Elucidating the Role of Conserved Glutamates in H (cid:1) -pyrophosphatase of Rhodospirillum rubrum*

H (cid:1) -pyrophosphatase (H (cid:1) -PPase) catalyzes pyrophos-phate-driven proton transport against the electrochemical potential gradient in various biological membranes. All 50 of the known H (cid:1) -PPase amino acid sequences contain four invariant glutamate residues. In this study, we use site-directed mutagenesis in conjunction with functional studies to determine the roles of the glutamate residues Glu 197 , Glu 202 , Glu 550 , and Glu 649 in the H (cid:1) -PPase of Rhodospirillum rubrum (R-PPase). All residues were replaced with Asp and Ala. The resulting eight variant R-PPases were expressed in Escherichia coli and isolated as inner membrane vesicles. All substitutions, except E202A, generated enzymes capable of PP i hydrolysis and PP i -energized proton translocation, indicating that the negative charge of Glu 202 is essential for R-PPase function. The hydrolytic activities of all other PPase variants were impaired at low Mg 2 (cid:1) concentrations but were only slightly affected at high Mg 2 (cid:1) concentrations, signifying that catalysis proceeds through a three-metal pathway in contrast to wild-type R-PPase, which employs both methyl)methylamino]-2-hydroxypropanesulfonic acid; TES, 2-{[2-hydroxy-1,1-bis(hydroxymethyl)ethyl]amino}ethanesulfonic acid; Tricine, N -[2-hydroxy-1,1-bis(hydroxymethyl)ethyl]glycine.

Proton pumping pyrophosphatase (H ϩ -PPase) 1 is an integral membrane protein that utilizes the energy released upon the hydrolysis of PP i to transport protons across the membrane against the electrochemical potential gradient (1)(2)(3). H ϩ -PPases represent a distinct class of ion translocases with no sequence similarity to ubiquitous ATP-energized pumps such as the F-, V-, or P-type ATPases or ABC transporters (4). The distribution of H ϩ -PPase among the species is wide, but sporadic. H ϩ -PPase has been identified in most plants, some algae, protists, bacteria, and Archaea, but not in mammals. In prokaryotic species, H ϩ -PPase resides in the cytoplasmic membrane and pumps protons away from the cytoplasm, whereas in eukaryotic species the enzyme acidifies internal organelles such as vacuoles in plants (2,3) and acidocalcisomes in protozoa (5). Moreover, H ϩ -PPases have recently been identified in the plasma membranes of protozoa (6,7). Both hydrolytic and proton translocation activities are associated with a single polypeptide of 66 -90 kDa (8 -11), which possibly forms a dimer (10,12). H ϩ -PPase is a highly hydrophobic protein, as evident from the 14 -16 transmembrane spans predicted by computer modeling (Fig. 1). Sequences from different species display at least 30% identity, and most of the 60 conserved residues are clustered in three hydrophilic regions comprising cytosolic loops that probably form the active site. These residues include four Glu residues ( Fig. 1) that are present in all 50 of the known H ϩ -PPase sequences.
H ϩ -PPase acts specifically on PP i and requires Mg 2ϩ for activity (13)(14)(15). Mg 2ϩ binds and activates both free enzyme and PP i . Some H ϩ -PPases additionally require potassium for activity (2,16). A common property of H ϩ -PPases that distinguishes them from their soluble counterparts is high sensitivity to competitive inhibition by aminomethylenediphosphonate (AMDP) and relative insensitivity to inhibition by fluoride (17). Moreover, H ϩ -PPases are highly sensitive to sulfhydryl reagents, such as mersalyl and N-ethylmaleimide, and the carboxyl group reagent dicyclohexylcarbodiimide (2).
