Dissociation of hexameric Escherichia coli inorganic pyrophosphatase into trimers on His-136-->Gln or His-140-->Gln substitution and its effect on enzyme catalytic properties.

Each of the five histidines in Escherichia coli inorganic pyrophosphatase (PPase) was replaced in turn by glutamine. Significant changes in protein structure and activity were observed in the H136Q and H140Q variants only. In contrast to wild-type PPase, which is hexameric, these variants can be dissociated into trimers by dilution, as shown by analytical ultracentrifugation and cross-linking. Mg2+ and substrate stabilize the hexameric forms of both variants. The hexameric H136Q- and H140Q-PPases have the same binding affinities for magnesium ion as wild-type, but their hydrolytic activities under optimal conditions are, respectively, 225 and 110% of wild-type PPase, and their synthetic activities, 340 and 140%. The increased activity of hexameric H136Q-PPase results from an increase in the rate constants governing most of the catalytic steps in both directions. Dissociation of the hexameric H136Q and H140Q variants into trimers does not affect the catalytic constants for PPi hydrolysis between pH 6 and 9 but drastically decreases their affinities for Mg2PPi and Mg2+. These results prove that His-136 and His-140 are key residues in the dimer interface and show that hexamer formation improves the substrate binding characteristics of the active site.

Phosphoryl transfer enzymes form one of the largest classes of enzymes (Knowles, 1980), yet their mechanisms of action are still not fully understood (Herschlag and Jencks, 1990). This class includes soluble inorganic pyrophosphatases (EC 3.6.1.1; PPase), 1 which hydrolyze inorganic pyrophosphate (PP i ) to inorganic phosphate (P i ). These enzymes, essential in both bacteria (Chen et al., 1990) and yeast (Lundin et al., 1991), are ubiquitous and play an important role in energy metabolism, providing a thermodynamic pull for biosynthetic reactions such as protein, RNA, and DNA synthesis (Kornberg, 1962). According to Peller (1976), nucleic acid synthesis would be energetically impossible in vivo if it were not coupled to the PP i hydrolysis catalyzed by PPases. The two best-studied soluble PPases are those from the yeast Saccharomyces cerevisiae and Escherichia coli: each accelerates the rate of PP i hydrolysis by a factor of 10 10 compared with the rate in solution. Detailed understanding of their catalytic mechanisms is important for understanding the class of phosphoryl transfer enzymes as a whole.
E. coli PPase is homohexameric (Wong et al., 1970) and contains 175 amino acid residues per subunit (Lahti et al., 1988). Its three-dimensional structure, recently been determined at 2.5-2.7-Å resolution (Kankare et al., 1994;Oganessyan et al., 1994), is very like that of S. cerevisiae PPase (Kuranova et al., 1983;Terzyan et al., 1984), in accord with the conservation of active site residues and mechanism in the two enzymes (Cooperman et al., 1992;Kankare et al., 1994). E. coli PPase requires four Mg 2ϩ ions per active site for catalysis, as described in Scheme I. This scheme, which fully accounts for the overall catalysis of PP i :P i equilibration by E. coli PPase, is a slightly modified version of the one proposed by Baykov et al. (1990) following the approach developed for S. cerevisiae PPase (Springs et al., 1981;Welsh et al., 1983).
The cloning and sequencing of the E. coli ppa gene that encodes PPase (Lahti et al., 1988) together with the analysis of the conservation of functional residues between yeast and E. coli PPase (Lahti et al., 1990a) have made it possible to study the structural and functional relationship of E. coli PPase by site-directed mutagenesis (Lahti et al., 1990b). All 17 polar residues located in the active site cavity have been substituted (Lahti et al., 1990b;Lahti et al., 1991;Cooperman et al., 1992) and some of the variant PPases have been characterized in detail Kä pylä et al., 1995).
Earlier chemical modification studies of Samejima et al. (1988) had implicated histidines as being involved in the activity of E. coli PPase. The absence of His residues from the active site cavity (Kankare et al., 1994) and the lack of conserved His residues in soluble PPases (Cooperman et al., 1992) make it clear that His residues have no direct role in catalysis, leaving open the possibility that the chemical modification results arise from an indirect effect. In this work we demonstrate that substitution of each of the five His residues in E. coli PPase with Gln results in substantial retention or even enhancement of catalytic activity, confirming their nonessential character.
