Deceleration of the E1P-E2P Transition and Ion Transport by Mutation of Potentially Salt Bridge-forming Residues Lys-791 and Glu-820 in Gastric H+/K+-ATPase*

A lysine residue within the highly conserved center of the fifth transmembrane segment in PIIC-type ATPase α-subunits is uniquely found in H,K-ATPases instead of a serine in all Na,K-ATPase isoforms. Because previous studies suggested a prominent role of this residue in determining the electrogenicity of non-gastric H,K-ATPase and in pKa modulation of the proton-translocating residues in the gastric H,K-ATPases as well, we investigated its functional significance for ion transport by expressing several Lys-791 variants of the gastric H,K-ATPase in Xenopus oocytes. Although the mutant proteins were all detected at the cell surface, none of the investigated mutants displayed any measurable K+-induced stationary currents. In Rb+ uptake measurements, replacement of Lys-791 by Arg, Ala, Ser, and Glu substantially impaired transport activity and reduced the sensitivity toward the E2-specific inhibitor SCH28080. Furthermore, voltage clamp fluorometry using a reporter site in the TM5/TM6 loop for labeling with tetra-methylrhodamine-6-maleimide revealed markedly changed fluorescence signals. All four investigated mutants exhibited a strong shift toward the E1P state, in agreement with their reduced SCH28080 sensitivity, and an about 5–10-fold decreased forward rate constant of the E1P ↔ E2P conformational transition, thus explaining the E1P shift and the reduced Rb+ transport activity. When Glu-820 in TM6 adjacent to Lys-791 was replaced by non-charged or positively charged amino acids, severe effects on fluorescence signals and Rb+ transport were also observed, whereas substitution by aspartate was less disturbing. These results suggest that formation of an E2P-stabilizing interhelical salt bridge is essential to prevent futile proton exchange cycles of H+ pumping P-type ATPases.

The ubiquitous Na,K-ATPase and the gastric H,K-ATPase belong to the P IIC subgroup of the extensive class of P-type ATPases, which use ATP hydrolysis for active transport of cations. The reversible phosphorylation of a highly conserved Asp residue, a hallmark of all P-type ATPases, is coupled to the transition between two principal conformational states (E 1 and E 2 ) and the corresponding phosphointermediates (E 1 P and E 2 P). Na,K-and H,K-ATPases share several similarities. The catalytic ␣-subunits are highly homologous (ϳ60% sequence identity), whereas the accessory ␤-subunits have lower sequence identity but nevertheless share similar basic structural features (1). Furthermore, Na,K-and H,K-ATPases are the only known K ϩ -countertransporting P-type pumps in eukaryotes.
Despite these common features, there are also important differences between the two enzymes. For example, the asymmetrical transport stoichiometry of Na ϩ and K ϩ exchange (3 versus 2) by all known Na,K-ATPase isoforms results in a net electrogenic transport (2), whereas gastric and non-gastric H,K-ATPases operate strictly electroneutral with 2:2 (or 1:1) stoichiometry (3,4). Although the transport cycle of the H,K-ATPase is net electroneutral, evidence has been accumulated that several partial reactions of the cycle are electrogenic. Electrophysiological experiments using purified H,K-ATPasecontaining membrane fragments on planar lipid bilayers (5,6) have shown that the proton branch of the cycle, which involves the E 1 P-E 2 P conformational change investigated in the current study, includes an electrogenic step. As it is the case for the electrogenic Na,K-ATPase, the E 1 P-E 2 P distribution might be driven by the redistribution of cations within intraor extracellularly oriented high-field access channels to the transport sites. The electrogenicity during the H ϩ branch of the cycle is most likely counterbalanced by another partial reaction of reversed electrogenicity in the K ϩ branch to bring about net electroneutrality (7).
Notably, mutagenesis studies have shown that Na,K-ATPase and H,K-ATPase extrude Na ϩ and H ϩ (or H 3 O ϩ ), generally utilizing the same conserved carboxylic acids of their respective cation binding sites in the transmembrane domains TM4 2 to TM6 (see Fig. 1). This raises not only the important question of how the different stoichiometries are achieved on a molecular level but also why these carboxyls with an expected pK around 3-5 can release protons at a lumenal pH of ϳ1 in the case of the gastric H,K-ATPase. Finally, a remarkable particularity of the gastric H,K-ATPase (and also some other P-type proton pumps like the plant/fungal H ϩ -ATPase) is its inability to run backward and synthesize ATP (8,9), whereas the reverse operation has readily been observed for Na,K-ATPase (10 -16) and Ca 2ϩ -ATPases (17)(18)(19)(20) under sufficiently steep ion gradients and high ADP/ ATP ratios in presence of P i .
Notably, all these unique H,K-ATPase properties have been linked to the presence of a lysine residue in the fifth transmembrane segment of the catalytic ␣-subunit (Lys-791 in gastric H,K-ATPase (shown in blue in Fig. 1 and in the sequence alignments in Fig. 9). This lysine in the otherwise highly conserved (K/S)NIPEIT sequence motif is replaced by an uncharged serine in Na,K-ATPases (see P-type ATPase alignments in Fig. 9). Remarkably, it is the only positively charged amino acid in the whole TM region of the H,K-ATPase ␣-subunit. Homology modeling of the cation binding pocket based on the SERCA (sarco(endo)plasmic reticulum calcium ATPase) structure in the E 2 -state together with mutagenesis studies have predicted an E 2 -conformation-specific salt bridge between the side chains of Lys-791 (in TM5) and Glu-820 (in TM6, highlighted in brown in Figs. 1 and 9) of the cation binding pocket (21). Koenderink et al. (21) concluded that this interhelical interaction could contribute to the inherent E 2 preference of the gastric H,K-ATPase, which in turn could be relevant for preventing a backward running of the pump (see "Discussion"). Moreover, molecular dynamics simulations by Munson et al. (22) have shown that the positively charged side chain of Lys-791 probably reorients during the E 1 P 7 E 2 P transition as a consequence of a change in the relative positions of the TM5 and TM6 helices. The reorientation may move the NH 3 ϩ group of the side chain closer to the plane of the ion binding site in E 2 P (Fig. 1), thereby lowering the effective pK a of the putative H 3 O ϩ -coordinating carboxylates. This change in pK a could enable proton release at low lumenal pH (22,23).
