Site-directed Mutagenesis of ATP Binding Residues of Biotin Carboxylase

Acetyl-CoA carboxylase catalyzes the first committed step in fatty acid synthesis in all plants, animals, and bacteria. The Escherichia coli form is a multimeric protein complex consisting of three distinct and separate components: biotin carboxylase, carboxyltransferase, and the biotin carboxyl carrier protein. The biotin carboxylase component catalyzes the ATP-dependent carboxylation of biotin using bicarbonate as the carboxylate source and has a distinct architecture that is characteristic of the ATP-grasp superfamily of enzymes. Included in this superfamily are d-Ala d-Ala ligase, glutathione synthetase, carbamyl phosphate synthetase,N 5-carboxyaminoimidazole ribonucleotide synthetase, and glycinamide ribonucleotide transformylase, all of which have known three-dimensional structures and contain a number of highly conserved residues between them. Four of these residues of biotin carboxylase, Lys-116, Lys-159, His-209, and Glu-276, were selected for site-directed mutagenesis studies based on their structural homology with conserved residues of other ATP-grasp enzymes. These mutants were subjected to kinetic analysis to characterize their roles in substrate binding and catalysis. In all four mutants, theK m value for ATP was significantly increased, implicating these residues in the binding of ATP. This result is consistent with the crystal structures of several other ATP-grasp enzymes, which have shown specific interactions between the corresponding homologous residues and cocrystallized ADP or nucleotide analogs. In addition, the maximal velocity of the reaction was significantly reduced (between 30- and 260-fold) in the 4 mutants relative to wild type. The data suggest that the mutations have misaligned the reactants for optimal catalysis.

Enzyme-biotin ϩ MgATP ϩ HCO 3 Ϫ L | ; Mg 2ϩ Enzyme-biotin-CO 2 Ϫ ϩ MgADP ϩ P i Enzyme-biotin-CO 2 Ϫ ϩ Acetyl-CoA º Malonyl-CoA ϩ Enzyme-biotin REACTIONS 1 AND 2 The Escherichia coli form of this enzyme consists of three separable components. The biotin carboxylase component catalyzes the first half-reaction, which involves the phosphorylation of bicarbonate to form a carboxyphosphate intermediate, followed by the transfer of the carboxyl group to the 1Ј nitrogen of biotin (2). The carboxyltransferase component catalyzes the second half-reaction. In vivo the biotin molecule is linked to the biotin carboxyl carrier protein through an amide bond to a specific lysine residue. Both biotin carboxylase and carboxyltransferase retain activity in the absence of the other two components and will also use free biotin as a substrate (3). The crystal structure of the biotin carboxylase component has been solved and is the only three-dimensional structure of a biotindependent carboxylase, making it the paradigm for structurefunction analysis of this class of enzymes (4). Two years after the solution of the crystal structure, Artymiuk et al. (5) observed that biotin carboxylase had a strong structural homology to glutathione synthetase and D-Ala D-Ala ligase. Despite the remarkable similarity in the three-dimensional structures of biotin carboxylase, D-Ala D-Ala ligase, and glutathione synthetase, there is only an 11% primary sequence identity between the three enzymes (5). Although biotin carboxylase is metabolically unrelated to these two enzymes, all three enzymes are mechanistically homologous in that they catalyze the ATP-dependent ligation of a carboxylate-containing substrate to an amine-containing substrate via formation of an acylphosphate intermediate (5,6). Structural similarity between the three enzymes includes a common three-domain architecture in which the flexible central domain extends away from the main body of the protein. The crystal structure of biotin carboxylase was originally determined in the absence of any ligands or substrate analogs (4), and its central domain (known as the B-domain) was in the "open" conformation, extending far out from the main body of the enzyme. In contrast, the structures of D-Ala D-Ala ligase and glutathione synthetase were solved in the presence of ADP and ATP, respectively, which revealed that the central domain forms a "lid" that clamps down over the active site upon nucleotide binding (7,8). Using the structures of D-Ala D-Ala ligase and glutathione synthetase, Artymiuk et al. (5) identified several active-site residues of biotin carboxylase as potentially important for catalysis; among these were Lys-116, Lys-159, His-209, Lys-238, Glu-276, Glu-288 and Asn-290. Soon after the observations of Artymiuk et al. (5), the three-dimensional structure of carba-myl phosphate synthetase was reported and found to be homologous to biotin carboxylase, D-Ala D-Ala ligase, and glutathione synthetase (9). The structural and mechanistic similarity of all four enzymes suggested they were linked through evolution, and thus, they became the charter members of the ATP-grasp family of enzymes. The name "ATP-grasp" derives from the novel nucleotide binding fold observed in these enzymes. The ATP-grasp family of enzymes expanded even further to include several enzymes involved in purine biosynthesis based on a position-specific iterative BLAST sequence alignment (6,10). The three-dimensional structures of two of these enzymes, N 5 -carboxyaminoimidazole ribonucleotide synthetase (11) and glycinamide ribonucleotide transformylase, (12) have been determined with nucleotides bound.
