Ambiguities in Mapping the Active Site of a Conformationally Dynamic Enzyme by Directed Mutation

On the basis of ligated crystal structures, Asn21, Asn38, Thr42, and Arg419 are not involved in the chemical mechanism of adenylosuccinate synthetase from Escherichia coli, yet these residues are well conserved across species. Purified mutants (Asp21 → Ala, Asn38 → Ala, Asn38 → Asp, Asn38 → Glu, Thr42→ Ala, and Arg419 → Leu) were studied by kinetics, circular dichroism spectroscopy, and equilibrium ultracentrifugation. Asp21 and Arg419 are not part of the active site, yet mutations at positions 21 and 419 lowerk cat 20- and 10-fold, respectively. Thr42 interacts only through its backbone amide with the guanine nucleotide, yet its mutation to alanine significantly increasesK m for all substrates. Asn38hydrogen-bonds directly to the 5′-phosphoryl group of IMP, yet its mutation to alanine and glutamate has no effect onK m values, but reduces k catby 100-fold. The mutation Asn38 → Asp causes 10–57-fold increases in K m for all substrates along with a 30-fold decrease in k cat. At pH 5.6, however, the Asn38 → Asp mutant is more active, yet binds IMP 100-fold more weakly, than the wild-type enzyme. Proposed mechanisms of ligand-induced conformational change and subunit aggregation can account for the properties of mutant enzymes reported here. The results underscore the difficulty of using directed mutations alone as a means of mapping the active site of an enzyme.

IMP ϩ L-aspartate ϩ GTP 7 adenylosuccinate ϩ GDP ϩ phosphate REACTION 1 Primary sequences of AMPSase (2-10) are 40% identical for any pairwise comparison, indicating a strong tendency to preserve a primordial gene throughout evolution (10). The enzyme putatively facilitates the formation of 6-phosphoryl-IMP by the nucleophilic attack of the 6-oxo group of IMP on ␥-phosphate of GTP, and then the formation of adenylosuccinate by displacement of the 6-phosphoryl group by L-aspartate (11,12). AMP-Sase from Escherichia coli is a monomer at physiological concentrations (1 M), but dimerizes when nucleotide ligands are present (13,14). The dimer is probably the physiologically active form of the enzyme, having a K m corresponding to intracellular concentrations of IMP.
The disordered active site of unligated AMPSase becomes ordered in the presence of substrates and substrate analogs (15)(16)(17)(18)(19)(20). The largest conformational change is a 9-Å movement of the loop 42-53 (40s loop), which folds against the guanine nucleotide. Loop 120 -131 (120s loop), which interacts with IMP, and loop 299 -303 (300s loop), which interacts primarily with analogs of L-aspartate in crystal structures, become ordered in the presence of ligands. The conformational changes above putatively exemplify induced fit, a concept introduced by Koshland some three decades ago (21). The ligand-enzyme interactions, however, which contribute most (in terms of a thermodynamic driving force) to the observed conformational changes have yet to be identified, nor can we exclude the possibility of energy contributions from interactions between protein residues well removed from the active site.
On the basis of ligated crystal structures of adenylosuccinate synthetase, Asp 21 , Asn 38 , Thr 42 , and Arg 419 do not interact with atoms of substrates involved with the chemistry of phosphotransfer or nucleophilic attack by L-aspartate (17,19). In fact, Asp 21 and Arg 419 do not interact with ligands (16,17,19,20). The backbone amide of Thr 42 forms a hydrogen bond with the ␣-phosphoryl group of the guanine nucleotide, but its side chain forms only a weak hydrogen bond to that same ␣-phosphoryl group (20). Asn 38 provides one of five hydrogen bonds to the 5Ј-phosphoryl group of IMP; implicating Asn 38 in the ground-state stabilization of the IMP-enzyme complex (17,19). Nevertheless, mutations of Asn 38 have little effect on the K m of IMP, but a major impact on k cat , mutations of Asp 21 and Arg 419 destabilize the transition state, and the mutation of Thr 42 to alanine reduces affinities for all substrates. The effects of each of the mutations above can be understood in terms of their influence upon two interdependent, dynamic mechanisms in AMPsase: (i) ligand-induced reorganization of the active site and (ii) ligand-induced dimerization of the enzyme. The above mutations also illustrate the importance of a sound understanding of structure and dynamics of an enzyme before assigning functional roles to side chains.

