Contribution of linear free energy relationships to isozyme- and pH-dependent substrate selectivity of glutathione S-transferases: Comparison of model studies and enzymatic reactions

Article abstract of DOI:10.1021/ja980816o

A novel application of linear free energy relationships is described in which the substrate selectivities and pH dependencies of glutathione S-transferases (GSTs) are correlated to the pK(a), of glutathione (GSH) at the active site. To determine whether t

Full text of DOI:10.1021/ja980816o

J. Am. Chem. Soc. 1998, 120, 6651-6660
6651
Contribution of Linear Free Energy Relationships to Isozyme- and
pH-Dependent Substrate Selectivity of Glutathione S-Transferases:
Comparison of Model Studies and Enzymatic Reactions
Brenda S. Nieslanik and William M. Atkins*
Department of Medicinal Chemistry, Box 357610, UniVersity of Washington,
Seattle, Washington 98195-7610
ReceiVed March 11, 1998
Abstract: A novel application of linear free energy relationships is described in which the substrate selectivities
and pH dependencies of glutathione S-transferases (GSTs) are correlated to the pKa of glutathione (GSH) at
the active site. To determine whether the variation in the thiol pKa of GSH at the active sites of GST isozymes
can contribute to their differential selectivity for electrophilic substrates, model studies were performed with
4-substituted thiobenzenes, with pKa values ranging from 4.5 to 7.5. Second-order rate constants were
determined for the specific base-catalyzed reaction of each thiol with a diverse range of GST electrophilic
substrates. Brønsted coefficients (ânuc) for these reactions in 10% DMF:90% H2O were determined for each
electrophile; ânuc ranged from 0.16 to 0.93. In 30% DMF:70% H2O, the ânuc values increased relative to 10%
DMF and ranged from 0.29 to 1.04. Numerical simulations demonstrate that these ranges of ânuc values along
with the isozyme-dependent variation in GSH pKa could account for a 7.5-fold difference in relative turnover
rates for GST catalysis of some electrophilic substrates. To challenge the predictions of this Brønsted analysis,
electrophiles for which chemical steps are rate limiting in enzyme turnover were used as a substrate in reactions
with a series of GSTA1-1 mutants with variable GSH pKa. ânuc values were determined to be 0.16 ( 0.05 for
cumene hydroperoxide (CHP) and 0.25 ( 0.06 for 1-chloro-2,4-dinitrobenzene, in excellent agreement with
the model studies. Furthermore, the dependence of the relative rates of CHP turnover on GSH pKa was well
correlated, at pH 6.5, 7.4, and 8.0 with the relative rates predicted by the Brønsted analysis. Thus, even for
a reaction characterized by a low ânuc value, variation of the pKa of enzyme-bound GSH leads to changes in
the intrinsic reactivity of the nucleophilic GS-, according to the Brønsted free energy relationship. In principle,
variation of the pKa of GSH may contribute to isozyme-dependent substrate selectivity.
The cytosolic glutathione S-transferases (GSTs)1 are a family
of detoxication enzymes that catalyze the conjugation of
glutathione (GSH) with various endogenous and xenobiotic
electrophiles. Due to their primary role in drug metabolism
and tumor drug resistance and their potential role in bioreme-
diation, GSTs have been the recent focus of intense mechanistic
and structural research.2 The mammalian cytosolic GSTs are
represented by five gene classes (A, P, M, K, and T) that exhibit
overlapping, but distinct, selectivities for structurally diverse
electrophilic substrates. High-resolution X-ray structures of
isozymes of each gene class, in various ligand states, have
facilitated structure/function comparisons between isozymes and
classes.3 Each of the cytosolic GSTs has a catalytic tyrosine
or serine at the active site that hydrogen bonds to the thiol of
GSH. In addition, spectroscopic and kinetic data yield a pKa
for enzyme-bound GSH of 6.5-7.4, in contrast to the pKa of
thiols in solution, 9.3.4 The catalytic advantage appears obvious
because the thiolate anion is a more reactive nucleophile than
the protonated thiol. In addition, it was appreciated long ago
that the selectivity for electrophilic substrates varies between
GST isozymes,5 and the available X-ray structures clearly
suggest that the active site topologies of GST isozymes are likely
to contribute to relative activities toward different substrate
electrophiles. In light of results recently obtained, however,
(3) (a) Armstrong, R N. Chem. Res. Toxicol. 1997, 10, 2. (b) Sinning,
P.; Kleywegt, G. J.; Cowan, S. W.; Reinemer, P.; Dirr, H. W.; Huber, R.;
Gilliland, G. L.; Armstrong, R. N.; Ji, X.; Board, P. G.; Olin, B.; Mannervik,
B.; Jones, T. A. J. Mol. Biol. 1993, 232, 192. (c) Wilce, M. C. J.; Board,
P. G.; Feil, S. C.; Parker, M. W. EMBO J. 1995, 14, 2133. (d) Ji, X.; Zhang,
P.; Armstrong, R. N.; Gilliland, G. L. Biochemistry 1992, 31, 10169. (e)
Bjornestedt, R.; Stenberg, G.; Widersten, M.; Board, P. G.; Sinning, I.;
Jones, T. A.; Mannervik, B. J. Mol. Biol. 1995, 247, 765.
(4) (a) Graminski, G. F.; Kubo, Y.; Armstrong, R. N. Biochemistry 1989,
28, 3562. (b) Kong, K. H.; Takasu, K.; Inoue, H.; Takahashi, K. Biochem.
Biophys. Res. Commun. 1992, 184, 194. (c) Dietze, E. C.; Ibarra, C.;
Dabrowski, M. J.; Bird, A.; Atkins, W. M. Biochemistry 1996, 35, 11938.
(d) Liu, S.; Zhang, P.; Ji, X.; Johnson, W. W.; Gilliland, G. L.; Armstrong,
R. N. J. Biol. Chem. 1992, 267, 4296. (e) Huskey, S.-E.; Huskey, W. P.;
Lu, A. Y. H. J. Am. Chem. Soc. 1991, 113, 2283. (f) Liu, S.; Ji, X.; Gilliland,
G. L.; Stevens, W. J.; Armstrong, R. N. J. Am. Chem. Soc. 1993, 115,
7910.
* Corresponding author phone: (206) 685-0379. Fax: (206)685-3252.
E-mail: winky@u.washington.edu.
(1) Abbreviations: CHP, cumene hydroperoxide; CDNB, 1-chloro-2,4-
dinitrobenzene; DMF, dimethylformamide; E, an electrophile reactant of
glutathione conjugation; EPNP, 1,2-epoxy-3-(p-nitrophenoxy)propane; EtOH,
ethanol; FAB, fast atom bombardment mass spectrometry; trans-PBO, trans-
4-phenyl-3-buten-2-one; GST, glutathione S-transferase; GSH, glutathione;
GS-, the thiolate anion of GSH; [GST‚GSH‚E], the ternary complex formed
from GST, GSH, and an electrophile; [GST‚GS-‚E], the ternary complex
of GST, GS-, and E; 2-NP, 2-nitropropane; MES, 2-(N-morpholino)-
ethanesulfonic acid; 1, 4-methoxybenzenethiol; 2, 4-methylbenzenethiol;
3, 4-hydroxybenzenethiol; 5, 4-chlorobenzenethiol; 6, 4-nitrobenzenethiol.
