SCH)E2⅐MgF structure represents a hybrid conformation, with the relative orientation of the cytoplasmic domains similar to that in a typical luminal-closed type conformation (Figs. 3 and 7) and the TM region assuming a luminalopen conformation (Fig. 5).

DISCUSSION
In this paper we describe two identical cryo-EM structures of H ϩ ,K ϩ -ATPase in the (SCH)E2⅐AlF and (SCH)E2⅐MgF states. Including these newly determined structures, all E2P-related structures of H ϩ ,K ϩ -ATPase were crystallized in the presence of phosphate analogs with or without the synthetic antagonist SCH28080. The phosphate analogs used in these studies are well characterized for SERCA (20,21) and are also applicable for H ϩ ,K ϩ -ATPase (17,19); thus, they likely act as close mimetics of reaction substeps of the phosphorylated form of the enzyme. The K ϩ -competitive antagonist, SCH28080, binds to the luminal cavity, which connects with the cation binding site in the middle of the TM helices (18). The arrangement of the TM helices in the SCH28080-bound structures of H ϩ ,K ϩ -ATPase is surprisingly similar to that in the E2⅐BeF structure of SERCA in which the cation binding site is exposed to the luminal solution (21,22). Therefore, it is most likely that the luminal-open TM arrangement imposed by SCH28080 mimics the situation in which the luminal gate is opened to exchange H ϩ /K ϩ to the luminal solution. As protein crystal structures with bound artificial antagonists or inhibitors are reported to reveal a snapshot of the protein at work, it is not surprising that an artificial antagonist utilizes the binding site exposed when the protein is in a certain conformation (46 -48). We, therefore, conclude that the molecular conformations induced by XFs and their combination with SCH28080 might occur in the native enzyme.
As discussed below, H ϩ ,K ϩ -ATPase may inherently prefer the luminal-closed conformation because, even when using BeF, which drives SERCA into the luminal-open conformation, H ϩ ,K ϩ -ATPase closes its luminal gate (17). This makes structural investigation of luminal-gating mechanism of H ϩ ,K ϩ -ATPase particularly difficult if using only XFs as a conformational modulator. Using SCH28080 to impose the luminal-open conformation allows us to address the mechanistic rationale for the luminal gating of H ϩ ,K ϩ -ATPase. Systematic comparison of the molecular conformations reveals several key components and how they regulate the luminal gating of H ϩ ,K ϩ -ATPase (Fig. 9).
To our surprise the newly determined (SCH)E2⅐MgF structure likely represents a chimeric conformation between the luminal-open (SCH)E2⅐BeF (18) and luminal-closed (Rb ϩ )E2⅐ AlF (19) states (Fig. 9), a conformation that has until now been inaccessible for SERCA and other P-type ATPases. For ATPdriven P-type ATPases, however, a conformational change of the ATP-hydrolyzing cytoplasmic domains must in principle be tightly coupled to that in the cation-transporting TM helices. This principle is supported by several crystal structures of SERCA in E2P analog states in which the luminal gate closure is coupled with a 20°rotational motion of the A domain via a connecting unwound structure, the A-M2 linker (16,21). In this respect, the molecular conformation in the observed hybrid structure of the H ϩ ,K ϩ -ATPase (SCH)E2⅐MgF state is uncoupled in terms of the conformational relationship between the cytoplasmic domains and the TM region. Comparison of the hybrid structure with the luminal-open and luminal-closed structures revealed that two contradicting motions, a 20°rotational motion of the A domain on the cytoplasmic side that drives the luminal-closed TM arrangement (Fig. 3) and the luminal-open arrangement in the TM helices fixed by the SCH28080-binding (Fig. 5), are both transmitted to the connecting A-M2 linker (Fig. 7). As a result, this linker region (more specifically, its boundary with the M2 helix; Fig. 8) is constrained in the hybrid conformation of the (SCH)E2⅐MgF  Fig. 9 for details). state (Fig. 9). Hence, the A-M2 linker acts as a switching element between two portions of the enzyme, thereby mediating an important role for the correct coupling of the ATP hydrolysis and luminal gating in physiologic situations.