However, in contrast to wild-type enzyme, two variants, H136Q and H140Q, are found to dissociate readily into trimers, providing valuable insight into the details of subunit:subunit interaction.

EXPERIMENTAL PROCEDURES
Enzymes-Wild-type and the His 3 Gln mutant PPases were expressed and purified using the overproducing E. coli strains HB101 (Lahti et al., 1990b) and MC1061/YPPAI(⌬ppa) , respectively. The enzyme concentration was estimated on the basis of a subunit molecular mass of 20 kDa (Josse, 1966) and an A 280 1% of 11.8 (Wong et al., 1970).
The hexameric and trimeric forms of H136Q-and H140Q-PPases were obtained by varying the composition of their stock solutions in order to shift the hexamer … trimer equilibrium in the desired direction (see "Results"). Stock solutions of hexameric PPase contained 225-1000 M enzyme, 1 mM (for H136Q) or 50 mM (for H140Q) MgCl 2 , and 0.15 M Tris-HCl, pH 7.2. Dilutions were made with the same buffer containing 20 mM (for H136Q) or 50 mM (for H140Q) MgCl 2 . Because the H140Q-PPase dissociated so rapidly, we had to add 0.3 mM PP i to the dilution medium for this enzyme and could keep the diluted solution for no longer than 15 s, during which PP i was still present. The stock solution of trimeric H136Q-PPase contained 0.9 M enzyme, 1 mM MgCl 2 , and 0.15 M Tris-HCl, pH 8.5; that of trimeric H140Q-PPase contained 2.25 M enzyme, 1 mM MgCl 2 , and 0.15 M Tris-HCl, pH 7.2. All enzyme solutions contained, in addition, 50 M EGTA, and 0.5 mg/ml bovine serum albumin was added to the dilution media. All stock solutions were preincubated for at least 30 min before use.
Sedimentation-Sedimentation velocity measurements were carried out at 60,000 rpm, 20°C, in a Spinco E analytical ultracentrifuge with scanning at 280 nm. Sedimentation coefficient was calculated with a standard procedure (Chervenka, 1972).
Other Methods-The initial rates of PP i hydrolysis were estimated from continuous recordings of P i liberation obtained with an automatic P i analyzer (Baykov and Avaeva, 1981). The initial rates of the reverse reaction (PP i synthesis) were determined by measuring the concentration of PP i released from the enzyme using ATP-sulfurylase and luciferase as coupling enzymes (Nyren and Lundin, 1985;Baykov and Shestakov, 1992). The rate constant k 2 was calculated from these data using the following equation: where k s is the catalytic constant for synthesis (see "Results"). Rates of oxygen exchange between P i and H 2 O were measured by mass spectrometry as described by Baykov et al. (1990). Other kinetic and binding measurements as well as calculations of various rate and binding constants were carried out as described by Kä pylä et al. (1995). Unless otherwise indicated, the media used were buffered with Tris-HCl, the concentration of which was varied to maintain the ionic strength at 0.15-0.20 M. EGTA (typically 50 M) was included in all solutions containing enzyme. All experiments reported were carried out at 25°C.

RESULTS
Preliminary Characterization of Variant PPases-Mutations of the five histidine residues in E. coli PPase to glutamine yielded three variants (H60Q, H110Q, and H119Q) that were unchanged as compared with wild-type enzyme with respect to catalytic activity (measured at pH 8), Nile Red fluorescence (measuring protein surface hydrophobicity; Salminen et al., 1995), and migration as a hexamer on native gel electrophoretic analysis. In contrast, each of the remaining two variants, H136Q and H140Q, showed increased catalytic activity (see below) and much increased Nile Red fluorescence compared with wild-type PPase, as well as the presence of a fastermigrating dissociated form in addition to a hexamer on native gel electrophoresis. Accordingly, only the latter two variants were studied further.