(2) (a) Rushmore, T. H.; Pickett, C. B. J. Biol. Chem. 1993, 268, 11475.
(b) Armstrong, R. N. AdV. Enzymol. Relat. Areas Mol. Biol. 1994, 69, 1.
(c) Hayes, J. D.; Pulford, D. J. CRC Biochem. Mol. Biol. 1995, 30, 445.
(5) (a) Habig, W. H.; Pabst, M. J.; Jakoby, W. B. J. Biol. Chem. 1974,
249, 7130. (b) Mannervik, B.; Danielson, U. H. CRC Crit. ReV. Biochem.
1988, 23, 283. (c) Mannervik, B. AdV. Enzymol. Mol. Biol. 1985, 57, 357.
S0002-7863(98)00816-6 CCC: $15.00 © 1998 American Chemical Society
Published on Web 06/23/1998

6652 J. Am. Chem. Soc., Vol. 120, No. 27, 1998
Nieslanik and Atkins
we examine here the possibility that differential reactivity of
GST isozymes toward different electrophiles, E, may result also
from different inherent nucleophilicity of the thiolate anion in
the ternary complex [GST‚GS-‚E], as predicted by Brønsted
behavior. This analysis yields a novel perspective of GST
substrate specificity and pH vs rate profiles based on a ‘classical’
free energy relationship.
Several investigators6 have pointed out that the hydrogen bond
between GS- and active site residues leaves the enzyme with a
paradox: according to Brønsted relationships, the thiolate that
is generated from GSH with a reduced pKa is predicted to be
less nucleophilic than the analogous thiolate generated from a
GSH with a ‘normal’ pKa. The Brønsted equation describing
nucleophilic reactivity (eq 1), which is a variation of the original
Brønsted theory for acid-base catalysis,7 is one example of a
linear free energy relationship in which equilibrium constants
for protonation/deprotonation of a series of acids vary linearly
with the rate constants, k, for nucleophilic reactions of their
conjugate bases:
reaction rates for various electrophiles bound at the active sites
of individual GSTs, according to the Brønsted relation sum-
marized above. As shown within, Brønsted relationships predict
that there is an optimal thiol pKa for each electrophile. (3) The
pKa that yields optimal rates for each electrophile will change
with pH. To the extent that pH is variable in experiments in
vitro, it must be considered as part of a complete Brønsted
analysis for GST-dependent processes. (4) The ‘nonaqueous’
nature of GST active sites must be considered. It is well
appreciated that solvent markedly affects ânuc values by dif-
ferential effects on ground vs transition states. It is generally
observed for reactions between thiolate anions and common
GST substrates that, with decreasing solvent polarity, ânuc values
increase.6c,9 On the basis of several high-resolution crystal
structures of different GSTs and solvent isotope effects,4e it is
clear that in the presence of electrophilic substrates the nucleo-
philic thiolate is likely to experience a decreased solvent polarity
as compared to aqueous solution. In turn, relevant ânuc values
may be larger than observed for analogous reactions in bulk
aqueous solvent. Therefore, the solvent composition was
explicitly varied here, and the dependence of ânuc on macro-
scopic dielectric constant was determined for thiol attack on
electrophilic GST substrates.
log k ) ânucpKa + C
(1)
That is, as the enzyme generates more of the nucleophilic anion
by reducing the pKa of GSH, the intrinsic nucleophilic reactivity
of this anion also decreases by an amount determined by the
ânuc value of the Brønsted relation. The situation for GST is
directly analogous to the case considered many years ago for
serine proteases.8 As discussed previously, even if active site-
imposed steric constraints were not present, then the observed
rate of the enzymatic product formation, Vmax, between GS-
and electrophile bound in the ternary complex, [GST‚GS-‚E],
will be a complex function of the pKa of the enzyme-bound
GSH, the pH, and the ânuc value for each particular GS-‚E pair,
where ânuc is the Brønsted coefficient for nucleophilic attack
in eq 1.
Notably, attempts to determine ânuc values for GST-catalyzed
reactions with the electrophile CDNB have been made with site-
directed mutants for which bound GSH has different pKa
values4d,f and with synthetic GSH analogues.6b In some cases,
‘abnormal’ Brønsted behavior has been observed, i.e., negative
ânuc values are obtained. These results have prompted wide-
spread interest10 in the factors controlling reactivity of the
thiolate nucleophile. On the basis of X-ray structures of GST
mutants,6a the abnormal Brønsted behavior is likely due to
changes in reaction coordinate geometry or solvation upon
amino acid substitutions in GST or GSH.
As described herein, we hypothesized that the variance in
pKa exhibited by GSH at the active sites of different GSTs and
large ânuc values expected for some electrophilic substrates under
the nonaqueous solvation conditions could be sufficient to cause
differences in apparent Vmax rates for GSH conjugation with
various electrophiles at the active sites of different GST
isozymes. In principle, the pKa of GSH bound to an individual
GST could be optimized for reaction with a specific electrophile,
at the expense of other substrates, if ânuc values differed
sufficiently among electrophiles. Moreover, on the basis of the
observation that the pH vs rate profiles vary with electrophile
and with GST variants having different GSH pKa values, we
hypothesized that Brønsted behavior contributes to the pH vs
rate profiles. That is, if GSTs sample a sufficient range of
â-pKa-rate space, then the pKa will be a determiant of substrate
selectivity. To test these hypotheses, however, it was necessary
to determine representative values of ânuc for reactions in which
thiols of varying pKa attack chemically distinct electrophiles.
The range of ânuc values for this series of thiols provides
constraints on the â-pKa-rate space for GST-catalyzed con-
jugation of glutathione with different electrophilic substrates.
The model reactions studied here are summarized in Figure 1,
and they include thiols that exhibit a pKa range from 4.5 to 7.5,
which spans the range of pKa values observed for GSH bound
to GSTs. These p-substituted thiophenols provide control of
Four components of the classic Brønsted relation as it applies
to GST catalysis warrant reconsideration: (1) Structurally
diverse electrophiles are substrates for GSTs, and hence the
range of relevant ânuc values may be large. Nucleophilic attack
of aliphatic thiols on electrophiles in aqueous solution is
associated usually with low ânuc values.6 Qualitatively, there-
fore, an increase in the fraction of thiolate as compared to
protonated thiol that is achieved by hydrogen bonds at the active
sites of GSTs is expected to offset any decrease in reactivity of
the resulting thiolate anion. However, a quantitative analysis
has not been performed. (2) There is a range of pKa values for
GSH bound at the active sites of individual GST isozymes rather
than a single value. For example, the pKa of GSH bound to
the rat A 1-1 GST is 7.4, in marked contrast to the values
reported for the M and P class enzymes, pKa ) 6.5-6.9.4
Apparently, the pKa of GSH bound to different GST isoforms
varies by nearly an entire pKa unit. Therefore, even if ânuc
values for GST-catalyzed reactions are small, this large range
of pKa values would, in principle, contribute to differences in
(6) (a) Xiao, G.; Liu, S.; Ji, X.; Johnson, W. W.; Chen, J.; Parsons, J.
F.; Stevens, W. J.; Gilliland, G. L.; Armstrong, R. N. Biochemistry 1996,
35, 753. (b) Chen, W. J.; Graminski, G. F.; Armstrong, R. N. Biochemistry
1988, 27, 194. (c) Douglas, K. T. Reactivity of Glutathione in Model
Systems for Glutathione S-Transferase and Related Enzymes. In Glutathione
Conjugation; Seis, H., Ketterer, B., Eds.; Academic Press: New York, 1988;
pp 1-41.
