The anomalous pKa of Tyr-9 in glutathione S-transferase A1-1 catalyzes product release.

The pKa of the catalytic Tyr-9 in glutathione S-transferase (GST) A1-1 is lowered from 10.3 to approximately 8.1 in the apoenzyme and approximately 9.0 with a GSH conjugate bound at the active site. However, a clear functional role for the unusual Tyr-9 pKa has not been elucidated. GSTA1-1 also includes a dynamic C terminus that undergoes a ligand-dependent disorder-to-order transition. Previous studies suggest a functional link between Tyr-9 ionization and C-terminal dynamics. Here we directly probe the role of Tyr-9 ionization in ligand binding and C-terminal conformation. An engineered mutant of rGSTA1-1, W21F/F222W, which contains a single Trp at the C terminus, was used as a fluorescent reporter of pH-dependent C-terminal dynamics. This mutant exhibited a pH-dependent change in Trp-222 emission properties consistent with changes in C-terminal solvation or conformation. The apparent pKa values for the conformational transition were 7.9 +/- 0.1 and 9.3 +/- 0.1 for the apoenzyme and ligand-bound enzyme, respectively, in excellent agreement with the pKa for Tyr-9 in these states. The Y9F/W21F/F222W mutant, however, exhibited no such pH-dependent changes. Time-resolved fluorescence anisotropy studies revealed a ligand-dependent, Tyr-9-dependent, change in the order parameter of Trp-222. However, no pH dependence was observed. In equilibrium and pre-steady-state ligand binding studies, product conjugate had a decreased equilibrium binding affinity (KD), concomitant with increased binding and dissociation rates, at higher pH values. Furthermore, the recovered pKa values for the pH-dependent microscopic rate constants ranged from 7.7 to 8.4, also in agreement with the pKa of Tyr-9. In contrast, the Y9F/W21F/F222W mutant had no pH-dependent transition in KD or rate constants for ligand binding or dissociation. The combined results indicate that the macroscopic populations of "open" and "closed" states of the C terminus are not determined solely by the ionization state of Tyr-9. However, the rates of transition between these states are faster for the ionized Tyr-9. The ionized Tyr-9 states provide a parallel pathway for product dissociation, which is kinetically and thermodynamically favored. In silico kinetic models further support the functional role for the parallel dissociation pathway provided by ionized Tyr-9.

A structural feature common to nearly all cytosolic GSTs is a conserved hydrogen bond between the hydroxyl group of catalytic Tyr or Ser and the thiol group of GSH. This interaction contributes to the lowering of the GSH pK a from ϳ9.3 in solution to 6.5-7.4 at the active site of various GSTs. In the A-class GSTs, the catalytic Tyr has an unusually low pK a (9,10). The spectroscopic pK a values of Tyr-9 in the human and rat A1-1 isoforms are 8.1 and 8.3, respectively, compared with 10.3 for Tyr in solution. In the GST A4-4 isoform, the pK a of the corresponding residue is even lower (11). Upon discovery of the anomalous pK a of Tyr-9, a general base mechanism was considered for the A-class GSTs, wherein Tyr abstracts a proton from GSH prior to attack of the resulting GS Ϫ at the electrophiles (9). However, solvent deuterium studies indicate that the catalytic Tyr in GSTM1-1, which has a "normal" pK a , does not undergo a general base catalysis (12). Subsequently, kinetic studies on hGSTA1-1 substituted with fluorotyrosine-9 analogs suggested that the pK a of Tyr-9 in the native enzyme is not sufficiently low to support general base catalysis (13). Thus, a functional role of the anomalous pK a of catalytic Tyr-9 in Aclass GST has not been identified.
A structural feature unique to GSTA1-1 is the presence of a dynamic C-terminal helix (residues 208 -222) that covers the active site upon binding of ligand at the H-site or both sub-sites of the enzyme (Fig. 1). As described previously (14 -16) for hGSTA1-1, wild-type rGSTA1-1, and several mutants, the binding of product GSH conjugate (GS-EA) follows a biphasic mechanism, which includes a bimolecular docking step (k 1 and k Ϫ1 ) and a slower unimolecular isomerization step (k 2 and k Ϫ2 ) that corresponds to the ordering and closure of the C terminus ( Fig. 1). Indirect evidence from several laboratories suggests an intimate link between the ionization state of Tyr-9 and the dynamics of the C terminus (14,15,17,18). One possible implication is that ligands will preferentially bind to or disso-ciate from the ionized state of the enzyme. Although substitution of Tyr-9 with Phe has demonstrated the importance of this residue in ligand binding, there is no direct evidence that has addressed whether the ionization state of Tyr-9 alters ligand binding kinetics. Similarly, no studies have probed the mechanism of "communication" between Tyr-9 and the C terminus. The objectives of the present studies were to determine directly whether the local conformational status of the C terminus is in fact controlled by the ionization properties of Tyr-9, and whether the extent and rate of ligand binding and release are dependent on the ionization state of Tyr-9. Our results indicate that the rates of the C-terminal disorder-to-order transition are faster when Tyr-9 is ionized, and equilibria for C-terminal opening and release of ligand are modestly shifted toward the dissociated state. Therefore, these results provide a clear, albeit modest, functional role for the unusual pK a of Tyr-9.