The development of expression systems for plant H ϩ -PPases in yeast Saccharomyces cerevisiae (9) and bacterial H ϩ -PPases in Escherichia coli (18) has paved the way for site-directed mutagenesis of these proton pumps. This approach was successfully employed to identify membrane-embedded charged residues contributing to dicyclohexylcarbodiimide binding in H ϩ -PPases (19), reactive Cys residues (18,20), and those residues that determine the potassium requirement of H ϩ -PPases (16). The role of conserved residues possibly located in the cytosol is more difficult to analyze, because their substitution results in an inactive enzyme (21). Here we perform site-directed mutagenesis of all four conserved glutamate residues, Glu 197 , Glu 202 , Glu 550 , and Glu 649 , in the H ϩ -PPase of Rhodospirillum rubrum. All of the residues were functionally substituted with aspartates or alanines except Glu 202 , which was only replaceable by aspartate. All of the glutamates appeared to be essential for catalytic efficiency at low Mg 2ϩ concentrations. Our data indicate that Glu 197 and Glu 202 contribute to substrate binding, whereas Glu 550 and Glu 649 control R-PPase conformation.

EXPERIMENTAL PROCEDURES
Plasmid Construction and R-PPase Expression-The construct comprising the full-length R-PPase gene (22) cloned into the pET22b(ϩ) vector under the control of the T7 lac promoter was described previously (18). Mutagenesis was performed by an overlapping PCR technique using the Stratagene QuikChange TM kit and the primers listed in Table  I. All mutations were verified by DNA sequencing. Wild-type and variant R-PPases were expressed in E. coli C43(DE3) cells, and inner membrane vesicles (IMVs) were isolated as described previously (18). Protein concentrations in E. coli IMV were estimated using the Bradford assay (23).
Activity Measurements-PP i hydrolysis was assayed by continuously recording P i liberation with an automatic P i analyzer (24) at 25°C. Reactions were performed in an initial volume of 25 ml including pH buffer, calculated amounts of MgCl 2 and Na 4 PP i , and 5-50 l of IMV suspension. The reaction was initiated by the addition of IMVs in measurements of rate as a function of [Mg 2ϩ ] and [Mg 2 PP i ] or the simultaneous addition of MgCl 2 and Na 4 PP i in all other measurements. P i liberation was monitored for 3 min. The following 0.1 M buffers were used at a pH equal to their pK a values: MES, pH 6.1; MOPSO, pH 6.9; MOPS, pH 7.2; TES, pH 7.4; TAPSO, pH 7.6; HEPPSO, pH 7.8; Tricine, pH 8.1; TAPS, pH 8.4; CAPSO, pH 9.6; and CAPS, pH 10.4. The pH was adjusted using KOH, and the indicated amounts of EGTA were added as follows: 50 -100 M, pH 6.1-8.1; 10 M, pH 8.4; and 1 M, 9.6 -10.4. The resulting ionic strength of the buffers was 0.1 M, and the K ϩ concentration was 50 mM. In experiments measuring the inhibitory effects of aminomethylenediphosphonate, methylenediphosphonate, or mersalyl, IMVs were preincubated with these reagents for the times indicated in Figs. 4 and 5. H ϩ translocation across the IMV membrane was assayed with acridine orange as described in a previous report (18).
Calculations and Data Analysis-The amounts of MgCl 2 and PP i required to achieve the desired concentrations of free Mg 2ϩ and the Mg 2 PP i complex in the presence of 50 mM K ϩ were calculated using the apparent dissociation constants for the magnesium and potassium complexes of PP i at pH 7.2 (25) in which E represents enzyme, M is Mg 2ϩ , PP is PP i , K M1 , K M2 , K A1 , and K A2 are dissociation constants, K m is the Michaelis constant, and k h and k h Ј depict Mg 2ϩ concentration-independent catalytic constants. Scheme I assumes that enzyme species lacking substrates are in equilibrium with one another, similar to the case with all enzyme-substrate species. This is a minimal scheme in that the addition of more enzyme species did not significantly affect the quality of the fit, whereas omitting any species, except the dimagnesium metal complex (EM 2 ), resulted in a poor fit. The EM 2 complex was stoichiometrically significant for only two variant R-PPases. This scheme is principally similar to that derived for authentic R-PPase in R. rubrum chromatophore membranes (26). The Mg 2ϩ concentration dependences of k h app and k h app /K m app are specified by Equations 2 and 3, shown here, which were used to calculate the initial estimates of all parameters in Scheme I. These initial estimates were further refined by simulta- where k h,ind is the pH-independent value of k h , and K a1 and K a2 are the ionization constants of essential base and acid, respectively. The time courses of the slow phase of R-PPase inactivation by mersalyl were fitted to Equation 5, where A 0 is the residual activity observed after the "instant" inactivation phase, A is activity at time t, k is the second-order rate constant, and where k E,mers and k EM,mers are the second-order rate constants for inactivation at 0 and infinite Mg 2ϩ concentrations, respectively, and K M,mers refers to the dissociation constant for the R-PPase-Mg 2ϩ complex (26). Competitive inhibition of PP i hydrolysis by diphosphonates was analyzed with Equation 7, FIG. 1. The topological model of R-PPase as predicted from the amino acid sequence using the programs HMMTOP (36) and TMHMM (37). The approximate positions of identical amino acid residues in the H ϩ -PPase family are depicted in black. The glutamate residues substituted in this study and mersalyl-reactive cysteine residues (18) are indicated. Similar models have been suggested by other investigators (19,21,22). Single letter amino acid abbreviations are used with position numbers.
