Identification of Domain-Domain Docking Sites within Clostridium symbiosum Pyruvate Phosphate Dikinase by Amino Acid Replacement* 210

Potential domain-domain docking residues, identified from the x-ray structure of the Clostridium symbiosum apoPPDK, were replaced by site-directed mutagenesis. The steady-state and transient kinetic properties of the mutant enzymes were determined as a way of evaluating docking efficiency. PPDK mutants, in which one of two stringently conserved docking residues located on the N-terminal domain (Arg219 and Glu271) was substituted, displayed largely unimpeded catalysis of the phosphoenolpyruvate partial reaction at the C-terminal domain, but significantly impaired catalysis (>104) of the ATP pyrophosphorylation of His455 at the N-terminal domain. In contrast, alanine mutants of two potential docking residues located on the N-terminal domain (Ser262 and Lys149), which are not conserved among the PPDKs, exhibited essentially normal catalytic turnover. Arg219 and Glu271 were thus proposed to play an important role in guiding the central domain and, hence, the catalytic His455 into position for catalysis. Substitution of central domain residues Glu434/Glu437 and Thr453, the respective docking partners of Arg219 and Glu271, resulted in mutants impaired in catalysis at the ATP active site. The x-ray crystal structure of the apo-T453A PPDK mutant was determined to test for possible misalignment of residues at the N-terminal domain-central domain interface that might result from loss of the Thr453-Glu271 binding interaction. With the exception of the mutation site, the structure of T453A PPDK was found to be identical to that of the wild-type enzyme. It is hypothesized that the two Glu271 interfacial binding sites that remain in the T453A PPDK mutant, Thr453 backbone NH and Met452 backbone NH, are sufficient to stabilize the native conformation as observed in the crystalline state but may be less effective in populating the reactive conformation in solution.

Pyruvate phosphate dikinase (PPDK) 1 catalyzes the interconversion of ATP, P i , and pyruvate with AMP, PP i , and PEP (1) using two separate active sites linked by a mobile domain containing the phosphoryl group carrier His 455 (2). At the first active site, to which ATP, P i and Mg(II) are bound, His 455 attacks the ␤-P of ATP, forming a pyrophosphorylhistidine enzyme intermediate (E-PP) and AMP (Scheme 1). The P i ligand then reacts with the terminal phosphoryl group of E-PP to form a phosphorylhistidine intermediate (E-P) and PP i . The phosphorylhistidine of E-P then moves to the second active site where pyruvate, Mg(II), and a monovalent cation (K ϩ or NH 4 ϩ ) are bound. Upon transfer of the phosphoryl group to pyruvate to form the final product PEP, the His 455 residue returns to the first active site to initiate a new catalytic cycle.
The x-ray crystal structure (2) of the three-domain, 96-kDa subunit of the apo-PPDK homodimer from Clostridium symbiosum is represented at the top of Fig. 1. A Mg(II) cofactor and the ATP and P i substrate ligands have been modeled into the structure at the N-terminal "ATP grasp" domain (residues 2-430). The locations of the ligand binding sites (comprising active site 1) within the N-terminal domain are supported by the results from earlier affinity labeling and mutagenesis studies (4,5). As His 455 is located in close proximity to active site 1, the PPDK crystal structure, with some local adjustments, corresponds to a conformer competent to catalyze the ATP/P i partial reaction (Scheme 1). Henceforth, this conformer will be referred to as "conformer 1." Affinity labeling and mutagenesis studies have shown that the pyruvate ligand binds at a second active site, located on the ␣/␤ barrel C-terminal domain (residues 534 -874) (6,7). A model of an enzyme conformer (illustrated at the bottom of Fig.  1; henceforth referred to as "conformer 2") judged to be effective in catalysis of the pyruvate partial reaction, was generated from the original structure by transferring the central domain (residues 390 -504) from its binding site at the concave surface of the N-terminal domain to the concave surface of the Cterminal domain (2). Mg(II) and PEP were modeled into active site 2 of conformer 2 (2) as pictured in Fig. 1 (bottom). In this model, the PEP phosphoryl group is aligned for nucleophilic attack by the catalytic His 455 residue.