MATERIALS AND METHODS
Chemicals and Instrumentation-GSH, EA, MES, HEPES, CAPSO, and GSO 3 Ϫ were obtained from Sigma. GS-EA was synthesized and its purity analyzed as described previously (14). Stopped-flow experiments were performed with BioLogic SFM/QFM fluorimeter. Steady-state fluorescence was performed with Aminco SLM-8100 spectrofluorimeter.
Time-correlated Single Photon Counting Anisotropy Decay-A coherent 701 rhodamine 6 G dye laser is pumped with about 1 watt of 532 nm radiation from an Antares 76-s CW mode-locked Nd:YAG laser. The 701 dye laser is cavity-dumped at 3.8 MHz. The pulses have an autocorrelation of about 10 ps frequency width at half-maximum. Excitation of tryptophan at 295 nm was achieved by focusing the dye laser pulses with a 5-cm lens onto a crystal of potassium diphosphate. Fluorescence is collected at right angles through a polarizer mounted at 54.7°to the excitation polarization and then passed through an ISA H-10 monochromator with a 160nm bandpass or through cut-off filters. A Hamamatsu 2809u microchannel plate, amplified by a Minicircuits ZHL-1042J, and a FFD 100 EG & G photodiode provide the start and stop signals, respectively. Constant-fraction discrimination of these signals is performed by a Tennelec TC 455 and time-to-amplitude conversion by an ORTEC 457. Data are stored in a Norland 5500 multichannel analyzer before transfer to and analysis with a PC. The instrument function of this system has a frequency width at halfmaximum of 100 -150 ps. The polarized fluorescence traces used to obtain fluorescence anisotropy decay parameters are collected to a maximum of 16,000 counts in the peak channel. Samples were 5 M protein, at the indicated pH, in 35 mM MES, 35 mM Tris, 35 mM CAPSO, with either no ligand or 50 M S-hexyl-GSH.
Global Data Fitting-Time-resolved fluorescence data, acquired at various wavelengths, were subjected to a global fitting procedure found in Spectra Solve TM . A local 2 is calculated for each local fit. After all curves are fit (typically 6 -11), a global 2 is calculated. The global parameters are varied, and the whole process is repeated to calculate a new global 2 . This process is repeated until a minimum is reached for 2 . The quality of the fit is determined by visual inspection and residuals.
Site-directed Mutagenesis-The construction of W21F/F222W rGSTA1-1 has been described previously (14). The triple mutant Y9F/W21F/F222W was constructed by subcloning the available W21F/F222W insert, which was acquired from BglII and SalI digests, into a Y9F/W21F plasmid that had been linearized with a similar set of restriction enzymes. Originally, the single and double Y9F mutants were generated by PCR-based amplification of a fragment spanning the EcoRI and BglII restriction sites contained in the linearized pKKGTB34 or pKKGTB34-W21F plasmid (19). All mutants employed in the present studies were validated by DNA sequencing.
Protein Expression and Purification-The expression and purification of wild-type rGSTA1-1 and its various mutants have been described previously (20). The purity of each protein was assessed by SDS-PAGE. Enzymatic activity of each mutant protein was assessed as described previously (21).
Determination of Conformational Transition in GSTA1-1-Emission spectra were monitored at excitation wavelength of 295 nm, with slit widths of both monochromators set to 4 nm. In contrast to the previously employed tyrosine titration method (9), the emission properties of tryptophan were exploited to determine the nature of the surrounding environment of this residue. The pH dependence of Trp emission was monitored in 2 M GST prepared in mixed buffer of 35 mM MES, 35 mM Tris, 35 mM CAPSO at 25°C. The pK a values were determined by fitting the spectral center of mass, emission peak intensity, relative intensity at 305 nm, and relative intensity at 345 nm versus pH to an equation describing single residue ionization via Enzfitter. Fluorescence titrations were also performed in the presence of 200 M S-hexylglutathione.
Equilibrium Binding Studies-The equilibrium dissociation constants (K D ) for the Michael addition product GS-EA were determined by the decrease in the fluorescence of W21F/F222W, Y9F/W21F/F222W, and ⌬209 -222 upon ligand binding. During the titration experiments, serial dilutions of equilibrated [GST⅐GS-EA] complex, consisting initially of 0.5 M GST and 100 M GS-EA in mixed buffer system at pH 6.5 and 9, were made. The steady-state emission of each protein at 25°C was monitored at an excitation wavelength of 295 nm, using slit widths of 4 nm. The observed fluorescence intensity at each dilution was corrected for the emission intensity of the reference buffer. The mole fraction of GS-EA bound to total protein was determined from a ratio of change in protein intensity when a ligand is bound and emission intensity of free enzyme. This fraction is further corrected for the dilution factor. K D values were then derived from Equation 1 via DeltaGraph 4, where is the mole fraction of GS-EA bound to protein, and P and C are the upper plateau and lower plateau, respectively, of the versus [GS-EA] free plot. Ligand Binding and Dissociation Kinetics-Stopped-flow fluorescence studies were performed to determine the rates of ligand binding and dissociation as described in previous studies (14 -16, 20). The rate constants for the binding reaction were measured by monitoring the decrease in protein fluorescence after rapid mixing of an equal volume of 2 M GST and 6.25-100 M GS-EA at 15°C and at various pH values. The multicomponent buffer used in these pH-dependent studies was 35 FIG. 1. Location of the C terminus in GSTA1-1 and ligand binding mechanism. The dimeric structure of [GSTA1-1⅐GS-EA] is shown with the C-terminal helix colored blue. The GS portion of the