(7) (a) Brønsted, J. N.; Pedersen, K. Z. Phys. Chem. 1924, A108, 185.
(b) Brønsted, J. N.; Guggenheim, E. A. J. Am. Chem. Soc. 1927, 49, 2554.
(8) (a) Bruice, T. C.; Fife, T. H.; Bruno, J. J.; Brandon, N. E. Biochemistry
1962, 1, 7. (b) Jencks, W. P.; Gilchrist, M. J. Am. Chem. Soc. 1962, 84,
2910.
(9) (a) Bruice, P. Y.; Bruice, T. C.; Yagi, H.; Jerina, D. M. J. Am. Chem.
Soc. 1976, 98, 2973. (b) Conlon, P. R.; Sayer, J. M. J. Org. Chem. 1979,
44, 262.
(10) (a) Zheng, Y. J.; Ornstein, R. L. J. Am. Chem. Soc. 1997, 119, 1523.
(b) Zheng, Y. J.; Bruice, T. C. J. Am. Chem. Soc. 1997, 119, 3868.

Brønsted BehaVior in Glutathione S-Transferases
J. Am. Chem. Soc., Vol. 120, No. 27, 1998 6653
Figure 1. Summary of model reactions studied. The series of 4-substituted benzenethiols with variable pKa was used in nucleophilic reactions with
the indicated electrophiles, CHP, trans-PBO, CDNB, EPNP, and 2-NP. The R groups affording different thiol pKa values are summarized to the left
of the reactions and include p-methoxy (1), -methyl (2), -hydroxy (3), unsubstituted (4), -chloro (5), and -nitro (6).
the pKa of the conjugate acid of the reactive nucleophile without
complication due to differential solvation or steric effects at
the nucleophilic atom. The electrophiles used represent common
GST substrates and include the aryl-halide CDNB, an epoxide,
a nitroalkane, a hydroperoxide, and two substrates that contain
an R,â-unsaturated carbonyl. Together, these reactions provide
a quantitative analysis of the theoretical contribution of linear
free energy relationships to isozyme-dependent substrate selec-
tivity of GSTs and their pH vs rate behavior.
Results
Brønsted Analysis: The General Case. For the general
enzymatic reaction catalyzed by GST with any electrophile, E,
Vmax ) kcat[GST‚GS-‚E] or kcat[GST‚ GS-E], where kcat is the
first-order rate constant for the chemical conjugation step or
diffusion-controlled product release. The analysis below relates
to cases in which the chemical step is rate limiting. Assuming
that the protonated complex [GST‚GSH‚E] is not catalytically
competent and that the thiol and thiolate complexes are in rapid
equilibrium, then for a specific pH the fraction of total GST in
the active form varies with the pKa of bound GSH. Therefore,
at any pH the relative Vmax, (Vmax)rel, may be defined as the
fraction of the optimal rate at complete ionization of GSH,
To link the model studies with the enzymatic system,
experiments were performed with a series of GSTA1-1 mutants
having variable pKa in the [E‚GSH] complex. These studies
demonstrated that the linear free energy relationships described
by the Brønsted relationship often will not be expressed in GST
reactions at steady state due to the prevalence of rate-limiting
physical steps. However, when rates of the chemical conjuga-
tion step can be observed, the intrinsic reactivity of enzyme-
bound GS- is a function of the GSH pKa, precisely as predicted
by Brønsted behavior. On the basis of these results, we suggest
that the heterogeneity of the pKa of GSH bound at the active
sites of different GSTs may contribute to the ‘substrate diversity’
of the GST family and hence contribute to the function of these
detoxication enzymes.
(Vmax opt
) , that would be obtained if the GSH pKa was sufficiently
low, and assuming that kcat is independent of the pKa:
(Vmax rel
)
) Vmax/(Vmax opt
)
) kcat[GST‚GS-‚E}/
kcat{[GST‚GSH‚E] + [GST‚GS-‚E]} (2)
As the pKa changes, perhaps through evolution or in vitro
mutagenesis, then the change in Vmax will be readily predicted
from the change in the fraction of complexed GSH in the thiolate

6654 J. Am. Chem. Soc., Vol. 120, No. 27, 1998
form, f, according to eq 3 or eq 4, where (Vmax rel
pH ) pKa + log {[GST‚GS-‚E]/[GST‚GSH‚E]} (3)
Nieslanik and Atkins
)
) (kcat)‚f:
or
pH ) pKa + log {[[f]/[1 - f]}
(4)
With this simple ionization model, enzyme turnover is controlled
only by the fraction of GST complexed with the thiolate GS-
at any pH and pKa.
However, according to classic descriptions of free energy
relationships, as the enzyme lowers the pKa of GSH, the
resulting GS- becomes less reactive, according to eq 1, where
k or kcat is the intrinsic rate constant for the uncatalyzed or
enzyme catalyzed reaction, respectively. In this case, competing
effects will be operative. As evolution or mutagenesis changes
the pKa of GSH, the rate of product formation will be determined
not only by the fraction of enzyme in this form but also by the
kcat associated with new pKa, as determined by eq 1. If ânuc
from the Brønsted relation (eq 1) is sufficiently low, as suggested
previously, then kcat will be insensitive to changes in GSH pKa
and Vmax will depend, in the limiting case, on the fraction of
GST complexed as [GST‚GS-‚E], as in the simple ionization
model. However, if ânuc is sufficiently large to cause changes
in kcat as the GSH pKa changes, then Vmax becomes a complex
function of ânuc, pH, and GSH pKa.
Because the goal of the present analysis was to explore the
role of variable GSH pKa in electrophilic substrate selectivity
of GSTs rather than to provide any detailed comparison of
transition state structures, the constant term C in eq 1 may be
eliminated. As long as C in eq 1 is independent of pKa, a
convenient general expression for the relatiVe rate of product
formation, (Vmax)rel, for each electrophile as a function of the
pKa of GSH in the enzyme complex is readily obtained and
accounts for the dependence of kcat on pKa. To do this, we
define the parameter (Vmax)f, which is the rate of product
formation at a GSH pKa that affords (kcat)f and fractional
ionization of the GSH thiol defined above as f. Similarly, the
optimal rate, (V
optimal rate by balancing pKa-dependent kcat and f. Thus, if
(Vmax)f is normalized to (Vmax opt, then we obtain (Vmax)rel, the
fraction of optimal Vmax at any pKa:
max)opt, is now the Vmax at a pKa that yields
)
(Vmax rel
)
) (Vmax)f/(Vmax opt
(Vmax rel ) [f]‚(kcat)rel
and (kcat)rel are unitless. When the expression
)
) [(kcat)f‚f]/[(kcat)opt] (5)
)
(6)
Thus (Vmax rel
)
for f from eq 4 and the expression for kcat from eq 1 are
substituted into eq 6, we obtain an expression for (Vmax rel in
)
terms of GSH pKa, pH, and â:
) {10(â)(pK )‚[10(pH-pKa)]}/[1 + 10(pH-pKa)] (7)
a
(Vmax rel
)
For convenience, the dependence of (Vmax)rel on pH, pKa, and
Figure 2. Plots of relative Vmax vs GSH pKa at variable ânuc, as
predicted by eq 7. As ânuc increases from 0.1 to 0.9, the pKa optimum
for the reaction increases. The precise optimum is also a function of
pH. The ânuc for the symbols used are (, 0.1; +, 0.3; 2, 0.5; ×, 0.7;
b, 0.8. (a) pH 6.5, (b) pH 7.4, (c) pH 8.0.
