Probing Essential Water in Yeast Pyrophosphatase by Directed Mutagenesis and Fluoride Inhibition Measurements*

The pattern of yeast pyrophosphatase (Y-PPase) inhibition by fluoride suggests that it replaces active site Mg2+-bound nucleophilic water, for which two different locations were proposed previously. To localize the bound fluoride, we investigate here the effects of mutating Tyr93and five dicarboxylic amino acid residues forming two metal binding sites in Y-PPase on its inhibition by fluoride and its five catalytic functions (steady-state PPi hydrolysis and synthesis, formation of enzyme-bound PPi at equilibrium, phosphate-water oxygen exchange, and Mg2+ binding). D117E substitution had the largest effect on fluoride binding and made the P-O bond cleavage step rate-limiting in the catalytic cycle, consistent with the mechanism in which the nucleophile is coordinated by two metal ions and Asp117. The effects of the mutations on PPi hydrolysis (as characterized by the catalytic constant and the net rate constant for P-O bond cleavage) were in general larger than on PPi synthesis (as characterized by the net rate constant for PPi release from active site). The effects of fluoride on the Y-PPase variants confirmed that PPase catalysis involves two enzyme·PPi intermediates, which bind fluoride with greatly different rates (Baykov, A. A., Fabrichniy, I. P., Pohjanjoki, P., Zyryanov, A. B., and Lahti, R. (2000)Biochemistry 39, 11939–11947). A mechanism for the structural changes underlying the interconversion of the enzyme·PPi intermediates is proposed.

Inorganic pyrophosphatase (EC 3.6.1.1; PPase) 1 catalyzes reversible phosphoryl transfer from pyrophosphate (PP i ) to water, a metabolically important reaction chemically similar to that catalyzed by numerous ATPases and GTPases. Yeast PPase is a homodimer containing 286 amino acid residues/ monomer (1) and requiring three or four divalent metal ions for catalysis, with Mg 2ϩ conferring the highest activity (2)(3)(4)(5). Two divalent metal ions (M1 and M2) per active site have been identified in the "resting" enzyme by x-ray crystallography and four metal ions (M1-M4) and two phosphates (P1 and P2) in the product complex of Y-PPase (6,7). PP i hydrolysis by PPase occurs via direct attack of water without formation of a phosphorylated enzyme intermediate (8). Modeling the transition state of the chemical step from the structure of the enzyme-product complex Y-PPase⅐Mn 2 (MnP i ) 2 has led to two models that differ in the identity of the water nucleophile placed between metal ions M1 and M2 in the model of Heikinheimo et al. (6) or in the vicinity of Tyr 93 in the model of Harutyunyan et al. (7,9). Although the former model requires an additional "relaxation" step, in which a new water molecule displaces P i oxygen from the position between M1 and M2, it has an advantage of providing an efficient mechanism for nucleophile activation through combined action of the two metal ions and an adjacent Asp 117 residue (see Fig. 1).
In aqueous solution and even in crystalline state, proteins are surrounded with water shells, making identification of function-related water molecules a difficult task. Use of fluoride, a potent and most specific inhibitor of cytoplasmic pyrophosphatase, provides a convenient approach to detect such water molecules because molecules of HF and H 2 O are isoelectronic and of similar size, as are the anions derived therefrom. Fluoride inhibition of yeast PPase during PP i hydrolysis and synthesis involves a rapid and slow phases (10), which refer to F Ϫ binding to enzyme-P i and enzyme-PP i intermediates, respectively (Scheme I). The rapid binding decelerates PP i hydrolysis 10-fold at pH 7.2, whereas the slow binding arrests it completely. These characteristics of the inhibition are consistent with fluoride replacing an essential metal-bound water molecule/OH Ϫ ion, acting as nucleophile in the PP i hydrolysis step (11).
In the present work, we employed fluoride inhibition in combination with site-directed mutagenesis to identify the fluoride binding site and, hence, the essential water molecule in the active site of PPase. Because fluoride inhibition of Y-PPase is closely associated with Mg 2ϩ binding (10,12), the effects of mutating five amino acid residues forming the M1 and M2 sites (Glu 48 , Asp 115 , Asp 120 , Asp 117 , and Asp 152 ) on fluoride inhibition were studied. A Y93F variant was also included in this list, because in the PPase mechanism suggested by Harutyunyan et al. (9) a water molecule associated with Tyr 93 is assumed to be the nucleophile and, hence, might be replaced by F Ϫ . X-ray crystallographic analysis of three relevant variants, D117E Y-PPase (13) and D65N and D70N Escherichia coli PPase (Asp 65 and Asp 70 in E. coli PPase correspond to Asp 115 and Asp 120 in Y-PPase, respectively) (14), indicated no marked structural changes induced by the mutations.

