The Electrophilic and Leaving Group Phosphates in the Catalytic Mechanism of Yeast Pyrophosphatase*

Binding of pyrophosphate or two phosphate molecules to the pyrophosphatase (PPase) active site occurs at two subsites, P1 and P2. Mutations at P2 subsite residues (Y93F and K56R) caused a much greater decrease in phosphate binding affinity of yeast PPase in the presence of Mn2+ or Co2+than mutations at P1 subsite residues (R78K and K193R). Phosphate binding was estimated in these experiments from the inhibition of ATP hydrolysis at a sub-K m concentration of ATP. Tight phosphate binding required four Mn2+ ions/active site. These data identify P2 as the high affinity subsite and P1 as the low affinity subsite, the difference in the affinities being at least 250-fold. The time course of five “isotopomers” of phosphate that have from zero to four 18O during [18O]Pi-[16O]H2O oxygen exchange indicated that the phosphate containing added water is released after the leaving group phosphate during pyrophosphate hydrolysis. These findings provide support for the structure-based mechanism in which pyrophosphate hydrolysis involves water attack on the phosphorus atom located at the P2 subsite of PPase.

Inorganic pyrophosphatase (EC 3.6.1.1; PPase) 1 is a ubiquitous enzyme catalyzing interconversion of PP i and P i . Soluble PPase provides a thermodynamic pull for biosynthetic reactions by removing PP i formed when nucleoside 5Ј-triphosphates are converted to the corresponding monophosphates (1). The PPase reaction involves, in the direction of hydrolysis, PP i binding, isomerization of the resulting complex, PP i hydrolysis, and the stepwise release of two P i molecules (Scheme I) (2). The hydrolysis step proceeds via direct attack of water on PP i without formation of a covalent intermediate (3). Numerous mechanistic studies of PPase have been carried out (for reviews, see Refs. 4 and 5), making this enzyme the best characterized among the catalysts of phosphoryl transfer from various polyphosphates (including ATP and GTP) to water.
X-ray crystallographic studies of PPase complexed with phosphate have identified two P i binding subsites, P1 and P2, within the active site (6, 7) (Fig. 1). In addition, an activated water molecule, placed between two metal ions in the vicinity of P2, was considered the only logical candidate for nucleophile (6). This supposition was confirmed by the structure of the F --inhibited complex (8), which showed a fluoride ion replacing that water molecule. Solution confirmation of this interpretation has been harder to achieve, however. The oxygen exchange measurements done in the presence of Mg 2ϩ as the activator have shown unequivocally that the P i containing the electrophilic phosphorus is released first after PP i hydrolysis (9). In terms of the structure-based mechanism (6,8,10), this would mean that P i is first released from P2, where it is more buried than at P1. Resolving this issue requires knowledge of the relative affinities of P1 and P2 for P i because it is logical to expect faster release from a weaker binding subsite. Earlier attempts to compare the affinities of P1 and P2 used three types of data (5), each of which has been subject to criticism. First, preferential binding of sulfate, an analog of P i , occurred at P1 in the PPase crystals grown in the presence of sulfate (11,12). However, the crystallization medium in these studies contained no metal ions, the major P i ligands at the P2 subsite ( Fig. 1). Second, x-ray data indicate more extensive hydrogen bonding between the enzyme and P i at P1 than at P2 (Fig. 1), which was thought to provide greater binding strength at P1. We will show below that this is not the case for Mn 2ϩ and Co 2ϩ as cofactors. Third, protection of Arg 78 , located at P1, against chemical modification upon binding of 1 mol of P i /mol of subunit in the presence of Mn 2ϩ was interpreted as showing that P1 binds first (13). However, because of the close proximity of P1 and P2, the protection could equally result from binding to P2. We will show below that P2 does in fact exhibit a far greater affinity for P i in the presence of Mn 2ϩ and Co 2ϩ .
We addressed the correct assignment of the phosphate binding subsites by employing site-directed mutagenesis of P i ligands together with P i binding and P i -water oxygen exchange measurements in the presence of Mn 2ϩ and Co 2ϩ . By comparison with Mg 2ϩ , these cations induce a much greater difference in the binding affinities of P1 and P2, as evidenced by the fact that only one P i binding site/subunit was observed over a wide range of P i concentrations (13,14). In the presence of Mg 2ϩ , the affinities of P1 and P2 for MgP i differ only 1-9-fold (9,15,16); for comparison, macroscopic binding constants for two sites with equal microscopic constants differ 4-fold (17). The results reported below demonstrate that the relative affinities of P1 and P2 and the order in which these sites release P i during PP i hydrolysis depend on the nature of the metal ion cofactor used and provide support for the mechanism currently proposed for this reaction.

