Reciprocal Effects of Substitutions at the Subunit Interfaces in Hexameric Pyrophosphatase of Escherichia coli

A homohexameric molecule of Escherichia coli pyrophosphatase is arranged as a dimer of trimers, with an active site present in each of its six monomers. Earlier we reported that substitution of His136 and His140 in the intertrimeric subunit interface splits the molecule into active trimers (Velichko, I. S., Mikalahti, K., Kasho, V. N., Dudarenkov, V. Y., Hyytiä, T., Goldman, A., Cooperman, B. S., Lahti, R., and Baykov, A. A. (1998) Biochemistry 37, 734–740). Here we demonstrate that additional substitutions of Tyr77 and Gln80 in the intratrimeric interface give rise to moderately active dimers or virtually inactive monomers, depending on pH, temperature, and Mg2+ concentration. Successive dissociation of the hexamer into trimers, dimers, and monomers progressively decreases the catalytic efficiency (by 106-fold in total), and conversion of a trimer into dimer decreases the affinity of one of the essential Mg2+-binding sites/monomer. Disruptive substitutions predominantly in the intratrimeric interface stabilize the intertrimeric interface and vice versa, suggesting that the optimal intratrimeric interaction is not compatible with the optimal intertrimeric interaction. Because of the resulting “conformational strain,” hexameric wild-type structure appears to be preformed to bind substrate. A hexameric triple variant substituted at Tyr77, Gln80, and His136 exhibits positive cooperativity in catalysis, consistent with this model.

Inorganic pyrophosphatase (EC 3.6.1.1; PPase) 1 catalyzes the interchange between pyrophosphate and orthophosphate and is essential for life (1,2). Because of its relative simplicity and high efficiency (k cat /K m ϭ ϳ10 9 M Ϫ1 s Ϫ1 ), PPase has become a paradigm for mechanistic and structural studies of enzymatic phosphoryl transfer from phosphoric acid anhydrides to water. The best studied PPases are those from Escherichia coli and Saccharomyces cerevisiae (3,4).
A molecule of E. coli PPase (E-PPase) is formed by six identical subunits, 20 kDa each, arranged with D 3 symmetry in two layers of trimers (Fig. 1A). The three or four contact zones (see below), which are not well separated, cover about 24% of the accessible surface area of a monomer. The intratrimer contacts are of a circular "head-to-tail" type, meaning that each monomer has two different intratrimer contact regions ( Fig. 1B) (5,6). These involve a mixture of hydrophilic and hydrophobic interactions that include Tyr 77 and the backbone NH of Gln 80 (5); the total surface area buried per monomer is about 1300 Å 2 (6).
There are also two different interfaces between trimers. The smaller one (140 Å 2 ) includes a Tyr 77 -Asn 24Ј hydrogen bond (primed and unprimed numbers refer to different monomers). Tyr 77 acts as a link between the intratrimer interface, the minor intertrimer interface, and the major intertrimer interface. The larger intertrimer interface in E-PPase (640 Å 2 ) chiefly involves ␣-helix A, including an ion-triple formed between His 140 , Asp 143 , and His 136Ј (5,8). Replacing either His 136 or His 140 with Gln destabilizes the E-PPase hexamer (9), whereas replacing both makes trimers the dominant species in solution even at millimolar protein concentrations (10). Another important interaction occurs through Mg 2ϩ bound at the intertrimeric interface. The Mg 2ϩ ion is octahedrally associated with six water molecules, which in turn hydrogen bond to the side chains of Asn 24/24Ј and Asp 26/26Ј , as well as to the backbone carbonyls of Asn 24 and Ala 25 (8,11). Substitution of Asp 26 with Asn or Ser eliminates Mg 2ϩ binding to the intertrimeric site and somewhat decreases hexamer stability but hardly affects catalysis (12). Hexamer dissociation into trimers greatly decreases the rate constant for substrate binding to enzyme but has no effect on the catalytic constant (10,12).
Here we describe the effects of substitutions of Tyr 77 and Gln 80 , predominantly at the intratrimeric interface, on the quaternary structure and catalytic activity of E-PPase. This interface mainly consists of hydrophobic contacts between strands and contains few hydrogen bonds between monomers (Fig. 1C). The side chains of Tyr 77 and Gln 80 are thus in a rather hydrophobic environment. We show that Y77D and Q80E substitutions markedly destabilize the intratrimeric contact and, in combination with substitutions at the intertrimeric contact, yield moderately active dimeric and virtually inactive monomeric E-PPase. The results of this study also shed light on the interactions between different subunit contacts in E-PPase.