Systematic comparison of molecular conformations in SCH28080-bound E2P-related structures of H ϩ ,K ϩ -ATPase revealed that the azimuthal position of the A domain in each XF-bound structure reflects the coordination geometries of each bound XF in SCH28080-bound conditions (Fig. 9), as observed in the SERCA structures (21,22). Closure of the luminal gate is accompanied by a lateral shift of the M3M4 helices (Fig. 5, C and D), and a vertical shift of the M1M2 helices toward the luminal side (Fig. 7, A-D). In particular, the latter shift induces a conformational change of the A-M2 linker and subsequently forces the A domain into a 20°rotated position like in the E2⅐AlF and (Rb ϩ )E2⅐AlF states (Fig. 3). When the bound SCH28080 imposes a luminal-open form on the TM helices, the A-M2 linker might be loosened (Figs. 7 and 8) because the M1M2 helices must move vertically to a more cytoplasmic side than in the luminal-closed form. Therefore, in the SCH28080bound forms, the A domain can freely rotate in response to the coordination geometry of each bound XF. This might be why the azimuthal position of the A domain in each different XFbound structure reflects the coordination geometry of the respective XF in the presence of SCH28080. In the BeF-bound form and probably in the genuine phosphate-bound form as well, the A domain remains at a 0°rotated position (Fig. 3A), which permits tight coordination of the aspartyl phosphate (or its analog, Asp-BeF) by the TGES loop when SCH28080 is bound to the luminal cavity, thereby allowing the loosened A-M2 linker to assume an ␣-helical structure extended from the M2 helix (Figs. 7A and 8A). In contrast, in the case of the AlF-or MgF-bound forms, even if SCH28080 is bound, the A domain remains in the 20°rotated position (Fig. 3A) because these XFs sterically restrain the A domain in the 0°rotational position, in which the expected position of the TGES loop seems too close to coordinate with them appropriately.
Notably, H ϩ ,K ϩ -ATPase tends to accumulate in the luminalclosed conformation (Fig. 5C) in which the A domain is in a 20°r otated position relative to the (SCH)E2⅐BeF structure (Fig. 3, C and D). In contrast to SERCA, even if BeF is used as the closest analog of aspartyl phosphate in the absence of SCH28080 (17), H ϩ ,K ϩ -ATPase adopts an identical structure in the luminalclosed E2⅐AlF state. These observations corroborate the notion that the TM helices of H ϩ ,K ϩ -ATPase prefer the luminalclosed conformation to reduce the risk of H ϩ reverse flow from the highly acidic gastric lumen (13,14). Because of tight coupling between the TM helices and cytoplasmic domains, this preference of the TM helices may be strong enough to influence the arrangement of cytoplasmic domains, especially the A domain, via the A-M2 linker in the case of H ϩ ,K ϩ -ATPase. Therefore, we speculate that the molecular conformation of H ϩ ,K ϩ -ATPase in a genuine E2P state is mainly the luminalclosed conformation like in the E2⅐AlF and E2⅐BeF states to prevent the reverse reaction with the isolation of the proton binding site from the highly acidic environment (13). Subsequent K ϩ binding, however, requires the formation of a physical pathway from the luminal bulk solution to the cation binding site. The hybrid structure of (SCH)E2⅐MgF described in the    Bound SCH28080 and Rb ϩ are shown as black sticks and blue circle, respectively, and the A-M2 linker is highlighted by a dark color in each reaction state. Two thin cylinders indicate M1M2 helices. Arrows indicate the conformational difference when each state is compared with the (SCH)E2⅐BeF state. The lower table indicates conformational features of key components for each reaction state. A domain, the azimuthal positions of the A domain in each state; that of (SCH)E2⅐BeF is set as 0° (Fig. 3). A-M2 linker, conformation of the A-M2 linker (Fig. 7). The asterisk indicates loosened A-M2 linker in the (SCH)E2⅐MgF state (Fig. 7B). M1M2, vertical location of the M1M2 helices. Up and Down represent the vertical location of the helices on the relatively cytoplasmic side and luminal side, respectively, in each state (Fig. 7). M4M5, arrangement of the luminal half of the M3M4 helices (Fig. 5)  present paper suggests the entity of such a luminal-open conformation of the TM helices without displacement of the A domain, even for a short time. Therefore, for the subsequent dephosphorylation of E2P in the transport cycle, K ϩ may enter when the thermal fluctuation of the TM helices induces a conformation such as (SCH)E2⅐MgF to allow for physical access of K ϩ to the cation binding site from the luminal solution.