MATERIALS AND METHODS
The expression and purification of wild type Y-PPase and its active site variants from overproducing Escherichia coli XL2blue b strain transformed with suitable plasmids were carried out as described by Heikinheimo et al. (18). Enzyme concentration was calculated on the basis of a subunit molecular mass of 32 kDa (19) and an A 280 1% equal to 14.5 (20).
ATP hydrolysis was assayed luminometrically. The assay mixture of a 0.2-1-ml volume contained 1 M ATP, 83 mM TES/KOH (pH 7.2), 17 mM KCl, and varied amounts of potassium phosphate and MnCl 2 . Assays done with Co 2ϩ employed a 100 mM MOPS/KOH buffer (without KCl) because TES was reported to bind Co 2ϩ appreciably (24). Bovine serum albumin (0.01 mg/ml) was also added for enzyme stability in the assays done with R78K-PPase. The reaction was initiated by adding PPase. Aliquots (20 l) of the assay mixture were withdrawn at 2-3min intervals over 20 -25 min and added to 0.18 ml of 0.1 M Tris acetate buffer (pH 7.75) containing luciferin/luciferase reagent (Sigma ATP assay mix), 10 mM magnesium acetate, 2 mM EDTA, 1 mg/ml bovine serum albumin, and 1 mM dithiothreitol. The luminescence was then measured with an LKB model 1250 luminometer. The concentration of the luciferin/luciferase reagent was sufficient to produce a 100-mV signal for samples containing 1 M ATP.
The procedures used to measure enzyme-bound PP i formation at equilibrium (25) and enzyme-catalyzed P i -water oxygen exchange (26) were as described previously. The media used in the incubations for oxygen exchange and synthesis of enzyme-bound PP i were prepared by mixing appropriate volumes of 100 mM potassium phosphate, 100 mM MOPS/KOH buffer (pH 7.2), and 100 or 3 mM CoCl 2 , respectively. Although the solubility products for CoHPO 4 (1.9⅐10 Ϫ7 M 2 (Ref. 27)) and MnHPO 4 (1.4⅐10 Ϫ13 M 2 (Ref. (28)) were exceeded in many of the incubations, no incipient precipitation occurred in the concentration ranges of the metal ions and P i used in this study. That the precipitation in these systems is quite slow was confirmed by the following data. First, no decrease in P i concentration was detected when solutions containing 1 mM Mn 2ϩ plus 200 M P i or 20 mM Mn 2ϩ plus 20 M P i were incubated for 30 min and centrifuged for 15 min at 4,000 ϫ g. Similarly, no change in Co 2ϩ concentration, measured with Arsenazo III (28), was detected at 0.5 mM Co 2ϩ and 5 mM P i after a 15-min incubation and centrifugation as above. At higher concentrations of the metal ions and P i , the solutions became opalescent, and a decrease in P i or Co 2ϩ concentration was observed after centrifugation.
All measurements were performed at 25°C. The concentrations of free Mn 2ϩ and Co 2ϩ ions and of their complexes with P i were calculated using dissociation constants of 3.6 and 9.8 mM, respectively, which were measured with Arsenazo III (29) under the conditions used in the present work. 2 ATP hydrolysis in the presence of P i is described by Scheme II, where k cat is the catalytic constant and K p is the dissociation constant of the enzyme⅐P i complex. Y-PPase converts ATP into ADP and P i , and further hydrolysis of ADP to yield AMP and P i proceeds at a negligible rate (30).
The rate of ATP hydrolysis obeyed Equation 1, where t is time, K m is the Michaelis constant, and [E] is free enzyme concentration. In the absence of P i (added externally or formed from ATP), [E] could be equated to the total enzyme concentration, [E] t , because the Michaelis constant for ATP is in excess of 100 M, a value much greater than its concentration in the reaction medium. In the presence of P i , [E] could be calculated from Equation 2, where [P] t is the total P i concentration in the reaction medium. Equation 2 was obtained by solving the equation for K p (Equation 3) together with Equations 4 and 5, describing mass balance for enzyme and inhibitor, respectively.
It should be noted that [P] t in Equations 2 and 5 refers to the sum of the added and enzymatically generated P i and is therefore given by Equation 6, where [P] 0 and [ATP] 0 are the initial concentrations of P i and ATP, respectively.