Methods-The initial rates of PP i hydrolysis were measured by a continuous P i assay (16). The reaction was initiated by adding enzyme and was carried out for 3-4 min at 25°C. No appreciable conversion among enzyme forms was observed during the assay, as evidenced by nearly linear product formation curves. Analytical ultracentrifugation, chemical cross-linking, and electrophoresis were performed as described previously (9).
Data Analysis-Equations 1 and 2, derived from Scheme I, describe time courses of activity (A) resulting from enzyme (E 2 ) dissociation into two equal parts (E) (i.e. hexamer to trimers or dimer to monomers) and the reverse reaction, as well as the equilibrium activity (at t ϭ ϱ, d␣ A /dt ϭ 0) as a function of enzyme concentration (10).

SCHEME I. Hexamer-trimer or dimer-monomer equilibration
A A and A D are the specific activities of the associated and dissociated enzyme, respectively; ␣ A is the fraction of the associated enzyme at time t; [E] t is total enzyme concentration expressed in monomers; k a and k d are the apparent rate constants for the association and dissociation, respectively; K d ϭ k d /k a ; n is 3 for hexamer-trimer and 1 for dimermonomer conversions. Equations 1 and 2 were simultaneously fit to data with the program SCIENTIST (MicroMath).

Effect of Substitutions on Quaternary
Structure-The effects of substitutions at the subunit interfaces on the quaternary structure of E-PPase were studied as a function of pH. Direct evidence for changes in the quaternary structure was provided by sedimentation data (Table I). The data were collected at 2-4°C, because both intratrimeric (see below) and intertrimeric (17) interactions are weakened at low temperature. At pH 7.2, all variants having no substitutions at the intertrimeric interface (Y77D, Q80E, and Y77D/Q80E) and the Q80E/H136Q variant retained their hexameric structure (s 20,w Ն 6.0 S). Lowering the pH to 3.8 -4.6, which disrupts the intertrimeric contact (12,18), decreased s 20,w for these variants to either 3.7-4.2 or 2.6 -3.1 S. Based on the s 20,w values previously measured for wild-type and variant E-PPase trimers whose structures were confirmed by independent methods (9,10,12,19,20), the corresponding enzyme forms were identified as trimer and dimer, respectively. The variants with modified intertrimeric interface behaved as dimers (Y77D/H136Q and Y77D/H136Q/H140Q) or as a mixture of hexamer and dimer/ monomer (Y77A/Q80E/H136Q and Y77D/Q80E/H136Q) at pH 7.2. One variant (Y77D/Q80E/H140Q) exhibited an s 20,w value as low as 2.0 S and was tentatively identified as monomer.
The designation of the enzyme species with s 20,w of about 3.0 S as a dimer is supported by several lines of evidence. First, the molecular mass of Y77D/Q80E/H136Q/H140Q-PPase determined from sedimentation equilibrium was 41 Ϯ 1 kDa under conditions (20°C, pH 7.2, 10 M enzyme, 1 mM MgCl 2 ) where the s 20,w is 2.8 S (Table I). At 1 M initial enzyme concentration, the average molecular mass was estimated to be 29 Ϯ 2 kDa (1 mM Mg 2ϩ , pH 7.2), indicating that the quadruple variant exists as a mixture of dimer and monomer under these conditions. Second, SDS-polyacrylamide gel electrophoresis of the Y77D/ Q80E/H136Q/H140Q variant cross-linked with glutaraldehyde revealed two major protein bands attributable to monomer and dimer (Fig. 2). The minor band corresponding to trimer was much less intense than that from H136Q/H140Q-PPase, which is essentially trimeric (10), and may have resulted from inter- molecular cross-linking. These data indicate that the trimer is either absent or present in only small quantities in the Y77D/ Q80E/H136Q/H140Q variant. The identification of different enzyme species in the variant PPases is further supported by activity data presented in the next section.
