Mutational analysis of conserved residues in the GCN5 family of histone acetyltransferases.

GCN5 is a critical transcriptional co-activator and is the defining member of a large superfamily of N-acetyltransferases. GCN5 catalyzes the transfer of an acetyl group from acetyl-CoA to the epsilon-amino of lysine 14 within the core H3 histone protein. Previous biochemical analyses have indicated a fully ordered kinetic mechanism. Recent structural studies have implicated several conserved residues in catalysis and substrate binding. Here the roles of Glu-173, His-145, and Asp-214 in yeast GCN5 have been evaluated using site-directed mutagenesis, steady state and pre-steady state kinetics, pH analysis, isotope partitioning, and equilibrium binding studies. The results with wild type and E173Q, H145A, and D214A mutants are consistent with chemical catalysis being rate-determining in turnover. All mutants exhibited K(d) values (3.5-8.5 microm) for AcCoA that were similar to wild type enzyme, indicating no functional role for these residues in AcCoA binding. The E173Q mutant demonstrated a approximately 500-600-fold decreases in k(cat) and k(cat)/K(m),(H3), consistent with Glu-173 acting as the general base catalyst as proposed previously. No significant effect was observed on substrate binding steps. His-145 was identified as a residue in the peptide binding cleft that must be unprotonated (pK(a) = 5.8) for peptide binding and likely hydrogen-bonds to the Ser-10 hydroxyl of histone H3. His-145 also contributes to lowering the pK(a) value (by 0.8 units) of general base Glu-173 through a water-mediated hydrogen bond to the carboxylate side chain. Analysis of D214A revealed an obligate protein isomerization step that occurs after AcCoA binding and permits efficient peptide binding. Asp-214 is part of a conformationally flexible loop that mediates the isomerization by stabilizing distinct conformers of the protein.

zymes capable of transferring an acetyl group from acetyl-CoA to an acceptor histone protein substrate. The acceptor site is the ⑀-amino group of lysine side chains within the aminoterminal tails of the core histones, H2A, H2B, H3, and H4. At least four gene families of HATs have been identified in mammals (1,2). The largest family includes the defining member GCN5 whose catalytic domain is well conserved from yeast to humans. The GCN5 family of HATs are part of a larger superfamily of enzymes capable of acetyl transfer to amine-containing substrates (referred to as GNATs, for GCN5-like N-acetyltransferases). PCAF (p300/CBP-associating factor) (3)(4)(5)(6) is also a member of the GCN5 family of HATs, displaying similar substrate specificity within the catalytic domain. In vitro, GCN5 displays a strong preference for Lys-14 of histone H3, although other acetylation sites on H3 and H4 have been observed (7)(8)(9)(10).
Yeast GCN5 has been known for some time to be essential for full transcriptional activation in a subset of genes (11)(12)(13). After the discovery of HAT activity in a related Tetrahymena enzyme (14), the causal link between histone acetylation and gene activation was codifying and has led to an explosion of HAT discoveries (reviewed in Ref. 15). In yeast, GCN5 promotes both global histone acetylation (i.e. genome wide) and gene loci-specific histone acetylation (9, 16 -18). In mammals where global acetylation is less pronounced than in yeast, gcn5 appears to be an essential gene (19).
Given the importance of these enzymes in controlling transcription by both global and loci-specific acetylation, understanding the molecular mechanisms of catalysis, regulation, and substrate specificity has become a requisite part of understanding the biological functions of these enzymes and the development of specific inhibitors. To date, GCN5 is the best characterized HAT in terms of its catalytic mechanism and structure. With both GCN5 and PCAF, the basic mechanism was shown to involve the initial formation of a ternary complex before any chemical events occur (20 -23). This implies the lack of an acetyl-enzyme intermediate that may form from the initial reaction of acetyl-CoA with enzyme. Also, attempts to trap an enzyme intermediate have not been successful (21,22). 2 The binding of substrates (histone and acetyl-CoA) and the release of products (acetylated histone and CoA) appear to be strictly ordered (21)(22)(23). It was demonstrated that conserved Glu-173 (yeast GCN5 numbering) is functioning as the general base catalyst, abstracting a proton from the ⑀-amino group of the bound lysine-containing substrate (20). This effectively increases the nucleophilicity of the amine nitrogen for efficient attack at the carbonyl carbon of bound acetyl-CoA, thus transferring the acetyl moiety to histone.
Several recent NMR (24) and x-ray structures (25,26) have * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
§ Supported by American Cancer Society Grant RSG-01-029-01-CNE and National Institutes of Health Grant GM59785. To whom correspondence should be addressed: Oregon Health Sciences University, Dept. of Biochemistry and Molecular Biology, 3181 SW Sam Jackson Park Road, Portland, OR 97201-3098. Tel.: 503-494-0644; Fax: 503-494-8393; E-mail: denuj@ohsu.edu. 1 The abbreviations used are: HATs, histone acetyltransferases; Ac-CoA, acetyl coenzyme A; CoA, coenzyme A; DTT, dithiothreitol; PCAF, p300/CBP-associating factor; PAGE, polyacrylamide gel electrophoresis; tGCN5, tetrahymena GCN5; yGCN5, yeast GCN5; Bis-Tris, 2-[bis(2-hydroxyethyl)amino]-2-(hydroxymethyl)propane-1,3-diol. corroborated these initial mechanistic conclusions. The structures of PCAF and GCN5 from several species revealed an ␣/␤ globular fold where two roughly orthogonal hydrophobic troughs were found along the surface of the protein. The structure of the dead-end complex of Tetrahymena GCN5 with CoA and peptide identified these troughs as the binding sites for H3 histone peptide and coenzyme A (25). The conserved glutamic acid resides within a deep hydrophobic pocket, near the junction where Lys-14 of substrate and the thiol of CoA meet. The crystal structures are consistent with the biochemical evidence that demonstrated that the reaction is ordered and that a conserved glutamate residue acts as the general base. However, many mechanistic questions remain.