EXPERIMENTAL PROCEDURES
The expression and purification of wild type Y-PPase and its active site variants from overproducing E. coli XL2blue b strain transformed with suitable plasmids were carried out as described by Heikinheimo et al. (15). PP i hydrolysis was measured continuously with an automatic P i analyzer (16). The assay medium contained, except as noted, 0.32 mM (at pH 7.2) or 0.28 mM (at pH 8.5) total PP i (corresponding to 0.2 mM Mg 2 PP i complex), 5.5 mM MgCl 2 (corresponding to 5 mM free Mg 2ϩ ), 0 -20 mM F Ϫ (added as NaF) and buffer. The following pH buffers were used (0.1 M ionic strength, 50 mM K ϩ ): 83 mM TES/KOH, 17 mM KCl, 50 M EGTA (pH 7.2), or 90 mM TAPS/KOH, 5 M EGTA (pH 8.5). Enzymebound PP i formation was assayed luminometrically with ATP-sulfurylase and luciferase (17,18) using 60 -260 M Y-PPase concentration. PP i synthesis in solution was measured continuously by the same coupled enzyme assay (11). P i -H 2 O oxygen exchange was measured by gas chromatography/mass spectrometry (19). The assay medium used to measure fluoride effects on PPase in the presence of P i contained 13.4 mM MgCl 2 (5 mM free Mg 2ϩ ), 20 mM total P i , 55 mM TES/KOH buffer, 11 mM KCl, and 33 M EGTA. Mg 2ϩ binding was assayed by equilibrium microdialysis in combination with atomic absorption spectroscopy to measure Mg content in the dialysis chambers (20). All experiments were performed at 25°C.
Effects of fluoride on PP i hydrolysis were analyzed in terms of Scheme II, a simplified version of Scheme I. K F1 is the dissociation constant governing rapid fluoride binding in the presence of PP i , k i, app and k r are the second-order and first-order rate constants for slow fluoride binding and release, respectively.
Equations 1 and 2 describe product (P) formation curves at a fixed fluoride concentration, where ␣ is the fraction of enzyme that has not yet undergone slow conversion into E inactive , and v 0, app is the initial velocity of product formation.
Fitting [P] as a function of both time and [F] was accomplished by making the following substitutions (v 0 and v 0 Ј are the initial velocities of product formation observed at [F] equal to zero and infinity, respectively).
Equations 1-4 were simultaneously fit, with the program SCIENTIST (MicroMath), to sets of product formation curves, each represented by 90 -200 pairs of [P] and t values. The calculated and measured curves agreed within 3%. Values of the dissociation constants K M1 and K M2 for Mg 2ϩ binding to two sites on Y-PPase were estimated by fitting equilibrium dialysis data to Equation 5, where n measures the number of Mg 2ϩ ions bound per monomer (21).

RESULTS
Mg 2ϩ Binding-The effects of conservative mutations of Tyr 93 and five amino acid ligands to Mg 2ϩ (Fig. 1) on the binding of two activating metal ions in Y-PPase active site were studied by equilibrium dialysis at pH 7.2 ( Fig. 2). Interestingly, every substitution suppressed Mg 2ϩ binding to the low affinity site, the lowest (1.8-fold) effect seen with Y93F, and the highest (147-fold) with E48D substitutions ( Table I). Binding of Mg 2ϩ to the high affinity site was markedly suppressed in both Asp 120 variants (34-and 16-fold effects with D120E and D120N, respectively) and less markedly (4.0-fold effect) in the D152E variant. These data are consistent with M1 being the high affinity site and M2 being the low affinity site, as measured by equilibrium dialysis.