a The mutated amino acid codons are underlined.

RESULTS
Hydrolytic and Proton-pumping Activities of R-PPase Variants-Each of the four conserved glutamate residues in R-PPase was individually substituted with aspartate and alanine, and the resulting variants were expressed in E. coli C43(DE3). Western analysis of E. coli IMVs with R-PPase antibodies (22) revealed that the expression levels of all the variants were similar to that of the wild-type enzyme (data not shown). The hydrolytic activities of R-PPase variants were assayed in IMVs at saturating substrate and Mg 2ϩ concentrations. Specifically, the E197D, E197A, E202D, E550D, E550A, E649D, and E649A variants displayed 60, 60, 35, 50, 50, 45, and 20% of the wild-type R-PPase activity, respectively. However, following E202A substitution, activity declined to a level indistinguishable from the background in our system (ϳ 5% of wild-type R-PPase activity). Interestingly, the replacement of Glu 197 or Glu 550 by alanine did not induce a further decrease in activity as compared with substitution with aspartate. Moreover, all variants capable of PP i hydrolysis were additionally active in PP i -energized H ϩ translocation in E. coli IMV (Fig. 2). The magnitude of acridine orange fluorescence quenching correlated with the relative activity of H ϩ -PPases, suggesting that the substitutions did not affect the coupling between PP i hydrolysis and H ϩ translocation.
Steady-state Kinetic Analysis of Mg 2ϩ Activation-To further investigate the role of the conserved glutamates in H ϩ -PPases, a detailed kinetic analysis of the Glu 3 Asp variants was performed. The dependence of the PP i hydrolysis rate on substrate (Mg 2 PP i ) concentration obeyed Michaelis-Menten kinetics at a fixed free Mg 2ϩ concentration (data not shown). The apparent catalytic constant (k h app ) and its ratio with the Michaelis constant (k h app /K m app ) derived from these dependences are depicted in Fig. 3 as functions of Mg 2ϩ concentration. The k h app profiles are similar for the wild-type and all variant enzymes, whereas the k h app /K m app profiles are different. Thus, the k h app /K m app value decreases with increasing Mg 2ϩ concentrations for the wild-type enzyme, increases to a saturating value for the E197D and E202D variants, and displays a maximum for the E550D and E649D variants.
The simplest kinetic model quantitatively accounting for the dependences shown in Fig. 3 is presented in Scheme I. This model implies that substrate (Mg 2 PP i ) binds to a preformed enzyme-metal complex (EM). The dimagnesium-enzyme complex, EM 2 , is included to explain the decrease in k h app /K m app at the high Mg 2ϩ concentrations observed with the E550D and E649D variants. This species is stoichiometrically insignificant for wild-type R-PPase and the E197D and E202D variants. The scheme contains three enzyme-PP i complexes of different Mg 2ϩ content, with only two being catalytically competent. Numerical values for all the parameters in Scheme I that were obtained by fitting Equations 1, 2, and 3 to the rate data, as described under "Experimental Procedures," are listed in Table II. As EM 3 PP is always dominant over EM 2 PP due to cooperative Mg 2ϩ binding by the enzyme-PP i complex (K A1 Ͼ K A2 ; Table II), the quality of the fit, as characterized by the sum of the squares of residuals, was rather insensitive to the value of k h Ј with the wild-type enzyme. Thus, varying k h Ј in the range of 0.4 -3 mol min Ϫ1 mg Ϫ1 (i. e. from 50 to 400% k h ) increased the sum of the residual squares by only 4%. However, setting k h Ј to 0 increased the sum by 80%, resulting in an unsatisfac-  Table II. The symbols on the curves are as follows: f, wild-type R-PPase; E, E197D; ‚, E202D; ƒ, E550D; and Ⅺ, E649D. The S.E. bars for variant data are not shown as they are smaller than the data symbols. tory fit. Therefore, we assumed that k h Ј ϭ k h , as was shown previously for authentic R-PPase isolated from R. rubrum chromatophores (26). In contrast, setting k h Ј to 0 was required to obtain an acceptable fit for all variant R-PPases.