A comparison of s 20,w values indicates that WT-PPase is fully trimeric by pH 5.5 but that lower pH is required for full conversion of the Q80E variant to trimer. This was somewhat unexpected, as Gln 80 is in the intertrimeric interface. The stabilizing effect of the Q80E substitution on the intertrimeric contact also explains why H136Q-PPase is trimeric, whereas Q80E/H136Q-Pase is hexameric at pH 7.2 (Table I). These data also indicate that the Q80E substitution stabilizes the enzyme over a wide pH range, not only below pH 7. Conversely, the H136Q substitution, which destabilizes the intertrimeric contact (9), clearly strengthens the intratrimeric contact; the s 20,w value for the Y77D/Q80E/H140Q variant is lower than for the Y77D/Q80E/H136Q/H140Q variant (Table I). These conclusions are corroborated by the activity data described below.
Effect of Substitutions on Activity-Earlier we showed that rates and equilibria of hexamer-trimer interconversion in E-PPase can be monitored by activity measurements at low (20 M) substrate concentration, because the Michaelis constant is drastically greater in trimer than in hexamer and interconversion between these species occurs slowly on the time scale of the enzyme assay (12). The data in Fig. 3 confirm this for WT-, Q80E-, and Q80E/H136Q-PPase and further suggest that this approach can be used to monitor hexamer-dimer interconversion in Y77D-and Y77D/Q80E-PPase. Similarly to the s 20,w values, the activities of WT-, Y77D-, and Y77D/Q80E-PPases drop to their lower levels at pH 5.5, but Q80E-PPase requires pH Ͻ 5.0 to be converted to its low activity form (Fig. 3A). The activity of Q80E/H136Q-PPase, which is substantially greater than that of H136Q-PPase at pH 7.2 ( Fig. 3B), further demonstrates the hexamer-stabilizing effect of the Q80E substitution in accordance with Table I. The activities presented in Fig. 3 represent equilibrium values, no changes being observed during longer incubations. Furthermore, the inactivation observed at low pH could be reversed by at least 80% in all cases by adjusting the pH in the enzyme solution to 7.2 and incubating it for 30 min at 25°C.
As shown in Fig. 4, the Q80E substitution appears to stabilize the hexamer at equilibrium (Table I and Fig. 3) by slowing down its dissociation. Surprisingly, the hexameric Y77D and Y77D/Q80E variants also exhibit much greater kinetic stability than WT-PPase under the conditions tested. At first glance, this contradicts the data in Fig. 3, which indicate inferior thermodynamic stability of these variants compared with WT-PPase. However, unlike WT-PPase, these variants dissociate to dimers (Table I), making hexamer reformation a slow trimolecular reaction.
The activities of the variants carrying the Y77D substitution decrease with decreasing enzyme concentration in pre-equilibrated stock solutions (Fig. 5) in a manner consistent with an equilibrium between active dimers and inactive monomers (Scheme I). Equilibrium between dimers and monomers in the quadruple variant is supported by the sedimentation equilibrium data presented above. In addition, consistent with Scheme I, concentrating a 0.5 M solution of the quadruple variant to 20 M (Centricon 10 ultrafilter, Amicon) reversed by at least 60% the inactivation resulting from dilution at pH 7.2. K d values obtained by fitting Equations 1 and 2 to the activity versus enzyme concentration profiles (Table II) indicate, in agreement with the sedimentation data, that replacing Gln 80 in the Y77D/H136Q/H140Q variant destabilizes the intratrimeric contact. For the Y77D/Q80E/H140Q variant, no K d value characterizing dimer-monomer equilibrium could be obtained in this way because activity rises sharply at enzymes concentrations above 10 M, probably because of hexamer formation, which is not prevented by substituting only one His residue in the intertrimeric interface (9). Nevertheless, the profile for the Y77D/Q80E/ H136Q/H140Q variant at low enzyme concentration is clearly shifted to the left compared with the Y77D/Q80E/H140Q variant. This indicates that the H136Q substitution stabilizes the dimer, in accordance with the sedimentation data described above. The K d values measured for the Y77D/Q80E/H136Q/H140Q variant indicate that the intratrimeric contact is stabilized with increasing pH, Mg 2ϩ concentration, or temperature (Table II).