Importantly, all the variants tested still exhibited appreciable affinity for Mg 2ϩ , such that both sites were almost saturated at 5 mM Mg 2ϩ concentrations used in the fluoride inhibition studies described below. The E48D variant, which exhibited the highest K M2 value (5.0 mM), is expected to be 50% saturated at the M2 site in the absence of substrate, but keeping in mind that substrate strengthens metal ion binding to the corresponding E-PPase E20D variant (22) and WT-E-PPase (23), at least, 90% saturation is expected in the presence of 200 M Mg 2 PP i used in fluoride inhibition studies. This rules out a possibility that the effects of the mutations on fluoride inhibition reported below result from incomplete metal binding. Fluoride added at a 10 mM concentration suppressed binding of M1 and stimulated binding of M2 approximately 2-fold in WT-Y-PPase (Table I).
Catalytic Properties-Values of the catalytic constant for PP i hydrolysis (k h ), rate of PP i synthesis (v s ), rate of P i -H 2 O oxygen exchange (v ex ), the distribution of five P i "isotopomers" during the exchange (as characterized by the partition coefficient P c ϭ k 4 /(k 5 ϩ k 4 ) (19), and the amount of enzyme-bound PP i at equilibrium with medium P i (f epp ) were measured for six variants at pH 7.2 and 8.5 (Table II). Such measurements were not performed for the D120N variant, which displayed very low hydrolytic activity (Ͻ0.005 s Ϫ1 ). These data allowed evaluation of two important parameters:  (21). Equations 6 (11) and 7 (21) relate k 3 Ј and k pp, off to the rate constants shown in Scheme I (K AB ϭ k A /k B ). In fact, k 3 Ј is the lower limit for k 3 (corresponding to K AB ϭϱ), and k pp, off is the lower limit for both k 2 and k B . For WT-PPase, K AB is equal to 25-65, and the values of k 3 Ј and k 3 differ by less than 4% at pH Ն 7.2 (11), but the values of K AB for the variant PPases are unknown. The advantage of using k 3 Ј and k pp, off instead of v ex /[E] t and v s /[E] t is that the former parameters are independent of the degree of enzyme saturation with P i in corresponding measurements.
Fluoride Inhibition of PP i Hydrolysis-According to the inhibition pattern, all the variants could be classified into three groups. Typical examples are shown in Fig. 3. Group I variants (E48D, Y93F, D115E, and D117E) resembled WT-Y-PPase (11) by exhibiting a virtually instant decrease in the initial slope of the P i production curve, followed by a slower inactivation step. The group II variant (D120E at pH 7.2) exhibited only slow inactivation, without the instant decrease in the initial slope, and the group III variant (D152E) exhibited only instant inactivation. At pH 8.5, D120E-PPase also behaved like a group III member, whereas other variants did not change their behavior. It should be noted that the P i productions curves look very similar for group I and group II variants (Fig. 3), and they could be only distinguished based on the fitting results shown below; the best-fit value of K F1 tended to approach infinity for D120E-PPase, clearly indicating no change in v 0,app in the presence of fluoride. Values for the rate constants k i and k r describing the slow phase of the inhibition and for the equilibrium constant K F1 describing the fast step of the inhibition (Scheme II) were obtained by simultaneous fitting of Equations 1-4 to series of product formation curves measured at different fluoride concentrations. The most drastic effects were observed on k i with the D117E and D152E variants (the latter exhibited no slow inactivation step at any pH value), on k r with the D115E and D120E variants, and on K F1 with the D117E variant (Table III).
Fluoride Effect on Enzyme-bound PP i Formation-The Y-PPase active site is preformed to bind PP i . Although the concentration of PP i present at equilibrium with 20 mM P i and 5 mM Mg 2ϩ at pH 7.2 is Ͻ1 M, i.e. the ratio [P i ]/[PP i ] Ͼ 20,000 (5,24,25), the fraction of enzyme containing bound PP in these conditions is as high as 16% (Table II), in agreement with previous estimates (5,26,27). This remarkable feature of the active site was completely retained in D117E and D120E variants, partially retained in the D48E, Y93F, and D115E variants and almost completely lost in the D152E variant (Table II and Fig. 4).
As shown in Fig. 4, addition of 2 mM fluoride to E48D, Y93F, and D115E variants at pH 7.2 caused a rapid decrease in enzyme-bound PP i , not resolved in time (F Ϫ binding to enzyme-P i complex in Scheme I), followed by a slow increase to a higher level (F Ϫ binding to enzyme-PP i complex) (Fig. 4), like in wild type Y-PPase (11). The time course of the slow increase could be described by Equations 8 and 9 (11), where f epp fast refers to the amount of enzyme-bound PP i after completion of the fast phase of the interaction with F Ϫ (i.e. the zero time point for the slow phase), and f epp slow is the amount of the enzyme-bound PP i accumulated at time t during the slow phase. Fluoride binding was assumed to be a second-order reaction (see also below).