conjugate is green, and the ethacrynic acid portion is red. With product GSH conjugates bound, this helix is ordered and covers the active site. In the absence of ligand, this helix becomes crystallographically invisible or disordered. The binding of GSH conjugate GS-EA follows a biphasic mechanism, with the slow step corresponding to closure of the C terminus. The protein structure is from Ref. 10 where a 1 and a 2 are the amplitudes of the two observed rate constants k obs1 and k obs2 ; t is the time; and C is the offset. In cases where two rate constants were experimentally observed, the rapid association and dissociation rates, k 1 and k Ϫ1(calc) , were fitted to Equation 4 that describes the linear dependence of k obs1 on GS-EA concentration, It should be noted that the k Ϫ1(calc) values derived from the binding experiment and the above equation are different from k Ϫ1(exp) values that were directly determined from the dissociation experiments. The association and dissociation rates of the slower step of the binding reaction, k 2 and k Ϫ2(calc) , were determined from a hyperbolic dependence (Equation 5) of k obs2 on GS-EA concentration, Alternatively, a single observed rate, k obs , could be plotted against GS-EA concentration to yield a hyperbolic curve fit (Equation 6), The dissociation of product from the [GST⅐GS-EA] complex was measured by monitoring the increase in fluorescence intensity after rapid mixing of equal volume of 2 M complex and 4 mM of the trapping agent, GS-sulfonate (GSO 3 Ϫ ). The dissociation rate constants, k Ϫ1(exp) and k Ϫ2(exp) , were measured from a single (Equation 1) or double (Equation 2) exponential fit to the raw data.
Kinetic Simulations-Simulations were performed with Kinecyte, using the following expressions (Equation 7) for the time-dependent concentration of relevant species. The specific rate constants and initial concentrations used are indicated under "Results."

Characterization of C-terminal Dynamics by Fluorescence-
Our previous studies on the anomalous pK a value of catalytic Tyr-9 in rGSTA1-1 suggest that the ionization state of the tyrosyl hydroxyl group influences the dynamics of the C terminus (20,18). In human and rat isoforms, the only tryptophan residue in the native GSTA1-1, Trp-21, near the domain-domain interface serves as a probe of global microconformational changes that occur upon ligand binding (22,23). However, to investigate directly the localized structural changes at the C terminus, a Trp residue was engineered at the ultimate position of rGSTA1-1. The kinetics of ligand binding and dissociation for this mutant, W21F/F222W, are comparable with the true wild-type, indicating that global changes in protein con-formation are coupled to local conformational changes at the C terminus (16). Thus, the W21F/F222W serves as a faithful mimic of ligand binding to the wild-type.
In order to determine the relationship between ionization state of Tyr-9 and C-terminal dynamics, here the W21F/F222W mutant was titrated between pH 7 and 10 to monitor the fluorescence emission properties of the terminal Trp. The emission of Trp, which absorbs at 275-295 nm and emits at 340 -353 nm, is very sensitive to its local environment and is often used as a direct probe of conformational changes in proteins (23). The changes in the emission spectra of Trp-222 in the site-directed mutants reflect the polarity of the surrounding environment. Fig. 2 summarizes the typical emission spectra for the W21F/F222W and Y9F/W21F/F222W at several pH values. The emission intensity decreases, and the spectrum is red-shifted up to 7 nm with increasing pH in the W21F/F222W mutant. In contrast, the emission intensity of the Y9F/W21F/ F222W mutant remained relatively constant, except at the very high pH values, at which the maximal spectral red shift was ϳ2 nm. Clearly, W21F/F222W exhibits a large pH-dependent change in environment only when Tyr-9 is present. From the fluorescence titrations employed, the pK a values for this transition could be determined by fitting the spectral peak intensity, relative intensity at 305 nm, or relative intensity at 345 nm against pH via single ionization model. The average pK a values recovered from all curve fittings based on intensity change are 7.9 and 10.5 for W21F/F222W and Y9F/W21F/ F222W, respectively (Table I).
A more sensitive measurement of polarity at the active site of the enzyme is the pH dependence of spectral center of mass, which closely resembles the trend observed in the pH dependence of peak intensity (Fig. 2). However, the spectral center of mass in W21F/F222W does not reach a plateau after pH 9.5. In fact, the center of mass continues to decrease even after pH 10. Presumably, at high pH, tyrosine residues in the protein become ionized and contribute to the additional spectral red shift as observed in the control Y9F/W21F/F222W mutant. Whereas tyrosine or tyrosinate would not contribute significantly to the  (Table I) are for the spectral transition based on peak intensity. change in intensity at the emission maximum of Trp-222, their spectral signals would contribute to the overall spectral center of mass. Therefore, an accurate pK a value for the conformational transition based on Trp-222 parameters could not be obtained. Although the spectral center of mass data could not be fitted to an equation yielding a pK a , they clearly demonstrate a change in Trp polarity in the pH range of 7.5-8.5, which does not occur in the triple mutant. In addition, the overall pK a recovered from the intensity data of W21F/F222W most likely reflects the pK a of active site Tyr-9. Although the catalytic Tyr in the triple mutant had been removed, the apparent pK a of 10.5 is probably due to seven other tyrosines in the protein that become ionized at very high pH.