EXPERIMENTAL PROCEDURES
Materials-GTP, IMP, L-aspartate, phenylmethylsulfonyl fluoride, and bovine serum albumin were from Sigma. Restriction enzymes were from Promega. Pfu DNA polymerase and E. coli strain XL-1 blue were obtained from Stratagene. E. coli strain H1238 (purA Ϫ ) was a gift from Dr. D. Bachman (Genetic Center, Yale University, New Haven, CT). Phenyl-Sepharose CL-4B came from Amersham Pharmacia Biotech.
Other reagents and chemicals came from Sigma if not otherwise specified.
Overexpression of Wild-type and Mutant AMPSases-Owing to the undetectable levels of expression for some mutants using the PMS204 plasmid expression system, we used here a more efficient prokaryotic expression system, based on the pTrc99A vector (Amersham Pharmacia Biotech).
The purA gene was amplified by polymerase chain reaction using the following primers to incorporate NcoI and PstI restriction sites at either end of the purA gene.
TrcN: 5Ј-GATGCCATGGGTAACAACGTCGTCG-3Ј ͑NcoI site underlined) TrcC2: 5Ј-AACTGCAGTCTGCCAGGCGTACCACA-3Ј ͑PstI site underlined) The NcoI restriction site was incorporated into the N terminus of the purA gene, before the first Met codon (ATG), and the PstI restriction site was introduced after the stop codon. Pfu DNA polymerase was used in the polymerase chain reaction to ensure high fidelity amplification (Fig. 1).
The amplified and subcloned purA gene containing the 1.3-kilobase pair fragment was sequenced twice in both directions using the chain termination method (22) at the Nucleic Acid Facility, Iowa State University. Sequence analysis showed 100% identity with the E. coli purA sequence deposited in GenBank™, except at position 415, a GAT (Asp) is replaced by GGT (Gly), which agrees with published crystal structures (15). The purA gene was subcloned into the pTrc99A vector and the final construct, pTrpA, was transformed into E. coli strain H1238 (purA Ϫ ) for overexpression of AMPSase.
Preparation and Kinetics of Wild-type and Mutant AMPSases-The wild-type and mutant enzymes were purified as described elsewhere (23)(24)(25) with the following modifications. The enzyme was eluted from a phenyl-Sepharose CL-4B column by a series of buffered solutions, 0.6 M, 0.4 M, and 0.2 M in (NH 4 ) 2 SO 4 and 50 mM in potassium P i (pH 7.0). All the enzymes eluted at 0.2 M (NH 4 ) 2 SO 4 . The enzyme fractions were concentrated and dialyzed against 50 mM potassium P i (pH 7.0), then further purified using a DEAE-TSK high performance liquid chromatography column. Enzyme purity was monitored by SDS-polyacrylamide gel electrophoresis (26), and protein concentrations were determined using Bradford reagent (Bio-Rad) (27). All the mutant enzymes were subjected to circular dichroism analysis as described elsewhere (22)(23)(24). From 1 to 500 g/ml enzyme was used in kinetic assays, depending on the activity of each mutant. Absorbance changes at 290 nm and 25°C were monitored with a GBC model 918 UV-visible spectrophotometer equipped with a Peltier-Effert temperature controller.
pH Effects on AMPSase Activity-MES and HEPES were chosen as buffers for the pH range 5.5-8.5, and acetate buffer was used at pH 5.2. Buffer concentrations were held at 20 mM for all assay solutions. Assay solutions contained 300 M GTP, 5 mM MgCl 2 , 5 mM L-aspartate for the wild-type enzyme and 1.5 mM GTP, 7 mM MgCl 2 , 10 mM L-aspartate for the Asn 38 3 Asp mutant enzyme. IMP concentrations varied from 25 to 750 M for assays of the wild-type enzyme, and 25 M to 10 mM for assays of the Asn 38 3 Asp mutant. 500 g/ml of the Asn 38 3 Asp mutant and 1 g/ml of wild-type AMPSase were used in assays. 2-mm quartz cuvettes were used to compensate for the high absorbance due IMP. Longer assay time (4 min) was used to compensate for the low activity of the Asn 38 3 Asp mutant.