Tris buffer, used in the incubation media for Fig. 5 and Tables I and II, has recently been shown to bind to the active site of E-PPase (21). Tris markedly decreases the k d for H136Q-PPase hexamer, thereby stabilizing the hexameric form (21). In contrast, Tris exerted only a minor effect on the trimer-dimer and dimer-monomer equilibria (data not shown), consistent with the severely decreased substrate binding affinity of these oligomeric forms (see below).
Functional Properties of Dimeric and Monomeric E-PPase-The activity versus substrate concentration profiles for WT-PPase or its variants with one or both of His 136 and His 140 at the intertrimeric interface indicated the presence of a small fraction of enzyme that is saturated at micromolar substrate concentration (Fig. 6). This is most likely due to trace amounts of hexamers, because it did not occur with the Y77D/H136Q/ H140Q and Y77D/Q80E/H136Q/H140Q variants (Fig. 6A), which do not form hexamers (10). For the variants that are dimeric or monomeric in stock solution (Table I), the bulk of the activity exhibited a nearly linear dependence on the concentration of Mg 2 PP i within its solubility range (Fig. 6A), indicating Michaelis constants of greater than 1.5 mM. This is at least 3 orders of magnitude weaker binding than for hexameric E-PPase (10). For the Y77D/Q80E/H136Q/H140Q variant, K m exceeded 1.5 mM over the pH range 7.2-10.0 (data not shown). In contrast, trimeric forms of wild-type and Q80E-PPase exhibited saturation kinetics with K m values of 350 Ϯ 80 and 160 Ϯ 20 M, respectively, and much greater activity values (Fig. 6B). The ratio of the catalytic constant to the Michaelis constant (k cat /K m ) (i.e. the second-order rate constant for substrate binding to enzyme), as calculated from the slope of the lines in Fig.  6A, ranged between 2 and 5 mM Ϫ1 s Ϫ1 for the dimeric variants and was only 0.2 mM Ϫ1 s Ϫ1 for the monomeric variant. Furthermore, the latter value may be overestimated because of presence of dimer in equilibrium with monomer.
The value of k cat /K m for dimers was maximal at pH 9.3, whereas the pH optima for the hexamer and trimer are 7.2 and 8.5, respectively (Fig. 7). The maximum values of k cat /K m for dimer, trimer (10), and hexamer (21) are 0.018, 0.3 and 100 M Ϫ1 s Ϫ1 , respectively. In contrast, the values of k cat are of the same magnitude: Ͼ70 s Ϫ1 for dimer, 230 s Ϫ1 for trimer (10), and 330 s Ϫ1 for hexamer (21).
In the absence of substrate, hexameric E-PPase binds 2.5  Table II. yhh, Y77D/H136Q/H140Q; yqhh, Y77D/Q80E/H136Q/H140Q; yqh, Y77D/Q80E/H140Q. Mg 2ϩ ions per monomer: one each at sites M1 and M2 located in the active site (K d values of 0.022 and 5.7 mM) and 0.5 (one per two subunits) between trimers (K d of 0.12 mM) (12,21). In the hexamer, substrate (Mg 2 PP i ) binding requires prior Mg 2ϩ binding to M1, but Mg 2ϩ binding to M2 is partially inhibitory (21). In trimers, the same is true at pH Ն 8.0, but at pH 7.2 Mg 2ϩ binding to both sites is required (10). The subunit interface site has no catalytic role in hexamers (12), is absent in trimers (10), and is most likely absent in dimers.
As measured by equilibrium dialysis, Mg 2ϩ interaction with the high affinity site is much weaker in the dimer than in the hexamer and trimer (Fig. 8). This is also indicated by the sigmoidal appearance of the activity versus [Mg 2ϩ ] profile at pH 7.2 ( Fig. 9), which did not occur for the hexamer or trimer in the same range of Mg 2ϩ concentrations. The ascending parts of the profiles shown in Fig. 9 are shifted to higher Mg 2ϩ concentrations compared with the hexamer and trimer, and no drop in activity (i.e. k cat /K m ) is observed at up to 50 mM Mg 2ϩ concentration. In contrast, k cat /K m is reduced 2.5-to 15-fold for the hexamer and trimer in the presence of 20 mM Mg 2ϩ at pH 9.3-9.5 because of competition between the metal ion bound to M2 and the second metal ion of Mg 2 PP i (10,21). This means that the M2 site also exhibits a markedly decreased affinity in the quadruple variant or that substrate binding becomes dependent on the presence of Mg 2ϩ at both M1 and M2 sites over the whole pH range examined.