The binding of S-hexylglutathione (S-hexyl-GSH) to GSTA1-1 has been shown to induce closure of the C terminus, with Tyr-9 pK a (ϳ9) remaining lower than the pK a of free tyrosine in solution (13,24,25). Here the emission spectra of W21F/F222W and Y9F/W21F/F222W were also obtained in the presence of saturating concentrations of S-hexyl-GSH. The Shexyl-GSH was used because it provides no interference with the emission in fluorescence measurements. If the pH-dependent changes observed with W21F/F222W correlate with ionization of Tyr-9 in the apoenzyme, and if ionization is a major determinant of helix dynamics, then spectral changes would still be observed in the presence of S-hexyl-GSH but at a higher pH. The recovered pK a values for the conformational transition (Table I) indicate an increase of 1.4 pK a unit for the ligandbound double mutant compared with the ligand-free double mutant, in excellent agreement with previously observed liganddependent increase in pK a of Tyr-9. The increase in Tyr-9 pK a of the conjugate-bound W21F/F222W suggests that, at equilibrium, only ϳ3% of the complex remains unprotonated at pH 7.5. Also, the triple mutant Y9F/W21F/F222W was titrated with ligand at each pH, in order to determine whether the observed pH dependence was related to Tyr-9 ionization. As expected for a mutant with no ionizable group at residue 9, the bound conjugate had little effect on the apparent pK a for the conformational transition of the triple mutant (Table I).
Together, these spectroscopic studies directly demonstrate for the first time that the local conformational equilibria of the C terminus are pH-dependent in addition to being ligand-dependent. Moreover, this pH dependence exhibits an apparent pK a equivalent to the pK a of Tyr-9 (ϳ8) in the apo form and pK a of ϳ9 for the ligand-bound form, and this dependence is eliminated upon removal of the Tyr-9 phenolic hydroxyl.
Time-resolved Anisotropy-The observed pH dependence in steady-state spectral properties of Trp-222 could be due to a pH-dependent shift in the macroscopic equilibrium between closed and open states of the C terminus. Alternatively, there could be a pH-dependent change in the characteristics of the "open" conformation or the "closed," without a significant change in the relative amount of each. In order to probe these possibilities, time-resolved anisotropy experiments were performed. The anisotropy would be expected to increase upon opening the C terminus to fully expose the C-terminal Trp-222. The anisotropy decay of W21F/F222W was measured for the ligand-free protein at two limiting pH values, 6.5 and 9, in the presence and absence of S-hexyl-GSH, as probes of the open and closed states. Identical experiments were done with the triple mutant in order to assess the contribution of Tyr-9 to any observed effects. The recovered rotational correlation times, and the order parameters, S 2 , are summarized in Table II. The data were analyzed globally according to a biexponential decay with a fixed rotational correlation time for the overall protein motion (20 ns). This was based on the assumption that pH or ligand would not change the overall structure sufficiently to alter global protein hydrodynamics. The recovered rotational relaxation time for the whole protein, 20 ns, is consistent with a protein of 48 kDa. In addition, for each sample a fast anisotropy decay component was detected with correlation times ranging from 0.8 to 1.8 ns, clearly reflecting local motion of the C terminus. The results clearly demonstrate the expected increase in the order of Trp-222 upon binding of S-hexyl-GSH, at both pH values. This provides a benchmark for the magnitude of the change in S 2 that is expected for closure of the C terminus. Furthermore, comparison of the double and triple mutants demonstrates that the Tyr-9 hydroxyl group is critical for maintaining the conformational integrity of the C terminus and the ligand-dependent switch. However, the moderate pH dependence observed for the ligand-free double mutant is observed also with the triple mutant. Based on these data, the ionization of Tyr-9 does not shift significantly the equilibrium between open and closed states, in contrast to the ligand, which does.
Rather, the open state, the closed state, or both are different in the ionized versus the un-ionized forms, yielding different steadystate emission spectra. However, these differences do not correspond to changes in the fast local motion of the C-terminal residue expected for closed versus open conformations.
Equilibrium Binding Affinity-Equilibrium binding studies were explored as an initial strategy to determine whether the pH-dependent microconformations of the C terminus differentially affected the binding of GSH conjugates. For several substrates, product release is rate-limiting for GSTA1-1 turnover, so the C-terminal conformational status is an important aspect of catalysis. These studies were performed at pH 6.5 and 9 at which the active site Tyr-9 in bound rGSTA1-1 (pK a ϳ9.3) is completely protonated and significantly unprotonated (ionized), respectively. The ligand employed in the equilibrium studies was GS-EA, which is a conjugate formed from ethacrynic acid and GSH via Michael addition, and for which we have previously characterized the binding mechanism in detail. Moreover, crystal structures demonstrate that the Cterminal helix is closed and ordered with GS-EA bound. To prevent the proteins from unfolding and the GSH conjugate from hydrolyzing to its parent compounds at highly basic conditions, equilibrium binding studies beyond pH 9 were not pursued. Throughout our studies, the initial stock of GS-EA was kept on ice until use to avoid hydrolysis, which was monitored in a separate experiment via high performance liquid chromatography (data not shown).