We anticipate that the rotation about the two interdomain linkers (residues 341-389 and 505-533), as required by the conformer 1 7 conformer 2 model of Fig. 1, occurs freely in solution, thus allowing the central domain to "swivel" between the two active sites. Full catalytic turnover on the enzyme will thus require that ATP, P i , and Mg(II) first bind to active site 1 of conformer 2 (see Scheme 2). The central domain must then dissociate from the C-terminal domain and bind with the Nterminal domain to form conformer 1. Despite the fact that the rotation about the linkers is intrinsically fast (ns time scale), the central domain must remain docked long enough for catal-ysis at active site 1 to be completed (millisecond time scale (Ref. 3)), but not so long as to impede the ensuing reaction of His(P) 455 with pyruvate at active site 2. The transition from conformation 1 to conformation 2 allows dissociation of the AMP and PP i product ligands from active site 1 as well prepares the enzyme for catalysis at active site 2, provided that this active site was "pre-loaded" with pyruvate and the Mg(II)/ NH 4 ϩ cofactors while the enzyme was in conformation 1. Following catalytic turnover at active site 2 in conformer 2, the central domain must dissociate from the C-terminal domain to allow release of the PEP product from conformer 1. If active site 1 contains ATP and P i , the next catalytic cycle can directly follow from this conformer.
The separate site catalysis of PPDK thus requires the transient formation of precisely oriented domain-domain complexes. An alignment made of the 17 known PPDK sequences (see Fig. 1 in supplemental materials, available in the on-line version of the journal) shows that a majority of the residues defining the proposed binding surfaces on the N-and C-terminal domains are stringently conserved as are many of the surface residues surrounding the catalytic His 455 on the central domain (see Fig. 2). In contrast, the surface residues on the opposite, solvent-exposed faces of the three domains are variable. The conservation of residues at the respective domaindomain interfaces of the observed and modeled conformers has been necessary throughout evolution for productive domain docking and accurate targeting of the catalytic His 455 .
Through inspection of the domain-domain interface region of the x-ray structure of apo-PPDK and of the alignment made of the PPDK sequences, two stringently conserved residues of the N-terminal domain (namely Arg 219 and Glu 271 ), whose side chains are positioned for favorable polar interaction with side chain and/or backbone atoms of the central domain, were identified. As is shown in Fig. 3, the guanidinium group of Arg 219 is within interaction distance of the carboxylate groups of the central domain residues Glu 434 (Glu or Asp conserved) and Glu 437 (not conserved). The carboxylate group of Glu 271 is within interaction distance of the hydroxyl group of the central domain Thr 453 and the backbone amide NHs of Thr 453 and Met 452 . In addition, N-terminal domain residues Ser 262 and Lys 149 , which are not conserved among all of the PPDK sequences, are also positioned for favorable interaction with the central domain. The hydroxyl group of Ser 262 is within Hbonding distance of the backbone CϭO of central domain residue Ala 398 . The side-chain ammonium group of Lys 149 is located 4.4 Å from the central domain Glu 430 carboxylate, which is too far for significant interaction. However, if the side chain of the Lys 149 is free to rotate from the conformation observed in the crystal, it can move within 3.5 Å of the carboxylate, close enough for favorable interaction to occur.
Since productive domain-domain binding is an integral component of PPDK catalysis (Scheme 2), an examination of the roles of Arg 219 , Glu 271 , Ser 262 , and Lys 149 in domain docking can be made by replacing these residues with amino acids that can not interact favorably and evaluating the catalytic efficiencies of the corresponding mutant enzymes. In the text that follows, we report on the steady-state and transient kinetic properties of the C. symbiosum PPDK mutants R219A, R219E, E434A/E437A, E271A, T453A, S262A, S262W, and K149A. The rate constants obtained for the mutants in catalysis of the full reaction and the E ϩ ATP 3 E-PP⅐AMP, E ϩ ATP ϩ P i 3 E-P⅐AMP⅐PP i , and E ϩ PEP 3 E-P⅐pyruvate partial reactions are compared with those determined in an earlier kinetic analysis of the wild-type PPDK (8). The results suggest that the productive docking of the central domain is controlled, in part, through the interaction of N-terminal domain residues Arg 219 ATP/P i Partial Reaction at Active Site 1 on N-terminal Domain:

Pyruvate Partial Reaction at Active Site 2 on C-terminal Domain:
(3) E-P ؉ pyruvate 3 E-P ؒ pyruvate 3 E-P ؒ PEP 3 E ؉ PEP SCHEME 1. The ATP/P i and pyruvate partial reactions of pyruvate phosphate catalysis (1, 3). Catalysis of the E ϩ ATP ϩ P i 3 E-P ϩ AMP ϩ PP i takes place at the N-terminal domain (green) active site in conformer 1 (top structure; ATP and P i ligands are blue, and Mg(II) is magenta), while catalysis of the E-P ϩ pyruvate partial reaction takes place at the C-terminal domain (blue) active site in conformer 2 (bottom structure; PEP ligand is red, and Mg(II) is magenta). Movement the catalytic His 455 residue (black) between the two active sites is perceived to occur through the interconversion of conformers 1 and 2 via the swivel of the central domain (yellow) about its linkers (pink). and Glu 271 with central domain residues Glu 434 /Glu 437 and Thr 453 /Met 452 , respectively. The N-terminal domain Ser 262 and Lys 149 residues, on the other hand, do not appear to play a vital role in orienting or stabilizing the domain-domain complex.