Apart from this proposed function as a pK a -modulating molecular device, Lys-791 is possibly also a major determinant for the electroneutral transport mode characteristic for H,K-ATPases, as suggested by mutagenesis studies on the non-gastric H,K-ATPase expressed in Xenopus oocytes. Upon mutation of the corresponding Lys-800 to Ala or Glu, the normally electroneutral Na ϩ /K ϩ exchange of wild-type toad bladder H,K-ATPase (sometimes also termed X,K-ATPase to differentiate it from the "gastric" H,K-ATPase) becomes electrogenic, thus, generating positive pump currents in the presence of extracellular K ϩ . Vice versa, a charge-introducing mutation of the respective serine to a positively charged arginine in toad Na,K-ATPase resulted in a loss of pump currents with minor effects on Rb ϩ transport activity (24). However, the interpretation of these ion transport studies on the non-gastric H,K-ATPase was complicated by the fact that both Na ϩ and H ϩ ions are transported in exchange for K ϩ and that the H ϩ (Na ϩ )/K ϩ transport stoichiometry is not known. Because the non-gastric H,K-ATPase is more similar to Na,K-ATPases in terms of transported ions (25,26) and the sensitivity toward ouabain (27) and SCH28080 (28,29), the implications of these findings for the gastric enzyme are unclear so far.
These limitations prompted us to investigate the functional relevance of Lys-791 for electroneutral transport of a bona fide H ϩ -transporting H,K-ATPase, the rat gastric H,K-ATPase expressed in Xenopus oocytes. We took advantage of a variant gastric H,K-ATPase ␣-subunit with a single cysteine replacement S806C in the extracellular TM5/TM6 loop (see Fig. 1) that enabled us to study several Lys-791 mutants also by voltage clamp fluorometry (VCF) upon site-specific labeling with tetramethylrhodamine-6-maleimide (TMRM), as described previously (30 -32). In addition to the information regarding steady-state transport, which was provided by Rb ϩ uptake measurements, this method allows the study of voltage-dependent conformational changes under presteady-state conditions. Moreover, VCF not only reveals the voltage-dependent distribution between E 1 P/E 2 P states but also the kinetics of the E 1 P 7 E 2 P conformational transition.

EXPERIMENTAL PROCEDURES
Molecular Biology and Oocyte Preparation-The cDNAs of the rat gastric H,K-ATPase ␤-subunit and a modified form of the ␣-subunit with a single cysteine replacement in the TM5/ TM6 extracellular loop (S806C, see Fig. 1) were subcloned into vector pTLN (33). This cysteine replacement is homologous to the N790C mutation in the TM5/TM6 loop of the Na,K-ATPase ␣-subunit (34,35). It has been shown previously that the S806C mutation enables site-specific labeling of H,K-ATPase with the environmentally sensitive fluorophore Shown is a homology model of the cation binding pocket of rat gastric H,K-ATPase based on the Na,K-ATPase crystal structure in the K ϩ occluded E 2 -state (PDB code 3B8E). Selected residues in TM4, TM5 and TM6 (yellow, red, and green cylinders) involved in cation coordination of the gastric H,K-ATPase are represented as colored sticks. Lys-791, which was mutated in the present study, is shown in blue; Glu-820, which potentially forms an E 2 -specific salt bridge with Lys-791, is illustrated in brown. The reporter site S806C in the TM5/TM6 loop used for labeling with TMRM (in magenta) is also indicated (in pink).
TMRM, and rubidium uptake measurements confirmed that the ␣S806C mutation did not affect the transport properties of gastric H,K-ATPase (32). Additional amino acid replacements at positions Lys-791 and Glu-820 were introduced into the ␣S806C reference construct (which is referred as "wildtype" or WT throughout the current study) using the QuikChange XL site-directed mutagenesis kit (Agilent Technologies) and verified by DNA sequencing (Eurofins MWG Operon).
Xenopus oocytes were obtained by collagenase treatment after partial ovarectomy from Xenopus laevis females. cRNAs were prepared using the SP6 mMessage mMachine kit (Applied Biosystems). A 50-nl aliquot containing 20 -25 ng of H,K-ATPase ␣-subunit cRNA and 5 ng of H,K-ATPase ␤-subunit was injected into each cell. After injection, oocytes were kept in ORI buffer (110 mM NaCl, 5 mM KCl, 2 mM CaCl 2 , 5 mM HEPES, pH 7.4) containing 50 mg/liter gentamycin at 18°C for 2 days. Before all experiments, which were usually carried out 2-3 days after injection at 21-24°C, oocytes were preincubated in solutions containing 100 M ouabain (Sigma) to inhibit the endogenous Xenopus Na,K-ATPase.
Rb ϩ Uptake Measurements-Oocytes were incubated for 15 min in Rb ϩ flux buffer (5 mM RbCl, 85 mM tetramethylammonium chloride, 20 mM tetraethylammonium chloride, 5 mM BaCl 2 , 5 mM NiCl 2 , 10 mM MES, pH 5.5, 100 M ouabain). After 3 washing steps in Rb ϩ -free washing buffer (90 mM tetramethylammonium chloride or NaCl, 20 mM tetraethylammonium chloride, 5 mM BaCl 2 , 5 mM NiCl 2 , 10 mM MES, pH 5.5) and 1 wash in water, each individual oocyte was homogenized in 1 ml of Millipore water. For inhibition experiments, the K ϩ -competitive inhibitor SCH28080 (Sigma) was added to the preincubation solution and Rb ϩ flux buffer (to final concentrations as indicated by the respective figure legends to Figs. 3, 6, and 8). For vanadate inhibition experiments, the respective oocytes were injected with 50 nl of a solution containing 100 mM sodium orthovanadate (buffered with 10 mM Hepes at pH 7.4) ϳ30 -40 min before Rb ϩ uptake measurements. Assuming an oocyte volume of about 1 l, this corresponds to a final intracellular concentration of ϳ5 mM. Whereas the inhibitor vanadate interacts specifically with the E 2 conformational state, SCH28080 is specific for both E 2 and E 2 P conformations (36 -38). Therefore, testing the inhibition efficiency by these compounds can be used to probe conformational preferences of H,K-ATPase variants (38,39).