The maximal emission intensities of W21F/F222W and Y9F/ W21F/F222W at pH 6.5 were measured at 333 and 330 nm, respectively, at varying ratios of GS-EA:protein. In the presence of GS-EA at low pH, these emission maxima blue-shifted to 321 nm, as shown previously (25). At pH 9, only the W21F/ F222W mutant exhibited red-shifted emission spectra. Equilibrium ligand binding studies were also performed in the Cterminal truncation mutant, ⌬209 -222, for comparison. Typical binding isotherms, from which K D values were determined, are shown on Fig. 3. The recovered K D values are reported in Table III. At pH 6.5, the W21F/F222W mutant yielded a K D of 3.3 M, which is comparable with a previously determined K D of wild-type protein (14). The K D for GS-EA significantly increased by 2.5-fold at pH 9, indicating lower affinity at equilibrium when Tyr-9 is ionized. In previous studies, the C-terminal truncation mutant, ⌬209 -222, has been shown to have Tyr-9 pK a that is slightly lower than the wildtype (20). Moreover, with an open, solvent-accessible active site, GS-EA was found to bind to the ⌬209 -222 mutant (K D ϳ6.4 M) with 3-fold increased K D compared with wild-type (14). In the present studies, the measured K D for GS-EA in this mutant is 8.4 M at pH 6.5, which is 2.5-fold greater than W21F/F222W. Thus, physical truncation of the C terminus or increased pH lowers the ligand affinity to a similar extent. As expected for a mutant lacking the C terminus to restrict the access of solvent and ligand at the active site, the ⌬209 -222 mutant did not show a pH dependence of K D (Table III). Similarly, the Y9F/W21F/F222W mutant showed no apparent pH dependence, as K D values at pH 6.5 and 9 remained essentially unchanged. The lack of pH dependence of K D for GS-EA, along with spectroscopic pK a of 10.5 for the conformational transition, suggests that ionization of Tyr-9 contributes to the ligand affinity through some type of communication with the C terminus. This is amplified by the observed pH dependence of K D in the W21F/F222W mutant. However, when access to the active site is no longer limited by the C terminus, such as in the ⌬209 -222 mutant, facile ligand binding and release follow. As a result, K D is no longer dependent on pH in the absence of the C terminus.
Kinetics of Ligand Binding and Dissociation-A further direct approach to examine the functional role of ionized Tyr-9 in the dynamics of the C terminus is a study of the pH dependence of individual rate constants that determine ligand binding. Stopped-flow fluorescence was employed to measure the individual rate constants that compose the binding mechanism of GSTA1-1 (Fig. 1). As described previously, the binding of GS-EA to rGSTA1-1 follows a two-step process, which includes a bimolecular docking step and a slow isomerization step that corresponds to the ordering and closure of the C terminus. The binding and dissociation of GS-EA were monitored by the decrease and increase in protein fluorescence, respectively, as we have demonstrated previously (20, 14 -16). The rate constants for the association and dissociation of GS-EA were determined by fitting to appropriate equations as described under "Materials and Methods." As we have shown previously, the raw data for wild-type fit best to double exponential equation (Equation 3), whereas the raw data for Y9F fit best to single exponential equation (Equation 2). Notably, the approach to equilibrium, described by k 2 and k Ϫ2 , is dramatically slower in the Y9F-containing mutant. The docking step could not be experimentally observed, presumably due to a fast k 1 . Fig. 4 depicts the dependence of observed rate constants (k obs ) on GS-EA concentration in wildtype and Y9F at pH 6 and 9. Although only a single rate constant could be obtained from the raw data, the concentration dependence of k obs in the Y9F mutant yielded a hyperbolic curve, consistent with a two-step mechanism. Moreover, the rates of ligand binding were significantly increased at higher pH in the wild type. In contrast, ligand binding at the extreme pH values, 6 versus 9, is essentially identical in the Y9F mutant. As with the spectral changes for Trp-222 and the equilibrium ligand binding titration, the pH dependence observed in the double mutant is lost in the triple mutant. Apparently, Tyr-9 controls this pH dependence.
Upon discovery of the pH dependence of GSH-conjugate binding and dissociation rates for the wild type, other stoppedflow experiments were undertaken to determine whether apparent pK a values could be extracted from the recovered rate constants as a function of pH. The rate constants for the docking and isomerization steps of the binding reaction (Table IV) showed a pH dependence and fit well to a single ionization model, with recovered pK a values ranging from 8.2 to 8.8 (Fig.  5). From this set of rate constants, calculated equilibrium dissociation constants (K D(calc) ) could be obtained. Of course, the calculated K D values are pH-dependent, with an apparent pK a of 7.7 (Fig. 6). Apparently, all of the recovered pK a values (Table IV) reflect the ionization of Tyr-9, further suggesting that ligand docking and dissociation are favored in the open conformation of the enzyme. Due to difficulties in obtaining significant amounts of the Y9F proteins, which bind poorly to GSH affinity column, combined with the need for high protein concentrations in stopped-flow experiments, the kinetics of li-TABLE II Anisotropy decay of Trp-222 1,r and 2,r are the rotational correlation times with fractional "zero time" limiting anisotropy values of r 1 (0) and r 2 (0), respectively. The anisotropy at time t is r(t) ϭ r 1 (0)e (Ϫt/r1) ϩ r 2 (0)e (Ϫt/r2) , 1 and 2 are the rotational correlation times of the individual components, and S 2 is the order parameter, defined as S 2 ϭ r 2 (0)/(r 1 (0) ϩ r 2 (0)).  (Table III) were determined from nonlinear regression fits to mole fraction of GS-EA bound to protein versus concentration of free GS-EA via Equation 1 ("Materials and Methods").  (Table IV). The apparent calculated K D values for GS-EA in Y9F were not significantly different at any given pH (Fig. 6). Clearly, there is no indication that binding and dissociation of GS-EA in Y9F mutant depends on the ionization of a protein residue in this pH range. It is important to note that the recovered microscopic rate constants at pH 9 are lower limits for the true rate constants when Tyr-9 is fully ionized. At pH 9, the Tyr-9 is ϳ50% ionized, based on the experimentally measured pK a with GSH-conjugate bound. Presumably, if higher pH values were experimentally accessible, higher rates would have been observed, reflecting the true rates for Tyr-9 in the completely unprotonated form.