For the D117E and D120E variants, the initial decrease in f epp could be resolved in time. Moreover, no slow increase in f epp followed after its rapid drop in the D120E variant. Values of k i, app , k r , and f epp fast for the D117E variant were obtained with Equations 8 and 9 using the points collected at t Ն 2 min on the curve shown in Fig. 4. The curve for the D120E variant and the initial part of the curve for the D117E variant (t Յ 2 min) could be described by Equations 10 and 11 for a reversible binding, where r is the percentage of fraction of enzyme-P i complex containing bound fluoride and the primed rate constants refer to F Ϫ binding to and release from enzyme-P i complex. As the enzyme-P i complex binds two fluoride ions (see below), Equation 11 contains the second power of fluoride concentration. The effect of F Ϫ on PP i bound to D152E-PPase was quite small, if there was any.
The ratio of k i, app /f epp fast , the true rate constant for F Ϫ binding to enzyme-PP i complex (k i, app refers to total enzyme), is de- In the presence of 10 mM NaF.
creased only in the D117E variant by comparison with wild type Y-PPase (Table IV). Inactivation by Fluoride in the Presence of P i -Upon the incubation with P i and F Ϫ at pH 7.2, the variant PPases were diluted 900 -4500-fold into assay medium, and the initial ve-locity of PP i hydrolysis was estimated from the slope of the P i production curve 1 min after the dilution (the dead-time of the P i analyzer increased by comparison with Fig. 3 because of the P i present in the enzyme solution). The extensive dilution and the 1-min delay were expected to completely reverse the fast    fluoride binding step seen in Fig. 4 for the E48D, Y93E, and D115E variants and to partially reverse the slower initial binding step observed with the D120E and D117E variants. In agreement with the data reported above, the preincubation with fluoride did not affect the activity of the D152E variant, slightly inactivated the D115E and D120E variants and more deeply inactivated the E48D, Y93E, and D117E variants (Figs. 5 and 6). Low degree of the inactivation seen with the D115E variant is consistent with its high k r value (Table IV). For all the variants, the degree of the inactivation observed after 80 min was smaller than for wild type Y-PPase (11), consistent with lower k i, app or higher k r value (Table IV). The values of these parameters estimated with Equation 1 for the two variants exhibiting Ͼ30% inactivation in Figs. 5 and 6 (Y93F and D117E) were similar to those derived from Fig. 4 (Table IV). For the D117E variant, only points with t Ͼ 5 min were used in this estimation. Increasing the F Ϫ concentration decreased D117E-PPase inactivation (Fig. 6), a phenomenon previously observed with wild type Y-PPase and explained by binding of two F Ϫ ions to enzyme-P i complex (11). Such binding decreases the concentration of the enzyme⅐PP i complex (the true binding species) proportionally to [F] 2 , and this is only partly compensated by an increase in the fluoride binding rate, which is proportional to [F].
Fluoride Inhibition of D117E-PPase during PP i Synthesis-Studies of PP i synthesis provided further evidence for the ability of D117E-PPase to bind two F Ϫ ions in the presence of P i (Fig. 7). The effect of [F] on the initial velocity of PP i synthesis was only poorly described by Equation 3 (with vЈ 0 ϭ 0), implying binding of only one F Ϫ ion but was well described by Equation 12, implying binding of two F Ϫ ions (Fig. 7, right  panel).
The nonlinearity of the luminescence versus time curves (Fig. 7, left panel) was similar at different [F] and could be fully explained by fluoride-independent inactivation of luciferase (11). This means that the slow fluoride binding step is not evident over the 1-4.5-min time interval in the presence of P i , consistent with the data in Fig. 4, showing no significant increase in f epp during this time interval.