Atomic Absorption Spectrometry-Aliquots of 20 l from oocyte homogenates were analyzed by atomic absorption spectrometry using an AAnalyst800 TM spectrometer (PerkinElmer Life Sciences) equipped with a transversely heated graphite furnace using a temperature protocol according to manufacturer's procedures (conditions available on request). Absorption was measured at 780 nm using a rubidium hollow cathode lamp (Photron, Melbourne, Australia). After Zeeman-background correction, Rb ϩ contents were calculated by comparison with standard calibration curves (measured between 0 and 50 g/liter Rb ϩ ).
Western Blot Analysis of Isolated Plasma Membranes-The procedures for isolation of plasma membranes and total cellu-lar membranes from Xenopus oocytes, gel electrophoresis, and immunoblotting were performed according to the protocols in Dürr et al. (31). The polyclonal antibody HK12.18 (Merck) (40) was used for detection of gastric H,K-ATPase ␣-subunit. We have shown previously that the plasma membrane fraction obtained by this procedure does not contain protein from intracellular membranes, as the H,K-ATPase ␣-subunit was not detected in this fraction when it had been co-expressed without ␤-subunit, although the ␣-subunit protein was clearly present in the total cellular membrane fraction (32). Furthermore, this study also demonstrated that the characteristic glycosylation pattern of the H,K-ATPase ␤-subunit present in isolated plasma membranes is another supportive indicator for the purity of the preparation (see Fig. 2A in Ref. 32).
Voltage Clamp Fluorometry-Site-specific labeling of H,K-ATPase-expressing oocytes was achieved by incubating oocytes in HK 7.4 buffer (90 mM NaCl, 20 mM tetraethylammonium chloride, 5 mM BaCl 2 , 5 mM NiCl 2 , 10 mM MOPS/Tris, pH 7.4) containing 5 M TMRM (Molecular Probes) for 5 min at room temperature in the dark followed by extensive washes in dye-free HK 7.4 buffer. Voltage clamp fluorometry measurements were carried out under high extracellular Na ϩ /K ϩ -free conditions for the characterization of H,K-ATPase mutants in HK 5.5 buffer (90 mM NaCl, 20 mM tetraethylammonium chloride, 5 mM BaCl 2 , 5 mM NiCl 2 , 10 mM MES/Tris, pH 5.5, 100 M ouabain). Due to the aforementioned electrogenicity of the H ϩ branch of the catalytic cycle, characteristic fluorescence changes are observed in response to voltage pulses under these conditions, reflecting the voltage dependence of the E 1 P-E 2 P conformational transition (30 -32). Several control measurements demonstrating the specificity of the voltage jump-induced fluorescence signals are presented in Fig. 2. Details on setup components, data acquisition, and analysis are given in Dürr et al. (31). For analysis of the voltage-dependent fluorescence signals, the fluorescence data traces were fitted with a single exponential function, and the stationary amplitudes from the fits were plotted against the membrane potential to be approximated with a Boltzmann-type function, in which F min and F max are the minimal and maximal fluorescence amplitude from the fit, z q is the equivalent charge (slope factor), V 0.5 is the half-maximal voltage, F is the Faraday constant, R is the molar gas constant, and T is the absolute temperature in K. The limited voltage range accessible in two-electrode voltage clamp experiments on oocytes together with the low voltage dependence of the H,K-ATPase did not allow direct determination of the saturation values of the voltage-dependent fluorescence amplitudes. However, already the slight deviations of the voltage-dependent distributions from linearity impose sufficient constraints on fits of a Boltzmann function to the data that consistent saturation values are obtained independent of the choice of start parameters. This is also valid for data sets, which cover the Boltzmann distribu-tion only to about the V 0.5 value. Any variation of the saturation values from the parameters determined by the fits resulted in significantly increased 2 values. After fitting the data with a Boltzmann function, an offset corresponding to F min was subtracted, and the resulting fluorescence amplitudes were normalized to the value of F max Ϫ F min . The resultant 1 Ϫ ⌬F/F distributions presumably reflect the voltagedependent distribution of the H,K-ATPase between E 1 P and E 2 P states, in analogy to the interpretation of the fluorescence responses of the homologous Na,K-ATPase mutant N790C (35). Homology Model of the Rat Gastric H,K-ATPase in the E 2 State-A rough molecular model of the gastric H,K-ATPase structure in the E 2 state was built using the crystal structure of the pig renal Na,K-ATPase (PDB structure entry 3B8E) in the K ϩ -occluded E 2 state (41) to illustrate the location of re-porter site S806C and positions of selected residues of the putative cation binding pocket, including Lys-791 and Glu-820 ( Fig. 1). After a few manual adjustments to an alignment between the rat gastric H,K-ATPase ␣-subunit and the pig Na,K-ATPase ␣ 1 -isoform (according to the P-type ATPase A and B, shown are fluorescence signals of an individual oocyte expressing the ␣S806C construct together with the H,K-ATPase ␤-subunit before (A) and after (B) the addition of the specific inhibitor SCH28080 (100 M). C and D, shown are voltage jump-induced fluorescence responses from two oocytes from the same batch, which were either injected with the cRNA of the ␣S806C construct alone (D) or co-injected with the H,K-ATPase ␤-subunit (C). E, shown is normalized cell fluorescence of TMRM-labeled oocytes, which were either uninjected or expressed the H,K-ATPase ␣S806C construct together with the wild-type ␤-subunit. Data were obtained from three batches of cells, and the fluorescence intensity was normalized to the mean fluorescence of uninjected oocytes in each batch. For the measurement of cellular fluorescence, all cells from each batch of oocytes were placed into the perfusion chamber of the experimental microscope and illuminated with constant excitation light intensity. A constant region of interest was chosen for all cells using a circular iris aperture that allowed measurement of the fluorescence of about 90% of the illuminated cell surface. a.u., arbitrary units.  DECEMBER (42,43) was used for the generation of a homology model and subsequent minimization. The structural model of TMRM was prepared with ChemDraw Ultra 11.0. The program VMD (44) was used for the graphical representation in Fig. 1.