In theory, the value for K p could be obtained from the time course of ATP hydrolysis measured in the absence of added P i ([P] 0 ϭ 0), thus evaluating the effect of P i produced enzymatically. However, much more accurate estimates for this parameter were obtained from fittings that simultaneously employed the time courses measured in the absence and presence of added P i (Fig. 2). When such time courses were measured at varied [E] t , it was treated as a variable along with t and [P] 0 .
The rate of P i -water oxygen exchange, v ex , was calculated as 4[P] t ln(E 0 /E)/t, where E 0 and E are the average 18 O enrichments of P i before and after incubation with Y-PPase, and t is the time of the incubation. Values of the partition coefficient P c were calculated using the program written by Hackney (31).
All the fittings were performed using the program SCIENTIST (MicroMath).

RESULTS
P i Inhibition of Mn 2ϩ -supported ATP Hydrolysis-P i was found to be a very potent inhibitor of wild type Y-PPase in the presence of Mn 2ϩ (Fig. 2). The inhibition constant calculated from the time courses of ATP hydrolysis, as described under "Materials and Methods," decreased with increasing [Mn 2ϩ ], approaching a constant level of about 0.04 M (Fig. 3). The mutations decreased the affinity of Y-PPase to P i , the effect being moderate with R78K and K193R variants and quite large with the K56R, and especially Y93F variants (Fig. 3).
The Mn 2ϩ concentration dependences of K p shown in Fig. 3 could be well described by Scheme III. This implies that two enzyme species, EM 2 and EM 3 , bind MnP i with different affinities. Y-PPase is known to have three binding sites for divalent metal ions in the absence of substrate and P i (9,32,33), but the affinities of two of these sites are quite high (see next section), and they are always saturated at the Mn 2ϩ concentrations used in the present study. Fitting of the data in Fig. 3 to Equation 7 derived for Scheme III yielded the parameter values shown in Table I. The fittings indicated that the EM 3 P species is present in insignificant amounts with wild type Y-PPase and its R78K variant over the P i concentration ranges used (0 -5 and 0 -7 M, respectively). In terms of Scheme III, the decrease in K p with increasing [Mn 2ϩ ] is caused by accumulation of MP (the actual binding form of P i ) and EM 3 (a better binding form by comparison with EM 2 ).
Values of k cat /K m , also obtained from Equations 1, 2, and 6 using the time courses of ATP hydrolysis (Fig. 2)  measurements of Mn 2ϩ binding in the presence of P i provided direct support for Scheme III. In these experiments, Y-PPase was equilibrated with 50 M MnCl 2 at varied concentrations of P i , and the manganese content of the two chambers separated by a dialysis membrane was measured by atomic absorption spectroscopy. As shown in Fig. 4, the stoichiometry of Mn 2ϩ binding approached 2 in the absence of P i , indicating nearly full occupancy of the two high affinity sites, consistent with earlier data (13,14,34), and was above 3 at the highest P i concentration. The theoretical curve obtained using Equation 8 with K p M4 ϭ 0.041 M, K p M3 ϭ ϱ, K M3 ϭ 1.7 mM (Table I) Table I.
EM 4 P SCHEME III. Phosphate and metal ion binding to Y-PPase.
͓MP͔ ϭ ͓M] t Ϫ ͓M] (Eq. 10) P i Inhibition of Co 2ϩ -supported ATP Hydrolysis-P i binding by Y-PPase and its variants in the presence of 0.5 mM Co 2ϩ was also estimated from the inhibition of ATP hydrolysis. The binding seemed to be somewhat less strong (Table II), but the effects of the mutations closely paralleled those obtained with Mn 2ϩ (Fig. 3). The effects on k cat /K m were also very similar with Mn 2ϩ (Table I) and Co 2ϩ (Table II).