Activity versus Enzyme Concentration Profiles-Specific activities of the H136Q and H140Q variants pre-equilibrated at 25°C at different enzyme concentrations were measured to define conditions that would favor enzyme dissociation or association. In these experiments, only the composition of enzyme stock solution was varied; the medium for the assays remained the same (20 M Mg 2 PP i , 20 mM Mg 2ϩ , 0.15 M Tris-HCl, pH 7.2). The preincubation was carried out for 40 -60 min until activity stabilized. The results (Fig. 2) indicate that the specific activity of the enzymes, with the exception of H136Q-PPase at high Mg 2ϩ concentration, decreases with decreasing enzyme concentration. The activity of WT-PPase is not changed on analogous incubations. The effects shown in Fig. 2 are reversible, i.e. the activity could be brought to the levels corresponding to new conditions on changing pH in either direction or increasing Mg 2ϩ concentration from 1 to 50 mM.
These results were analyzed in terms of Scheme II: using the following relationships (Kurganov, 1982): where A is the observed specific activity, A T and A H are the specific activities of trimeric and hexameric enzyme, respectively; ␣ H,eq is the fraction of hexameric enzyme at equilibrium (␣ H,eq ϭ 6[H]/([E] 0 ); [E] 0 is total enzyme concentration in terms of monomer. Fitting these equations to the data ( Fig. 2) yielded the K d values shown (Table I). When fitting the H140Q data obtained at 1 mM Mg 2ϩ , the value of A H was fixed at 191 s Ϫ1 , the limiting value of activity at infinite enzyme concentration as determined from the profile obtained at 50 mM Mg 2ϩ for this enzyme.
As seen from Fig. 2 and Table I, increasing the pH destabilizes the H136Q variants but stabilizes the H140Q variant in the presence of 1 mM Mg 2ϩ . Measurements of the equilibrium activity at a fixed enzyme conentration (0.11 M) indicated that pH 7.2 is optimal for the stability of H136Q-PPase while the stability of H140Q-PPase increases monotonically up to pH 10.1 (data not shown).
Direct Assessment of the Quaternary Structure of H136Qand H140Q-PPases-The variant PPases (and WT-PPase as a control) were cross-linked at pH 8.5 and subjected to electrophoresis in the presence of SDS. Only a faint band corresponding to hexameric protein is observed with the variant PPases while WT-PPase yields appreciable amounts of cross-linked hexamers ( Fig. 1). Consequently, the variant PPases seem to be predominantly dissociated at pH 8.5 and 17.5 M enzyme concentration used in cross-linking, in accord with the data in Fig. 2.
Other data are also consistent with the existence of a hexamer-trimer equilibrium in these two variant PPases and support the assumption made above that the effects of enzyme and Mg 2ϩ concentrations and pH on their specific activities result from shifts in this equilibrium. Firstly, the sedimentation coefficient (s 20, w ) changes with changing pH and Mg 2ϩ concentration (Table II), and this effect correlates with the activity data ( Fig. 2). It should be noted that for H140Q-PPase, the sedimentation profiles were generally broader than for H136Q-PPase, suggesting that both hexamers and trimers are present in significant, although not equal, amounts. Accordingly, most s 20,w values are somewhat lower for hexamer and greater for trimer versus H136Q-PPase (Table II). Secondly, analyzing the electrophoretic mobilities of the H136Q and H140Q variants as functions of polyacrylamide concentration (pH 9.5, in the absence of Mg 2ϩ ) (Hedrick and Smith, 1968) indicated a 2-fold decrease in their molecular weights compared to that of WT-PPase (data not shown). By contrast, a D97E variant, whose specific activity is independent of enzyme concentration, does not show such a decrease .