Rb ϩ Uptake and Cell Surface Expression of Lys-791
Mutants-Because mutation of Lys-791 in the non-gastric H,K-ATPase produced interesting effects on pump stoichiometry resulting in electrogenic pump activity, we first examined whether the corresponding amino acid replacements in rat gastric H,K-ATPase would give rise to K ϩ -stimulated pump currents upon expression in Xenopus oocytes. However, we were unable to measure any K ϩ -induced pump currents for all investigated Lys-791 mutants (data not shown). Therefore, we also assessed the Rb ϩ uptake activity of these mutants under saturating Rb ϩ concentrations (5 mM) by atomic absorption spectrometry. As shown in Fig. 3A, all Lys-791 variants (including the charge-conserving K791R mutant) exhibited substantially reduced transport activities in Rb ϩ uptake measurements, reaching only about 15-30% of the wild-type activity. Of note, the Lys-791 mutants also displayed a largely reduced sensitivity toward the E 2 -specific H,K-ATPase inhibitor SCH28080 (37), in agreement with previous observations on crude membrane preparations from HEK293 cells expressing the K791A mutant gastric H,K-ATPase (45). Unfortunately, however, because of the low Rb ϩ uptake activities of these mutants, a more detailed quantification of the effect (e.g. determination of IC 50 values) was impossible.
Because none of the previous studies on Lys-791 variants addressed the question of whether the plasma membrane targeting of the H,K-ATPase mutants is affected, we had to exclude that the reduced Rb ϩ uptake activities could be a simple consequence of an impaired cell surface delivery. Western blot analysis of isolated plasma membranes from Xenopus oocytes expressing these Lys-791 variants showed similar protein levels for all Lys-791 mutants compared with the wildtype, except for the charge-inverting K791E mutant, for which the amount of protein in plasma membrane fractions was reduced ( Fig. 3, B, upper panel, and C). However, the reduced cell surface expression of the K791E protein cannot fully account for the strongly reduced Rb ϩ uptake of this mutant, and also for the other Lys-791 mutants the strongly reduced Rb ϩ transport activity cannot be due to impaired plasma membrane targeting. E 1 P/E 2 P Conformational Distribution and Kinetics of E 1 P 7 E 2 P Transition of Lys-791 Mutants-To investigate whether these mutations influence the kinetics or voltage de-pendence of the E 1 P/E 2 P conformational distribution, we inserted the Lys-791 mutations into the backbone of H,K-ATPase mutant S806C, which carries a reporter cysteine for site-specific fluorescence labeling. This strategy enabled us to measure conformation-dependent fluorescence changes of mutant H,K-ATPases in response to voltage pulse protocols (insets in Fig. 4 A and F). Notably, all Lys-791 mutants showed substantially altered fluorescence signals compared with the wild-type pump (Fig. 4, A-G). Voltage jumps to hyperpolarizing potentials, which drive the enzyme into the E 1 P state, resulted in smaller relative fluorescence changes for all four mutants compared with the wild-type (Fig. 4A), whereas the fluorescence responses induced by jumps to positive potentials (favoring conversion to E 2 P) were larger and profoundly slower than the corresponding signals of the wild-type protein. A noticeable feature of the fluorescence changes observed for these variant pumps is the presence of small sharp peaks that occur immediately after an abrupt change in the membrane potential (for both on and off pulses, see the small arrows in Fig. 4, B and C). Most likely, this peak reflects a rapid initial change in the microenvironment of the fluorophore that cannot be resolved due to the limited time resolution in two-electrode voltage clamp experiments. This environmental change, which is not observed in the S806C wild-type H,K-ATPase, is possibly a specific consequence of the charge-neutralizing Lys-791 mutations in the vicinity of the TMRM labeling site S806C in the TM5/TM6 loop (see Fig. 1). Of note, this effect was much less pronounced for replacements by charged amino acids (K791E and K791R in Fig.  4, F and G).
The above described alterations in the fluorescence signals of the Lys-791 mutants indicate a conformational shift toward the E 1 P-state, which can be further quantified by plotting the steady-state amplitudes (from monoexponential fits to the fluorescence changes) at each membrane potential (Fig. 4, I-L). In fact, the resulting Boltzmann curves of all investigated mutants are strongly shifted to more positive potentials compared with the wild-type (V 0.5 values from fits of a Boltzmann function to the shown curves are given in each panel of Fig. 4, I-L). Interestingly, the Boltzmann parameters obtained for the two charge-neutralizing amino acid replacements (K791S and K791A) and the charge-inverting K791E mutant showed substantially higher slope factors z q (values between 0.43 and 0.67 compared with ϳ0.30 for the reference construct S806C), whereas this parameter was not significantly changed for the charge-conserving K79IR variant. The higher z q values for the former mutants hint at an increased voltage dependence of the E 1 P 7 E 2 P conformational transition. A  DECEMBER 10, 2010 • VOLUME 285 • NUMBER 50 more detailed kinetic analysis of the Lys-791 mutants (see supplemental Appendix B for details) revealed that the increased z q values of these mutants can be directly attributed to a significantly increased voltage dependence z q1 of the forward rate constant k f for the E 1 P 3 E 2 P conformational transition (see calculated z q1 and z q2 values in Table 1).

Lys-791/Glu-820 Mutations Slow Turnover of Gastric H,K-ATPase
The aforementioned changes in the time course of the fluorescence signals at jumps to positive membrane potentials likely reflect severely altered kinetics of the E 1 P 3 E 2 P conformational transition. To quantify the effect, we also determined the voltage-dependent reciprocal time constants from fits of a single exponential function to the fluorescence traces (Fig. 4H). This panel shows that for all investigated Lys-791 mutants the reciprocal time constants were significantly smaller than the wild-type values. Although the whole voltage range is affected, the most pronounced reductions of the 1/ values occur at positive potentials, which readily explains the strong shifts of the E 1 P/E 2 P conformational distribution toward the E 1 P state (as follows). The observed reciprocal time constants 1/ are the sum (k tot ) of the rate constants for the forward (k f ) and the backward reaction (k b ) of the E 1 P 7 E 2 P conformational transition. Therefore, the individual contributions of the forward and reverse rate constant can be calculated at each membrane potential using a simple two-state kinetic model to describe the E 1 P 7 E 2 P conformational transition (for details, see supplemental Appendix A). As shown in Fig. 5, the forward rate constants k f (E 1 P 3 E 2 P, red squares in Fig. 5, B-E) obtained by this procedure are between 5-and 20-fold decreased for all four Lys-791 variants compared with the wild-type (red squares in Fig. 5A). In contrast, the reverse rate constants k b (E 2 P 3 E 1 P, blue squares in Fig. 5, B-E) are much less affected by the mutations, with the remarkable exception of mutant K791R, which showed a strongly reduced k b as well. Notably, for all Lys-791 variant pumps, the observed shift toward E 1 P (destabilization of E 2 P) is the consequence of a reduced rate constant for E 2 P formation but not of an accelerated E 2 P decay toward E 1 P.