The structure determinations have provided some additional insight into the possible roles played by other conserved residues that line the active site and the substrate-binding grooves. We have undertaken a detailed kinetic analysis to provide biochemical and kinetic evidence for the functional role of several invariant residues. Here we examine the functions of His-145, Asp-214, and Glu-173 in yeast GCN5 using site-directed mutagenesis, pH analysis, steady state and pre-steady state kinetics, equilibrium dialysis, and isotope partitioning methods. Our findings are discussed in terms of previous structural and biochemical data.

EXPERIMENTAL PROCEDURES
Materials-Histone H3 peptide, ARTKQTARKSTGGKAPRKQLC (representing the 20 amino-terminal residues of human histone H3 and an additional carboxyl-terminal cysteine), and the corresponding H3 peptide with phosphoserine at position 10 were synthesized by the Protein Chemistry Core Lab at the Baylor College of Medicine.
[ 3 H]Acetyl-CoA (1.88 Ci/mmol) was from PerkinElmer Life Sciences. Acetyl-CoA was purchased from Roche Molecular Biochemicals. P81 phosphocellulose disks were from Life Technologies, Inc. Dispo-Equilibrium Dialyzers were from Amika Corp. All other reagents were from Sigma, Pierce, or Fisher.
Expression and Purification-Mutations in the catalytic domain of yGCN5 were generated using the Bio-Rad Muta-gene method, and oligonucleotides harboring the desired mutations are as follows: 5Ј-GAAAACAATTTGTGCGAATTC-3Ј for E173Q; 5Ј-AGCGTAATTAGCT-GCATATGT-3Ј for D214A; 5Ј-AGCGTAATTATTTGCATATGT-3Ј for D214N; 5Ј-CATGGAAAGAGCACTTCGATC-3Ј for H145A; and 5Ј-ATC-GAAAGGTTTATATGTTAT-3Ј for R164K. The mutations were verified by DNA sequencing. The catalytic domain (amino acids 99 -262) of yGCN5, and single amino acid substituted forms, were recombinantly expressed in BL21-DE3 bacteria by isopropyl-␤-D-thiogalactopyranoside induction for 4 h at 25°C using the T7 polymerase-based expression system as described (27). Following chromatography on S-Sepharose, fractions containing yGCN5 or the desired mutant protein (assessed by SDS-PAGE and HAT activity) were pooled and concentrated and then subjected to size-exclusion chromatography on G-75 Sephadex. Purified fractions (greater than 90% purity), determined by SDS-PAGE, were pooled and concentrated and stored at Ϫ20°C until use. Protein concentrations were determined by the method of Bradford (28).
Enzymatic Assays for yGCN5-yGCN5 HAT activity was measured using [ 3 H]acetyl-CoA and histone H3 peptide or the phosphorylated (Ser-10) H3 peptide as substrates using a radioactive P81 filter binding assay as described previously (20).
Bi-substrate Kinetic Measurements-Bi-substrate kinetic analyses were performed on the E173Q, D214A, and H145A mutants at AcCoA concentrations spanning 0.1 to 40 M and H3 peptide concentrations spanning 25 to 600 M. The data were fit to the sequential (ternary complex) mechanism equation (Equation 1), using the algorithms of Cleland (29) and the computer program KinetAssyst (IntelliKinetics, State College of Pennsylvania), using a nonlinear least squares approach.
Determination of the Dissociation Constant for AcCoA Binding to yGCN5 via Equilibrium Dialysis-Equilibration conditions were 50 mM Tris, pH 7.5, and 15-20 M yGCN5. Equilibration was performed using Dispo-Equilibrium Dialyzers (Amika Corp.), which contain two 75-l chambers separated by a 5-kDa cut-off dialysis membrane. The K d value for AcCoA in the presence of yGCN5 was determined by aliquoting 5-400 M AcCoA (20 -40 cpm 3 H/pmol) into the buffer chamber, and yGCN5 into the sample chamber. After 40 h of equilibration on a level shaker, samples were recovered from each chamber and counted by liquid scintillation to determine the amount of radioactivity in the buffer chamber ([AcCoA] free ) and the sample chamber ([AcCoA] free ϩ [AcCoA⅐yGCN5]) in order to determine the amount of bound AcCoA. The data were fit to Equation 2 and are presented in hyperbolic form.
Isotope Trapping Experiments-Pulse-chase experiments were done in order to obtain information regarding both kinetic and equilibrium aspects of the yGCN5⅐AcCoA (30 -32). Experiments were performed at 24 Ϯ 1°C, in 5 mM DTT, 50 mM Tris, pH 7.5 and 9.5, by preparing solutions containing 37.5 l of 1.0 M [ 3 H]acetyl-CoA and 20 M wild type yGCN5 (pulse) and 20 l of 10 mM unlabeled AcCoA and 0.0 -4.0 mM histone H3 peptide (chase). Under these conditions, the unlabeled AcCoA and histone H3 peptide react slowly yet spontaneously and therefore must be mixed immediately before use. Additionally, the equilibrium conditions for the pulse solution were chosen such that nearly all of the [ 3 H]acetyl-CoA is bound to yGCN5 as determined by equilibrium dialysis. 5-10 s after preparing the chase solution, 12.5 l of the chase was added to the pulse solution to obtain 50 l of 15 M yGCN5, 0.75 M [ 3 H]acetyl-CoA, 2.5 mM AcCoA, and 0 -1.0 mM histone H3 peptide. The reactions were rapidly mixed for ϳ5 s with a pipette; 40 l of the reaction volume was spotted onto P81 phosphocellulose, and the amount of [ 3 H]acetyl-H3 peptide was determined as described above (20).