Kinetic Simulations-Because the extent of ionization is low for the ligand-bound species, it is reasonable to consider whether the ionized state could have any effect on the rate of ligand binding and dissociation. What magnitude of rate change could be obtained with such a low extent of ionization? In order to relate the pH dependence of the experimentally determined microscopic rate constants to steady-state turnover of the enzyme, kinetic models were constructed, and simulations were performed. The kinetic scheme shown in Fig. 1 was used as a template but was expanded to include a parallel pathway of ligand binding and C-terminal ordering, via the state with unprotonated Tyr-9. The expanded model is shown in Fig. 7 and is oriented such that the reaction starts with the bound state, with the C terminus closed, [Tyr-9-OH⅐L] closed . The C terminus opens and closes, for the protonated state, with rate constants k Ϫ2 and k 2 , as above. For the unprotonated state the analogous rate constants are k Ϫ2 Ј and k 2 Ј. Similarly, the ligand egress and association rate constants are k Ϫ1 and k 1 for the protonated state and k Ϫ1 Ј and k 1 Ј for the unprotonated state. The rate constants in Table IV for the wild-type enzyme at pH 6 were used directly to model the protonated pathway. The rate equations describing the concentration of each species are provided under "Materials and Methods." For the unprotonated pathway, each of the rate constants in Table IV for pH 9 was doubled, in order to mimic the expected increase in rate constant at full ionization; based on the pK a of 9 in the ligandbound state, the experimental values are lower limits, as described above. The equilibria defined by each pK a value were modeled with a ratio of rate constants that yielded the appropriate ratio of protonated to unprotonated species, at equilibrium pH 7.4, i.e. 3% unprotonated for the closed, ligand-bound species ([Tyr-9-O Ϫ ⅐L] closed ; pK a ϭ 9. The intent of these simulations is not to quantitatively reproduce the experimental results, but rather to qualitatively model the potential effect of Tyr-9 ionization on the rates of ligand dissociation. Unlike the actual experiments, the in silico stopped-flow experiment contains no trapping agent (GSO 3 Ϫ ). Still, the time scale for approach to equilibrium in silico, with a half-time of ϳ3 s, is in good agreement with the present and past experimental results (14). Fig. 7A shows the simulated results for decay of the starting species, [Tyr-9-OH⅐L] closed , for the simple model with no Tyr-9 ionization (dashed curve), and with the parallel pathway available through the unprotonated species (solid curve). As expected, the decay is faster for the later case. The half-life for the process is 2.4-fold shorter when the unprotonated pathway is available. The results in Fig. 7B  -OH⅐L] open . At its peak concentration, there is slightly less of this intermediate when the parallel unprotonated, pathway is available (solid curve) than when it is excluded (dashed curve). This is expected, because the parallel pathway competitively removes some of the available enzyme. Also, the rate of decay of this intermediate is faster when the parallel pathway is available. This is also expected as this intermediate has two pathways through which to decay, rather than one. Fig. 7C shows the build up of the ligand free open state for the protonated species. This behavior is interesting in as much as the rate of accumulation of ligand free [Tyr-9-OH] is faster at early times when the parallel pathway through the unprotonated states is available (solid curve). However, it is slower at later times and reaches a lower equilibrium concen- and single exponential Equation 2, respectively. The rate constants k obs1 and k obs2 for wild-type showed linear dependence (Equation 4) and hyperbolic dependence (Equation 5), respectively, on GS-EA concentration (R 2 ϭ 0.99 and R 2 ϭ 0.92). Meanwhile, the single rate constant for Y9F mutant showed hyperbolic dependence (Equation 6) on GS-EA concentration (R 2 ϭ 0.98). Note that at pH 6 (squares) and pH 9 (circles), the rate constants for GS-EA binding to wild-type rGSTA1-1 showed apparent dependence on pH, although the rate constants for the Y9F mutant were essentially identical at the extreme pH values. Kinetic parameters are reported in Table IV. tration, when the parallel pathway is available. This behavior makes sense and validates the accuracy of the model. The equilibrium concentration of [Tyr-9-OH] is reduced by 24% because of population of the unprotonated species, [Tyr-9-O Ϫ ]. When the simulation is run to 20 s, the difference in concentration is exactly 24% (not shown). The fact that the rate of appearance of unbound [Tyr-9-OH] is faster at early times reflects the faster release of ligand when the parallel pathway is available. In fact, Fig. 7D demonstrates this directly. Fig. 7D is the rate of appearance of free ligand. When the parallel pathway is available (solid curve), the free ligand accumulates faster. The half-life for approach to equilibrium is 2.4-fold faster (0.9 versus 2.1 s) when the unprotonated species contribute to the flux. These simulated results match reasonably well with the difference in experimental k obs at pH 6 versus pH 9, which differs by 4-fold. Based on these simulations, it is clear that even a minor extent of ionization of the initial ligand bound, closed, state can allow flux through the faster, unprotonated pathway with a modest increase in the rate of ligand dissociation.