Cooperativity in Catalysis by Hexameric Y77D/Q80E/ H136Q-PPase-Catalysis by wild-type and variant PPases of different oligomeric states follows Michaelis-Menten kinetics, consistent with the notion that all six active sites function independently. The two exceptions were hexameric Y77D/ Q80E/H136Q-and Y77A/Q80E/H136Q-PPases, which exhibited positive cooperativity among subunits, as illustrated in Fig.  10 for one of the variants. The profiles shown were analyzed in terms of Scheme II, which assumes two equal populations of Four lines of evidence indicate that the apparent cooperativity seen in Fig. 10 does not result from interconversion between different oligomeric forms in the assay medium. First, the enzyme was clearly hexameric in its stock solution from which aliquots were added to the reaction mixture. This is indicated by s 20,w values of 6.6 Ϯ 0.2 and 6.0 Ϯ 0.2 S measured at 10 and 100 M enzyme concentrations, respectively, and by the independence of measured specific activity on the concentration of stock enzyme solution over the range of 0.05-50 M (the partial dissociation seen at low temperature (Table I) was avoided in this case by keeping the stock enzyme solution at 25°C). Second, product formation curves were linear in all cases, indicating no interconversion between dimer/trimer and hexamer during measurement. Third, the specific activity of Y77D/Q80E/H136Q-PPase remained constant when its concentration was varied in the range of 0.25-1.75 nM in the reaction medium. Finally, the profiles of rate versus substrate concentration do not show enzyme heterogeneity similar to that in Fig. 6, as would have occurred, for example, with a mixture of hexamer and dimer/trimer with different K m values.

Disrupting Subunit Interactions in E-PPase-
The results of this study identify Tyr 77 as an important residue in the intratrimeric interface of E-PPase. Replacing Tyr 77 with Asp is enough to dissociate hexamers into dimers at pH Յ 5.5 (Table  I). The Q80E substitution also destabilizes this interface (Table  II), but the effect is smaller and does not result in dimers or monomers under the same conditions (Table I). It is also, however, true that the change from Tyr 77 to Asp is larger than from Gln 80 to Glu, which is isosteric. The effects of the Y77D and Q80E substitutions on the intratrimeric interface may result because the charged side chains introduced in these substitutions are not favored in a relatively hydrophobic environment (Fig. 1C). Cavity formation and loss of the Tyr 77 -Asn 24Ј hydrogen bond in the minor intertrimeric interface are two other important factors that may contribute to the effects of Tyr 77 substitutions. With the Q80E substitution, electrostatic repulsion of charged Glu 80 side chains may play a role, as the distances between Gln 80 side chains within trimer are only 3.8 Å (22).
In addition, hexamers cannot be dissociated even by extensive substitutions at the intratrimeric interface unless the intertrimeric contact is broken. This can be achieved by incubating the enzyme at pH Ͻ 6 (12,18) or by substituting at least one His residue/monomer in the intertrimeric contact (10). Thus, the Y77D and Y77D/Q80E variants remained hexameric at pH 7.2 but could be dissociated into dimers by adjusting the pH to Յ 5.5 or by making additional mutations of His 136 and/or His 140 (Table  I). Consequently, the monomer-monomer interactions building the dimer are not from the intertrimeric interface, which is completely destabilized under these conditions (10,12).
The intertrimeric interaction in Bacillus stearothermophilus PPase is weak even at neutral pH, which allowed Shinoda et al. (23) to dissociate this enzyme into monomers by making a single substitution (V75D or V75K) at the intratrimeric interface. However, the variant enzymes contained some of the trimeric form, which may explain their rather high activity (9 -28% of wild-type PPase) compared with ours (Table II) charge, and activate the nucleophile and the leaving group (4,8,24). Because each PPase monomer carries a full active site capable of binding all these reactants, the changes in catalysis caused by disrupting PPase quaternary structure with point mutations presumably result from distortions of the active site cavity. The major effects of changes to the quaternary structure of E-PPase are on substrate and metal ion binding.