Rate and Equilibrium Constants for Catalysis by Hexameric
Variant PPases-Following the approach devised for S. cerevisiae PPase (Springs et al., 1981), we determined the rate and equilibrium constants of Scheme I for the H136Q-and H140Q-PPases by combining data from equilibrium formation of enzyme-bound PP i with data from the kinetics of PP i hydrolysis, PP i synthesis, and P i /HOH oxygen exchange. From the dependences of hydrolysis rate on [Mg 2 PP i ] and [Mg 2ϩ ] (not shown),

FIG. 2. Specific activities of H136Q-PPase (A) and H140Q-PPase (B)
pre-equilibrated at different enzyme and Mg 2؉ concentrations and pH. q, pH 7.2, 50 mM Mg 2ϩ ; E, pH 7.2, 1 mM Mg 2ϩ ; Ç, pH 8.5, 1 mM Mg 2ϩ . The incubation media contained 1 mg/ml bovine serum albumin. Following the incubation, enzyme activity was assayed at pH 7.2 in the presence of 20 M Mg 2 PP i and 20 mM Mg 2ϩ . The lines are drawn according to Equations 2a-2c, using parameter values given in Table I. H136Q-PPase H140Q-PPase we calculated values for the catalytic constant k h , the Michaelis constant K m,h , and the metal dissociation constants K m2 and K ma (Table III). 2 We also calculated values of K 3 , K 5 , and K 7 using the dependence of enzyme-bound PP i formation on MgP i concentration ( Fig. 3 and Equation 3).
The results of oxygen exchange measurements at two saturating Mg 2ϩ concentrations (Table IV) were used to calculate the catalytic constant for exchange k ex and the partition coefficient P c ϭ k 4 /(k 4 ϩ k 5 ) (Hackney and Boyer, 1978). The k ex values were calculated by extrapolating v ex to infinite [MgP i ] with Equation 3, using the K 3 , K 5 , and K 7 values in Table III. Finally, Fig. 4 shows the effects of Mg 2ϩ concentration on the synthesis and liberation of PP i to the medium (medium PP i synthesis). This is the only activity of E. coli PPase inhibited by excess Mg 2ϩ (Baykov et al., 1990). The asymptotic values of v s at 20 mM Mg 2ϩ and above were assumed to be the corresponding catalytic constants k s . From the above parameters (Table  III), we calculated the rate constants k 1 -k 8 in Scheme I (Table  V). Special care was taken to ensure that the variant PPases were hexameric during these measurements (see "Experimental Procedures"). Additionally, hydrolytic activity was measured at 20 M Mg 2 PP i (20 mM Mg 2ϩ , pH 7.2) to verify that the hexameric structure is retained by the end of enzyme incubation with P i and Mg 2ϩ in the oxygen exchange and enzymebound PP i measurements.
pH-rate studies (see below) showed that the maximal activity of the H136Q variant, approximately twice its activity at pH 7.2, occurs at pH 8.0 -8.5. Therefore, we also determined the rate constants k 1 -k 8 for this variant at pH 8.0. As for WT-PPase, 3 its metal binding affinities at pH 8.0 were greater than at pH 7.2. The metal binding constants could not, however, be determined accurately because the H136Q variant dissociates rapidly at pH 8.0 and above if the concentration of metal ion and/or substrate is low, as is required for such an analysis. The values of most, if not all, rate constants for H136Q-PPase are higher than those of WT-PPase at pH 8.0 (Table V).
Characterization of the Hydrolytic Activities of Trimeric H136Q-and H140Q-PPases-We could not determine every parameter for the variant PPases in their trimeric forms because they rapidly reassociated during incubations with P i and Mg 2ϩ at the high protein and Mg 2ϩ concentrations (Ͼ2.5 M and 20 mM, respectively) required to measure enzyme-bound PP i synthesis and 18 O exchange. Hence, only PP i hydrolysis parameters were obtained (Table III). Hexameric and trimeric . e Calculated as k 1 /k 2 (see Table V). f Value in parentheses calculated from K 7 ϭ K pp /K 1 K 3 K 5 , where K pp ϭ 199 M is the equilibrium constant for the reaction 2MgP i … Mg 2 PP i in solution at pH 8.0 .