Importantly, the forward rate constant k f of the E 1 P 7 E 2 P conformational transition might be rate-limiting for steadystate ion transport of the gastric H,K-ATPase under saturating K ϩ /Rb ϩ concentrations. 3 Accordingly, the markedly de-creased forward rate constant of the E 1 P 7 E 2 P conformational transition rationalizes the substantially reduced Rb ϩ uptake activity of all Lys-791 substitutions (Fig. 3A). E 2 P Stabilizing Interaction of Lys-791 to Glu-820-Because homology modeling combined with biochemical studies has suggested an E 2 P-stabilizing salt bridge between the charged side chain of Lys-791 and the adjacent glutamate residue Glu-820 (21), we examined amino acid replacements of Glu-820 as well. If a salt bridge between these two residues could indeed be formed for the sake of E 2 P stabilization, a disruption of the interaction from either side should result in a destabilization of the E 2 P state by essentially the same mechanism.
Although cell surface expression for all investigated Glu-820 substitutions was unaltered (see Fig. 6, A and B and Fig. 8,  A and B), only two Glu-820 mutants displayed fluorescence changes that were large enough to analyze their voltage-dependent properties (Fig. 7, A and B). At first glance, the voltage dependence of the fluorescence signals observed for mutants E820D and E820K appears quite similar to the wild-type signals (compare Fig. 7, A and B to Fig. 4A), but the signals are slightly (E820D) or profoundly (E820K) slower. The resulting Boltzmann curves of the fluorescence amplitudes exhibited a small hyperpolarizing shift, indicating a slightly increased preference for the E 2 P-state compared with the wild-type (Fig.  7, C and D). Kinetic analysis of the fluorescence signals revealed that the reciprocal time constants of both mutants are significantly smaller compared with the wild-type values over the whole voltage range (Fig. 7, E and F). In case of the charge-conserving E820D mutant, the 1/ values are essentially voltage-independent. Calculation of the k f and k b values (Fig. 7G) shows that this is due to the fact that the reverse rate constants k b are more reduced than the k f values, altogether explaining the slight preference of the mutants for the E 2 P state. The charge-inverting E820K mutant exhibited ϳ10-fold slower 1/ values, and the calculation of k f and k b showed (Fig.  7H) that also in this mutant the formation of E 2 P from E 1 P is drastically slowed, as previously observed for the Lys-791 variants. However, for E820K a shift in the conformational distribution toward E 1 P is not observed because the reduction of the k f values is nearly compensated by similar changes in k b . These considerations indicate that the E 1 P 7 E 2 P conforma-3 K. Dü rr and T. Friedrich, unpublished observations.

TABLE 1 Calculated z q values for the forward (k f ) and reverse (k b ) reaction of the E 1 P-E 2 P conformational transition (see supplemental Appendix B for details) and their sum z q(tot) in comparison to the experimentally obtained values z q
Note that the z q1 values (which describe the voltage dependence of the forward rate constant k f ) of mutants K791A, K791S, K791E, and K791R are by a factor of 2-4 increased compared to the wild-type, whereas the respective z q2 values are much less affected by the mutations. Therefore, the observed change in z q(tot) of the Boltzmann distributions (shown in Fig. 4, I-L) can be mainly attributed to an increased electrogenicity of the forward rate constant k f of the two variant pumps.

Lys-791/Glu-820 Mutations Slow Turnover of Gastric H,K-ATPase
tional change is impaired in both directions by the E820K mutation.
Notably, the observed differences in the k f values are qualitatively in good agreement with the mutants' reduced Rb ϩ uptake activities, which are probably rate-limited by the E 1 P 3 E 2 P partial reaction. Whereas the E820D mutant still retains 50 -60% of the wild-type transport activity (Fig. 6C), the charge inversion of the E820K mutant results in a residual activity of ϳ25% (Fig. 8C).
As shown in Fig. 6, C and D, charge-neutralizing substitutions by glutamine or alanine not only eliminated voltageinduced fluorescence signals but also had strong effects on the enzyme Rb ϩ uptake activities. An interesting property of these two mutants is revealed by comparing Rb ϩ uptake measurements at two different extracellular pH values. Whereas Rb ϩ uptake of the wild-type and E820D mutant are significantly increased upon a change in extracellular pH from 7.4 to 5.5, the behavior is inverted for pumps carrying the mutation E820A or E820Q (compare light gray and gray bars in Fig. 6D).
We have demonstrated recently that the increased Rb ϩ uptake activity observed for the gastric H,K-ATPase at pH ex ϭ 5.5 is caused by a slight intracellular acidification (ϳ0.5 pH units) as a consequence of the acidic extracellular pH, because a similar pH change of the cell interior, which can be achieved at pH ex ϭ 7.4 by adding 40 mM butyrate to the extracellular solution, results in very similar Rb ϩ uptake in the case of the wild-type protein (compare light gray and gray-shaded bars in Fig. 6D). The increased availability of intracellular protons at the cytosolic cation binding site presumably accelerates the phosphorylation reaction and the subsequent E 1 P 3 E 2 P conformational transition, which is rate-limiting for the transport activity at saturating Rb ϩ concentrations. Notably, this stimulation of Rb ϩ transport activity by intracellular acidification is maintained for E820D, but it can no longer be observed when Glu-820 is replaced by neutral residues (compare gray and gray-shaded bars in Fig. 6D). Compared with the Rb ϩ uptake activity observed at pH ex ϭ 7.4 in the presence of butyrate (which generates a small outwardly directed H ϩ gradient, ⌬pH ϳ 0.5), the transport activity of mutant E820Q and E820A is significantly diminished in the presence of a larger, inwardly directed pH gradient at pH ex ϭ 5.5 (ϳ⌬pH 1.5-2). In contrast, the magnitude of the ⌬pH has apparently no effect on ion transport of the wild-type and E820D variant pumps. These observations indicate that the charge-neutralizing mutations result in an increased competition of extracellular H ϩ (or H 3 O ϩ ) with Rb ϩ ions at the extracellular binding sites. Therefore, Glu-820 could be crucial for determining K ϩ (hence, also Rb ϩ ) selectivity in the E 2 P state, which is especially important at steep H ϩ gradients.