To determine the thermodynamic parameters for SCH28080 binding, we evaluated the effect of the temperature on the SCH28080 binding to E2P or the XF-inhibited E2P analog states. Overall changes in enthalpy (⌬H) and entropy (⌬S) accompanied by SCH28080 binding were obtained from the van't Hoff relationship (Fig. 6, D and E). SCH28080 binding studies revealed that the E2P or analogous states of E2P fall into two groups. E2P and E2⅐BeF show high affinity with both ⌬H and ϪT⌬S favorable for binding free energy. In contrast, E2⅐AlF and E2⅐MgF show lower affinity with an unfavorable ϪT⌬S term. Because an unfavorable entropy term usually reflects a large degree of flexibility, this could explain the observed poor density for bound SCH28080 in the (SCH)E2⅐AlF and (SCH)E2⅐MgF states (Fig. 6B). The SCH28080 binding pocket structures (i.e. arrangements of TM helices) in all SCH28080bound E2P states were indistinguishable from their EM maps (Fig. 5, A and B). We speculated that the different conformations of the A domain contribute to the different affinities and stabilities of the molecular conformation in each state. In the (SCH)E2⅐AlF (not shown) and (SCH)E2⅐MgF states, the A domain locates at a 20°-rotated position, which drives luminal gate closure by the unwinding of the A-M2 linker and subsequent downward movement of the M1M2 helices when SCH28080 is dissociated. Thus, although the TM helices in (SCH)E2⅐MgF are forced into the luminal-open conformation by bound SCH28080, the TM helices may always face conflicting pressure of the gate closure from the cytoplasmic A domain. Such a conformational dynamics may be related to the lower affinity of SCH28080 in the (SCH)E2⅐MgF state. Because the enthalpy-driven binding mode often involves conformational changes of the protein, the determined thermodynamic parameters may reflect conformational changes of the enzyme upon SCH28080 binding rather than its simple binding without conformational changes, like a lock and key. Hence, the different binding parameters observed in the two groups (E2P and E2⅐BeF versus E2⅐AlF and E2⅐MgF) suggest that SCH28080 binding induces different conformations.
Here we highlighted the important contribution of the A-M2 linker in gating the ion pathway on the luminal side of H ϩ ,K ϩ -ATPase by determining a hybrid conformation and comparing it with luminal-open and luminal-closed conformations. The accumulation of examples of the conformational changes in SERCA and in other P-type ATPases allows for their comparison (49 -52). A recently reported structure of the high affinity Na ϩ ,K ϩ -ATPase-ouabain complex with bound genuine phosphate (36) shows close similarity to the H ϩ ,K ϩ -ATPase (SCH)E2⅐BeF structure (18), suggesting that these two closely related ATPases have a similar mechanism for high affinity inhibitor binding as well as for the conformational change accompanied by luminal gate opening, which was well predicted from the structure of MgF-bound Na ϩ ,K ϩ -ATPase with bound K ϩ and ouabain (33). Structural comparison of the E2⅐BeF and E2⅐AlF states of CopA, a Cu 2ϩ -transporting PIBtype ATPase, also shows a very similar conformational rearrangement of the A domain and A-M2 linker induced by the binding of these two respective XFs in H ϩ ,K ϩ -ATPase, although a unique luminal transport pathway has been proposed (52). Such surprising similarities over the different subfamilies of P-type ATPases suggest that the basic mechanism for luminal gating coupled with conformational rearrangement of the cytoplasmic domains is likely conserved for P-type ATPase in general.