EXPERIMENTAL PROCEDURES
Preparation of C. symbiosum PPDK Mutants-Mutant genes were prepared from the plasmid pACYC184-D12 (9) using polymerase chain reaction techniques analogous to those described in (4). Mutagenic primers, 18 -22 base pairs in length, were synthesized by Life Technologies, Inc. BglII and BstXI restriction sites were employed in the construction of the Lys 149 , Arg 219 , Ser 262 , and Glu 271 mutants, whereas BstXI and KpnI sites were used for the Glu 434 /Glu 437 and Thr 453 mutants. The sequences of the isolated mutant genes were determined at the Center for Agricultural Biotechnology at the University of Maryland. Wild-type and mutant PPDK genes were expressed in Escherichia coli JM101 cells, and the protein products were purified to homogeneity (judged by SDS-polyacrylamide gel electrophoresis analysis) in yields of 20 -25 mg/g cell as described in Ref. 4. The chromatographic properties of the mutant proteins as well as their stability to storage in buffered solutions at 4 C were observed to be very similar to the chromatographic properties and stability of the wild-type enzyme.
Steady-state Kinetic Analysis-The spectrophotometric assay described in Ref. 10 was used to monitor the initial velocity of the PPDK-catalyzed reaction of AMP, PP i , and PEP to ATP, P i , and pyruvate. Initial velocities were measured as a function of the concentration of the varied substrate (in a range of 0.5-10-fold K m ) at fixed, saturating concentrations of cosubstrates (0.5 mM AMP, 0.5 mM PEP, 1 mM PP i ) and metal ion cofactors (5 mM MgCl 2 and 40 mM NH 4 Cl) in 20 mM imidazole (pH 6.8, 25°C). The initial velocity data were analyzed using Equation 1 and the computer programs of Cleland (11 [B] t is the product concentration at time t, [B] f is the product concentration at equilibrium, and k obs is the observed rate constant for the reaction. Single-turnover reactions of 40 M wild-type or mutant PPDK, 10 M [ 14 C]ATP, 2.5 mM CoCl 2 , and 10 mM NH 4 Cl in 50 mM K ϩ Hepes (pH 7.0, 25°C) were carried out in the presence or absence of 11 mM P i using a rapid quench device as described previously. The protein was removed from the acid quenched samples by centrifugation through a 500-l filter (cut-off molecular mass of 30 kDa) purchased from Pall Gelman Inc. The unconsumed [ 14 C]ATP and [ 14 C]AMP product were separated by HPLC using a Beckman ultrasphere C18 reversed-phase analytical column and 25 mM KH 2 PO 4 , 2.5% triethylamine, and 5% methanol as the mobile phase. The fractions containing ATP and AMP were analyzed for 14 C content by liquid scintillation counting. The time course for a single-turnover reaction was constructed, and the k obs was calculated, in the manner described in the previous section.