DISCUSSION
As already noted, there are many indirect data that indicate a functional role for Tyr-9 in ligand binding, which is mediated through the C terminus. For example, a thermodynamic coupling between Tyr-9 ionization and C-terminal dynamics is indicated by our previous pressure studies (9) in which we initially demonstrated a pressure-dependent ionization of Tyr-9, which was sensitive to C-terminal mutations. Also, the rates of ligand binding and dissociation are dramatically affected in the Y9F mutant compared with wild-type (15-17, see below), further implicating a role for Tyr-9 in C-terminal dynamics. However, there has been no direct evidence regarding a The pK a values recovered from each observed rate constant and equilibrium constant as a function of pH. b K 1 is determined from the ratio of k 1 /k 1 . c K 2 is determined from the ratio k Ϫ2 /k 2 . d Equilibrium dissociation constants were determined from K D(calc) ϭ K 1 (k Ϫ2 /(k 2 ϩ k Ϫ2 )). e K 1 and k Ϫ2 were determined from Equation (6).
FIG. 5. pH dependence of observed rate constants for GS-EA binding (A and C) and dissociation (B and D) in rGSTA1-1. Solid lines are nonlinear regression fits to a single ionization model (Enzfitter). The recovered pK a values from each fit (Table III)  the functional significance of coupling between C-terminal conformational status and Tyr-9 ionization.
Although various approaches have been used to address the pH dependence of ligand binding to GSTs, the focus in those studies had been on the ionization state of GSH and how its binding and release were subsequently affected (26 -29). In those cases, the rates of ligand binding and release were dependent on whether or not the GSH at the active site of several GST isoforms was unprotonated. Apparently, the rates of GSH binding and release were faster at lower pH, where the ligand is mostly protonated. Importantly, the available crystal structures indicate that the C terminus of GSTA1-1 does not close upon binding of GSH in the absence of an H-site ligand (30), so any pH dependence observed in those studies would not address C-terminal dynamics.
In contrast, the emphasis of our studies is on the ionization state of active site Tyr-9 and how it controls the rates of binding and release of GSH conjugates, for which crystallographic models clearly indicate ligand-induced closure of the C terminus.
Because release of the product conjugate is rate-limiting for several substrates, these studies reveal new information about kinetically relevant steps of the reaction cycle. No previous studies have directly investigated the role of Tyr-9 pK a in the product dissociation mechanism.
To examine directly the relationship between the ionization state of Tyr-9 and the conformational status of the C terminus, here we have exploited four approaches as follows: 1) the pH dependence of conformational dynamics of the C terminus, as reported by an engineered Trp-222; 2) pH-dependent equilibrium binding studies; 3) stopped-flow studies in which binding and dissociation of ligand were directly measured at variable pH; and 4) kinetic simulations that mimic the dissociation of a GSH conjugate, based on experimentally derived parameters.
On the basis of emission properties of Trp-222, the local environments of the double and triple mutants were indeed different from each other. The decrease in the emission intensity and the red shift in the spectrum of W21F/F222W from pH 7 to 9.5 are indicative of polar, solvent-accessible environment at higher pH, or some type of conformational change. This pH-dependent emission profile was lost in the triple mutant containing Y9F. Moreover, the dependence of Trp-222 spectral intensity on pH yielded an apparent pK a of 7.9 in the W21F/ F222W, which is remarkably close to the pK a for the ionization of Tyr-9 at the active site. In contrast, the pH dependence of spectral intensity in the Y9F/W21F/F222W resulted in a pK a of 10.5, which likely reflects the ionization of other tyrosines in the protein at the extreme pH values. In fact, no appreciable Tyr ionization of analogous mutant (Y9F:W21F) was observed at pH Ͻ11 (9). Presumably, the C terminus of Y9F/W21F/ F222W mutant is significantly less solvated than W21F/F222W at pH values above 8.0, suggesting that the C terminus in the former likely adopts a different conformation. This conclusion is further supported by similar studies in the presence of ligand known to induce closure of the C terminus and alter the pK a of Tyr-9. The increase in Tyr-9 pK a of W21F/F222W, as a result of S-hexyl-GSH binding, is not only consistent with our previous studies but also lends support to our hypothesis that the ionization of Tyr-9 also occurs to a limited extent even with GSH conjugate bound. Indeed, the pH-dependent spectral properties of W21F/F222W in complex with S-hexyl-GSH are shifted to pK a 9.1, in agreement with the Tyr-9 pK a in this complex. Importantly, these are the first spectroscopic data that correlate local conformation of the C terminus with ionization of Tyr-9.
One possible scheme consistent with the steady-state spectral properties of the two mutants is that the C terminus is open and disordered when Tyr-9 is ionized, and it is closed and static when Tyr-9 is protonated, corresponding to the two macroconformations observed in crystal structures. However, the anisotropy decay results are not consistent with this scheme. If the open and closed states of the C terminus corresponded directly with the ionized and neutral forms of Tyr-9, then a much larger pH dependence of the fluorescence anisotropy order parameter would be expected. Obviously, the local dynamics of the ultimate residue in the C terminus are more sensitive to addition of ligand than to ionization of Tyr-9 (Table  II). Whereas the anisotropy results nicely confirm the liganddependent disorder-to-order transition observed in the crystal and results from NMR, the results also indicate that a simple two-state model, with only open and closed states, is insufficient. The anisotropy data also confirm the importance of the Tyr-9 hydroxyl group, in as much as the ligand-dependent effects are lost upon substitution of Tyr-9 with Phe.