To determine the amount of radioactivity incorporated into [ 3 H]acetyl-H3 peptide after the initial partitioning event, control experiments were performed at 24 Ϯ 1°C, in 5 mM DTT, 50 mM Tris, pH 7.5 and 9.5, by preparing solutions containing of 3.33 mM unlabeled AcCoA, 1.0 M [ 3 H]acetyl-CoA, and 20 M yGCN5 (pulse) and 20 l of 0.0-4.0 mM histone H3 peptide (chase). Under these conditions, both the [ 3 H]acetyl-CoA and the unlabeled AcCoA are in rapid equilibrium with yGCN5, such that the labeled substrate is effectively diluted with unlabeled substrate. After pre-equilibrating the reaction solutions to 24 Ϯ 1°C, 12.5 l of the chase was added to the pulse solution to obtain 50 l of 15 M yGCN5, 0.75 M [ 3 H]acetyl-CoA, 2.5 mM AcCoA, and 0 -1.0 mM histone H3 peptide, the identical reaction conditions as the actual pulse-chase experiment. As before, the reactions were rapidly mixed for ϳ5 s with a pipette; 40 l of the reaction volume was spotted onto P81 phosphocellulose, and the amount of [ 3 H]acetyl-H3 peptide was determined. The amount of labeled product generated in these control experiments was used to correct for the amount of labeled product obtained in the partitioning samples after the initial partitioning event due to enzymatic turnover of the diluted labeled substrate. The percent of the total labeled substrate partitioned to product in the initial partitioning event versus initial [H3 peptide] was plotted, and the data were fit to Equation 3, Step, Pre-steady State Analysis-The rate-limiting step for wild type yGCN5 was examined using a Hi-Tech quench-flow device (Hi-Tech Ltd., Salisbury, UK). Experiments under single turnover conditions were performed using 20 and 30 M yGCN5, 0.5 M AcCoA, and 200 M H3 peptide at 24 Ϯ 1°C, pH 7.5 (concentrations given are post-mixing). Two concentrations of wild type yGCN5 were used to ensure that maximal rate was achieved for the single turnover reaction. After various reaction times between 0.01 and 10 s, the reactions were quenched with 2 N HCl, and the amount of [ 3 H]acetyl-H3 peptide was determined using the filter binding assay by spotting 50 l of the 200-l reaction slug on P81 phosphocellulose paper and washing away the free acetyl-CoA in sodium bicarbonate (20). Variations in the dilution of the slug were corrected versus an internal control of total 3 H radioactivity by spotting 50 l of the reaction slug onto P81 phosphocellulose paper and allowing it to dry without washing prior to scintillation counting. Control experiments with yGCN5 and [ 3 H]acetyl-CoA quenched prior to the addition of H3 peptide yielded insignificant counts, consistent with the lack of an acetyl-enzyme intermediate being involved in the reaction. Data were fit to a first-order exponential using Kaleidagraph version 3.0.5 by Abelbeck Software: where A is the amplitude of the exponential first-order rate of product formation; k is the first-order rate constant; t is time, and B is the total amount of product.
Quench flow experiments under multiple turnover conditions were performed using 5-10 M yGCN5 and 50 M [ 3 H]acetyl-CoA, which were rapidly mixed with 200 -500 M H3 peptide at 24 Ϯ 1°C, pH 7.5 (concentrations given are post-mixing). After various reaction times between 0.01 and 4.0 s, the reactions were quenched with 2 N HCl, and the amount of [ 3 H]acetyl-H3 peptide was determined as described above. Due to a linear increase in product formation and the lack of an apparent burst or lag phase, data were fit to a linear least squares fit.
Because of the slower catalytic rates of the mutated forms of yGCN5, pre-steady state experiments were conducted on the bench top without the use of the quench-flow device. Single turnover experiments were performed using 20 -50 M yGCN5 mutants, 0.5 M AcCoA, and 500 M histone H3 peptide at 24 Ϯ 1°C, pH 7.5. As with wild type yGCN5, increased concentrations of enzyme were used to guarantee that the maximal reaction rate was achieved. The reactions were quenched after various reaction times ranging from 0 to 4 h and were analyzed as described for wild type yGCN5. Multiple turnover experiments were performed using 5-10 M mutated yGCN5, 100 M AcCoA, and 500 M histone H3 peptide for reaction times ranging from 0 to 30 min. As a control to ensure enzyme stability, the mutated forms of yGCN5 were assayed for HAT activity after preincubation at 24 Ϯ 1°C for the length of time of the longest experimental data point.
pH Profile of the yGCN5-catalyzed HAT Reaction-At different pH values spanning 5.6 -10, the k cat /K m , H3 values for wild type yGCN5 and the H145A mutant were determined at saturating levels of AcCoA (100 M), and varying histone H3 peptides (25-300 M). In addition, at pH values spanning 6.5-9.5, the k cat /K m, AcCoA values were determined at near-saturating levels of H3 peptide (1.5 mM) while varying AcCoA (0.5-2 M). A 50 mM Tris, 50 mM Bis-Tris, and 100 mM sodium acetate (TBA) buffer was employed to keep the ionic strength constant over the pH range examined. Typically full saturation curves were fit to the basic Michaelis-Menten equation. For the pH profiles however, initial velocities were measured at varied sub-saturating concentrations for the substrate of interest. Consequently, the data were fit to a modified Michaelis-Menten equation (Equation 4) in order to directly obtain statistically valid values for k cat /K m , H3 and k cat /K m , AcCoA .