â will be referred to as the Brønsted GST model. Equation 7
was used to perform a numerical simulation (Figure 2), which
illustrates several important features of the dependence of rate
on ânuc and pKa. In Figure 2, the rates have been determined
for pH 8.0, 7.4, and 6.5 and normalized in each case to the
maximal value obtained for each ânuc value, (Vmax opt
)
. The
is an optimal pKa for GSH bound at the active site of GST. If
the pKa is below this optimum, then the decrease in intrinsic
reactivity associated with Brønsted-type linear free energy
relation ‘outweighs’ the gain in f that results from lowering the
normalized rates, (Vmax)rel, provide a measure of the sensitivity
of the rate of turnover for each electrophile as the thiol pKa
changes. For each value of ânuc and at any specific pH, there

Brønsted BehaVior in Glutathione S-Transferases
J. Am. Chem. Soc., Vol. 120, No. 27, 1998 6655
pKa. Also, the steepness of the pKa vs rate profile is greater on
the high pKa side for low ânuc values. However, as ânuc
approaches 1 the curves become less steep on the high pKa side
and more sensitive on the low pKa side. Note that, because
each curve is normalized within the data for a given electrophile,
comparison of rates for different electrophiles at a specific pKa
is not informative. However, the results emphasize that,
according to Brønsted model, the dependence of the reaction
rates on GSH pKa differs dramatically with the ânuc values and
with pH.
Determination of Brønsted Factors in Chemical Models.
To compare the effect of variable GSH pKa on the rate of
reaction with different electrophiles, it was necessary to
determine ânuc values for a representative range of electrophilic
substrates of GSTs. That is, the relevance of Figure 2 for GST
catalysis is unclear in the absence of ânuc values for electrophilic
GST substrates. Furthermore, because the dielectric environ-
ment of GST active sites is uncharacterized, it was necessary
to explore the effect of solvent hydrophobicity on ânuc values.
Therefore, the rate constants for the reactions of a series of thiols
with different pKa values with several electrophiles (Figure 1)
were determined under different solvent conditions. For each
thiol-electrophile conjugate, the rate of product formation was
determined spectrophotometrically by measuring loss of chro-
mophoric thiol reactant, as described in the Experimental
Section. To validate this method for determination of rate
constants and to ensure that thiol oxidation was not contributing
to the thiol consumption, the rate constants for reactions were
determined also by monitoring product formation for the series
of thiols with the electrophile CDNB. CDNB is a ‘universal’
GST substrate, and CDNB conjugates are easily quantitated by
colorimetric assay. Thus, synthetic standards were prepared for
each thiol-CDNB conjugate, and the extinction coefficients
were determined (Experimental Section) to ensure precise
measurement of rate constants. The rate constants and ânuc value
obtained for this series of reactions were identical whether
determined by following loss of thiol or by production of CDNB
conjugate. Thus, the former method is assumed to yield accurate
ânuc values with the other electrophiles, without contribution
from thiol oxidation. Typical raw progress curves are shown
in Figure 3, which demonstrates the expected concentration
dependence on rate of thiol consumption. The Brønsted plots
for a subset of electrophiles and the series of thiols also are
summarized in Figure 3. The recovered Brønsted values for
each electrophile are summarized in Table 1. Reaction with
each electrophile shown in Figure 1 was studied in 10% DMF.
Additional reactions were run in 30% DMF. There has been
significant debate concerning the relevant dielectric constants
for enzyme active sites, and it is not the goal of these studies
to accurately model the dielectric environment of GST active
sites. Rather, the experiments are intended to emphasize that
Figure 3. (a) Raw progress curves for reaction of thiobenzene (4) with
CDNB. The effect of varying the concentration of thiol on the rate is
shown. The offset in absorbance values at time 0 is due to different
extents of reaction occurring before initiating measurements. The slopes
(fitted lines) yield kobs at each concentration. See methods for details.
(b) Representative Brønsted plots for various electrophiles. Both C and
ânuc of eq 1 vary with electrophile. The full range of ânuc values are
summarized in Table 1.
Table 1. Brønsted Coefficients for Electrophilic GST Substratesa
ânuc
electrophile
90:10, H2O:DMF
70:30, H2O:DMF
CHP
0.16 ( 0.04
0.46 ( 0.08
0.51 ( 0.08
0.59 ( 0.01
0.78 ( 0.10
0.93 ( 0.08
Ɖnuc ) 0.77
0.29 ( 0.04
0.66 ( 0.01
0.63 ( 0.05
CDNB
EPNP
EA
trans-PBO
2-NP
â
nuc values are almost certainly greater for active site processes
than in aqueous solution for many electrophiles. As expected
on the basis of previous studies of thiol attack on electrophilic
centers, ânuc values increase with increasing hydrophobicity of
the solvent.
1.04 ( 0.13
Ɖnuc > 0.68
range
a See Experimental Section for reaction conditions. Plots used to
determine ânuc values were obtained with triplicate kinetic runs. Typical
progress curves are shown in Figure 3a, and representative Brønsted
plots are shown in Figure 3b. For all electrophiles other than CHP and
t-PBO, regression analysis of the Brønsted plots yielded r2 > 0.95.
For CHP and t-PBO, r2 values were 0.90 and 0.91, respectively.
In cases where direct comparison is possible, the recovered
values are in good agreement with ânuc values reported for thiol
attack on various electrophiles.6c,9 We do note that, for CDNB
and EPNP, the values obtained here are modestly higher than
reported reactions with aliphatic thiols; this is likely a result of
using arylthiols rather than alkanethiols and from inclusion of
the hydrophobic solvent as observed for other electrophiles (and
vide infra). Also, it should be noted that thiolate reactions with
unbranched nitroalkanes proceed via a SN2 nucleophilic mech-
anism, whereas R-substituted nitroalkanes may proceed by a
radical-anion chain mechanism, SRN1 mechanism.11,12 Indeed,
multiple mechanisms may be operative with GST catalysis.

6656 J. Am. Chem. Soc., Vol. 120, No. 27, 1998
Nieslanik and Atkins
a
Obviously, our results are relevant only for nitroalkanes that
react via a nucleophilic, SN2 mechanism.
Application of the Model to GST-Dependent Reactions.