In order to accommodate all of the data, we propose that there is a conformational distribution of open states and a Although only ϳ3% of the Tyr-9 is ionized in the ligand-bound state (left) at pH 7.5, the rates of C-terminal transitions (k 2 /k Ϫ2 and k 2 Ј/k Ϫ2 Ј) and product egress (k 1 /k Ϫ1 versus k 1 Ј/k Ϫ1 ) are faster for the ionized state. In particular, k Ϫ2 Ј is at least 10-fold faster than k Ϫ2 , and k Ϫ1 Ј is at least 4.5-fold greater than k Ϫ1 . In addition the equilibria defined by K 2 /k Ϫ2 versus k 2 Ј/k Ϫ Ј, and by k 1 /k Ϫ1 versus k 1 Ј/k Ϫ Ј lie further toward the open and ligand-free states for the ionized form. See text for details. The rate constants use for simulation are as follows: k 1 ϭ 3.08 M Ϫ1 s Ϫ1 ; k Ϫ1 ϭ 24.7 s Ϫ1 ; k 2 ϭ 16.4 s Ϫ1 ; k Ϫ2 ϭ 0.63 s Ϫ1 ; k 1 Ј ϭ 16.8 M Ϫ1 s Ϫ1 ; k Ϫ1 Ј ϭ 212 s Ϫ1 ; k 2 Ј ϭ 108.6 s Ϫ1 ; k Ϫ2 Ј ϭ 13.36 s Ϫ1 . distribution of closed states. This is supported by the crystal structures with different ligands bound. Different GSH conjugates, or the inhibitor ethacrynic acid in the absence of GSH, yield readily definable electron density for the C terminus corresponding to the closed macroconformation. However, the exact location of the C terminus is different in each case, defining several possible microconformations within the closed state. Presumably, there are more microconformations sampled in the open state than in the closed, but these experiments do not address that issue. Together, all of the fluorescence data suggest that, whereas the ionization state of Tyr-9 is not the major determinant of the distribution between open and close states (based on anisotropy), it does perturb the local conformational distribution of each state (based on emission properties and kinetic behavior). Alternatively, it is formally possible that even when the majority of the helix is in the closed state, the ultimate Trp residue remains isotropic and indistinguishable from the open state.
Regardless of these details, the equilibrium and pre-steadystate binding experiments indicate a modest, but clear, functional advantage for the ionized form of Tyr-9. To the extent that product release is rate-limiting for many substrates, retention of product conjugates is a kinetic "problem" for GSTA1-1. Furthermore, product inhibition is commonly observed as a result of high affinity of GSH conjugates for the enzyme. The equilibrium binding experiments reveal a modest, 3-4-fold, decrease in affinity of ligand as the Tyr-9 becomes unprotonated. Also, all of the individual rate constants describing the biphasic binding and dissociation reactions are increased upon deprotonation of Tyr-9. In particular, the microscopic rate constant k Ϫ2 , which corresponds to the first step of rate-limiting product dissociation, and which we propose corresponds to opening of the helix, is increased by 10-fold upon changing the pH from 6 to 9. Furthermore, although it is experimentally challenging to determine a residue pK a by stopped-flow, the variable pH experiments yield pK a values remarkably close to that of Tyr-9, and this pH dependence is lost in the triple mutant.
It is instructive to compare these results with a recent study by Armstrong and co-workers (31). They observe a similar modest, ϳ3-fold, increase in the k obs for product release of a mutant relative to wild-type GSTM1-1, based on burst phase analysis of substrate turnover. The increase in the rate of product release is correlated with a disruption of an intramolecular hydrogen bonding network that restricts segmental motion. Upon disruption of these interactions, product release is accelerated modestly. The similar magnitude in rate increase for product release upon alteration of conformational dynamics of the two proteins provides a benchmark for the extent to which diffusive steps are controlled by enzymes; perhaps, very large rate enhancements are inaccessible by fine-tuning ligand binding and dissociation when they are coupled to conformational changes.
The new information obtained from our studies clarifies a functional role for Tyr-9 ionization in GSTA1-1 catalysis. To the extent that the Tyr-9 pK a remains low (ϳ9.0) even with a GSH conjugate bound, a small fraction will be unprotonated even at physiological pH. Similarly, our previous results suggest a persistent disordered state, wherein even at saturating ligand and pH 7.4 the C-terminal helix is not completely closed and ordered. Thus, the equilibria involved yield small popula-tions of deprotonated or disordered helix even at saturating concentration of conjugate. Apparently, several microconformational states are nearly isoenergetic. Our results indicate that several microconformations compose the closed state, and deprotonation of Tyr-9 alters this ensemble in favor of microconformations that promote faster C-terminal opening and release of product. As long as the microconformational states corresponding to the protonated and unprotonated Tyr-9 are in rapid equilibrium, even a small population of the unprotonated form could provide a kinetically favored, parallel, pathway for product release, as depicted in Fig. 7. Based on this, the ionized form of Tyr-9 provides a modest catalytic advantage by lowering the kinetic barrier to C-terminal opening and product release, and by decreasing the extent of product inhibition.