Formation of Enzyme-bound PP i in the Presence of Co 2ϩ -
The Y-PPase active site contains two P i binding subsites. P i binding to both of these subsites can be accessed by measuring the fraction of the enzyme containing bound PP i (f epp ) as a function of P i concentration in solution (9) because PP i synthesis requires occupancy of both subsites. Values of f epp measured in the presence of 0.5 mM free Co 2ϩ ion exhibited a hyperbolic dependence on [P] t (Fig. 5), allowing calculation of the limiting value of f epp at infinite [P] t (0.18 Ϯ 0.03) and the dissociation constant for P i binding (3 Ϯ 1 mM). The latter value exceeds K p (Table II) by a factor of 240 and therefore characterizes binding of the second P i molecule. P i -Water Oxygen Exchange in the Presence of Co 2ϩ -Y-PPase catalyzes rapid exchange of oxygen between P i and water in the presence of Mg 2ϩ and other metal ions that activate PP i hydrolysis (30,(35)(36)(37). This exchange results from a dynamic reversal of the steps characterized by k 3 and k 4 in Scheme I (36,37): one oxygen originally present in P i is released as water when the two bound P i molecules dehydrate forming PP i (k 4 step) and is subsequently replaced by an oxygen from water when the PP i is converted back into P i (k 3 step). The exchange is conveniently followed by mass spectrometry, starting from  Table III shows two examples of the observed distributions of the five P i species that have from 0 to 4 exchanged oxygens in the Co 2ϩ -supported reaction catalyzed by wild type Y-PPase. Such distributions are characterized by two parameters: the exchange rate v ex and the partition coefficient P c , which equals the probability of bound phosphate conversion into EM 4 PP versus its release into solution in Scheme I (31). As shown in Table III and Fig. 6A, P c exhibited a strong dependence on [P] t in the Co 2ϩ -supported reaction, changing from less than 0.3 at low [P] to ϳ0.9 at its saturating concentration. By contrast, P c is independent of [P] t in the Mg 2ϩ -supported reaction (9,15,31). v ex /[E] t increased with [P] t , reaching a maximum at about 1.5 mM P i , and dropped slightly at higher [P] t (Fig. 6B). By   dividing v ex /[E] t by 4P c /(4 Ϫ 3P c ) (the average number of the exchanged oxygens in each P i molecule leaving the enzyme), one obtains the rate of release of the phosphate that contains exchanged oxygens. This parameter dropped hyperbolically with increasing P i concentration (Fig. 6B), yielding a dissociation constant of 1.2 Ϯ 0.4 mM, a value not much different from that for the binding of the second P i molecule, as derived above from f epp measurements. DISCUSSION Five amino acid residues interact through their side chains with P i in the PPase active site (Fig. 1): three (Arg 78 , Tyr 192 , and Lys 193 ) at subsite P1 and two (Lys 56 and Tyr 93 ) at subsite P2. The remaining interactions are through four metal ions. We have demonstrated previously that the Y-PPase variants used in the present work retain 2-20% activity against PP i (18), indicating that the active site remains catalytically competent. It is noteworthy that the mutations used are highly conservative and do not change the overall charge of the active site. The absence of large changes in structure outside active site in the R78K variant has been demonstrated by x-ray crystallography (38). In this variant, the positive charge on the introduced lysine is displaced by about 3 Å compared with arginine.
The P i binding capacity of four of these variants was estimated here from the inhibition of ATP hydrolysis in the presence of Mn 2ϩ and Co 2ϩ as cofactors. The advantage of these cations over Mg 2ϩ is 3-fold. First and most important, the binding affinities of the subsites P1 and P2 differ much more in the presence of Mn 2ϩ or Co 2ϩ (13,14). Second, these cations support hydrolysis of ATP (Mg 2ϩ does not), a substrate with a high K m value. At 1 M ATP, more than 99% of Y-PPase is substrate-free in the assay medium. Therefore, the true P i binding constant can be estimated directly from the effect of P i on activity at this fixed ATP concentration. Third, the structure shown in Fig. 1 was, in fact, determined for M ϭ Mn 2ϩ , although similar radii and coordination chemistry for Mn 2ϩ and Mg 2ϩ suggest that this structure should be largely preserved with Mg 2ϩ .
P2 Is the Tighter Binding Subsite for Phosphate in the Presence of Mn 2ϩ and Co 2ϩ -The data in Fig. 3 and Tables I and II indicate that the two mutations at subsite P2 have a much greater effect on P i binding to Y-PPase than the two mutations at subsite P1. The measured binding clearly refers to the tighter binding subsite because occupancy of any one of the two subsites by P i will suffice for inhibition. Furthermore, the data in Fig. 5 demonstrate that the dissociation constant for the weaker binding subsite is as high as 3 mM in the presence of Co 2ϩ , 240 times larger than for the tighter binding site (Table  II). The weaker binding site was thus essentially empty during K p measurements, consistent with the earlier equilibrium dialysis and P i titration data revealing only one P i binding subsite in the presence of Mn 2ϩ and Co 2ϩ and up to 0.1 mM P i (13,14).