For the three active Glu 3 Ala variants (E197A, E550A, and E649A), the k h app /K m app values were lower by a factor of 3.5-9 at 0.1 mM Mg 2ϩ as compared with that at 2 mM Mg 2ϩ (data not shown). In this respect, alanine variants were similar to aspartate mutants (Fig. 3B). The absolute values of k h app /K m app were similar to those measured for the corresponding Glu 3 Asp variants.
Dependence of k h app on pH-The pH dependences of k h app for the wild-type and variant R-PPases were bell-shaped (data not shown), indicating that one basic and one acidic group is necessary for the substrate conversion step. In this respect, R-PPase is similar to the H ϩ -PPase of the plant Vigna radiata (25). The two corresponding pK a values, along with the pHindependent values of the catalytic constant, k h,ind , derived from these dependences with Equation 4 are listed in Table II. The effects of the mutations on the pK a values were insignificant except for the observed increase in pK a2 by 0.6 in the E202D variant, the increase in pK a1 by 0.6, and the decrease in pK a2 by 1.1 in the E649D variant. Interestingly, the pH-independent k h values for the E649D variant and the wild-type enzyme were similar within the combined experimental error, indicating that the low k h app value for the E649D variant observed at pH 7.2 is due to the effect of the substitution on the pK a of ionizing groups rather than to intrinsic catalytic efficiency.
Mersalyl Inactivation and Mg 2ϩ Binding-R-PPase contains three mersalyl-reactive Cys residues (18). Modification of Cys 222 occurs nearly instantly and decreases enzyme activity slightly (typically by ϳ10%), whereas modification of Cys 185 and Cys 573 can be resolved in time and render R-PPase completely inactive. Cys 185 and Cys 573 modifications are protected by Mg 2ϩ binding. To compare the effects of the mutations on mersalyl inactivation, an additional C185G substitution corresponding to the naturally occurring isoform was introduced to block Mg 2ϩ -dependent inactivation resulting from Cys 185 modification (18). The C185G substitution did not affect the hydrolytic and transport activities of the variant and wild-type R-PPase (data not shown) but made the slower inactivation step a first-order reaction at a fixed mersalyl concentration (Fig. 4), thus facilitating quantitative analysis.
The values of the second-order rate constants for inactivation in the absence of Mg 2ϩ (k E,mers ) and at saturating Mg 2ϩ concentrations (k EM,mers ) derived from the time-course experiments depicted in Fig. 4 were similar for wild-type, E197D, and E202D variants (Table II). In contrast, the E550D and E649D substitutions increased the k E,mers values by factors of 20 and 4 and the k EM,mers values by factors of 40 and 4, respectively. In all cases, Mg 2ϩ afforded substantial protection against mersalyl inactivation (k E,mers Ͼ k EM,mers ). In the E550D variant, the k E,mers /k EM,mers ratio decreased by a factor of ϳ2, which indicates that, upon Mg 2ϩ binding, the conformational change that inhibits the reactivity of Cys 573 in R-PPase toward SH reagents becomes partially compromised. The values of the dissociation constant governing Mg 2ϩ binding to R-PPase and its variants (K M,mers ) were estimated from the protection afforded by this cation against enzyme inactivation by mersalyl (Fig. 5). The dependences shown in Fig. 5 were simultaneously fit to Equations 5 and 6. The parameters k E,mers and k EM,mers were constrained in these fittings to the values obtained above. The resulting K M,mers values (Table II) indicate that E649D is the only variant in which Mg 2ϩ binding affinity is significantly affected (a 2-fold decrease).