Once the k cat /K m , H3 and k cat /K m , AcCoA values were obtained at the indicated pH, each parameter was fit to Equation 5 or 6.
where C is the pH-independent value, K a1 and K a2 are individual ionization constants, and H is the proton concentration.

RESULTS AND DISCUSSION
Mutational Analysis of yGCN5-Initial kinetic analyses of wild type yGCN5 and the highly homologous human PCAF were previously published (21)(22)(23). These analyses indicated an ordered Bi-Bi kinetic mechanism as follows: AcCoA binds to the free HAT enzyme; H3 histone binds second to form a ternary complex; chemistry occurs by facilitating the transfer of the acetyl group from AcCoA to the ⑀-amino group acceptor of Lys-14 of histone H3, generating CoA and Ac-Lys-14 H3 histone products; Ac-Lys-14 histone H3 is released from the ternary complex; and CoA is released to regenerate the free HAT (21)(22)(23). In addition, biochemical analysis of an E173Q mutant of yGCN5 implicated the invariant Glu-173 as the general base, abstracting a proton from Lys-14 of histone H3 to facilitate acetyl transfer from AcCoA (20).
The crystal structures of Tetrahymena GCN5 (tGCN5) bound to coenzyme A, and tGCN5 bound to both coenzyme A and a histone H3 peptide (representing the amino-terminal tail of histone H3) has recently been solved (25). These structures have provided valuable structural information that are consistent with Glu-173 functioning as the general base in catalysis and that have suggested the involvement of other conserved residues in catalysis or substrate binding. For example, the conserved Asp-214 of yGCN5 resides on a conformationally flexible loop. Its location near the presumed active site had suggested a potential role in catalysis, perhaps as a general base and/or general acid catalyst (Fig. 1) (24,25).
In addition to Glu-173 and Asp-214, which may be involved in acetyl transfer to histone H3, His-145 appeared to be within hydrogen bonding distance of Ser-10 of histone H3 peptide in the structure of the ternary complex ( Fig. 1) (25). Ser-10 of histone H3 has been demonstrated to be phosphorylated during mitosis (33,34) and transiently during mitogen stimulation of quiescent cells (35,36). Mutation of Ser-10 impairs condensation and segregation of chromosomes during mitosis (37). The This model was generated using the solved ternary co-crystal structure of Tetrahymena GCN5 (lower left) with bound CoA (lower right) and unmodified histone H3 peptide (across top) (25). The amino acid numbers refer to those of yeast GCN5. A serine residue was modeled into the peptide at position 10 (orange). The crystal structure suggests that both Glu-173 (violet) and Asp-214 (green) are in a position to facilitate acetyl transfer from AcCoA to Lys-14 of histone H3 peptide. The structure also suggested that His-145 (black) is within hydrogen bonding distance from Ser-10 of histone H3. By comparing the various crystal structures of Tetrahymena GCN5 (25), we observed that the His-145 (yGCN5) and the conserved general base Glu-173 (yGCN5) were within a distance of ϳ6.4 Å. His-145 and Glu-173 appear to be bridged by a water molecule in the apo-and CoA-bound structures or by the serine residue at position 10 of the H3 peptide, in CoA, peptide tGCN5 complex.
phosphorylation of Ser-10 on histone H3 has been demonstrated to be synergistically coupled to acetylation of Lys-14 and is mediated by an increased affinity for the phosphorylated substrate by wild type yGCN5 (36). Due to the proximity of His-145 to Ser-10 of histone H3 peptide, we hypothesized that this His residue may play a role in peptide binding and possibly in recognition of the phosphorylated substrate.
To elucidate the functions of Glu-173, Asp-214, and His-145, we generated and kinetically characterized the E173Q, D214A(D214N), and H145A mutated forms of yGCN5. As one of the conclusions from this analysis, we provide data for an isomerization step in the catalytic mechanism subsequent to AcCoA binding and prior to H3 peptide binding (Scheme 1). The experimental evidence for this isomerization will be discussed below, along with a discussion of the functions of Glu-173, Asp-214, and His-145.