Charge-inverting Lys-791/Glu-820 Amino acid Replacements-A famous example for an interhelical salt bridge in a membrane transport protein, which was initially proposed by biochemical studies (46 -48) and later confirmed by x-ray crystallography (49), is residue Asp-237 and residue Lys-358 in the lactose permease LacY from Escherichia coli. Notably, neutral substitutions for either of the two charged residues resulted in transport-defective carriers (46). How-  DECEMBER 10, 2010 • VOLUME 285 • NUMBER 50 ever, inversion of the polarity by mutual exchange of both residues maintained substantial transport activity (47). Similarly, for yeast plasma membrane H ϩ -ATPase, single chargeneutralizing mutations of Arg-695 and Asp-730 resulted in defects in folding and biogenesis and reduced ATPase activity, a phenotype that could be overcome when both residues were replaced by neutral amino acids or mutually exchanged (50).

Lys-791/Glu-820 Mutations Slow Turnover of Gastric H,K-ATPase
To test whether the inactive phenotypes of the here-investigated charge-neutralizing Lys-791 and Glu-820 mutants could be "rescued" by an analogous mechanism, we characterized the respective charge-inverting K791E/E820K double mutant. Yet, the Rb ϩ uptake of this variant was not significantly higher than that of the single replacements E820K (Fig.  8C) and K791E (Fig. 3A). Again, we confirmed that the pronounced reduction in Rb ϩ uptake activity of these mutants was not due to a reduced cell surface expression (Fig. 8A). Unfortunately, however, the double mutant displayed no discernable fluorescence changes in response to voltage jumps, not even with largely expanded pulse durations. This might imply that (possibly for kinetic reasons) the E 1 P/E 2 P equilibrium of the K791E/E820K mutant is no longer voltage-sensitive or that the deleterious effects of the single mutations on the conformational dynamics are additive rather than compensatory. Yet, the fact that inversion of the salt bridge polarity does not rescue function does not necessarily exclude that Lys-791 and Glu-820 in the wild-type proton pump interact in an E 2 P-stabilizing manner. Apart from the presence of opposite charges, structural constraints regarding side-chain length and geometry might be critical for the interaction, too. Notably, already the charge-conserving mutations E820D (Fig. 7C) and even more so K791R (Fig. 4, H and L) led to significant functional changes. The mutual exchange of the two charged residues might cause steric clashes, or the charge inversion could be unfavorable within the local electrostatic environment, thus, preventing reversed salt bridge formation in the double mutant.

DISCUSSION
In this study we addressed the functional significance of a positively charged lysine in the fifth transmembrane segment D, H,K-ATPase-mediated Rb ϩ uptake at 5 mM RbCl and pH ex ϭ 7.4 (dark gray bars), pH ex ϭ 5.5 (light gray bars), or at pH ex ϭ 7.4 in the presence of 40 mM butyrate, which causes a slight intracellular acidification (by ϳ 0.5 pH units, hatched dark gray bars), is shown. Results from uninjected control oocytes or oocytes injected with the reference construct HK␣S806C/␤wt or HK␣S806C/E820X (X ϭ Ala, Gln, Asp) are shown. Rb ϩ uptake was measured on individual cells by atomic absorption spectrometry (see "Experimental Procedures"). Data are the means Ϯ S.E. from four individual experiments with 15-20 oocytes, normalized to Rb ϩ uptake of the wild-type construct HK␣S806C/␤wt (corresponding to 23.9, 22.5, 21.2, and 24.9 pmol/oocyte/ min, respectively). Inset, shown is inhibition of Rb ϩ uptake by sodium orthovanadate for oocytes expressing either the wild-type or the E820Q mutant proton pump at 5 mM RbCl and pH 7.4 in presence of 40 mM butyrate (black bars, see "Experimental Procedures" for details). One representative experiment is shown.

Lys-791/Glu-820 Mutations Slow Turnover of Gastric H,K-ATPase
of the gastric H,K-ATPase, Lys-791, as it was shown previously that the homologous residue in non-gastric H,K-ATPase is crucial for electrogenicity (24). In contrast to a study on a X,K-ATPase from toad bladder in which ouabainsensitive, K ϩ -stimulated stationary currents for two different amino acid replacements of the aforementioned lysine in TM5 (24) were reported, we did not observe any pump currents for the corresponding Lys-791 mutants of the gastric H,K-ATPase (K791A and K791E). However, our voltage clamp fluorometric characterization for several Lys-791 mutants revealed significant changes in the z q parameter, which reflects the electrogenicity of a partial reaction that occurs in conjunction with the E 1 P-E 2 P conformational change of the pump cycle. Of note, all Lys-791 replacements, which neutralized or inverted the charge of the side chain, resulted in a significantly increased z q , whereas the parameter was unaffected for the charge-conserving K791R mutant (Fig. 4, I-L).
Furthermore, the mutants-altered voltage dependence of fluorescence amplitudes indicated a strong shift of the E 1 P/ E 2 P distribution toward E 1 P. This is in agreement with their reduced sensitivity toward the E 2 -specific inhibitor SCH28080 in Rb ϩ uptake measurements (Fig. 3A) and might provide an alternative mechanistic explanation for the severely reduced SCH28080 sensitivity of the K791A mutant found in biochemical studies (45), in contrast to a direct effect on the inhibitor binding site.