The sequence analysis studies identified several residues as being strictly conserved throughout the entire ATP-grasp family of enzymes. Not surprisingly, the conserved residues in biotin carboxylase were Lys-116, Lys-159, His-209, Glu-276, Glu-288, and Asn-290. Site-directed mutagenesis studies of Glu-288 and Asn-290 confirmed that these two residues were indeed important for catalysis (13). In fact, mutation of Glu-288 to lysine resulted in a completely inactive mutant (14). Recently, Thoden et al. (14) determined the crystal structure of the inactive mutant of biotin carboxylase, E288K, cocrystallized with ATP. The structure showed that the B-domain of biotin carboxylase does exhibit the characteristic "trap door" closure in the presence of nucleotide, with some atoms moving by more than 8 Å. As expected, comparison of the structure of the mutant biotin carboxylase-ATP complex with the structures of the other enzymes of the ATP-grasp superfamily revealed a significant degree of homology. For example, Lys-116 of biotin carboxylase and the residue homologous to Lys-116 in the other ATP-grasp enzymes were found to interact with ATP. However, there were some notable differences between the structure of biotin carboxylase and the structures of the other ATP-grasp enzymes. Namely, the biotin carboxylase crystal structure suggested that Lys-159, His-209, and Glu-276 did not interact with ATP, whereas the structures of the other ATPgrasp enzymes indicated that these residues did interact with ATP. Thus, the objective of this study is to test the hypothesis that residues Lys-116, Lys-159, His-209, and Glu-276 of biotin carboxylase are involved in binding ATP.

MATERIALS AND METHODS
Chemicals and Enzymes-Sodium bicarbonate labeled with 14 C was from Amersham Pharmacia Biotech and had a specific activity of 0.1 mCi/mmol. His binding resin was from Novagen. Pyruvate kinase was from Roche Molecular Biochemicals. Restriction grade bovine thrombin was from Enzyme Research Laboratories. Primers were synthesized by Life Technologies, Inc. All other reagents were from Sigma or Aldrich. The growth and purification of wild type and mutant forms of biotin carboxylase were performed as previously described (13).
Site-directed Mutagenesis-Site-directed mutagenesis of biotin carboxylase was carried out by the PCR method of overlap extension as previously described (13). The following mutants were constructed: H209A, E276Q, K159Q, K116Q, and K116A. The pairs of internal mutagenic primers used to make each site-directed mutant can be found in Table I. The entire gene of each mutant was sequenced to confirm that the desired mutation was made and that no other mutations were incorporated during polymerase chain reaction.
Enzymatic Assays-The rate of ATP hydrolysis by biotin carboxylase in the absence or presence of biotin was measured spectrophotometrically by coupling the production of ADP to pyruvate kinase and lactate dehydrogenase and monitoring the oxidation of NADH at 340 nm. Each assay contained 0.5 mM phosphoenolpyruvate, 0.2 mM NADH, 10 units of pyruvate kinase, 18 units of lactate dehydrogenase, and 100 mM HEPES at pH 8. To ensure the formation of the MgATP complex, MgCl 2 was included at concentrations at least twice that of the highest concentration of ATP in each assay. Since the K m for biotin is high (134 mM), the ionic strength of the reaction mixture was held constant with KCl when the initial velocities were measured as a function of biotin concentration.