Analytic Equilibrium Sedimentation Ultracentrifugation-Wildtype, Asn 38 3 Asp, and Asn 38 3 Glu mutant AMPSases were analyzed by analytical equilibrium sedimentation ultracentrifugation, in the absence and in the presence of ligands (5 mM MgCl 2 , 30 M IMP, 30 M GTP, 1 M hadacidin) at pH 7.7 and 5.6. Equilibrium ultracentrifugation experiments were performed as described previously (13), using hadacidin in place of aspartate. Enzyme concentrations were such that they gave an A 280 reading of 0.4. Monomer-dimer dissociation constants were determined as described previously (13).

Sequence Comparison and Crystallographic
Analysis-Positions corresponding to Asp 21 , Asn 38 , His 41 , Thr 42 , and Arg 419 are almost 100% identical across species (Table I). In the absence of ligands, Asn 38 has no specific interactions, but in the ligated synthetase it hydrogen-bonds with the 5Ј-phosphate of IMP ( Fig. 2; Ref. 17). His 41 , a putative catalytic acid (17,28), has three mutually exclusive sets of interactions. (i) In the unligated synthetase, His 41 hydrogen-bonds with Asp 21 (16,17). (ii) In the complex of GDP, NO 3 Ϫ , IMP, Mg 2ϩ , and hadacidin, His 41 hydrogen-bonds with the ␤-phosphoryl group of GDP (17). (iii) In the complex of GDP, 6-thiophosphoryl-IMP, Mg 2ϩ ,   (19). Asp 21 in the unligated synthetase hydrogen-bonds with His 41 , as noted above, but also makes a salt link with Arg 419 in ligated complexes (17). Thr 42 has no specific interaction in the unligated synthetase (16,17), but in ligated complexes its backbone amide interacts with the ␣-phosphoryl group of GDP and its backbone carbonyl hydrogen-bonds with a water molecule, which in turn hydrogenbonds with the 2Ј-OH of GDP and backbone carbonyl 417. In one of three ligated complexes of the synthetase (17,19,20), the side chain of Thr 42 may interact weakly with the ␣-phosphate of GDP (oxygen to oxygen, donor-acceptor distance of approximately 3.1 Å; Ref. 20). The interactions above are summarized in Table II. Expression, Purification, and Characterization of Wild-type and Mutant AMPSases-All proteins have an apparent monomer molecular mass of 48 kDa and purity greater than 95% as shown by SDS-polyacrylamide gel electrophoresis. pTrpA plasmids containing the mutations of this study complement the purA Ϫ auxotroph on M9 minimal medium, supplemented with threonine and arginine, indicating sufficient enzyme activity to sustain the transformed cells.
Secondary Structure Analysis-The CD spectra of Asp 21 3 Ala, Thr 42 3 Ala, Arg 419 3 Leu, and wild-type enzymes are identical from 200 to 260 nm, indicating the absence of conformational change because of mutation. The position 38 mutants showed small differences in their CD spectra relative to the wild-type spectrum (data not shown). The differences are probably a consequence of a perturbation in the conformation of the 40s loop and/or the monomer-dimer equilibrium of AMPsase (see below).
Kinetic Analysis of Wild-type and Mutant Enzymes-The K m values of the Asp 21 3 Ala, Asn 38 3 Ala, and Asn 38 3 Glu mutants are comparable to those of wild-type AMPSase (Table  III). Although k cat for the Thr 42 3 Ala mutant was the same as that of the wild-type enzyme, K m values showed 5-10-fold increases. The mutation of Asp 21 to alanine reduced k cat 20-fold and slightly increased K m GTP and K m IMP , whereas Arg 419 reduced k cat 10-fold, with a slight increase in K m aspartate , and 8-fold increases in K m GTP and K m IMP . Asn 38 3 Ala, Asn 38 3 Asp, and Asn 38 3 Glu mutants showed 30 -200-fold reductions in k cat relative to the wild-type enzyme. Of the three position Asn 38 mutants, only Asn 38 3 Asp exhibited increased K m values (K m IMP increased 80-fold).