Bi-substrate Kinetics-In order to elucidate the function of conserved amino acids in the catalytic mechanism of GCN5, initial bi-substrate saturation kinetics were performed to obtain the steady state kinetic parameters for the mutated forms of yGCN5: E173Q, D214A, and H145A. Initial velocities were determined as a function of AcCoA concentrations at different fixed H3 peptide concentrations (Fig. 2). These data were plotted in double-reciprocal form and, for all three mutated forms, displayed an intersecting line pattern that intersects to the left of the 1/velocity axis. Entire data sets were fit directly to Equation 1, using a nonlinear least squares method. The steady state parameters obtained from this bi-substrate analysis are summarized in Table I. The results for the mutant forms of yGCN5 were compared with the kinetic parameters obtained with wild type yGCN5. At pH 7.5, the E173Q mutant exhibited a 680-fold decrease in k cat and a 540-fold decrease in k cat /K m , H3 , with no observable change in the K m , H3 . Because the E173Q mutant could be saturated with AcCoA as low as 0.05 M AcCoA, the employed assays were not sensitive enough to quantitate the extremely low K m,AcCoA value. Consequently, with the E173Q mutant, an upper limit of 0.025 M was assigned for the K m,AcCoA value, and by dividing the measured k cat by the estimated K m,AcCoA , a lower limit of 1.0 ϫ 10 5 M Ϫ1 ⅐s Ϫ1 was therefore assigned for the k cat /K m , AcCoA value. The resulting values of K m,AcCoA and k cat / K m , AcCoA suggested that there may be little if any effect on AcCoA binding steps. Since there is an ϳ600-fold decrease in k cat , these results are fully consistent with Glu-173 acting only during chemical catalysis (k 7 ) and that Glu-173 functions as a general base by deprotonating the ⑀-amino group of the histone lysine, thereby facilitating nucleophilic attack of AcCoA (20).
For the D214A mutant, we observed an ϳ200-fold decrease in k cat and an ϳ150-fold decrease in k cat /K m , H3 . If Asp-214 functioned during chemical catalysis, perhaps as a general acid/base, then these results would be consistent with a simple 150 -200-fold decrease in the rate of chemistry (k 7 ). However, we also observed an ϳ340-fold decrease in the k cat /K m , AcCoA with little or no change in the K m,AcCoA . This observation is consistent with additional alteration(s) in the kinetic mechanism. To examine whether the decrease in k cat /K m , AcCoA might be the result of a pK a shift of GCN5, and therefore an apparent decrease in k cat /K m , AcCoA at pH 7.5, a k cat /K m , H3 pH analysis (from pH 6.5 to 10) for the D214A mutant was performed (data not shown). The magnitude change of k cat /K m , H3 for wild type yGCN5 versus the D214A mutant was equivalent at all the pH values evaluated (i.e. no change in pK a was observed). This suggests that the observed change in k cat /K m , AcCoA was not caused by a change in the equilibrium of the enzyme protonation state. However, a decrease in k cat /K m , AcCoA may suggest an effect on the dissociation constant (k 2 /k 1 , Scheme 1) for AcCoA binding.
The H145A mutated form of yGCN5 exhibited an ϳ15-fold decrease in k cat and an ϳ25-fold decrease in k cat /K m , H3 , which is again consistent with a simple 15-25-fold decrease in the rate of chemistry (k 7 ). However, as with the D214A mutant, we observed a 15-fold decrease in the k cat /K m , AcCoA with little or no change in the K m,AcCoA . These results are not consistent with a change in k 7 , since k cat /K m , AcCoA reflects AcCoA binding and any kinetic steps prior to the addition of peptide. The observed SCHEME 1. Proposed kinetic mechanism of yGCN5. The rate equations for the steady state parameters were derived for the proposed mechanism using the method of net rate constants at fixed pH (39). The reversibility of the reaction was analyzed previously for wild type yGCN5, and the turnover number for the reverse reaction (k 8 ) was estimated as 0.00002 s Ϫ1 , ϳ100,000-fold lower than the k cat of the forward reaction (22). In the rate equation derivation, k 9 and k 11 , product release steps are assumed to be rapid, and the reverse chemical step k 8 is slow.  Table I. decrease in k cat /K m , AcCoA may therefore suggest a decrease in AcCoA binding affinity. In addition to changes in k cat /K m , AcCoA , we observed a significant 2-fold increase in the K m , H3 suggesting that His-145 may contribute slightly to peptide binding.
Equilibrium Dialysis Binding Assays-The initial observations with the D214A and the H145A mutants had suggested that the AcCoA dissociation constant (K d ) for these mutants may have been significantly increased. In order to determine the AcCoA affinity of the yGCN5 mutants, binding assays were performed using equilibrium dialysis. The dissociation constant for AcCoA binding to wild type yGCN5 was previously determined in the absence of histone H3 peptide via equilibrium dialysis (22). In this study, equivalent experiments were performed on the E173Q, D214A, D214N, H145A, and R164K mutants. The dissociation constant (K d ) was determined by dispensing varying concentrations of [ 3 H]AcCoA into the buffer chamber and fixed concentrations of the various mutated forms of yGCN5 into the sample chamber. The data were collected and analyzed as described under "Experimental Procedures." A representative binding experiment is shown in Fig. 3. The average K d values from duplicate experiments are summarized in Table II. The mutations E173Q, D214A(D214N), H145A, and R164K appeared to have little or no effect on AcCoA binding (values ranged from 3.5 to 8.5 M). In addition, the data were in excellent agreement with the calculated enzyme concentrations determined by the method of Bradford (28).
The equilibrium binding results for the E173Q mutant are consistent with Glu-173 playing no direct role in AcCoA binding. Since there is no observable change in k cat /K m , AcCoA for the E173Q mutant, the observed decrease in k cat is consistent with an ϳ600-fold decrease in k 7 (Scheme 1), and in agreement with the conclusion that Glu-173 functions as the general base in catalysis. Also, the lack of an effect with the R164K mutant is consistent with Arg-164's proposed role in phosphoserine 10 recognition on H3 substrate. The steady state parameters of R164K were indistinguishable from those of wild type enzyme when unmodified H3 peptide is the substrate (36). For the D214A and H145A mutated forms, these results strongly suggest that changes in the steady state kinetic parameters, namely k cat /K m , AcCoA , may be caused by affecting steps in the kinetic mechanism (such as enzyme isomerization) prior to H3 peptide binding but distinct from initial AcCoA binding (k 1 and k 2 ).