Substrate binding, as characterized by k cat /K m , is progressively weakened from hexamer to monomer (100, 0.30, 0.018, and Ͻ0.0001 M Ϫ1 s Ϫ1 for hexamer, trimer, dimer, and monomer, respectively). At least for the trimer, this results from decreased substrate binding affinity, the consequence of a decreased forward rate constant for substrate binding with little or no change in the backward rate constant (12). The similarity of the k cat /K m values for wild-type trimer (Fig. 6B) and H136Q/ H140Q trimer (10) indicates no direct effect of the H136Q and H140Q substitutions on active site. In contrast, the difference in k cat /K m between trimer and dimer may be a direct effect of the Y77D substitution not its effect on oligomeric structure. This is because K m value increases in the hexameric Y77D and Y77D/ Q80E variants. If so, the trimer to dimer conversion by itself has a smaller effect on catalytic properties than the hexamer to trimer and dimer to monomer conversions. This may result because one of the two identical monomer-monomer contacts existing in the trimer still remains in the dimer. Metal ion binding to the active site appears to be unchanged between hexamer and trimer (12) but is markedly weakened in the dimer (Fig. 8).
In terms of three-dimensional structure, differences in substrate binding between trimer and hexamer are explained by destabilization of ␣-helix A (residues 128 -140), which, via His 136 and His 140 , makes very important contributions to the intertrimeric contact and also forms an essential part of the active site cavity (5). Dissociation into dimers and monomers probably mispositions a long excursion (residues 22-53), which contains residues from the intratrimeric contact (Fig. 1C), as well as two basic residues, Lys 29 and Arg 43 , which make important contacts with the substrate (4,24). In addition, Leu 79 , Val 84 , and Ile 85 , all close to Tyr 77 , contribute both to the hydrophobic core and to the intratrimeric hydrophobic interface in E-PPase.
The behavior of the subunit interface variants thus supports the idea that hexamerization of prokaryotic PPases is important for forming catalytically competent active sites. Alternatively, nearly optimal active site conformation(s) can be generated in lower oligomeric forms of E-PPases upon substrate binding, as indicated by only small variations in k cat between hexamer, trimer, and dimer. This, however, takes up most of the substrate binding energy, as suggested by the dramatic decrease in k cat /K m in the trimer and dimer.
Interdependence of the Subunit Contact Zones-Surprisingly, the disruptive Q80E and Y77D substitutions in the intratrimeric interface stabilize the intertrimeric interaction both at acidic and neutral pH values (Table I and Figs. [3][4][5]. This result is especially surprising with the Y77D substitution, which eliminates the intertrimeric Tyr 77 -Asn 24Ј hydrogen bond. Similarly, the disruptive H136Q substitution in the intertrimeric contact zone strengthens the interaction within trimers, although to a smaller degree. A likely explanation is that the subunit contact zones in E-PPase are "strained," so that the optimal intratrimeric interaction is not compatible with the optimal intertrimeric interaction. The surface area buried upon oligomerization cannot be rigorously ascribed to a single interface (6), suggesting that the very closeness of the interfaces to each other (Fig. 1B) may be why optimal intratrimer and intertrimer interactions cannot co-exist. The strain also may result from the conformational changes of ␣-helix A and excursion 22-53 associated with oligomerization leading to changes in the hydrophobic core of the molecule (see above). Partial relaxation accompanied by some distortion to active site may occur in the subunit interface variants even when hexameric. The K m value is increased significantly in almost all such variants, which can form hexamers (see also Ref. 9). The largest effect is observed with the hexameric Y77D/Q80E/ H136Q and Y77A/Q80E/H136Q variants, for which K m values are greater than for other hexameric variant E-PPases and even greater than for the trimeric wild-type E-PPase.
If, as proposed, hexameric wild-type E-PPase is preformed into the substrate/product binding conformation, all six active sites should function independently if the conformational changes during catalysis are small. The same model explains why hexameric Y77D(A)/Q80E/H136Q variants with partly impaired subunit interactions and, consequently, distorted active site cavity exhibit strong positive cooperativity in substrate binding (Fig. 10). In these variants, substrate binding at one monomer causes a conformational change that brings a contacting monomer (presumably in the other trimer) into substrate-binding conformation. An association between cooperativity and an increase in the Michaelis constant has also been reported for variant tyrosyl-tRNA synthetases (25) and glutathione transferases (26).