In light of the model studies that indicate the Ɖnuc range is
large enough to contribute to electrophile-dependent differences
in Vmax, it is of interest to determine their predictive value in
relation to the enzyme-catalyzed reactions. For example, the
Brønsted analysis suggests that a GST complexed with GSH
having pKa of 8 should be more reactive with electrophiles
characterized by large ânuc values than a GST with a cofactor
pKa of 6.5. In contrast, reactions characterized by low ânuc
values should be more efficiently catalyzed by GSTs that afford
a GSH pKa of 6.5 than 8. Moreover, at higher pH the rate of
reaction with a low ânuc value will be a more sensitive function
of the pKa of GSH than at low pH. The extent to which these
predictions are observable depends on whether the experimen-
tally monitored kinetic parameter reflects the microscopic rate
constant for the chemical conjugation step. If the chemical step
for enzymatic turnover is not rate limiting, then the free energy
relationship will be masked in steady-state experiments.
b
To determine whether chemical steps are rate limiting for
enzymatic turnover of the electrophiles used in the model
studies, enzymatic reactions were performed in the presence of
varying concentrations of viscogen. When physical steps such
as ligand association or dissociation are rate limiting, the overall
turnover rate decreases with increasing viscogen. In contrast,
if chemical steps are rate limiting, the steady-state rate is
insensitive to viscogen. The use of viscogens as probes of
segmental motion has been described for several enzymatic
systems including GSTs.13 No turnover was detected with
t-PBO or EPNP with this particular GST isozyme, and an
enzymatic assay for 2-NP turnover is not readily available. The
influence of viscogen on reaction rates for wild-type GSTA1-
1, at pH 7.4, with CHP, CDNB, and EA is shown in Figure 4
(a). These results clearly demonstrate that chemical conjugation
is cleanly rate limiting only for CHP (m ) 0.01). Physical steps
are cleanly rate limiting for EA (m ) 0.92). Intermediate slopes
for plots of this type (m ) 0.12), as with CDNB, may reflect a
partially rate-limiting chemical step. Therefore, CHP and
CDNB were used as substrates with a series of mutants,
previously described, including F220Y, F220E, F220I, and
F220L.4c,14 These mutants are catalytically comparable to the
wild type but exhibit variable pKa for the [GST‚GSH] complex.
Importantly, Phe-220 does not directly contact the electrophile
binding site nor the sulfur atom of the GS- nucleophile but
rather provides part of the immediate environment of the
catalytic Tyr-9, which in turn modulates the pKa of GSH at the
active site via an indirect effect. Together with wild type, these
mutants provide a limited set of GSTs with variable GSH pKa
(pKa 7.0-9.0), with minimal structural variation expected in
the immediate environment of the GS- thiolate or the electro-
phile. Chemical steps are cleanly rate limiting for each GST
variant when CHP is the substrate (Figure 4b, m ) 0.007-
0.059), but physical steps contribute differentially with these
variants when CDNB is the substrate, as indicated by the slopes
Figure 4. Dependence of steady-state turnover rates on viscogen at
pH 7.4. (a) Rates of enzymatic reaction vs viscogen concentration are
shown for wild-type rat GSTA1-1 with CHP, CDNB, and EA. Only
the reaction with CHP is limited cleanly by chemical steps. (b) Rates
of reaction vs viscogen concentration for CHP and mutant GSTs with
variable GSH pKa. Mutation-induced changes in pKa do not change
the rate-limiting step for this substrate. For each data set, the fitted
lines were obtained from k0/k ) n/n0 where k0 and k are the rates in
the absence and presence of viscogen and n0 and n are the viscosities.
The slopes obtained with EA, CDNB, and CHP are 0.92, 0.12, and
0.012, respectively (a). The slopes obtained for CHP remained <0.05
for each of the mutants (b).
ranging from 0.062 to 0.34 (not shown). Thus, Brønsted
behavior will likely be masked partially in steady-state experi-
ments with CDNB. On the basis of these results, experimental
challenge of the model defined by eq 7 with steady-state Vmax
rates (or kcat/KM) for the EA substrate is not possible and will
require more detailed pre-steady-state kinetic analysis that yields
microscopic rate constants for chemical conjugation steps.
Brønsted analysis was performed for enzymatic turnover of
CDNB and CHP, at three pHs. The ânuc values for the
enzymatic turnover of CHP and CDNB were determined from
(11) Benn, M.; Meesteres, A. C. J. Am. Chem. Soc. Chem. Commun.
1977, 597.
(12) (a) Bowman, W. R. Chem. Soc. ReV. 1988, 17, 283. (b) Bowman,
W.; Richardson, G. Tetrahedron Lett. 1981, 22, 1551.
(13) (a) Caccuri, A. M.; Antonini, G.; Nicotra, M.; Battistoni, A.; Lo
Bello, M.; Board, P. G.; Parker, M. W.; Ricci, G. J. Biol. Chem. 1997,
272, 2968. (b) Johnson, W. W.; Liu, S.; Ji, X.; Gilliland, G. L.; Armstrong,
R. N. J. Biol. Chem. 1993, 268, 11508. (c) Sampson, N.; Knowles, J.
Biochemistry 1992, 31, 8488. (d) Adams, J.; Taylor, S. S. Biochemistry
1992, 31, 8516.
plots of (Vmax lim vs GSH pKa, where (Vmax)lim is the limiting
)
rate at high pH, where all of the complexed GSH is ionized.
This method has been previously described,4f,6b and the Brønsted
plots are shown in Figure 5. Notably, the recovered ânuc value
for the enzymatic reaction with CHP, 0.19 ( 0.08, is in excellent
agreement with the values obtained in the model systems. The
recovered value for CDNB, 0.25 ( 0.07, is in reasonable
(14) Atkins, W. M.; Dietze, E. C.; Ibarra, C. Protein Sci. 1997, 6, 873.

Brønsted BehaVior in Glutathione S-Transferases
J. Am. Chem. Soc., Vol. 120, No. 27, 1998 6657
a
b
Figure 5. Brønsted plots for enzymatic turnover with CHP and CDNB.
(Vmax
max vs pH for each mutant. The resulting plots were fit to the equation
log Vmax ) log{(Vmax
lim/[(1 + [H+]/Ka]}, where Ka is the acid ionization
constant of GSH bound to each GST. Values of (Vmax recovered
)lim at complete GSH ionization was determined from plots of log
V
)
)
lim
from these analyses for each protein are plotted vs the pKa of complexed
GSH to obtain C (y-intercept) and ânuc (slope). The recovered values
are C ) -1.16 and ânuc ) 0.19 ( 0.08 for CHP; C ) 2.61 and ânuc
0.25 ( 0.14 for CDNB.
)
agreement with the model studies, although slightly lower.
Presumably, the contribution of diffusion-limited processes to
the rate-limiting step with CDNB (Figure 4) contributes to
masking of the true ânuc value, and CDNB is likely to fit the
steady-state Brønsted model less well than CHP.
Further analysis with CHP was performed in order to
determine whether the linear free energy relationship predicted
by eq 7 was operative. The experimentally determined ânuc and
C values for CHP (Figure 5) were used with eq 1 to calculate
Figure 6. Comparison of the simple ionization model and the Brønsted
model. The experimentally determined (Vmax
) values (2) are plotted
rel
for the mutants with variable GSH pKa. The (Vmax)rel values are obtained
from normalization to (Vmax opt for the Brønsted model. The solid lines
)
represent the fitted curves according to the Brønsted GST model, in
which kcat varies with pKa, as described by eq 7 in the Results. Open
circles are experimetentally determined (Vmax)rel values using the (Vmax)opt
from eq 2 or the simple ionization model. The dashed line represents
the fitted curves for the simple ionization model, in which kcat is
insensitive to changes in pKa. Even for electrophiles with low ânuc
values, the experimental rates are perturbed from the simple ionization
model as predicted by the Brønsted GST model.