For experiments in which bicarbonate was varied, all solutions (except for coupling enzymes, which were diluted into degassed buffers) were degassed to lower the level of endogenous bicarbonate (13) and stored in septum vials capped with rubber septa. All assay reactions were performed in a total volume of 1 ml and included the following components: 60 mM biotin, 0.5 mM phosphoenol pyruvate, 0.2 mM NADH, 21 units of pyruvate kinase, 35 units of lactate dehydrogenase, and 100 mM HEPES at pH 8.
For experiments in which the concentration of magnesium was varied, the N-terminal His tag on biotin carboxylase was removed by thrombin cleavage to eliminate the possibility of magnesium binding to the His tag. A ratio of two units of thrombin per unit of biotin carboxylase was used (13).
The rate of ATP synthesis from MgADP and carbamyl phosphate was determined spectrophotometrically. The formation of ATP was coupled to hexokinase and glucose-6-phosphate dehydrogenase, with the production of NADPH monitored at 340 nm. Each assay contained 0.5 mM glucose, 0.4 mM NADP, 2.5 units of hexokinase, 2.5 units of glucose-6phosphate dehydrogenase, 100 mM KCl, and 100 mM HEPES at pH 8. To ensure the formation of the MgADP complex, MgCl 2 was included and held at concentrations at least twice that of the highest concentration of ADP in the assay.
Initial velocities were measured using a Uvikon 810 (Kontron Instruments) spectrophotometer interfaced to a PC equipped with a data acquisition program. The temperature of the reactions was maintained at 25°C by a circulating water bath. All reactions were initiated by the addition of enzyme. Kinetic parameters were calculated per active site using a molecular mass of 50,000 daltons for the biotin carboxylase monomer (biotin carboxylase exists as a homodimer).
To determine if there was a stoichiometric production of ADP and carboxybiotin, the amount of carboxybiotin produced by biotin carboxylase was determined using a 14 C fixation assay and compared with the production of ADP as previously described (13). The reaction mixtures contained 20 mM ATP, 70 mM bicarbonate, 100 mM biotin, 50 mM MgCl 2 , and 100 mM HEPES at pH 8 in a total volume of 0.5 ml.
Data Analysis-The K m and V max parameters were determined by nonlinear regression analysis of the velocity versus [substrate] data to the Michaelis-Menten equation using the program Enzfitter.

RESULTS
Bicarbonate-dependent ATPase Reaction-In the absence of biotin, biotin carboxylase from E. coli catalyzes a bicarbonatedependent ATP hydrolysis (Reaction 1) as follows. REACTION 3 This reaction has been proposed to occur via formation of carboxyphosphate, which rapidly decomposes in the absence of biotin (15). The Michaelis constants for ATP and the maximal velocity of this partial reaction were determined for the wild type enzyme and four mutant enzymes of biotin carboxylase (Table II). All four mutants showed no significant change in V max when compared with the wild type enzyme. However, the a The underlined bases indicate the nucleotide positions that were changed.
K m for ATP for all four mutants increased between 40-and 90-fold.
The K m for bicarbonate for each of the mutants was determined at fixed, nonsaturating levels of ATP and biotin. Since it was not possible to achieve saturation with biotin or with ATP, the K m values for bicarbonate are apparent K m values. The apparent K m values for wild type biotin carboxylase and the four mutants are shown in Table II. The mutations did not significantly affect the apparent K m for any mutant except that of H209A, where the K m was 15 times greater than that of wild type.
Biotin-dependent ATPase Reaction-In the presence of biotin, biotin carboxylase from E. coli catalyzes the phosphorylation of bicarbonate by ATP to form a carboxyphosphate intermediate. The carboxyl group is then transferred from carboxyphosphate to biotin to form carboxybiotin. When ATP hydrolysis activity was examined in the presence of biotin, the K m values for biotin for all mutants were not significantly different compared with the wild type enzyme (Table III). The K m values for biotin are apparent because it was not possible to saturate with ATP. The largest change in K m was exhibited by H209A but was less than 10-fold. However, V max for three of the four mutants decreased significantly between 30-and 200fold. The one exception is K116Q, for which a K m value could not be obtained; the activity of this mutant did not increase with the addition of biotin over a range of 1-300 mM. To further investigate the role of this lysine residue in the biotin-dependent ATPase reaction, a K116A mutant was constructed, and the initial velocity as a function of biotin concentration was measured. The K m for biotin for the K116A mutant was 147 Ϯ 12 mM, whereas the maximal velocity was 1.03 Ϯ 0.04 min Ϫ1 . The effect of this mutation was similar to that of the other three in that the V max rather than the K m was significantly altered with respect to wild type. These data suggest that the lack of stimulation of ATP hydrolysis by biotin for K116Q reflects a function of the mutant glutamine residue rather than of the role of the native lysine residue.