Additional kinetic analysis of the fluorescence signals showed that the observed E 1 P-shift of all these mutants is apparently a consequence of a markedly decreased forward rate constant of the E 1 P 7 E 2 P conformational transition (Figs. 4H and 5). The concomitantly reduced Rb ϩ transport activity suggests that this partial reaction step might critically limit the turnover rate of the enzyme.
The E 1 P-shifted phenotypes of the here-investigated Lys-791 mutants corroborate results from earlier mutagenesis studies on the gastric H,K-ATPase, which predicted an E 2specific, interhelical salt bridge between Lys-791 in TM5 and Glu-820 in TM6 (21). Homology modeling of the cation binding pocket (Fig. 1) suggested that the putative salt bridge is an exclusive feature of the E 2 -form due to the large distance between the two residues in an E 1 -state model of the gastric H,K-ATPase. If the salt bridge would energetically contribute to the stability of the E 2 or E 2 P states, a disruption of the interaction by introduction of charge-neutralizing or sterically unfavorable amino acids in either of the two positions is expected to selectively destabilize the E 2 (P) conformation. This notion is indeed supported by the alterations in the voltagedependent fluorescence changes and the reduced rate constants (k f ) for the formation of E 2 P, as observed for the Lys-791 and Glu-820 variants studied in this work.
From the observed shifts of the E 1 P/E 2 P distributions (⌬V 0.5 ϳ 160 mV) with a z q of about 0.5, the relative destabilization of E 2 P would be by more than ϳ3 kT (ϭ 0.08 electron volt (1 kT ϭ 0.025 electron volt)), which is well in the order of magnitude of strong H-bond interactions in proteins. Similar values (3-5 kcal/mol ϳ5-8 kT) were reported for the energy contribution of salt bridges in other proteins (e.g. T4 Lysozyme (51,52)).
Unfortunately, the lack of fluorescence signals did not allow us to reveal similar E 1 P-shifted phenotypes for charge-eliminating replacements of the partner residue (Glu-820) within this putative charge pair (mutants E820A and E320Q). However, it should be noted that these two mutations were shown to result in a constitutive (K ϩ -independent) ATPase activity of the proton pump expressed in Sf9 cells (53). This finding can actually provide an explanation for the small fluorescence changes observed for the Glu-820 mutants, as they are expected to undergo full ATPase cycles even under the K ϩ -free conditions of our VCF experiments. Under these conditions the wild-type pumps usually accumulate in E 2 P, a situation that allows a switch of the conformational equilibrium by voltage jumps, resulting in changes of fluorescence amplitudes. Due to the constitutive activity of these two Glu-820 mutants, the redistribution of the enzyme molecules over all reaction cycle intermediates apparently creates a situation in which the enzyme conformation is insensitive to changes in the transmembrane potential.
According to the afore-stated Sf9 cell studies, both constitutively active mutants also exhibited a significantly reduced sensitivity toward the E 2 -specific inhibitors SCH28080 and vanadate (38,54,55), which is in agreement with the results from our Rb ϩ uptake experiments ( Fig. 6C and the inset in D). Again, these findings hint at E 1 (P)-shifted phenotypes, which are possibly obscured by the constitutive activity of these mutants in the VCF experiments. Importantly, in the current study a charge-conserving replacement of Glu-820 by aspartate maintained the E 2 P preference of the proton pump. This further corroborates the notion of an E 2 P-stabilizing salt bridge between Lys-791 and Glu-820, although the individual energy barriers for the forward and backward transition, as inferred from the changes in the reciprocal time constants, are also modified by the Glu to Asp substitution. FIGURE 7. Voltage dependence of the E 1 P/E 2 P distribution and kinetics of E 1 P/E 2 P transitions of H,K-ATPase mutants E820D and E820K. A and B, shown are fluorescence responses of site-specifically labeled gastric H,K-ATPase under K ϩ -free conditions (90 mM NaCl, pH 5.5) upon voltage jumps from a holding potential of Ϫ40 mV to voltages between Ϫ180 and ϩ60 mV (same voltage protocols as in Fig. 4, A and F). Recordings originated from a representative oocyte coexpressing the wild-type HK␤ subunit with HK␣S806C,E820D in A or HK␣S806C,E820K in B, respectively.

Lys-791/Glu-820 Mutations Slow Turnover of Gastric H,K-ATPase
Unlike the observations for the aforementioned salt bridge in LacY, a charge-inverting double mutation of the proton pump (K791E/E820K) did not restore the wild-type phenotype but behaved rather similarly to the single amino acid replacements in Rb ϩ uptake experiments. However, the altered side-chain geometry or the charge inversion itself could be incompatible with salt bridge formation in the context of the local environment. In addition, it must be noted that in the case of the LacY, neither of the individual residues nor the salt bridge between them plays an important role in the transport mechanism, as simultaneous neutral substitutions of both residues that remove the salt bridge do not cause inactivation of the enzyme (47). This means that the two charges are only required for neutralizing each other, which is still possible if the residues are mutually exchanged. This is in strong contrast to the roles of Lys-791 and Glu-820, which are both part of the cation binding pocket of gastric H,K-ATPase. Thus, interchanging the residues might even allow the formation of an inverted salt bridge but at the same time interfere with other functions of the two side chains for ion transport (e.g. proton storage of Glu-820, see below).
On the other hand, one could argue that the pronounced effects of the Lys-791 and Glu-820 mutations on Rb ϩ transport activity might be a direct consequence of the involvement of the residues in cation binding. However, in case of the E822D mutant of the rabbit gastric proton pump (corresponding to mutation E820D here), Asano et al. (56) found a ϳ60% reduction in K ϩ -stimulated ATPase activity but at the same time demonstrated that the K ϩ affinity was unaffected. Notably, our VCF data revealed a significantly reduced forward rate constant of the E 1 P-E 2 P conformational transition for this mutant (Fig. 7E), therefore, nicely explaining the phenotype and corroborating our assumption that this partial reaction is rate-limiting under saturating K ϩ concentrations.