Roles for the Six Residues in Metal Binding and Catalysis-
There is a substantial increase in K M2 with five of the six variants under study (Table I), indicating the M2 site to be highly sensitive to changes in the delicate network of interactions in the active site of Y-PPase (Fig. 1). The most drastic effects on K M2 are seen with the E48D and D120E variants, which is consistent with the x-ray crystallographic data showing Glu 48 and Asp 120 to be most closely associated with M2 (Fig. 1). The M1 site is clearly much more resistant to amino acid substitutions than the M2 site; a significant effect is only seen with the D120E(N), D152E, and D117E variants. This is compatible with changes seen between the EM 2 and EM 2 (MPi) 2 complexes (6). In these structures, the M2 site showed itself to be more rigidly fixed than the M1 site, so that the loops carrying Asp 115 and Asp 152 swing in toward a rigid M2 binding site in the product complex structure. Consequently, variation at the M1 ligands Asp 115 and Asp 152 is more readily tolerated than variation at the M2 ligands Asp 120 and Glu 48 . However, it is unclear why D115E and D152E substitutions have much larger effect on M2 than on M1 binding (Table  I). One possible explanation is that the substitutions effectively misposition M1 and Asp 120 and so prevent the rigid M2 binding site from forming properly. Consistent with this notion is the fact that the changes seen at M2 for the D115E and D152E substitutions, although large, are smaller than that seen for directly mutating Asp 120 , as if the effects of the D115E and D152E substitutions were to misalign Asp 120 . As expected, mutating Tyr 93 , which is not involved in metal binding (Fig. 1), had no effect on K M1 and only a minor effect on K M2 (Table I).
In general, the effects of the mutations on PP i hydrolysis (as characterized by k h and k 3 Ј) are larger than the effects on PP i synthesis (as characterized by k pp, off ) ( Table II). All the mutations, with the only exception of E48D at pH 8.5, decrease k h , whereas k pp, off is increased in three variants (E48D, Y93F, and D115E) at pH 8.5. The latter effect appears to result from increased k B (see Equation 7), because the k B step is ratelimiting in PP i synthesis by wild type PPase (11). A similar stimulating effect of an E20D substitution on PP i synthesis was earlier observed with E. coli PPase (Glu 20 in E. coli PPase corresponds to Glu 48 in Y-PPase) (22).
Combining Equation 6 with the expression for k h (11), one obtains At pH 7.2, k A (840 -2200 s Ϫ1 ), k 3 Ј (1200 s Ϫ1 ), k 5 (800 s Ϫ1 ), and k 7 (960 -3300 s Ϫ1 ) are nearly equal for wild type Y-PPase (11), and (1 Ϫ P c ) does not differ much from unity (Table II). Therefore, a mutation decreasing only k 3 Ј by a factor of n would decrease k h by a factor approaching (n ϩ 3)/4. This criterion is met for most mutations, except for D120E and D152E, at pH 7.2 (Table II). For the latter mutations, the effect on k h is nearly equal to (D120E) or exceeds (D152E) the effect on k 3 Ј, which can only result if, at least, one of k A , k 5 , and k 7 is also decreased. The anomalous behavior of the D152E variant is preserved at pH 8.5. Moreover, despite a decrease in k 3 Ј, k h is slightly increased in the E48D variant at pH 8.5, suggesting an increase in k A , k 5 , and/or k 7 . Thus, both PP i hydrolysis (k 3 Ј) and P i release are decelerated in the D152E variant, PP i hydrolysis is decelerated, and P i release is accelerated in the E48D variant, and only PP i hydrolysis is affected in the other four variants. According to Equation 6, the effects on k 3 Ј may result from a decrease in k 3 or K AB . At present, we cannot distinguish between these alternatives. The invariance of k 5 in the Y93F, D115E, and D117E variants would mean that the decrease in  (18). The two panels refer to different time scales. Lines were obtained with Equations 8 and 9 using parameter values found in Table IV. their P c values (Table II) results from a drop in k 4 . This effect is the largest with the D117E variant (6 -13-fold) but, nevertheless, is much less than the effect on k 3 Ј.