Since the ATPase assay measured the production of ADP in the presence and absence of biotin, the question still remained as to whether carboxybiotin was being produced by the mutant enzymes. In other words, is there a 1:1 stoichiometry for the formation of ADP and carboxybiotin or is the hydrolysis of ATP uncoupled from the formation of carboxybiotin? If the 1:1 ratio were altered, this would suggest that the mutations had affected the carboxyl transfer step. The ratio of carboxybiotin to ADP produced during the ATPase reaction for wild type and the four mutants was determined (Table IV). All four mutants produced a ratio of carboxybiotin to ADP that was nearly 1:1. These results indicated that the mutations did not prevent the production of carboxybiotin, and therefore, the carboxyl transfer step had not been uncoupled from the hydrolysis of ATP.
ATP Synthesis Reaction-Biotin carboxylase from E. coli has been shown to catalyze the transfer of the phosphoryl group of carbamyl phosphate to ADP to form ATP and carbamate as follows.
MgADP ϩ Carbamyl phosphate º MgATP ϩ Carbamate REACTION 4 The carbamate rapidly decomposes to carbon dioxide and ammonia. Reaction 4 represents the reverse of Reaction 3, with carbamyl phosphate acting as an analog of the putative carboxyphosphate intermediate. Although biotin does not participate in the chemistry of this reaction, its presence does stimulate the rate of phosphoryl transfer (16).
The kinetic parameters for the ATP synthesis reaction were determined in the absence of biotin (Table V). The mutations did not have a significant effect on the K m for either carbamyl phosphate or ADP. However, a modest decrease in V max of 4-fold or less was observed.
To test the ability of biotin to stimulate the phosphoryl transfer reaction of the wild type and four mutants of biotin carboxylase, initial velocities were measured at a saturating concentration of ADP and carbamyl phosphate, both in the absence and presence of 60 mM biotin (Table VI). The degree of stimulation of the ATP synthesis activity by biotin was decreased 10-fold by the E276Q mutation, whereas the H209A mutant showed no significant decrease in stimulation. Both   K116Q and K159Q showed a decrease of ϳ2.5-fold compared with wild type (Table VI). Magnesium Assay-Biotin carboxylase requires two equivalents of magnesium for activity. One equivalent is complexed to ATP, whereas the role of the other equivalent is unknown. The effect of the four mutations on the ability of magnesium to stimulate the biotin-dependent ATPase activity of wild type and mutant biotin carboxylase was evaluated by measuring initial velocity as a function of [MgCl 2 ]. All four mutants exhibited a dependence on MgCl 2 similar to that of wild type. This suggests that these mutations did not affect the affinity of the enzyme for magnesium (Fig. 1). DISCUSSION The objective of this study was to test the hypothesis that four residues of biotin carboxylase, Lys-116, Lys-159, Glu-276, and His-209, were involved in binding ATP. Each of the corresponding site-directed mutants displayed an elevated K m value for ATP relative to the wild type value. This suggests that all four conserved active-site residues bind ATP. For the K116Q mutant, the increased K m value for ATP was consistent with the three-dimensional structure of biotin carboxylase bound to ATP as well as with the three-dimensional structures of other ATP-grasp enzymes bound to ADP or AMPPNP. 1 The crystal structure of biotin carboxylase complexed with ATP revealed an electrostatic interaction between the ⑀-amino group of Lys-116 and the ␣-phosphoryl oxygen of ATP (14). As shown in Table VII, the residues homologous to Lys-116 in enzymes of the ATP-grasp family also interacted with the oxygens of the ␣ or ␤ phosphates, as determined by crystallography. Moreover, mutation of the homologous residue in the carboxyphosphate domain of carbamyl phosphate synthetase resulted in a 5-fold increase in the K m for ATP (17).