Isotope Trapping Experiments, Evidence for Isomerization-To provide evidence for an enzyme isomerization, isotope trapping experiments (pulse-chase experiments) (30 -32) were performed on wild type yGCN5. This method provides a technique for evaluating (under various concentrations of second substrate) the relative amount of labeled [ 3 H]acetyl-CoA substrate from a [ 3 H]acetyl-CoA⅐yGCN5 complex that dissociates to a large pool of unlabeled AcCoA substrate (effectively irreversible) versus the amount of labeled substrate from the complex that is trapped as product (31). Since an equilibrium binding study does not preclude the possibility of a catalytically dysfunctional complex, and since improper binding is not easily revealed under steady state conditions, this method provides information that cannot be obtained from either a steady state or an equilibrium analysis. The ability to trap all of the [ 3 H]acetyl-CoA⅐yGCN5 complex at infinite concentrations of histone H3 peptide would suggest that all of the complex exists as a relevant Michaelis complex capable of catalysis, and from this, k off,AcCoA may be determined (30). Alternatively, the failure to trap any or all of the  Table II.  the ternary complex at a rate much faster than k cat (30).
As described under "Experimental Procedures," the equilibrium conditions for the pulse solution were chosen such that nearly all of the [ 3 H]AcCoA (Ͼ95%) is bound to yGCN5 as determined by equilibrium dialysis. In the event that all of the AcCoA⅐yGCN5 complex forms a competent Michaelis complex, and if the rate of AcCoA dissociation from either the binary or ternary complex is not extremely fast relative to k cat , then all of the [ 3 H]AcCoA may be partitioned to product at infinite concentrations of H3 peptide. For wild type yGCN5, the percent of labeled substrate partitioned to product from the AcCoA⅐yGCN5 complex was determined at pH 7.5 and pH 9.5 at varied concentrations of histone H3 peptide. The data were plotted as a function of initial [H3 peptide] and were fit to Equation 3 (Fig. 4).
The maximal percent of labeled substrate partitioned to product ([*P max ] ϭ 8.76 Ϯ 0.23%) by wild type yGCN5 at pH 7.5 proved to be saturable at low concentrations of histone H3 peptide, with a KЈ H3 of 9.42 Ϯ 1.35 M well below the K m , H3 under steady state conditions. In addition, when this experiment was performed at pH 9.5, similar values were obtained for [*P max ] (8.38 Ϯ 0.62%) and KЈ H3 (10.12 Ϯ 2.81 M) indicating that partitioning labeled substrate from the AcCoA⅐yGCN5 complex is independent of pH. The failure to trap all of the [ 3 H]AcCoA⅐yGCN5 at infinite concentrations of histone H3 peptide suggests that either [ 3 H]AcCoA dissociates from the binary or ternary complex at a rate much faster than k cat or that the complex can exist in more than one form, not all of which are competent Michaelis complexes (30). However, since this phenomenon is fully saturable at low concentrations of H3 peptide and is independent of pH (which has been demonstrated to affect k cat ), these results strongly support the conclusion that the AcCoA⅐yGCN5 complex exists in more than one isomeric form, not all of which are kinetically competent Michaelis complexes. The isomerization occurs after AcCoA binds and likely promotes the efficient binding of histone H3. This isomerization step has been made explicit in the kinetic mechanism for GCN5 (Scheme 1).
Asp-214 Provides Structural Stability to a Conformationally Sensitive Loop of yGCN5-The x-ray crystallographic structures of free (apo) tGCN5, of tGCN5 bound to AcCoA substrate, and of tGCN5 bound as a dead-end complex to CoA and a histone H3 peptide have provided insight into the role of many conserved residues in the GCN5 family of enzymes (25). Upon initial consideration of these structures, it has been observed that the ␤5-loop-␣4 region of GCN5 undergoes distinct conformational changes upon formation of the binary and ternary complexes (Fig. 5). The results of the isotope partitioning ex-periment suggest that this isomerization is an important step in the catalytic mechanism of yGCN5, resulting in a fully ordered mechanism (Scheme 1). Evidence for this isomerization is provided by the D214A mutant of yGCN5. Asp-214 is located directly in the center of the loop portion of the conformationally flexible ␤5-loop-␣4 region of GCN5 and appears to make a number of stabilizing hydrogen bonding interactions with neighboring residues (Fig. 5). As the ␤5-loop-␣4 portion of GCN5 shifts to accommodate substrate binding, Asp-214 actually swings out of the active site, and the side chain makes a new hydrogen bonding interaction with the backbone amide nitrogen of the neighboring Asn-215, providing additional conformational stability.
Based on structural data and our current biochemical investigation, mutation of Asp-214 to an alanine impedes the ability of the ␤5-loop-␣4 region of yGCN5 to isomerize toward a stable conformation that can effectively bind peptide. This conclusion is further corroborated by the results of the kinetic analysis performed on the D214A mutant. We observed an ϳ200-fold decrease in k cat and an ϳ150-fold decrease in k cat /K m , H3 both of which contain k 3 (isomerization) and k 7 (chemistry) rate constants (Scheme 1). We also observed an ϳ340-fold decrease in the k cat /K m , AcCoA , which reflects AcCoA binding (k 2 /k 1 ) and k 3 . Since AcCoA affinity (k 2 /k 1 ) has not been affected in the D214A mutant (Table II), and because changes in the steady state kinetic parameters were independent of pH in comparison to wild type yGCN5, this implies that k 3 must be the perturbed step. It should be duly noted, however, that changes in k 3 do not preclude the possibility that mutation of Asp-214 also influences the rate of other steps in the mechanism, such as peptide binding or chemistry. Pre-steady state reaction kinetics were required to probe this possibility.