1,1-Diphosphonate Binding-The binding affinities for two non-hydrolyzable substrate analogues (AMDP and MDP) were estimated by measuring their inhibitory activities on PP i hydrolysis (Fig. 6). The value of the inhibition constant, K i , derived from the data with Equation 7 remained unchanged for  both analogues in the D550E variant and increased by a factor of 3-8 in the other Glu 3 Asp variants (Table II). The K i values for AMDP measured with the Glu 3 Ala variants E197A, E550A, and E649A were 0.40 Ϯ 0.06, 0.20 Ϯ 0.03, and 0.44 Ϯ 0.09 M, respectively. Thus, the effects of the alanine substitutions on AMDP binding were similar except in the case of the E550A variant, which displayed a lower affinity for this substrate analogue than did the corresponding E550D variant.
Effects of Asp and Lys Substitutions-Nine aspartates (187, 191, 642, 669, 673, and 677) and three lysines (184, 195, and 646) in the putative cytoplasmic loops were similarly replaced with glutamates and arginines, respectively. These substitutions did not affect the expression of the variant enzymes in E. coli but, except for K195R, inhibited R-PPase activity measured at 2 mM Mg 2ϩ to a level indistinguishable from the background in our system (data not shown). K195R activity was 10% of that of the wild-type enzyme. The drastic decrease in PPase activity following these substitutions signifies that these mutated Asp and Lys residues are indispensable for catalysis but precludes further studies using the kinetic approaches employed for the Glu 3 Asp variants. DISCUSSION In the present study, we examined the roles of conserved Glu 197 , Glu 202 , Glu 550 , and Glu 649 in H ϩ -PPase of R. rubrum by site-directed mutagenesis and heterologous expression in E. coli. These residues, with the exception of Glu 202 , can be replaced with aspartates or alanines and display retention of at least 20% PPase activity measured at high Mg 2ϩ (2 mM) con-centrations. In view of the finding that the E202D variant retained significant activity, whereas in the E202A variant activity declined to a level indistinguishable from the background in our system (ϳ 5% wild-type R-PPase activity), we propose that the carboxyl at position 202 of R-PPase sequence is indispensable for enzyme function. All variants active in PP i hydrolysis were also capable of PP i -energized proton translocation with a magnitude of coupling similar to that of wild-type R-PPase. This finding is consistent with the tentative assignment of glutamates in hydrophilic regions to cytoplasmic loops, which implies that these residues are unlikely to be involved in proton transport. Interestingly, substitution of Glu 263 (corresponding to R-PPase Glu 197 ) in the H ϩ -PPase of the plant V. radiata to aspartate led to complete loss of PP i -energized proton translocation despite the retention of Ͼ30% hydrolytic activity (21). It is unlikely that this residue is related to proton transport in V. radiata H ϩ -PPase but not R-PPase. The E263D substitution possibly caused partial decoupling in the plant enzyme indirectly, perhaps by disrupting the more fragile overall structure. Indeed, a weak current arising from PP i -energized proton transport was recently detected in the E263D variant of V. radiata H ϩ -PPase by patch clamp analysis of giant yeast vacuoles (27). In contrast, our data on the D187E, D669E, D673E, D677E, and K196R R-PPase variants are consistent with the results for V. radiata H ϩ -PPase reported by Nakanishi et al. (21).
Detailed steady-state kinetic analysis performed at a wide range of Mg 2ϩ concentrations indicated that the common effect of all four Glu 3 Asp substitutions in R-PPase is reduced catalytic efficiency (signified by k h app /K m app ) as compared with that of wild-type R-PPase at low Mg 2ϩ concentrations (Ͻ 0.5 mM). In terms of Scheme I, this observation is explained by hydrolysis of PP i by the variant enzyme via the EM 3 PP complex only. In contrast, the wild-type enzyme additionally hydrolyzes PP i via the EM 2 PP complex. Thus, in the absence of conserved Glu, two metal ions are insufficient to form a catalytically competent active site structure that is restored upon binding of the third metal ion. Similar effects of active site substitutions were observed previously in soluble PPases. Whereas catalysis by wild-type E. coli and S. cerevisiae PPases proceeds through both three-and four-metal pathways (28,29), active site variants require the whole complement of four metal ions, but are inactive when only three metal ions bind (30,31). The similarities between the effects of the substitutions in two non-homologous PPase families imply that conserved Glu residues of R-PPase also form part of an active site.