pH-dependent Kinetic Studies of Wild-type and Asn 38 3
Asp-Of the three position 38 mutants, only K m values for the Asn 38 3 Asp mutant show large increases. Altered kinetic parameters for the Asn 38 3 Asp mutant could originate from electrostatic repulsion between the 5Ј-phosphoryl group of IMP and the side chain of Asp 38 . Protonation of the Asp 38 side chain, or the 5Ј-phosphoryl group of IMP, then, could restore k cat and K m to wild-type levels. k cat versus pH profiles for the wild-type and the Asn 38 3 Asp enzymes are bell-shaped (Fig. 3). The optimum pH values for wild-type and Asn 38 3 Asp enzymes are 7.8 and 5.6, respectively. In fact, the Asn 38 3 Asp mutant had higher catalytic activity than the wild-type enzyme at pH 5.6 (Table III) and 5.2 (data not shown). Although wild-type levels of activity were recovered in the Asn 38 3 Asp mutant by dropping the pH, K m IMP remains 100-fold higher than that of the wild-type enzyme.
Analytic Equilibrium Ultracentrifugation-IMP induces dimerization of AMPsase monomers (13,14). Furthermore, the dimer arguably has 100-fold higher affinity for IMP than the monomer. Hence, we examined the Asn 38 3 Asp mutant by  ultracentrifugation methods, in order to determine whether its elevated K m IMP originated from its properties of aggregation. In centrifugation studies reported here, the wild-type and Asn 38 3 Glu enzymes are controls, as each exhibits comparable K m values for IMP, and hence should be dimers in the presence of IMP. The monomer-dimer dissociation constants for the three enzymes in the absence and presence of ligands at pH 7.7 are in Table IV. At pH 5.6, the three enzymes precipitated during ultracentrifugation runs. The three AMPSases show similar K D values in the absence of ligands. In the presence of ligands, however, the K D of the Asn 38 3 Asp mutant (252 M) differs significantly from that of wild-type and Asn 38 3 Glu enzymes, both of which are vanishingly small (Table IV). Furthermore, the dissociation constant for Asn 38 3 Asp in the presence of ligands is 19 times higher than the dissociation constant for Asn 38 3 Asp in the absence of ligands. DISCUSSION For the rapid equilibrium random mechanism of AMPSase (1), k cat represents the breakdown of the quarternary complex of enzyme and substrates to enzyme and products. Thus, k cat for AMPsase is sensitive to the energy changes in the transition state. Kinetic data indicate destabilization of the transition state for the Asp 21 3 Ala and Arg 419 3 Leu mutants (k cat falls 20-and 10-fold, respectively, relative to the wild-type enzyme).
Thus, the Asp 21 -Arg 419 salt link must stabilize the transition state. As the hydrogen bond between Asp 21 and Arg 419 is about 13 Å from the bound Mg 2ϩ (approximate center of catalysis), the stabilization is necessarily a result of an indirect mechanism. The loss of the Asp 21 -Arg 419 salt link could destabilize interactions between the 40s and 400s loop. Indeed, in the unligated enzyme Arg 419 is disordered (15,16) and the 40s and 400s loop do not interact. In the ligated enzyme, backbone carbonyl 417 hydrogen-bonds with a water molecule, which in turn hydrogen-bonds to backbone carbonyl 42 and the 2Ј-OH of GDP (17).
An alternative mechanism by which the loss of the Asp 21 -Arg 419 salt link could influence the transition state is by way of a small perturbation of the P-loop (residues 8 -16). Asp 13 is a putative catalytic base (17,28). A displacement of its side chain by perhaps as little as 0.5 Å could lead to complete inactivation of the synthetase. The interaction of Asp 21 with Arg 419 could stabilize the position of helix H1, which lies on the C-terminal side of the P-loop. Mutations in the P-loop have only modest effects on K m GTP and no effect on the K m for other substrates (28), consistent with the lack of change in K m values for the Asp 21 3 Ala mutant.
The mutation of Arg 419 to leucine has a minor effect on the K m values for IMP and GTP. The increased K m GTP must be a result of a localized change in the 400s loop as the Asp 21 mutation has no effect on K m values. The mutation of Arg 419 could perturb the stacking of Pro 417 with the guanine base and hence lead to the increase in K m GTP . Given the observed binding synergism of IMP and GTP (13), a reduced GTP interaction should weaken IMP binding.