Rapid Quench Analysis of PPDK-catalyzed Single-turnover Reactions of [ 14 C]ATP in the Absence or Presence of P i -[
X-ray Crystallographic Analysis of T453A PPDK-The crystallization conditions used for the C. symbiosum T453A PPDK mutant were the same as those used for the wild-type enzyme (2). Single crystals An illustration of the free and phosphorylated forms of PPDK conformer 1 (E 1 , E 1 -P) and conformer 2 (E 2 , E 2 -P) (see Fig. 1) with the ligands that these enzyme forms bind and convert to products.  were obtained at 30°C by vapor diffusion in hanging drops. Protein drops were equilibrated against reservoir solutions containing 50 -55% saturated ammonium sulfate and 100 mM Hepes buffer (pH 7.0). The drops consisted of protein solution (10 mg/ml, 20 mM imidazole buffer (pH 6.5), 100 mM KCl, 0.1 mM EDTA, and 1 mM mercaptoethanol), diluted by an equal volume of reservoir solution. The crystals belong to space group P 2 , and within the accuracy of the data the unit cell dimensions are the same as those of the wild-type protein crystals (Table II). The crystal was transferred to a mother liquor solution containing 20% glycerol, and then flash-frozen using an Oxford cryosystem. X-ray diffraction data were collected on a Siemans area detector mounted on a three-circle goniostat, with a monochromated CuK ␣ x-ray supplied by a Rigaku Rotaflex RU200BH rotating anode generator. Data were processed with the XENGEN package (12). Data collection statistics are provided in Table VI. Structure refinement was performed with the program CNS (13). The simulated annealing slow-cooling protocol at 3000 K was followed by positional and temperature factor refinement cycles. Data up to 2.65 Å for which F Ն 2(F) were included. Adjustments to the model were made on a Silicon Graphics INDIGO II computer graphics workstation using the program TURBO-FRODO (14).

RESULTS AND DISCUSSION
Steady-state Kinetic Properties of PPDK Mutants-The PPDK mutants were first screened by measuring k cat and K m values for substrates and K i values for competitive inhibitors. The steady-state kinetic constants were determined for the AMP ϩ PP i ϩ PEP 3 ATP ϩ P i ϩ pyruvate reaction (with Mg 2ϩ and NH 4 ϩ serving as cofactors) using a continuous spectrophotometric assay to monitor pyruvate formation. PPDK adheres to an ordered, nonclassical bi (ATP, P i ) bi (PP i , AMP) uni (pyruvate) uni (PEP) kinetic mechanism (10). For conven-ience, the K m (and V m ) for each substrate was evaluated from initial velocity data measured with the concentrations of the two co-substrates held constant at saturating levels, while the concentration of the third substrate was varied at ϳ0.5-10-fold the K m value. The K i values of the ATP analog AMPPNP (versus AMP) and PP i analog PNP (versus PP i ) were determined from the initial velocity data measured in the absence and presence of the inhibitor.
The K i , K m , and k cat values obtained for wild-type PPDK and the PPDK mutants at pH 6.8 and 25°C are listed in Table I. Except for S262A and K149A, each of the mutants display significantly reduced catalytic efficiency compared with that of the wild-type PPDK. The k cat is deceased ϳ10-fold for the R219A, R219E, and E434A/E437A mutants, ϳ20-fold for T453A, ϳ100-fold for S262W, and Ͼ5000-fold for E271A. As will be described in the section that follows, the reduction in k cat derives from a reduction in the rate of the nucleotide partial reaction.
The abilities of the mutant enzymes to bind ATP and PP i are reflected by the K i values measured for the inert analogs AMP-PNP and PNP. AMPPNP competes with AMP for the same binding site on conformer 1 of the free and phosphorylated forms of the enzyme (see Scheme 2). Likewise, PNP competes with PP i for the same site on the AMP complex of the free or phosphorylated enzyme in conformation 1. An increase in the K i value may signify that the structure or electronic environment of the binding site has been altered by the mutation. Alternatively, if the ligand binding site is changed by domaindomain association, a mutation that affects domain binding can affect the K i of the inhibitor. Thus, an increase in the K i values may also indicate impaired domain-domain docking.