The fluoride inhibition data support the notion that PP i conversion to P i becomes rate-limiting in some variants. For D117E-PPase, values of k i in hydrolysis (Table III) and k i, app / f epp fast (Table IV) are the same. This is possible if EM 4 PP* and EM 4 PP predominate during hydrolysis and are at equilibrium with each other, which is equivalent to k 3 being the ratedetermining step in hydrolysis. This is consistent with high K F1 value for this variant (Table III), indicating that the fraction of EM 3 P is quite low during hydrolysis. That EM 3 P, when present, binds fluoride well is indicated by its large effect on PP i synthesis by D117E-PPase (Fig. 7) by comparison with wild type PPase (11). The effect of the Y93F substitution on K F1 is also larger at pH 7.2 (Table III), where k 3 Ј is nearly ratedetermining (Table II), than at pH 8.5, where k 3 Ј is not such. Interestingly, changes in pH had an opposite effect on k h for all the variants, except for D115E, by comparison with wild type Y-PPase (Table II). This appears to result from a substantial alkaline shift in the bell-shaped pH profile for k h (28). As a result, the pH 7.2 and 8.5 points are on the descending limb of the pH rate profile for wild type Y-PPase and on its ascending limb for most variants. By contrast, the D115E substitution leaves the pH profile for k h nearly unchanged (28). The shift may be explained by a change in the pK a of the nucleophilic water (28,29) or by inhibition of the three-metal pathway in a Parameters obtained with Equations 8 and 9 from the data shown in Fig. 4. b Parameters obtained with Equation 1 from the data shown in Figs. 5 and 6. c Parameters obtained with Equations 10 and 11 from the fast step of f epp decline shown in Fig. 4 (dimension for k i , app is mM Ϫ2 min Ϫ1 in this case; see text for details).  Table IV, the line for wild type Y-PPase is from Baykov et al. (11), and the other lines were drawn by eye. the variant PPases (21).
Fluoride Binding Site-The slow inhibition step results in the incorporation of one fluoride ion per substrate molecule, which remains intact and tightly bound to PPase (30), suggesting that fluoride replaces the nucleophilic water. Substitutions mispositioning its ligands are thus expected to affect fluoride binding. Computer modelling of the enzyme-PP i complex from the x-ray structure of Y-PPase and E-PPase suggested that the nucleophile is either Wat1, located between metal ions M1 and M2 and further liganded by Asp 117 (Fig. 1 and Ref. 6) or the water molecule associated with M2 and Tyr 93 (7). The results presented in Tables III and IV clearly favor Wat1 as the nucleophile, because they show that the D117E substitution does have the most drastic effect on F Ϫ binding, as characterized by k i , K F1, K F2 , and k i, app /f epp fast . Furthermore, a strong effect of the D120E substitution on F Ϫ binding is consistent with Asp 120 being important for exact placing M1 and M2 and, hence, Wat1 (Fig. 1). The substitutions of Asp 115 and Asp 152 , which are in the first coordination sphere of M1, and of Glu 48 , which is in the second coordination sphere of M2, apparently have smaller effects on Wat1 position, which correlates with much greater activity of the D115E, D152E and D48E variants by comparison with the D120E or D120N variants (Table II;  Several other lines of evidence support the identification of Wat1 as the water molecule substituted by fluoride. First, Wat1 is replaced with a Glu 117 carboxylate oxygen in D117E-PPase (13), showing the largest decrease in fluoride binding. Second, as measured by differential UV spectroscopy, fluoride binding to Y-PPase in the absence of substrates requires metal ions M1 and M2 (10), the ligands to Wat1. Third, F Ϫ ion becomes entrapped together with two Mg 2ϩ ions in the stable enzyme-F Ϫ -PP i complex (12). Fourth, F Ϫ inhibition of E-PPase, in which the carboxylate of Asp 67 (equivalent of Asp 117 in Y-PPase) is moved away from M1 and M2 by about 1 Å by comparison with Y-PPase (7), exhibits only a fast step. Moreover, a D67N substitution results in appearance of the slow inhibition step in E-PPase (31), which is similar in this respect to Y-PPase. In a series of D/N substitutions in the active site of E-PPase, the D67N substitution had the largest effect on fluoride binding (31). Fifth, other dimetal hydrolases, including enolase (32,33), aminopeptidase (34), purple acid phosphatase (35), and urease (36), presumably generating a hydroxide ion in the dimetal active site, are similarly inhibited by fluoride.