Other data support the identification of P2 as the tighter binding subsite in the presence of Mn 2ϩ and Co 2ϩ . First, the tighter binding site requires four Mn 2ϩ ions for optimal P i binding (Scheme III), as one would expect for P2 that has four metal ligands, rather than P1 that has only two (Fig. 1). Second, P i was observed only in P2 in the manganese structure of the R78K variant (38), whose P i binding affinity is similar to that of the wild type enzyme. Finally, mutations at the P2 subsite (K29R and Y55F) in E. coli PPase have also been shown to have greater effects on P i binding measured in the presence of 0.1 mM Mn 2ϩ than mutations at the P1 subsite (R43K, Y141F, and K142R). 4 Interestingly, the effects of the mutations on ATP binding, as characterized by k cat /K m (Tables I and II), parallel those on P i binding ( Fig. 3 and Table II). This indicates that ATP binding is dominated by interactions at subsite P2. One of the P1 mutations (R78K) increases k cat /K m for Mn 2ϩ -supported ATP hydrolysis nearly 5-fold (Table I), suggesting that the interac- tion of Arg 78 with the ␤-phosphate of ATP is destabilizing in the enzyme-substrate complex.
P2 Contains the Electrophilic Phosphorus Atom-One would expect faster release of the product P i from P1, where binding is much weaker, insofar as this P i faces solution, whereas the P2 P i is buried at the bottom of the active site (6,7). On the other hand, the P i containing the electrophilic center is released first in the presence of Mg 2ϩ (9). Therefore, either the electrophilic center is at P1, or the order of P i release is reversed with Mn 2ϩ and Co 2ϩ . A choice between these alternatives can be made on the basis of the oxygen exchange data.
The rationale for using this approach is as follows. Once bound to the electrophilic center, each P i molecule undergoes a series of synthesis/hydrolysis cycles, each resulting in exchange of one of the four oxygens. The longer the P i molecule stays in the active site, the more extensive is the exchange before it is released into solution. The average number of exchange cycles depends on the probability of P i forming PP i versus being released into solution, P c . If the electrophilic P i molecule is released first, P c is equal to k 4 /(k 4 ϩ k 5 ) (9, 31) and is thus independent of [P] t , as observed in the case of the Mg 2ϩ -supported reaction (9,15,31). If, however, the electrophilic P i molecule is released second, the time it stays in the active site would increase with increasing the occupancy of the other subsite and hence would depend on [P] t according to Equation 11 (9). At infinite [P] t , P i never leaves the enzyme, resulting in exchange of all of its oxygens (P c ϭ 1). At [P] t approaching 0, P c would also approach 0, corresponding to zero exchange. A thorough theoretical analysis of the P i -water oxygen exchange catalyzed by PPase was done by Hackney (31).
(Eq. 11) Fig. 6A shows that P c increases from about 0 to 0.88 with [P] t in the Co 2ϩ -supported reaction. Importantly, this effect is observed at the P i concentrations allowing appreciable binding to both P2 and P1 (Fig. 5). This provides strong evidence that the P i molecule acquiring oxygen from water during PP i hydrolysis is released second. This conclusion is consistent with v ex /[E] t going through a maximum at increasing [P] t (Fig. 6B). Indeed, the exchange rate depends on [EP] in this case, and [EP] passes through a maximum when P i concentration is increased. Accordingly, the rate of the release of phosphate-containing exchanged oxygens (equal to v ex (4 Ϫ 3P c )/4P c [E] t , where 4P c /(4 Ϫ 3P c ) is the average number of the exchanged oxygens in each P i molecule (23)) decreases with [P] t (Fig. 6B). In the Mg 2ϩ -supported reaction, P c is independent of [P] t , which is strong evidence that the P i molecule containing the oxygen that comes from water is released first, and the exchange rate, which is proportional to [EPP] in this case, always increases with [P] t (9,15,31). Because slower release of P i is expected from the tighter binding site, the electrophilic center is thus at P2.
As mentioned above, the independence of P c from [P i ] in the presence of Mg 2ϩ provides strong evidence that the P i containing the electrophilic phosphorus is released first after PP i hydrolysis (9). This implies that the order in which the two subsites release P i is reversed in the presence of Mg 2ϩ , which may mean that the P i binding affinities of P1 and P2 are also reversed.
It should be noted that the upper limit for P c in Fig. 6A is less than the value of unity predicted by Equation 7. A likely explanation is that the release of the two P i molecules formed is not a strictly ordered reaction in the presence of Co 2ϩ ; that is to say, a significant fraction of P i leaves P1 first. This explanation is consistent with the observation that the value of v ex (4 Ϫ 3P c )/4P c [E] t tends to reach a constant non-zero value with increasing P i concentration (Fig. 6B) rather than approach 0, as expected for a strictly ordered release.
To sum up, these data provide strong support for the mechanism of PP i hydrolysis (6,8), involving water/hydroxide addition to the phosphorus atom located in the site P2 of PPase.