Based on the AMPPNP K i values listed Table I, it is evident that the ATP binding affinity is not altered in the R219E, R219A, E343A/E347A, S262A, and T453A mutants, despite the fact that all but the S262A mutant has a significantly lower turnover rate. On the other hand, the AMPPNP K i values measured for the S262W and K149A mutants are ϳ4-fold larger than that measured for wild-type PPDK. The exact cause of reduced ATP binding in these mutants is presently unknown; however, since the Ser 262 and Lys 149 residues do not appear to function as docking residues (i.e., the k cat values of the Ala mutants are normal), we have not pursued the issue further.  Table II. The PP i binding affinity was found to be severely weakened in the R219E, R219A, E434A/E437A, S262W, and T453A PPDK mutants, as indicated by the comparatively large PNP K i values observed for the mutant enzymes (ϳ20 -100-fold larger than that of the wild-type PPDK). Only the K149A and S262A mutants, both of which display normal k cat values (and thus, unimpaired domain-domain docking), were unaffected. The PP i binding site lies outside of the regions mutated and, therefore, may not be directly effected by the amino acid replacements made. On the other hand, PP i binds at the entrance of active site 1, which becomes desolvated by the region of the central domain surrounding (and including) the His(P) 455 . A likely scenario is that the reduced PNP binding affinity observed with the mutant enzymes is a reflection of altered domain-domain binding. 2 Catalysis at the PEP/Pyruvate Active Site-The E⅐PEP Ј E-P⅐pyruvate partial reaction is catalyzed by PPDK in conformation 2 ( Fig. 1; Scheme 2). Thus, mutations made at the N-terminal domain are not expected to inhibit catalysis at the C-terminal domain. This was tested by measuring the k obs for single-turnover reactions of R219A, R219E, E271A, and S262W mutant PPDK ϩ [ 32 P]PEP 3 [ 32 P]E-P⅐pyruvate using rapid quench techniques. The time courses for these reactions are shown in Fig. 4, whereas the k obs values calculated from the rate data are reported in Table II. All four mutants displayed normal catalysis of the E⅐PEP 7 E-P⅐pyruvate partial reaction. The fact that the k obs values for the four mutants are not noticeably higher than that measured for the wild-type enzyme suggests that the population of conformer 2 has not been significantly increased and/or that the rate of conformer equilibration is not limiting. not direct binding interaction occurs between the bound P i ligand and the central domain, cannot be discerned on the basis of the data on hand. Nevertheless, it is reasonable to expect that the domain docking produces significant alterations in the P i environment. The PP i ligand displays a higher binding affinity for the E-P⅐AMP complex (K d ϭ 50 M) than does P i for the enzyme-ATP complex (8,10). PP i binding (k on ϭ 2 M Ϫ1 s Ϫ1 ; k off ϭ 100 s Ϫ1 (Ref. 8)) is followed by phosphohistidine 455 binding. It is likely that the environment of the PP i ligand is altered by domain-domain association and that the PP i binding constant is effected by mutations that impair the domain-domain docking reaction.   (8). Substitution of Co(II) for Mg(II) as cofactor in the reaction increases both ATP binding to the enzyme and the internal equilibrium, so that the amount of E-PP⅐AMP produced from reaction of 40 M PPDK with 10 M ATP increases from ϳ5% for the Mg(II)-activated enzyme to ϳ25% for the Co(II)-activated enzyme. For this reason, the abilities of the PPDK mutants to catalyze the E ϩ [ 14 C]ATP 3 E-PP⅐[ 14 C]AMP partial reaction were evaluated using Co(II) rather than Mg(II) as cofactor.
On the other hand, Mg(II) is the superior choice for kinetic analysis of the single-turnover reaction of E (20 M) ϩ ATP (5 M) ϩ P i (11 mM) 3 E-P⅐AMP⅐PP i . For this reaction, it is advantageous to minimize the level of AMP formed in the initial E⅐ATP P i 3 E-PP⅐AMP⅐P i step so that the total amount of AMP formed will coincide with the amount of AMP produced as the result of the second partial reaction: E-PP⅐AMP⅐P i 3 E-P⅐AMP⅐PP i , which, by mass action effect, drives the first reaction forward. For wild-type PPDK, the k obs for the first step is ϳ 300 s Ϫ1 while the k obs for the second step is ϳ6 s Ϫ1 (4).
If catalysis of the first step is impaired in the PPDK mutant, this will be reflected by the time course for [ 14 (Table IV). Thus, the first (E-PP⅐AMP forming) step is ratelimiting. The k obs values, therefore, are less than or equal to the rate constants for the second step catalyzed in the mutants.