the Vmax rate at the optimal pKa. This calculated rate, (Vmax opt
was used to calculate (Vmax rel from eq 2 and the experimentally
measured rates of Vmax for CHP with each protein at each of
the three pHs. Simulated curves of (Vmax rel vs GSH pKa, at
) ,
)
)
three pHs, were constructed according to the simple ionization
model (eq 2), and the Brønsted model, (eq 8) using the
experimentally determined ânuc and C values for enzymatic
turnover of CHP. Equation 8 is derived with the same strategy
as eq 7 was derived, except the constant term C in eq 1 is
retained. This is required because absolute rates and not
normalized rates are experimentally obtained:
obtained from the experimentally measured absolute rates also
are shown in Figure 6, normalized to (Vmax opt as predicted by
)
either the simple ionization model (open circles) or the Brønsted
model (closed triangles). Excellent agreement was observed
between the experimentally determined rates for the GST
variants and the model that incorporates Brønsted free energy
linkage. Indeed, statistical fitting of the two models, ionization
only, according to eq 2, vs eq 7, indicates that the experimental
data fit better when Brønsted behavior is included, particularly
at the lower pHs. The ø2 values for pH 6.5, pH 7.4, and pH
a
(Vmax rel
)
) {10(â)(pK )‚[10(pH-pKa)]}/[1 + 10(pH-pKa)] +
[10C‚10(pH-pK )]/[1 + 10(pH-pKa)] (8)
a
The simulated curves for the ionization model are shown as
dashed lines in Figure 6, and the simulated Brønsted models
are shown with a solid line. The calculated relative rates

6658 J. Am. Chem. Soc., Vol. 120, No. 27, 1998
Nieslanik and Atkins
8.0 are 0.074, 0.037, and 0.043, respectively, when the data
are fit to the Brønsted model; the corresponding ø2 values are
1.63, 0.41, and 0.050 when fit to the simple ionization model.
function of the fraction of ionization of GSH. The significance
of this conclusion is most apparent with a specific, theoretical,
example based in Figure 2. Figure 2 predicts that, for a reaction
with ânuc value of ∼0.1 at pH 7.4, the ratio of Vmax rates for a
GST complex with pKa of 6.5 to a complex with pKa 7.5
(Vmax6.5/Vmax7.5) will be 1.9, and there is a catalytic advantage
for the isozyme with the lower pKa. In contrast, the analogous
ratio for a reaction with a ânuc value of 0.8 will be 0.25. In
this case, the complex having the higher pKa will exhibit the
faster turnover for the Michael-type addition. Thus, there is a
7.7-fold difference in the selectivity of the two isozymes for
the electrophile with a low ânuc value relative to the ratio of
Discussion
Linear free energy relationships have been analyzed for a
series of thiol acids with variable pKa and the nucleophilic
reaction of their conjugate bases with a range of electrophiles.
The thiol acids of both small molecule models and GSH
complexed to GST variants have been studied experimentally.
Although linear free energy relationships, including Brønsted
analysis, have been utilized extensively in attempts to delineate
transition-state structure for enzyme-catalyzed reactions,15 the
goal of the experiments described here was significantly
different; the present model studies provide a basis for predicting
quantitatively the change in apparent rate of reaction for
conjugation of the nucleophilic GS- with several electrophilic
GST substrates, as a function of the pKa of the enzyme-bound
GSH. Because the pKa of GSH at the active site of different
GST isozymes (wild type) varies by as much as ∼1 pKa unit,
these models were developed in order to determine whether the
observed differences in GST isozyme selectivity for different
electrophiles are due partially to differences in intrinsic reactivity
of the individual [GST‚GS-‚E] complexes. We are cautious
to not over interpret the ânuc values for these model systems in
terms of detailed transition-state structures of the enzyme-
catalyzed reaction. Indeed, it is reasonable to question the utility
of ânuc values determined here for small molecule model systems
in understanding the extent of bond formation or charge
dispersion in the enzyme-catalyzed reactions. The question of
whether enzymes can control parameters that determine slopes
of plots derived from linear free energy relationships has been
discussed, and it is possible that ânuc values are ‘tuned’ by
enzymes.16 Solvent dramatically affects ânuc values, and enzyme
active sites provide ‘solvation’ different from bulk aqueous
phase. Presumably, then, ânuc values will differ for the two
environments (see below). However, it is unlikely that, for a
wide range of electrophiles, the ânuc values determined in bulk
solution would converge to a single value or narrow range of
values for active site processes. Importantly, some compression
of ânuc values is expected with increasing solvent hydrophobicity
because the theoretical upper limit is 1.0. With increasing
solvent hydrophobicity, therefore, the range of ânuc values may
become smaller, but their average magnitude will become larger
(Table 1). As emphasized above, the actual ânuc values are not
used here to interpret details of the enzymatic reaction mech-
anisms, and these values are likely to be crude approximations
given the unknown dielectric constant of the GST active site
and the use of aryl thiols. Rather, the model studies are intended
to demonstrate that ânuc values are not uniVersally low for all
thiol-electrophile pairs, as suggested previously. Regardless
of the actual values, the studies presented here indicate the
likelihood of a wide range of ânuc for GST-catalyzed conjugation
of GSH with different electrophiles.
their selectivities for the substrate characterized by the high ânuc
.
Also, depending on the electrophilic substrate, there will be
different optimal GSH pKa values that balance the increased
fraction of thiolate with the decreased nucleophilicity of the
resulting thiolate, and this optimal pKa will vary with solution
pH. Therefore, the results summarized in Figure 2 indicate that
the relative substrate selectivity for GST isozymes with different
GSH pKa values may be partially controlled by the linear free
energy relationship described in ‘traditional’ Brønsted analysis.
The extent to which these proposals are relevant to GST catalysis
depends on the range of ânuc values spanned by reactions with
different GST electrophilic substrates. The model studies
indicate that this range is large, with Ɖnuc as large as 0.5-0.6,
conservatively, depending on the hydrophobicity of the reaction
environment. Therefore, differences in GSH pKa associated with
different isozymes or due to mutagenic variation are likely to
cause modest differences in the inherent nucleophilicity of the
GS- nucleophile at their active sites.
There are several mechanisms by which the linear free energy
relationships may be masked in enzymatic reactions. Obviously,
active site architecture, or topology, also contributes to isozyme-
dependent substrate selectivity. Numerous examples of amino
acid substitutions that lead to altered substrate selectivity of
GSTs, including stereochemical selectivity, have been reported.17
Furthermore, differences in substrate selectivity can be intu-
itively rationalized in some cases by comparing the available
X-ray structures. Certainly, comparison of turnover rates for a
specific electrophile between GSTs with different GSH pKa
values will be complicated by differences in active site topology,
which are likely to dominate the isozyme-dependent substrate
selectivity profile.
A second determinant of isozyme-dependent substrate selec-
tivity, especially across class boundaries, results from the
presence of additional catalytic elements present in some, but
not all, GSTs. For example, M class GSTs have been shown
to utilize general acid catalysis via an active site tyrosine distinct
from the GSH hydrogen bond partner to the leaving oxygen
during epoxide conjugation.13b Obviously, even if chemical
steps do control Vmax in such cases, the additional catalytic
groups present in some isozymes will increase the apparent,
intrinsic, nucleophilicity of the GS- anion, and a comparison
of rates with isozymes from other classes that do not have a
similar acid catalyst is futile. Together, the differences in active
site topology and catalytic groups makes difficult a comparison
of the GSH pKa and substrate selectivity for GSTs belonging
to different classes. In contrast, our model studies have
eliminated these effects by design, and the analysis provided
here indicates that if other active site features are identical, then
the pKa of enzyme-bound GSH will be an important determinant
of the substrate selectivity of GST isozymes. This conclusion
Most importantly, the major conclusion obtained from these
model studies is that a difference of ∼1 pKa unit, as reported
for GSH bound to a GST M3-3 vs GST A1-1, should afford a
significant difference in kcat as the electrophile changes, such
that Vmax will change in accord with eq 7, and not be a simple
(15) (a) Toney, M. D.; Kirsch, J. F. Science 1989, 243, 1485. (b)
Schweins, T.; Geyer, M.; Kalbitzer, H. R.; Wittinghofer, A.; Warshel, A.