Pre-steady State Reaction Kinetics at pH 7.5-In order to elucidate further the function of specific amino acids in the catalytic mechanism of GCN5, pre-steady state experiments were used to determine whether substrate binding, chemical catalysis, or product release is rate-limiting for yGCN5 and its various mutated forms. Two forms of pre-steady state experiments were employed as follows: single turnover experiments ([enzyme] Ͼ [substrate]) and multiple turnover experiments ([enzyme] Ͻ [substrate]). Under single turnover conditions, the first-order exponential rate of product formation yields the rate of chemistry when substrate binding is not rate-limiting. Therefore, if the observed rate of product formation is equivalent to k cat , then chemistry is rate-limiting relative to substrate binding. If a pronounced lag exists in the initial stage of the reaction, causing multiphasic kinetics, then substrate binding or an enzyme isomerization step may be rate-limiting. Under multiple turnover conditions, if substrate binding or an enzyme isomerization is rate-limiting, a lag phase should again be evident in the kinetic trace, followed by a steady state linear phase. If chemistry is rate-limiting, then the rapid kinetic trace should be linear. If release of product is rate-limiting, an exponential burst phase should begin the kinetic trace followed by a linear steady state phase.
A quench flow apparatus was used to determine the ratelimiting step in the catalytic mechanism of wild type yGCN5 at pH 7.5. Under single turnover conditions, wild type yGCN5 (20 M) was rapidly mixed with a fixed concentration of H3 peptide (200 M) and substoichiometric amounts of [ 3 H]AcCoA (0.5 M). The reaction was quenched with 2 N HCl at various reaction times, and the concentration of [ 3 H]acetyl-H3 was determined as described under "Experimental Procedures." The concentration of product was plotted versus time and was fitted to a first-order exponential (Fig. 6) with an observed rate constant of 0.44 Ϯ 0.08 s Ϫ1 . This value is in excellent agreement with the FIG. 4. Isotope trapping experiment with wild type yGCN5. The percent of 0.75 M total [ 3 H]AcCoA partitioned to product during the first turnover of 15 M wild type yGCN5 relative to the amount that dissociates and is diluted into 2.5 mM unlabeled AcCoA was determined for 0 -1.0 mM histone H3 peptide at 24 Ϯ 1°C, in 5 mM DTT, 50 mM Tris, pH 7.5 (closed diamonds) and 9.5 (data not displayed). The data were plotted and fit to Equation 6. The experiments were performed in duplicate with a representative plot displayed. steady state turnover rate of 0.47 Ϯ 0.05 s Ϫ1 obtained under equivalent conditions (Table III). Similar to human PCAF (21,23), this suggests that chemistry (k 7 ) is the rate-limiting step in the reaction for wild type yGCN5.
In order to provide additional evidence for rate-limiting chemistry, multiple turnover experiments were also performed at pH 7.5 (Fig. 6). Wild type yGCN5 (5-10 M) was rapidly reacted with a fixed concentration of H3 peptide (200 M) at saturating concentrations of [ 3 H]AcCoA, and the formation of acetyl-H3 was determined as before. Over the course of 30 ms to 2 s, the data produced a linear kinetic trace starting at the origin, with no apparent burst or lag phase. The rate constant determined from the slope of the line (0.41 Ϯ 0.02 s Ϫ1 ) is entirely consistent with the values obtained under single turnover and steady state conditions (Table III), suggesting that for wild type yGCN5, chemistry (k 7 ) is rate-limiting relative to product release steps.
Similar experiments were performed to elucidate the ratelimiting step for the E173Q, D214A, D214N, and H145A mutated forms of yGCN5 (Fig. 6). Under single turnover conditions, the mutated forms of yGCN5 were rapidly reacted with a fixed concentration of H3 peptide (500 M) and substoichiometric amounts of [ 3 H]AcCoA (0.5 M) for varied amounts of time. Enzyme concentrations were varied (20 -50 M) to ensure that the experiments were performed under single turnover conditions (i.e. that all of the AcCoA is initially bound) such that increased enzyme concentrations did not cause an increase in the observed first-order rate constant. Multiple turnover experiments were also performed for the mutated forms of yGCN5 (5-10 M) using 500 M H3 peptide at saturating concentrations of [ 3 H]AcCoA. As with wild type yGCN5, the concentration of product was plotted versus time for both types of experiment. For all of the mutated forms of yGCN5, single turnover experiments produced first-order exponential traces, and multiple turnover experiments generated linear traces starting at the origin, with no apparent burst or lag phase (Fig.  6). In addition, the rate constants obtained from these experiments correlated well with the rate constants obtained under steady state conditions (Table III). As with wild type yGCN5, the results suggest that chemistry (k 7 ) is rate-limiting for the E173Q, D214A, D214N, and H145A mutants. It should be noted, however, that we cannot rule out the possibility that another step may partially limit turnover. In this case, a small lag would not be discernible with this approach.