Altered Active Site Versus Altered Domain-Domain Docking in PPDK Mutants-The kinetic properties measured for the R219E, R219A, E434A/E437A, E271A, T453A, and S262W PPDK mutants show that the amino replacements made inhibit catalysis of the E ϩ ATP 3 E-PP⅐AMP partial reaction. Because catalysis of this partial reaction depends both on the integrity of the N-terminal domain active site and on the accurate placement of the His 455 residue into the active site, impaired catalysis in these mutants can derive from an alteration in the active site environment and/or an alteration in the domain-domain docking process. Here, we argue that the Nterminal domain active sites in the PPDK mutants 3 are intact and that impaired catalysis derives from altered domain-domain docking.
First, we note that the mutation sites are located at the surface of the N-terminal or central domains. We do not anticipate that the amino acid substitutions will alter the domain fold (as is demonstrated for the T453A mutant by x-ray crystallographic analysis; see below), and in this way inhibit catalysis of the E ϩ ATP 3 E-PP⅐AMP partial reaction. Substitutions of Arg 219 , Glu 271 , Lys 149 , and Glu 434 /Glu 437 involve changes in the electrostatic properties of the domain surface, not in but near the active site. In previous (unrelated) studies, analogous electrostatic perturbations have been generated at the N-terminal domain surface in close vicinity of the active site, without alteration of catalytic activity (in the Lys 163 , Յ0.005 c a Average results of two reaction trials carried out using separately prepared enzymes. The marked error was the difference between the k obs values obtained from the two trials.
b The wild-type PPDK catalyzed ATP reaction reached more than 60% conversion within the dead time of the rapid quench instrument; therefore, the wild-type k obs was only an estimated value.
c Mutant E271A-and S262W-catalyzed ATP turnover reactions were too slow to be accurately quantitated due to considerable decomposition of nucleotides at the active sites of these mutants when incubation time exceeded 300 s. The lower limit of measurable rate constant is 0.005 s Ϫ1 . a Average results of two reaction trials carried out using separately prepared enzymes. The marked error was the difference between the k obs values obtained from the two trials. Lys 177 , Glu 168 , Glu 170 , and Glu 175 PPDK mutants examined) (5). Although it can be argued that these sites are not located at the domain-domain interface, we point out that two of the domain-domain interface residues examined in the present study, Ser 262 and Lys 149 have been replaced by an Ala residue without loss of catalytic activity.
Second, we note that mutation of Arg 219 to Glu has the same impact on catalysis at active site 1 that mutation of Arg 219 to Ala has. Thus, the loss of binding interaction between Arg 219 and the Glu 434 /Glu 437 residues is what is critical to catalysis and not the type of charge alteration made at position 219. Moreover, which partner of the paired docking residues, Arg 219 or Glu 434 /Glu 437 , is replaced in the mutagenesis experiment does not effect the outcome; the catalytic properties of the Glu 434 /Glu 437 mutant and the Arg 219 mutant are similar.
Finally, we note that disruption of the Glu 271 -Thr 453 /Met 452 docking site has more impact on catalysis than does disruption of the Arg 219 -Glu 434 /Glu 437 docking site. Unlike the Arg 219 -Glu 434 /Glu 437 docking site, the Glu 271 -Thr 453 /Met 452 docking site is close to the His 455 residue (see Fig. 3). The His 455 side chain is positioned on the N terminus of a short helix, just down stream from the Glu 271 docking residues Thr 453 and Met 452 . The backbone amide NHs of both of these residues as well as the side chain of the Thr 453 residue bind with the Glu 271 carboxylate group, thereby drawing the His 455 into the active site region. Consequently, the Glu 271 -Thr 453 /Met 452 docking site might play a more dominant role in the productive domaindomain binding than the Arg 219 -Glu 434 /Glu 437 site, which is further removed from the His 455 . Alternatively, the differential effects of mutating Arg 219 and Glu 271 may be attributed to the difference in the number of interdomain electrostatic interactions lost. Specifically, the present data indicate that the removal of a single interaction, regardless of the docking site involved, results in a ϳ10 4 -fold reduction in k obs . This may be a simple coincidence, or it may suggest that the contributions of the electrostatic interactions are additive, in which case one can rationalize the greater loss in activity for the E271A PPDK mutant in terms of the loss of three rather than one interaction.