g Binding is to a mixture of hexamer and trimer (see text). variants have about the same k h values, but the trimers have greatly increased K m,h , K m2 , and K ma values. Consequently, they display reduced activity upon dilution (Fig. 2) because the trimers are not saturated at the 20 M substrate concentration used in our standard enzyme assay. In fact, virtually no activity loss was observed with the H136Q variant incubated as for Fig. 2 but assayed at 1 mM Mg 2 PP i . The binding characteristics of trimeric H140Q-PPase (K m,h and K ma ) were somewhat improved at higher pH (Table III). Three lines of evidence rule out the possibility that the activities measured result from hexamer formation during the assay. First, hexamer formation is a slow reaction. For H136Q variant, the association rate constant is only 0.77 M Ϫ1 min Ϫ1 at pH 7.2 (see below), which corresponds to half-time of 9 days for hexamer formation at 0.1 nM enzyme concentration, which we routinely use in hydrolysis assays. Second, the product formation curves are strictly linear during 3 min of the assay, indicating that no interconversion between enzyme forms occurs during the assay. Third, the measured K m,h values are very different from those for the hexameric forms.
The Effects of pH on the Hydrolytic Activities of H136Q and H140Q Variants-Values of k h app and K m,h app for PP i hydrolysis by the variant PPases were determined at 20 and 50 mM Mg 2ϩ in the pH range 6 -9.6. The pH profile of k h app for H136Q-PPase is little changed between hexamer and trimer and is virtually the same at both Mg 2ϩ concentrations (Fig. 5A). In contrast, the values of k h app /K m,h app are approximately 50 -100-fold higher for hexamer versus trimer and, at pH Ͻ 6.5, are considerably higher at 50 mM Mg 2ϩ than at 20 mM Mg 2ϩ (Fig. 5B). By comparison, for WT-type PPase, k h app /K m,h app is practically constant in the pH range 6 -8.5  and shows little dependence on Mg 2ϩ concentration above 20 mM.
The results at 50 mM Mg 2ϩ were used to calculate apparent ionization constants for enzyme with substrate bound, indicated as ESH, and for enzyme lacking substrate, indicated as EH, as well as of pH-independent values for k h app and k h app /K m,h app at 50 mM Mg 2ϩ , according to Equations 4a and 4b.
Parameter values are presented in Table VI, along with those determined from similar data (not shown) for the H140Q hexamer (we were unable to determine pH-rate profiles for H140Q trimer). For the most part, values for the hexameric variants are quite similar to those for wild-type. The only notable differences are the 2-fold increase in k h for H136Q variant (see below) and the increases that both variants show in the values of pK EH2 and pK ESH2 . Mutations of active site residues also give rise increases in pK EH2 and pK ESH2 , but the magnitude of such increases is considerably greater . The H136Q trimer shows a still larger increase in pK EH2 as well as a large decrease in k h /K m .
Mg 2ϩ Binding to the H136Q and H140Q Variants-Mg 2ϩ

FIG. 4. Rates of PP i synthesis to medium by wild-type and variant PPases as functions of Mg 2؉ concentration.
MgP i concentration was fixed at 20 mM (E, Ç, Ⅺ) or 10 mM (å). Rates were measured by a coupled enzyme assay (Baykov and Shestakov, 1992). E, wild-type PPase, pH 7.2; Ç, H136Q-PPase, pH 7.2; å, H136Q-PPase, pH 8.0; Ⅺ, H140Q-PPase, pH 7.2. binding by the variant PPases in the absence of substrate was characterized in two ways. Firstly, equilibrium dialysis measurements  yielded dissociation constants for two Mg 2ϩ sites per subunit indistinguishable from those for WT-PPase (Table III). These experiments were conducted at 750 M enzyme concentration. In the Mg 2ϩ concentration range used (0.03-2.5 mM), the H136Q variant is predominantly hexameric while the H140Q variant is a mixture of hexameric and trimeric forms (Fig. 2). Secondly, by the protective effect of Mg 2ϩ ions on enzyme dissociation upon dilution. The time course of trimer formation from hexamers is given by Equations 5a and 5b (Kurganov, 1982).