Although substitutions at Asp-214 clearly have deleterious effects on the enzyme isomerization step (k 3 ), the pre-steady  (25). The reorientation of the loop upon AcCoA binding slightly pulls the conserved Asp-214 out of the active site, priming the enzyme for histone H3 peptide binding. E, upon histone H3 binding, the side chain of Asp-214 vacates the active site to make room for the substrate. The backbone amide acyl oxygen of Asp-214 remains in position to hydrogen-bond with the backbone amide nitrogens of residues 216 and 217. However, the side chain of Asp-214 swings out to make a new hydrogen bonding interaction with the backbone amide nitrogen of the neighboring Asn-215 (Asn-163 in tGCN5), thereby conferring additional stability. M Ϫ1 ⅐s Ϫ1 . The pH profile of the H145A mutant, however, indicated only a single critical ionization with a pK a value of 9.1 Ϯ 0.1, and a pH-independent k cat /K m , H3 value of 1.0 Ϯ 0.1 ϫ 10 4 M Ϫ1 ⅐s Ϫ1 . The data suggest that mutating His-145 to an alanine results in a loss of the critical ionization with a pK a value of 5.8 and a concurrent shift of pK a1 from 8.3 Ϯ 0.1 to 9.1 Ϯ 0.1. The shift in the pK a1 value from 8.3 to 9.1 suggests that His-145 may be responsible for effectively lowering the pK a of Glu-173 in the active site of wild type yGCN5. In addition, the pHindependent value for H145A is approximately 2-fold lower than the pH-independent value for wild type yGCN5 which is,  Table III. within error, the difference between the k cat /K m , H3 values obtained for wild type yGCN5 and H145A at pH 7.5 (Table I).
Due to the increased approximate K m , H3 value of H145A, obtaining a full k cat pH profile was not plausible. Instead, a single turnover pre-steady state control experiment was performed at pH 9.5 for both wild type yGCN5 and the H145A mutant (Fig. 8), in order to verify that the plateau in the k cat /K m , H3 pH profile (Fig. 7) was not a result of changing the rate-limiting step at high pH. The single turnover experiment was performed at pH 9.5 by mixing either wild type yGCN5 (20 M) or the H145A mutated form (20 M) with a fixed concentration of H3 peptide (50 M) and substoichiometric amounts of [ 3 H]AcCoA (0.5 M). As in the previous single turnover reactions performed at pH 7.5, the kinetic trace did not exhibit a lag phase, and the observed first-order exponential rate constants for acetyl-H3 peptide formation (0.97 Ϯ 0.05 and 0.43 Ϯ 0.02 s Ϫ1 for wild type yGCN5 and the H145A mutant, respectively) correlated well with rate values obtained under equivalent steady state conditions (1.01 Ϯ 0.08 and 0.42 Ϯ 0.08 s Ϫ1 for wild type yGCN5 and the H145A mutant, respectively) (Table  III). These data indicate that the observed pK a shift in the pH profile reflects an intrinsic change in the pK a of Glu-173 and not an apparent shift due to a change in the rate-limiting step for H145A catalysis. At pH 7.5, it appeared that the k cat /K m , AcCoA for the H145A mutant was ϳ17-fold less than that for wild type yGCN5. In order to assess whether this change was a result of the pK a shift, was caused by an inherent decrease in AcCoA binding, or was due to a change in the rate of isomerization (as with the D214A mutant), a pH profile was obtained for k cat /K m , AcCoA at near-saturating peptide concentrations over the pH range 6.5-9.5 (Fig. 9). The pH rate profiles for both wild type yGCN5 and the H145A mutant exhibited pH dependence for k cat /K m , AcCoA . Wild type yGCN5 demonstrated a critical residue with a pK a of 8.2 Ϯ 0.1 and a pH-independent k cat /K m , AcCoA value of 2.8 Ϯ 0.4 ϫ 10 6 M Ϫ1 ⅐s Ϫ1 . The H145A mutant demonstrated a critical residue with a pK a of 8.9 Ϯ 0.1 and a pH-independent k cat / K m , AcCoA value of 2.3 Ϯ 0.1 ϫ 10 6 M Ϫ1 ⅐s Ϫ1 . The pK a values obtained from the k cat /K m , AcCoA pH profiles of wild type yGCN5 and the H145A mutated form are equivalent, within error, to their respective pK a values obtained in the k cat /K m , H3 pH profiles. In addition, the pH-independent value for k cat /K m , AcCoA with H145A is approximately equivalent to the pH-independent value for wild type yGCN5, suggesting that at higher pH,   Table III. tained for wild type enzyme. Therefore, these residues play no significant functional role in AcCoA binding. These data also indicated that these substitutions had no gross effect on protein structure, allowing unencumbered interpretation of alterations on other steps on the reaction pathway. The E173Q mutant demonstrated ϳ500 -600-fold decreases in k cat and k cat /K m , H3 , consistent with Glu-173 acting as the general base catalyst as proposed previously (20). No significant effect was observed on either substrate binding steps. His-145 was identified as a residue within the peptide-binding cleft that must be unprotonated (pK a ϭ 5.8) for peptide binding and that likely hydrogenbonds to the Ser-10 hydroxyl of histone H3. His-145 also contributes to lowering the pK a value (by 0.8 units) of general base Glu-173 through a water-mediated hydrogen bond to the carboxylate side chain. Detailed analysis of the D214A mutant revealed an obligate protein isomerization step that occurs after AcCoA binding and permits efficient peptide binding, explaining the ordered addition of substrates (20 -23) and the inability to bind peptide efficiently in the absence of coenzyme (21). Asp-214 is located within a conformationally flexible loop that mediates the isomerization of distinct enzyme conformers.