Misalignment versus Decreased Residence Time by Impaired Domain-Domain Docking-Domain movement in PPDK occurs via bond rotation within the solvated linkers (Fig. 1). Domain-domain association occurs through the diffusion of the two tethered protein bodies through restricted solvent space. The binding equilibrium is controlled by the attraction between the two interfaced surfaces. Protein interfaces observed in protein dimers, protease-protein inhibitor complexes, and antibodyprotein antigen complexes (where in permanent association is desired) range from 700 to 4900 Å 2 (16). In PPDK, for which the binding of between the two domain surfaces is intended to be transient, the accessible area buried in conformation 1 (Fig. 1) is ϳ700 Å 2 (2).
Hydrophobic interactions deriving from the desolvation of nonpolar residues are often present at protein-protein interfaces (17,18). Inspection of the domain-domain interface within the PPDK structure reveals two matched sets of hydrophobic regions which may contribute to the intrinsic binding energy. The electrostatic interaction between Arg 219 and Glu 434 /Glu 437 and between Glu 271 and Thr 453 /Met 452 (NH) may also contribute to the domain-domain binding energy as well as provide specificity to the domain-domain interactions. A steering mechanism, based on the pairing of specific residues at the domain-domain interface, could reduce the number of collisions required to achieve productive binding of the catalytic His 455 in the N-terminal domain active site.
Proof that the electrostatic interaction between Arg 219 and Glu 434 /Glu 437 and between Glu 271 and Thr 453 /Met 452 (NH) is used for domain-domain steering is not easily obtained. Crys- where F o and F c are the observed and calculated structure factors). The stereochemical parameters are well within the range known from crystal structures of small peptides. The root-mean-square deviations from ideal values for bond lengths and bond angles of the standard geometry as compiled by Engh and Huber (19) are 0.012 Å and 1.7°, respectively. The model includes 869 of the 874 amino acid residues (the N-terminal methionine and residues 506 -509 are associated with regions of low electron density), 227 water molecules, and 2 sulfate ions. Within the accuracy of the structure determination, the overall structure is the same as that of the wild-type protein (2). tallization experiments of these mutant enzymes, which following structure determination would reveal if the relative orientation of the two domains has been altered, have been unsuccessful. Thus far, only the x-ray structure of the T453A mutant has been determined (see Fig. 7, Table VI, and "Experimental Procedures" for details). Within the accuracy of the structure determination, the overall structure is the same as that of the wild-type protein (2). The C ␣ atomic coordinates of the mutant and wild-type molecules superimpose with a rootmean-square deviation of 0.5 Å. The electron density map at the mutation site is consistent with the replacement of a threonine by an alanine residue at position 453. The interaction between the side chain of Glu 271 and the hydroxyl group of Thr 453 is therefore eliminated, but the remaining two interactions with the two NH groups of residues 452 and 453 are maintained (Fig. 7). In addition, the side chains of Met 452 and Leu 369 shift concertedly so that the Met 452 side chain occupies space that was partially occupied in the wild-type protein by the methyl group of Thr 453 , and the Leu 396 side chain shifts toward the position occupied by Met 452 in the wild-type protein.
These changes are compensatory and therefore are not expected to impact significantly on the catalytic efficiency of the enzyme. Thus, although the structure of T453A PPDK provides evidence that the reduction in catalytic efficiency is not caused by alteration in active site conformation, it does not provide any indication of misalignment of the domain-domain surfaces. This static picture may or may not provide an accurate view of the dynamics between protein species in solution. The absence of a particular conformer in the crystal does not mean that it is also absent in solution.

CONCLUSIONS
The current model of PPDK catalysis involves the swiveltype movement of a central, phosphoryl carrier domain between the ATP/P i active site of the N-terminal domain and the pyruvate active site of the C-terminal domain. We hypothesize that the stringently conserved residues at the respective domain-domain interfaces play an active role in guiding the catalytic His 455 of the central domain into the active sites and holding it there long enough for catalysis to be completed. The loss of catalytic activity observed for the mutants examined in the present study supports the proposed roles of the N-terminal residues Arg 219 and Glu 271 in the productive binding of the central domain for catalysis of the ATP/P i partial reaction.