Fitting Equations 5a and 5b, in combination with Equations 2b and 2c, to the data shown in Fig. 6A generated k d values shown in Fig. 6B. K d values paralleled with k d , indicating that the effect of metal concentration on k a is small. The dissociation constant for Mg 2ϩ of 2.4 mM was obtained by fitting k d to Equation 6: where k d , lim is k d at zero Mg 2ϩ concentration and K Mg is the metal binding constant. A similar analysis was carried out for the H140Q variant (data not shown). For this variant k d,lim was as high as 0.6 s Ϫ1 , and the data were more scattered. Nevertheless, the value of K Mg we obtained, 1.8 mM, was again close to the K m2 values derived above from hydrolysis kinetics and equilibrium dialysis measurements (Table III). Despite the high rate constant for hexameric H140Q dissociation, the P i liberation curves in the assay of residual activity were linear for at least 3 min, indicating that substrate profoundly stabilizes the hexameric form.

Hexamer Stability and the Interface between Monomers-
PPase can either be considered as a dimer of trimers or a trimer of dimers centered around a local 2-fold axis ( Fig. 7; Kankare et al. (1994)). The main components of the interface around the local 2-fold axis are the two antiparallel symmetry-related ␣-helices A (Fig. 7) and a short loop (residues 46 to 50) connecting strands 4 and 5 (data not shown). His-136 and His-140 are located at the C terminus of helix A (Fig. 7A), at the center of a network of intra-and intersubunit interactions, a network so clearly important that we predicted that mutating His-136 and His-140 would affect the oligomeric structure of the enzyme. Starting from His-136 in the first monomer (136 in the righthand monomer in Fig. 7B), the N⑀2 of His-136 forms an interchain hydrogen bond to O␦2 of Asp-143 in monomer 2 (143Ј in Fig. 7B). The same Asp-143 O␦2 also forms an intrachain hydrogen bond to N␦1 of His-140 in monomer 2 (140Ј in Fig.  7B). The above, of course, describes a half-interface: His-136 3 Asp-143Ј 3 His-140Ј; there is another half-interface that runs His-136Ј 3 Asp-143 3 His-140. The network of interactions is completed by hydrophobic contacts between the almost-parallel His-140 and His-140Ј which face each other, about 3.5 Å apart, across the intermolecular 2-fold (x in Fig. 7B). A similar arrangement of two antiparallel ␣-helices also forms the bulk of the homologous trimer-trimer interface in the crystal structure of Thermus thermophilus PPase (Teplyakov et al., 1994). His-136 and His-140 in E. coli PPase align with His-134 and Thr-138 in T. thermophilus PPase, respectively, suggesting that the  It seems very likely that, per monomer at neutral pH, a single positive charge is shared between His-136 and His-140. (The resolution of the current structure is not sufficient to tell which is the more likely His.) Consequently, replacing either His will either weaken or destroy the ion pair depicted in Fig.  7B. In both cases, the replacement Gln can orient to allow its ⑀-amido group to replace the lost His-Asp hydrogen bond, thus partially mitigating the effect of the substitution. The helixforming tendency of glutamine is slightly higher than that of histidine (O'Neil and DeGrado, 1990), thus making unlikely any destabilization of the ␣-helix A in the variant proteins.
Based on the structure of the subunit contact region, one would expect, and it is indeed observed (Table I), that the hexameric structure of the H136Q variant is more stable than that of the H140Q variant. His-136 is more exposed to solvent than His-140 so the Gln-136 side chain in the H136Q variant can swing out of the way to allow solvation both of Asp-143Ј and of Gln-136. Conversely, in the H140Q variant, the completely buried Gln140 will have unsatisfied hydrogen bond donors and acceptors that cannot be solvated (at least one His on the ⑀-amido group and lone pairs on the ⑀-carbonyl group). In addition, there is a loss of hydrophobic interaction between His-140 and His-140Ј in the H140Q variant. Finally, because 140 is the most buried residue, the conformation of Gln-140 will be much more constrained than the conformation of Gln-136 in the two respective variants: as a result, more entropy will be lost on forming H140Q hexamers than on forming H136Q hexamers.