Although all the available crystallographic and kinetic data implicate Lys-116 in binding an ␣ or ␤ phosphate oxygen, roles for the other three residues are not as well defined. Sitedirected mutagenesis data for biotin carboxylase and the crystal structures of other ATP-grasp enzymes suggest differing roles for Lys-159, His-209, and Glu-276. First, the ⑀-amino group of Lys-159 was not implicated in ATP binding based on the crystal structure of biotin carboxylase complexed with ATP. However, the residues homologous to Lys-159 in other ATPgrasp enzymes have been shown to interact with the nucleotide (Table VII), which supports the 90-fold increase in K m for ATP in the K159Q mutant of biotin carboxylase. Site-directed mutagenesis of the homologous residues of carbamyl phosphate synthetase (carboxyphosphate domain) and D-Ala D-Ala ligase revealed a K m for ATP that was 31-and 50-fold higher than the wild type value, respectively (17,18). Together these data implicate Lys-159 in binding to ATP. Second, the crystal structure of biotin carboxylase complexed with ATP showed that the imidazole group of His-209 is about 4 Å from the hydroxyl groups of the ribose of ATP, whereas the crystal structures of other ATP-grasp enzymes showed that the residue homologous to His-209 hydrogen bonded to the 2Ј and 3Ј hydroxyl groups of the ribose (Table VII). As with K159Q, the K m for ATP in the H209A mutant of biotin carboxylase was elevated 70-fold relative to wild type. Again, the structural data from other ATP-grasp enzymes and the site-directed mutagenesis results of biotin carboxylase strongly suggest that His-209 interacts with ATP, presumably with the hydroxyl groups of ribose.
Third, when the three-dimensional structure of biotin carboxylase with ATP was modeled with bound biotin, the ureido ring of biotin was located near the side chain of Glu-276 (14). However, three-dimensional structures of other ATP-grasp enzymes showed that the residue homologous to Glu-276 coordinated to either magnesium or manganese and that the divalent cation in each structure coordinated to the phosphoryl oxygens of the ␥ phosphate group (Table VII). Site-directed mutagenesis of Glu-276 in biotin carboxylase indicated no change in the K m for biotin but a nearly 40-fold increase in the K m for ATP. This

TABLE VI Stimulation of ATP synthesis reaction
The initial velocity of the ATP synthesis reaction was measured both in the presence and absence of 60 mM biotin. All reactions contained saturating levels of ADP, carbamyl phosphate, and magnesium. The stimulation factor is the ratio of the rate with biotin to the rate without biotin. result supports the hypothesis that Glu-276 coordinates to a magnesium ion in biotin carboxylase and that mutation of Glu-276 distorts the binding of magnesium and in turn the binding of ATP. It should be noted that although the mutation of Glu-276 altered the binding of magnesium ion to the residue, it did not appear to change the affinity of magnesium for the enzyme (Fig. 1). Additionally, when the homologous residue in the carboxyphosphate domain of carbamyl phosphate synthetase was mutated to alanine, the K m for ATP for the resulting mutant was 19-fold higher than wild type carbamyl phosphate synthetase (17). Again, the mutagenesis data for biotin carboxylase were consistent with the structures and mutagenesis data for the other ATP-grasp enzymes and inconsistent with the biotin carboxylase structure complexed with ATP.
The discrepancies described above can be explained by noting that the biotin carboxylase used in the crystal structure complexed with ATP contained a mutation of a critical residue (E288K). Although this mutation crippled the ability of the enzyme to hydrolyze ATP, it allowed crystals to be grown in the presence of ATP. The structural data from the other ATP-grasp enzymes in Table VII were obtained from wild type protein. Furthermore, the crystal structure of biotin carboxylase did not contain a divalent cation coordinated to ATP. Thus, the crystal structure of the E288K mutant form of biotin carboxylase complexed with ATP may not be a completely accurate description for the binding of ATP to this enzyme, which would explain the disagreement between the mutagenesis data and the crystal structure. Moreover, the evidence that ATP was bound incorrectly in the E288K mutant of biotin carboxylase may explain, at least in part, the lack of activity for the E288K mutant and provide insight into why the maximal velocities of the biotindependent ATPase activity for the four mutants characterized in this paper were significantly decreased.