Mechanism of Fluoride Inhibition and Catalysis-According to Scheme I, derived from studies of wild type Y-PPase (11), fluoride binds with different rates and stoichiometry to two enzyme intermediates. Effects of fluoride on Y-PPase variants provide further support to Scheme I. Thus, five of the six variants exhibit a biphasic transition in the amount of enzymebound PP i upon addition of fluoride ( Fig. 4; compare values for f epp and f epp fast in Tables II and IV), the rapid drop and slow rise corresponding to fluoride binding to EM 3 P and EM 4 PP*, respectively. These data are inconsistent with a consecutive binding model, in which rapid preassociation of a single enzyme form (EM 4 PP* in our case) with the inhibitor is followed by a slow conformational change, because such a model predicts an increase in the amount of enzyme-bound PP i during both rapid and slow phases. Besides, the involvement of two different intermediates is supported by the observation that the D120E (at pH 8.5) and D152E substitutions cancel the slow inhibition phase without significantly affecting the fast phase. These substitutions appear to greatly reduce the steady-state concentration of EM 3 P, which binds fluoride rapidly. The ability of EM 3 P to bind two F Ϫ ions, another important feature of Scheme I, is confirmed by the inactivation patterns seen in Figs. 6 and 7 for the D117E variant.
The fluoride binding data obtained for the E48D and D115E variants confirm the occurrence of two PP i -containing intermediates (EM 4 PP* and EM 4 PP in Scheme I) in PPase catalysis. For both variants, k i values (Table III) (Table IV), despite the fact that the former parameter refers to the total enzyme, of which only a fraction contains bound PP i during steady-state hydrolysis, and therefore underestimates the rate constant for fluoride binding. Importantly, the k i, app /f epp fast values are true rate constants for fluoride binding with the sum of EM 4 PP* and EM 4 PP. Like with wild type Y-PPase, this difference can be only explained by existence of two PP i -containing intermediates that react with fluoride at substantially different rates (11). Independent evidence for two PP i -containing intermediates was obtained in presteady-state measurements of PP i hydrolysis. 2 The nature of the structural changes accompanying the EM 4 PP* to EM 4 PP transition still remains to be determined. An attractive hypothesis presented below is based on the observation that the metals M1 and M2 are bridged by two water molecules in EM 2 (6,9) and by only one water molecule in EM 4 P 2 (6) and EM 4 PP. 3 We suggest that the two-water bridge is preserved upon initial substrate binding (i.e. in EM 4 PP*), but one water molecule is expelled during the transition to EM 4 PP (Scheme III), consistent with a decrease in the pK a of the essential basic group, attributable to Wat1, from Ͼ7.0 to 5.8 (21). Recent model studies of Kaminskaia et al. (37) have indicated no substantial increase in nucleophilicity for a bridging water versus terminally bound water. Thus, the M1-Wat1-M2 structure apparently helps to favorably position Wat1 with respect to the attacked P i residue. Scheme III differs from Scheme I in that Scheme I shows fluoride interaction with EM 4 PP* as one bimolecular reaction, whereas in Scheme III this interaction involves rapidly reversible binding of fluoride to EM 4 PP*, with a dissociation constant of KЈ F , and slow conversion of the resulting unstable intermediate, with a rate constant of k A , to a form having a fluoride ion between M1 and M2. However, the two mechanisms are kinetically equivalent, provided that the bottom left species in Scheme III is stoichiometrically insignificant, i.e. KЈ F Ͼ Ͼ [F]. In terms of Scheme III, k i is equal to kЈ A /KЈ F . Given the chemistry of fluoride, the M1-F-M2 complex is likely to be very stable, making the reverse reaction, which requires insertion of a water molecule, very slow. For this reason, it is not shown in Scheme III. On the other hand, the bound PP i cannot hydrolyze and leave as P i . This explains why fluoride stabilizes the enzyme-substrate intermediate to an extent allowing its isolation by gel filtration (30). At neutral pH, the bridging water molecule exists predominantly as OH Ϫ in EM 4 PP and, presumably, EM 4 P 2 (21), which explains their very slow, if any, binding of the negatively charged fluoride ion. The proposed structure of EM 4 PP* is consistent with the results of recent x-ray crystallographic studies of the fluoride-inhibited PP i complex of Y-PPase, showing a single electron density between M1 and M2. 3 By contrast, two water molecules are expected to be present between M1 and M2 in EM 3 P because fluoride binding to this species is relatively weak and proceeds in a rapidly reversible manner with a stoichiometry of two per active site. The unsurpassed stability of the F-H-F hydrogen bond (38) would favor the M1-F-H-F-M2 structure, despite the low pK a value (3.2) for hydrofluoric acid.