Another clear difference between the two variants is in their pH-stability profiles: the H136Q variant shows a pH optimum at about pH 7.2. This is consistent with two Asp Ϫ :HisH ϩ ion pairs stabilizing subunit association, with the loss of stability at lower and higher pH being due to protonation of Asp-143 and deprotonation of His-140, respectively. Here we note the report of Borshchik et al. (1986) that even WT-PPase will dissociate into trimers on incubation at pH 5. At present we have no cogent rationale for the pH stability profile of the H140Q variant.
Catalytic Properties of the Hexameric Forms-The kinetic parameters of WT and variant PPases are very similar (Table  III); neither substitution greatly affects the active site. The pH profiles of k h app are very similar to each other, and the differences are chiefly due to changes in the pH dependence of the rate constant for product release (k 7 ) between wild-type and the two variant PPases (Table V). In contrast, an active site D97E variant principally reduces the rate of the chemical catalysis step (k 3 ) .
Measured at pH 8.0, close to the optimum for both enzymes, k h and k ex are 2-3-fold greater for the H136Q variant than for wild-type PPase, as are most of the microscopic rate constants (Table V). At least three possible explanations can be suggested for why this is so. First, the mutation may somehow optimize active site conformation. Second, weakening intersubunit interactions may increase conformational flexibility of the protein molecule, which may be important provided that some reaction steps involve rate-limiting conformational changes. Finally, the trimers in WT-PPase may interact such that the perfect hexameric symmetry breaks down and one trimer is always inactive, while in the H136Q-PPase the trimers are  equivalent and active. Alternatively, the trimers could cycle: one "on" and one "off" as, for example, proposed for ATP synthase (Boyer, 1993;Abrahams et al., 1994). Hexameric H136Q-PPase would thus be more active than (hexameric) WT-PPase because its trimers would be more independent. We cannot, at present, select between the above possibilities. It is, however, intriguing that E. coli PPase crystallizes in two different forms. One has obligate perfect identity between all six monomers because the space group is R32 and there is only one monomer in the asymmetric unit . The other has two monomers in the asymmetric unit of a double-sized R32 unit cell and so the trimers, related by a noncrystallographic 2-fold axis (Kankare et al., 1994), are not constrained to be identical and, indeed, appear not to be. 4 Catalytic Properties of the Trimeric Forms-As the variant PPases dissociated easily, we could determine some of their catalytic properties as trimers. Although the substrate and Mg 2ϩ binding parameters for trimeric H136Q-and H140Q-PPases are somewhat different at pH 7.2 (Table III), the same trend is evident: the affinities for both ligands are drastically reduced on dissociation, while k h values remain essentially unchanged.
In terms of the three-dimensional structure, the binding changes are explained by the fact that the interactions discussed above (Fig. 7B) not only stabilize the monomer-monomer interface but also define a specific conformation for the 141-143 loop that immediately follows helix A. That loop contains one of the more important active site residues. Mutation of Lys-142 to Arg-142 results in a large increase in the K m for Mg 2 PP i  and in weakened binding for both MgP i and a competitive inhibitor of Mg 2 PP i . 5 Consequently, any destabilization of the 141-143 loop on hexamer dissociation might be expected to affect the active site and, in particular, the binding of PP i and P i . The greater affinity of hexameric PPase for Mg 2ϩ and Mg 2 PP i explains why these ligands stabilize it versus trimeric PPase.
We demonstrated (data not shown) that the increased K m,h is an intrinsic property of the trimers, as opposed to arising from an indirect effect of each mutation on the active site, by showing that WT-PPase trimer, formed by prolonged incubation at low pH following Borshchik et al. (1986), resembles the variants studied in this paper in having a k h value similar to that of the hexamer but much higher K m,h values. This latter result differs from that reported by Borshchik et al. (1986), who claimed that the K m,h values of trimer and hexamer were similar. It is, however, consistent with reports of activity loss following dissociation of the structurally related PPases of the thermophilic bacteria PS-3 (Hachimori et al., 1979) and Bacillus stearothermophilus (Schreier, 1980) into trimers. Our work suggests that this loss of activity is due to an effect on K m,h . Finally, by analogy with our present results, we speculate that the loss of E. coli PPase activity found by Samejima et al. (1988) on chemical modification of His residues arises from dissociation of hexamer into trimers.