Although the increase in the Michaelis constants of the four mutants of biotin carboxylase strongly suggests that these residues are involved in binding ATP, the concomitant decrease in their maximal velocities indicates that these residues also play a role in catalysis. If these four residues were solely involved in binding ATP, then the corresponding mutants should have the same maximal velocity as wild type biotin carboxylase. This brings us to the question of what role these residues could play in catalysis that would be consistent with their role in binding ATP. Any discussion of the catalytic roles of these residues must begin with the observation that biotin carboxylase exhibits the phenomenon of substrate-induced synergism with respect to biotin (13). That is, in the absence of biotin, the enzyme will cleave ATP into ADP and P i in a bicarbonate-dependent manner, albeit at a very slow rate. However, in the presence of biotin, the rate of ATP hydrolysis increases 1100-fold. Thus, the hydrolysis of ATP is synergistic with the binding of biotin. A possible explanation for substrateinduced synergism in biotin carboxylase that is consistent with the data presented in this paper is that in the absence of biotin ATP binds to the enzyme in a large number of nonproductive modes, which is manifested as a low maximal velocity. However, upon the binding of biotin to biotin carboxylase, the number of nonproductive binding modes of ATP is reduced, allowing for a more reactive alignment of the substrates. This phenomenon is manifested as a dramatic increase in the maximal velocity. Recently, a more sophisticated version of this concept has been proposed (19,20). In the current theory, the enzyme molecule pre-organizes the active site to allow the substrates to form near attack conformers. A near attack conformer refers to the juxtaposition of the substrates in the ground state such that they closely resemble the transition state. The effect of mutating any of the four residues in this study would be a shift in the active site geometry, which would possibly reduce the number of near attack conformers. This is manifested as a significantly reduced maximal velocity for the mutant enzyme. Biotin cannot properly cause the alignment of ATP for the reaction because of greater conformational flexibility of ATP in the active site due to the mutation. This concept of greater conformational flexibility of ATP is supported by both the increase in K m for ATP in each of the mutant enzymes and the presumably incorrect binding of ATP in the active site of the inactive E288K mutant. Recall that none of the mutants affected the carboxyl transfer from carboxyphosphate to biotin, yet the V max values were significantly decreased. Thus, the fact that these four ATP binding residues are conserved throughout the ATP-grasp superfamily of enzymes further attests to the notion that binding interactions and correct positioning of the substrates appear to play the dominant role in catalysis by biotin carboxylase. The question is now how does the binding of biotin to the enzyme position the substrates into a more reactive conformation. A conformational change (i.e. induced fit) in biotin carboxylase upon biotin binding could explain the large increase in rate for ATP hydrolysis. However, the major conformational change in biotin carboxylase occurs upon the binding of ATP (14), which binds to the enzyme first before bicarbonate and biotin (21). A crystal structure of biotin carboxylase with only biotin bound showed no difference in conformation compared with the unliganded structure of the enzyme (4). The lack of a large conformational change in biotin carboxylase upon binding biotin is consistent with the high K m for biotin (134 mM). Using this value as an apparent binding constant for biotin to calculate the binding energy, a relatively low value of 1.2 kcal/mol is obtained. The low binding energy of biotin to the enzyme is not suggestive of a large conformational change.
How then is biotin able to stimulate the rate of ATP hydrolysis if not via a conformational change? Perhaps biotin only promotes very small changes in the enzyme that result in the alignment of substrates for catalysis. To this end, recent studies on hydrogen tunneling in dehydrogenases have found a correlation between protein dynamics and enzymatic activity (22,23). Particularly intriguing is the case of isocitrate dehydrogenase, to which the binding of its substrate, isocitrate, induces shifts of less than an angstrom in the amino acid side chains of the active site. These seemingly insignificant changes in conformation are in fact, related to rate increases of many orders of magnitude (24,25). Thus, it may be that very subtle dynamic behavior of biotin carboxylase is enough to generate the large increase in the rate of ATP hydrolysis upon the binding of biotin. Further studies will be required to determine this aspect of the mechanism. In summary, the four active-site residues of biotin carboxylase, Lys-116, Lys-159, His-209, and Glu-276, were shown to be involved in binding ATP. Furthermore, these four residues have also been found to be involved in catalysis, and their role in catalysis is to orient ATP in a conformation that allows for optimal catalysis. Finally, the results also suggest that the crystal structure of the mutant biotin carboxylase, E288K complexed with ATP, may not be a completely accurate depiction for the binding of ATP to the wild type form of biotin carboxylase.