The side chain of Thr 42 binds weakly (at best) to the ␣-phosphoryl group of GTP, yet it is a conserved residue and its mutation to alanine increases K m values by approximately 10-fold. The above phenomenon may be a consequence of the 40s loop, which is in an equilibrium between two conformational states. The "open" conformer predominates in the absence of ligands, whereas the "closed" conformer appears when ligands are bound to the active site. The absence of an appreciable effect on k cat caused by the mutation of Thr 42 to alanine implies no significant perturbation of the 40s loop in the ligated  conformation of the synthetase. If however, the mutation to alanine stabilizes the open conformation of the loop, then ligands must spend a greater fraction of their binding energy to drive the 40s loop to its closed conformation. That the K m values increase for all substrates is a reflection of synergism in substrate binding (13). Asn 38 3 Ala and Asn 38 3 Glu have no influence on K m values but reduce k cat by 200-and 30-fold, respectively. Evidently, the interaction between Asn 38 and the 5Ј-phosphate of IMP does not enhance the affinity of IMP for the active site. Instead, the binding energy probably is diverted completely to the stabilization of the transition state. Asn 38 belongs to a short element (residues 38 -41) that immediately precedes the 40s loop. Within this structural element, the peptide link between residues 40 and 41 undergoes a significant change in conformation (17) and the ␣-carbon and side-chain of Asn 38 move toward the 5Ј-phosphate by 1 Å. These small conformational adjustments preceding the 40s loop are probably coupled to the 9-Å movement of the loop itself. Hence, the energy of interaction of Asn 38 with the 5Ј-phosphate of IMP may go to support the conformational change in the 40s loop.
The mutation of Asn 38 to aspartate presents a far more complex phenomenon. The introduction of a negative charge at position 38 probably elevates K m IMP by charge repulsion. All of the allowable conformations of Asp 38 bring its side chain close to the 5Ј-phosphate of IMP. (In contrast, allowable conformations of Glu 38 are possible, which bring about significant separation of the 5Ј-phosphate and the side chain.) By protonation of Asp 38 and/or the 5Ј-phosphate of IMP, a hydrogen bond is possible in the Asn 38 3 Asp mutant. Thus, k cat for the mutant at pH 5.6 is restored to wild-type levels, possibly because IMP can again drive the required conformational change in the 40s loop.
K m values for the Asn 38 3 Asp mutant are not restored to wild-type levels at low pH, however, suggesting that a pHindependent mechanism is responsible for the elevated K m values of the mutant. The wild-type enzyme is predominantly a monomer in the absence of ligands at concentrations used in assays (13). Furthermore, IMP induces dimerization of the wild-type enzyme at pH 7.7, its interaction with Arg 143 of a second monomer being essential in stabilizing the dimer. The mutation of Arg 143 to leucine or lysine does not influence k cat , but increases K m IMP 100-fold at pH 7.7, and concomitantly abolishes ligand-induced dimerization at low concentrations of IMP and enzyme (13). The high K m IMP values for the Asn 38 3 Asp mutant at pH 7.7 then may reflect the weak interaction of IMP with the mutant as a monomer, rather than the binding of IMP to a mutant dimer. This is supported by the monomer-dimer dissociation constants for Asn 38 3 Asp with and without ligands. The higher dissociation constant for the Asn 38 3 Asp in the presence relative to the absence of ligands shows that Asn 38 3 Asp favors a dimeric state in the absence of ligands. At pH 5.6, IMP may bind to the active site of the mutant as a monoanion, allowing formation of the hydrogen bond with the side chain of Asp 38 . However, the monoanionic state of IMP may not stabilize sufficiently its interaction with Arg 143 of a second subunit. Hence, the affinity for IMP remains low, because the mutant remains a monomer at low pH. The wild-type enzyme, on the other hand, selects the dianion state of IMP from solution at low pH and, as a consequence, dimerizes. Hence, the K m IMP for wild-type enzyme at pH 5.6 reflects the interaction of IMP with a synthetase dimer.
AMPsase is rife with ligand-induced changes, which in a broad perspective influence subunit dimerization and/or reorganization of its active site. As a consequence of the coupling of ligand-binding energy to dynamic processes that influence the conformation of loci remote from the active site, residues can bind directly to substrates and have no effect on binding affinity as measured by K m and residues remote from the active site can have substantial effects on the stability of the transition state or the binding affinity of substrates. Such phenomena underscore the pitfalls in using site directed mutations alone in assigning residues to the active site of an enzyme.