Biochemistry 1996, 35, 14225.
(16) (a) Burbaum, J. J.; Raines, R. T.; Albery, W. J.; Knowles, J. R.
Biochemistry 1989, 28, 9293. (b) Ellington, A. D.; Benner, S. A. J. Theor.
Biol. 1987, 127, 491.
(17) (a) Bammler, T.; Driessen, H.; Finnstrom, N.; Wolf, C. R.
Biochemistry 1995, 34, 9000. (b) Zhang, P.; Liu, S.; Shan, S.; Ji, X.;
Gilliland, G. L.; Armstrong, R. N. Biochemistry 1992, 31, 10185.

Brønsted BehaVior in Glutathione S-Transferases
J. Am. Chem. Soc., Vol. 120, No. 27, 1998 6659
is based in Figure 2, which highlights the dependence of rate
on ânuc and GSH pKa, and Table 1, which demonstrates that
Ɖnuc is likely to be large enough to contribute to rate differences
for some electrophiles.
substrates. Whether this is the case in the enzymatic reactions
remains to be determined.
An experimental challenge of the Brønsted model, provided
by CHP (Figure 6), clearly demonstrates that this model more
accurately describes the reactivity of GSTs than the simple
ionization model. Unfortunately, the only substrate we exam-
ined for which chemical steps are cleanly rate limiting, CHP,
is also characterized by a very low ânuc value. Obviously, a
more dramatic distinction between the two models would be
obtained with reactions characterized by higher ânuc values or
at lower pKa ranges. Neither case is readily attainable with the
available GST mutants and electrophiles surveyed here.
Although the contribution of the Brønsted free energy
relationships to substrate specificity is modest, the demonstration
that the pKa of GSH at the active sites of GSTs contributes to
rates of chemical steps naturally leads to several questions. Why
has nature varied the pKa among different GST isozymes? Have
different GSTs evolved to optimal GSH pKa values for different
electrophiles? Perhaps, it is useful to contrast the role of free
energy relationships for the detoxication catalysts, GSTs, with
the serine proteases, which also are considered to be ‘broad
specificity’ enzymes. The pKa of the nucleophilic serine in the
proteases is subject to the same ‘paradox’ described here for
GSTs. However, a critical difference between these catalysts
lies in their respective biological niches. Whereas the serine
proteases collectively hydrolyze peptides with remarkably
different sequence specificity, the local transition-state structures
and ânuc values will be nearly invariant for different peptide/
isozyme combinations. The relatively constant ânuc values are
expected because, regardless of the substrate peptide sequence,
an identical amide functional group is attacked by the serine.
The amino acid side chains of the substrate peptide that dictate
isozyme selectivity are remote from the reaction center, resulting
in nearly identical local transition states for all substrates. In
contrast, the electrophilic substrates for GST catalysis represent
a wide range of functional groups, which lead to dramatically
different transition states and ânuc values, each with an optimal
pKa for the GSH cofactor. In fact, it may be speculated that a
distribution of pKa values among different GST isozymes
provides a means for extending their collective substrate
diversity, by optimizing some GSTs for reactions characterized
by ânuc values in the range 0.3-0.4 and optimizing others for
reactions with ânuc values in the range of 0.5-0.6. The classic
linear free energy profile analysis used here reveals a simple
but novel mechanism by which substrate diversity may be
optimized within a family of detoxication enzymes.
The enzymatic experiments clearly indicate a third source of
deviation from the Brønsted GST model. The studies with
variable concentration of viscogen provide an essential reminder
that the contribution of linear free energy relationships to
substrate selectivity of GSTs with variable pKa will be masked
if only steady-state kinetic parameters are compared. For many
electrophile‚GST pairs, physical steps are rate limiting. There-
fore, Vmax and its pH dependence will not exhibit the variation
with GSH pKa predicted by the Brønsted analysis. The linear
free energy relationships will only be expressed fully for a subset
of [GST‚GS-‚electrophile] combinations, such as the [GSTA1-
1‚GS- ‚CHP] complex. Indeed, the results obtained for enzy-
matic turnover of CHP, summarized in Figure 6, demonstrate a
remarkable adherence to the pH and GSH pKa dependence that
is predicted by eq 7. Together the results demonstrate that the
pKa of GSH at the active site of GSTs plays a modest role in
the electrophile-dependent efficiency of chemical steps in the
reaction cycle.
The ânuc values obtained for GST-dependent turnover of CHP
and CDNB warrant some discussion. ânuc has been determined
to be 0.3 ( 0.2 for GSH conjugation to CDNB by an M class
GST.4f,6b The A1-1 GST yields an identical value within
experimental error of 0.25 ( 0.07. To whatever extent these
values are interpretable at the molecular level, the mechanisms
and transition states for this reaction catalyzed by the two
isozymes are apparently nearly identical. The value obtained
in our model studies, at 10%DMF:90%H2O, agrees reasonably
well this value. Presumably, the larger value obtained in the
model studies results from the use of aryl rather than alkyl thiols
and the masking of the linear free energy relationship (vide
infra). In contrast, the ânuc value obtained from the model
studies for CHP is nearly identical to the enzymatic result.
Notably, this is only the second substrate for which the ânuc
value has been measured for a GST-dependent reaction.
Together, these results suggest that the CHP reaction is
characterized by a lower ânuc value than the CDNB reaction,
both in solution and enzymatically. It appears that GST active
sites do not tune ânuc values significantly far from the solution
behavior as long as purely aqueous conditions are not used as
a reference.
Linear free energy relationships arise when ∆∆G for a series
of reference equilibria, such as ionization of GSH, is partially
or completely incorporated in ∆∆Gq for a series of reactions
involving components of the equilibria, where ânuc is propor-
tional to ∆∆Gq/∆∆G. Here, when ânuc is 1.0, any free energy
of destabilization of the GS- anion for the variant GSTs would
be expected to destabilize the ground-state anion relative to the
transition state for nucleophilic attack of the thiolate by the same
differential energy, i.e., ∆∆G ) ∆∆Gq, and the nucleophilic
reactions will be slower. In as much as the transition state is
‘unaffected’ by the changes that perturb ∆∆G, large ânuc values
are interpreted to indicate a late transition state that is very
different from the starting thiolate/electrophile complex. In
contrast, a low ânuc value is observed if the stabilization of
thiolate relative to thiol, ∆∆G, is also apparent in stabilization
of the transition state for reaction of the thiolate, such that ∆∆Gq
< ∆∆G. These data indicate that GST-dependent metabolism
of CHP and CDNB proceed through early transition states, as
in the nonenzymatic reactions. The model studies suggest,
however, that a late transition state is operative with other GST
Experimental Section
Materials and Characterization. Each of the p-substituted thiophe-
nols and electrophiles, other than CDNB, shown in Figure 1 were
purchased from Aldrich Chemical (Milwaukee, WI) and were used
without further purification. CDNB was purchased from Sigma
Chemical (St. Louis, MO). UV/vis absorbance spectra were recorded
on a Cary 3E UV/vis spectrophotometer. 1H NMR spectra were
obtained at 300 MHz using a Varian VXR 300 spectrometer. Low
resolution FAB and EI mass spectra were obtained using a micromass
70SEQ tandem hybrid mass spectrometer.