This suggestion is consistent with the effects of the substitutions on the binding of the non-hydrolyzable substrate analogues AMDP and MDP. The K i values reported here for AMDP and MDP are true binding constants, in contrast to the Michaelis constant for substrate, which is generally a complex combination of microscopic rate constants for individual reaction steps. AMDP is a 30-fold more potent inhibitor of H ϩ -PPase in comparison to MDP, which lacks an aminomethyl group in its structure. However, the effects of the substitutions were similar for AMDP and MDP, suggesting that enzyme interactions with the phosphate moieties of the inhibitors were affected. We thus interpret the reduced affinity of the E197D, E202D, and E649D variants for AMDP and MDP as possible evidence of impaired PP i binding in these variants. Interestingly, at positions 197 and 649, AMDP binding was similarly impaired by Glu 3 Asp and Glu 3 Ala substitutions, suggesting that in these cases aspartates are poor substitutes for glutamates with respect to inhibitor binding. Alternatively, at position 550, AMDP binding was impaired in the Glu 3 Ala but not the Glu 3 Asp variant. This finding may be explained by the considerable flexibility in the FIG. 5. Mg 2؉ concentration dependence of wild-type and variant R-PPase inactivation by mersalyl. IMVs were incubated with mersalyl for 10 min (wild-type, E197D, and E202D), 2 min (E649D), or 1 min (E550D) at the indicated Mg 2ϩ concentrations. Other details were as for Fig. 4. Lines are drawn in accordance with Equations 5 and 6 using the parameter values specified in Table II. vicinity of Glu 550 , such that the shorter aspartate residue maintains the same interactions as the longer glutamate residue.
The catalytic incompetence of the EM 2 PP complex and its decreased affinity to substrate analogues and, presumably, to substrate in the E197D, E202D, and E649D variants and to Mg 2ϩ in the E649D variant indicate that the active site structure is significantly distorted by the substitutions. However, the sources of this distortion are distinct in the different variants. In E550D and E649D, this may be a long-range conformational change as indicated by the increased mersalyl sensitivity that may result from increased accessibility of Cys 573 or other previously inaccessible Cys residues. Thus, Glu 550 and Glu 649 likely function in maintaining the R-PPase conformation. In contrast, Glu 197 and Glu 202 may be directly involved in metal cofactor binding. According to Scheme I, the enzyme has a metal-binding site (M1) to form EM and an additional metalbinding site (M2) to bind M 2 PP if one of the two metal ions that come with PP i bridges PP i and the enzyme in the resulting EM 3 PP complex as in the case of soluble PPases (32,33). Glu 197 may be part of the M1 site as its substitution markedly increases K M1 , but not K A1 and K A2 (Table II), whereas Glu 202 possibly belongs to both the M1 site and the M2 site because its substitution increases all of these parameters. The contradictory invariance of K M,mers in the E197D and E202D variants (Table II) may result from the fact that this parameter provides a measurement of overall metal binding, whereas the kinetically determined K M1 , K A1 , and K A2 parameters evaluate only functional binding. The flexibility of the protein structure allows for efficient metal binding in the variant enzymes, but the bulk of the resulting complex is non-productive because the metal is mispositioned. Notably, the lack of effect of similar substitutions (Asp 3 Glu) of metal ligands on thermodynamically controlled metal binding is well documented in soluble PPases (34,35).
In summary, all of the conserved glutamates in R-PPase are essential for catalytic efficiency at low Mg 2ϩ concentrations, which accounts for their conservation in H ϩ -PPases. In addition, Glu 197 and Glu 202 may contribute to substrate binding by controlling metal binding but do not affect the overall enzyme structure. On the other hand, Glu 550 and Glu 649 may be important for maintaining the catalytically competent conformation of R-PPase.