Apart from its possible participation in an interhelical salt bridge, our Rb ϩ uptake experiments hint at a potential role of the Glu-820 side chain in intracellular proton binding, as the sensitivity of the proton pump toward intracellular acidification is eliminated upon replacements by non-protonatable amino acids. A rather conservative replacement of Glu-820 by a protonatable aspartate, on the other hand, maintains the stimulatory effect of intracellular acidification but results in slightly reduced Rb ϩ uptake activity under all investigated pH conditions (Fig.  6D). This is in agreement with previous results from Sf9 cells, which revealed a reduced apparent ATP affinity for the E820D variant in ATP phosphorylation experiments. Because a pH change from 7.0 to 6.0 increased the apparent ATP affinity for both wild-type and the E820D mutant, the authors concluded that the reduced ATP affinity of the mutant can be attributed to a decreased H ϩ affinity, which in turn affects the phosphorylation kinetics (57). This is again in line with the reduced forward rate constant of the E 1 P 7 E 2 P conformational transition observed for the E820D mutant in our VCF experiments (Fig. 7E) and its diminished Rb ϩ uptake activity (Fig. 6D). Apparently, the microenvironment around the acidic group (which is critical for its effective pK a ) is slightly altered due to the shorter side chain of the aspartate. Of note, the corresponding residue of Glu-820 in all sodium potassium ATPases is an aspartate (e.g. Asp-804 in the sheep ␣ 1 isoform; see the alignments in Fig. 9). The reduced proton affinity of the E820D mutant described by Hermsen et al. (57) might, thus, provide an explanation for the fact that, in contrast to the case of the gastric H,K-ATPase, the acidic residues involved in ion binding of the Na,K-ATPase do not coordinate protons, at least not to a significant extent at physiological conditions (58). Therefore, the slightly longer side chain of Glu-820 together with a different local electrostatic environment (e.g. Lys-791, see below) seem to be important factors in determining H ϩ affinity of the gastric H,K-ATPase. The acidic side chain of Glu-820 could represent a site where protons are bound before being expelled to the extracellular space. The immediate formation of a salt bridge with Lys-791 after expulsion of the proton from this site could be crucial for preventing reprotonation of the site from the lumenal space at a physiological pH of ϳ1. The site probably stays occupied by the positive side chain of Lys-791 until the cation binding pocket is no longer exposed to the gastric lumen, i.e. after closure of the extracellular gate. To enable subsequent reloading of the site with a H ϩ from the cytosolic space, the salt bridge must be transiently disrupted. This most likely occurs during the E 2 -E 1 conformational transition, causing a rearrangement of the Lys-791 side chain. The here-proposed pumping model would provide a rationale for the concept that the salt bridge is E 2 conformation-specific.
Remarkably, similar interhelical salt bridges between charged residues in TM5 and TM6 have been observed in other proton pumping P-type ATPases as well. For example, mutagenesis studies on the plasma membrane H ϩ ATPase PMA1 from Saccharomyces cerevisiae hint at a charge pair in this area (50), which however involves other residues, i.e. Arg-695 in TM5 and Asp-730 in TM6 (highlighted in the sequence alignment in Fig. 9 in yellow and green, respectively). Of note, the corresponding Asp-684 (also highlighted in green in Fig. 9) of the plant plasma membrane H ϩ -ATPase from Arabidopsis thaliana (AHA2) probably interacts with yet another arginine in TM5 (Arg-655, highlighted in yellow in Fig. 9) according to the recent crystal structure (59). The authors suggested that this arginine could serve as a built-in counterion during phosphorylation, as the pump does not countertransport any other ion in exchange for protons. Furthermore, similar to the here proposed role of Lys-791 in the gastric H,K-ATPase, the arginine may serve as a positive plug that prevents extracellular protons from re-protonating Asp-684 (presumably the central proton donor/acceptor of the pump) in the E 2 P state, when the proton exit pathway opens to the extracellular space (59).
Although there is currently no experimental proof available that the aforementioned putative salt bridges in other H ϩ -transporting P-type ATPases are also specific for the E 2 P-state, the close similarity regarding their respective location is striking. Because putative salt bridge-forming charges in M5 and M6 are conserved among other H ϩ -translocating P-type ATPases (Fig.  9), a common functional importance of these salt bridges for H ϩ transport seems reasonable. Of note, all these H ϩ ATPases release their transported protons in the E 2 P state. Hence, the conserved salt bridge as a potential device for preventing reprotonation of the proton release site would be required in the E 2 P conformation but must be removed in E 1 to enable protonation from the cytoplasmic space.

CONCLUSIONS
According to the current study, mutations of Lys-791 in TM5 of gastric H,K-ATPase do not change the overall electrogenicity of ion transport. Rather, this lysine is essential for a rate-limiting partial reaction of the transport cycle, the E 1 P 3 E 2 P conformational transition, and thus, is crucial for transport activity. Our results suggest that Glu-820 in TM6 is a critical residue for intracellular proton binding and might serve as a putative partner for an electrostatic interaction to Lys-791, a salt bridge that could contribute to the inherent E 2 P preference of the gastric H,K-ATPase and prevent reprotonation of Glu-820 after proton release to the extracellular space. This salt bridge between TM5 and TM6 might be a universal feature of H ϩ -translocating P-type ATPases to avoid futile pump cycling, unlike other P-type ATPases, which can FIGURE 9. Alignment of TM5 and TM6 from P-type ATPases of the P IIC -and P IIIA -type subfamilies. Sequence alignments were adapted from Axelsen and Palmgren (60) and adjusted manually for comparison of P IIC -and P IIIA -type ATPases. The sequence of the rat gastric H,K-ATPase used for mutagenesis in the present study, and sequences from other P-type ATPases for which individual residues are explicitly discussed here are framed by red boxes. TM5/TM6 loops are underlaid in black, and charged residues that are involved in putative interhelical salt bridges between TM5 and TM6 are highlighted in different colors (color coding is analogous to Fig. 1). The position (␣S806C) used for site-specific TMRM-labeling of the gastric H,K-ATPase is highlighted in magenta. Residues of the rat gastric H,K-ATPases that are shown in Fig. 1 are underscored. be readily forced to run backward. This might be an essential requirement for efficient ion transport of H ϩ transporting ATPases, which face the steepest electrochemical gradients of all P-type ATPases.