Reaction Kinetics. Reactions were performed at 25 °C in 10 mM
MES buffer, pH 6.7, and 10% DMF, 2 mM EDTA, 0.2 mM NiSO4 to
prevent oxidation of thiols. This solvent system has been established
to afford negligible general acid-base catalysis for nucleophilic
reactions involving thiolate anions.18 Electrophile and thiol stock
solutions were prepared fresh every 4 h in EtOH and DMF, respectively,
(18) (a) Capozzi, G.; Modena, G. The Chemistry of the Thiol Group;
Patai, S., Ed.; Wiley: London, 1974; pp 785-833.

6660 J. Am. Chem. Soc., Vol. 120, No. 27, 1998
Nieslanik and Atkins
and stored under argon. To initiate reactions, 10 µL each of thiol and
electrophile was added to 680-880 µL of buffer and 100-300 µL of
DMF under pseudo-first-order conditions with excess electrophile.
Electrophile concentrations were 333 µM EPNP, 200 µM CDNB, 200
µM 2-NP, 200 µM CHP, and 100 µM trans-PBO, and thiol concentra-
tion was varied. For reactions with each electrophile, thiol consumption
was monitored at the following wavelengths where the numbers refer
to the different thiols as depicted in Figure 1: (1), ꢀ264 ) 34 900 M-1
cm-1; (2), ꢀ265 ) 18 900 M-1 cm-1; (3), ꢀ262 ) 13 900 M-1 cm-1; (4)
4-Hydroxybenzenethiol-CDNB Conjugate (Product of (3) and
CDNB). 1H NMR (acetone-d6): ∂ 8.99 (1H, d, Ja ) 2.5 Hz, Jb ) 8.7
Hz), 7.52 (2H, d, J ) 8.7 Hz), 7.16 (1H, d, J ) 9.0 Hz), 7.08 (2H, d,
J ) 8.7 Hz). EI/MS: (70 eV) 292 (M+, 43.2), 275 (67.5), 200 (23.6),
199 (17.7), 167 (base), 139 (40.1). ꢀ340 ) 9000 M-1 cm-1
.
Benzenethiol-CDNB Conjugate (Product of (4) and CDNB). 1H
NMR (acetone-d6): ∂ 9.01 (1H, dd, Ja ) 2.3 Hz, Jb ) 9.0 Hz), 7.74-
7.64 (5H, m), 7.16 (1H, d, J ) 9.1 Hz). EI/MS: (70 eV) 276 (M+,
41.0), 259 (6.1), 195 (9.1), 183 (18.2), 166 (base), 152 (18.3), 139
ꢀ265 ) 22 900 M-1 cm-1; (5), ꢀ274 ) 19 800 M-1 cm-1; (6), ꢀ416
)
(52.3), 77 (37.5). ꢀ340 ) 9000 M-1 cm-1
.
13 300 M-1 cm-1. For all reactions, the pH of the final solution was
measured and used to calculate the final concentration of nucleophile,
adjusted for the fraction of thiolate based on the experimentally
determined pKa values (below) and the standard Henderson-Hasselbach
equation. Because the formation of CDNB-thiol conjugates is
routinely quantitated by monitoring product formation rather than thiol
consumption, the thiol/CDNB reaction was also performed using the
thiol as excess reagent and monitoring formation of the thiol-CDNB
conjugate formation at 340 nm. This method yielded an identical ânuc
value for the reaction as when the thiol consumption was monitored.
For either method, pseudo-first-order rate constants were deconvoluted
with the known concentrations of electrophile or thiol to yield second-
order rate constants for each electrophile-thiol pair. For example, with
excess electrophile, E, and variable thiol concentration, kobs was obtained
from slopes of plots of rate vs [thiol]. The bimolecular rate constant,
k, for each pair of thiol and electrophile was then obtained from k )
4-Chlorobenzenethiol-CDNB Conjugate (Product of (5) and
CDNB). 1H NMR (acetone-d6): ∂ 9.01 (1H, d, J ) 2.5 Hz), 8.36 (1H,
dd, Ja ) 2.5 Hz, Jb ) 8.9 Hz), 7.77-7.65 (4H, m), 7.24 (1H, d, J )
9.01 Hz). EI/MS: (70 eV) 312 (M++2, 12.8), 310 (M+, 32.2), 288
(42.6), 286 (56.4), 211 (16.9), 202 (31.0), 200 (80.0), 143 (base), 139
(36.0), 108 (53.2). ꢀ340 ) 9200 M-1 cm-1
.
4-Nitrobenzenethiol-CDNB Conjugate (Product of (6) and
CDNB). 1H NMR (acetone-d6): ∂ 9.02 (1H, d, J ) 2.5 Hz), 8.43 (2H,
d, J ) 9.1 Hz), 8.36 (1H, dd, Ja ) 2.4 Hz, Jb ) 9.0 Hz), 8.02 (2H, d,
J ) 8.7 Hz), 7.390 (1H, d, J ) 9.0 Hz). EI/MS: (70 eV) 321 (M+,
86.7), 240 (42.8), 227 (21.1), 211 (93.1), 210 (56.8), 165 (61.7), 153
(53.0), 139 (base), 95 (29.7), 79 (33.0), 63 (62.5). ꢀ340 ) 10 700 M-1
cm-1
.
Enzymatic Reactions. Enzymatic reactions were performed at 25
°C, in 0.1 M potassium phosphate at the pHs indicated in Results and
the figures. Enzymatic activities of wild-type rat GSTA1-1 and site-
directed variants for CDNB, EA, and EPNP were determined spectro-
photometrically according to Habig et al.,5a in the presence of varying
concentrations of viscogen. CDNB-dependent reactions contained 1
mM CDNB and 1mM GSH. EA-dependent reactions contained 0.6
mM EA and 5 mM GSH. For EPNP turnover, reactions contained 5
mM EPNP and 5 mM GSH. Activity with CHP was determined by
the method of Lawrence and Burke.20 With this substrate, reactions
contained 1.5 mM CHP, 1 mM GSH, 0.3 units of GSSG reductase
(Sigma, St. Louis, MO), and 0.25 mM NADPH. Substrates were added
from concentrated stock solutions in EtOH. The final concentrations
(v/v) of EtOH in the reaction mixtures were 4% for CDNB and EA
and 5% for CHP and EPNP. At each pH and viscogen concentration,
it was demonstrated experimentally that the electrophile was at
saturating concentration. Solutions contained 0-30% sucrose (w/v).
Viscosities were determined with an Ostwaldt viscometer at 25 °C.
pKa values of GSH complexed with each of the proteins studied were
determined by UV spectroscopy, monitoring the absorbance at 239 nm,
in solutions containing protein and saturating GSH as described
previously.4c
kobs/[E]. Brønsted coefficients, ânuc, were obtained from plots of second-
order rate constant vs thiol pKa for each electrophile.
Determination of pKa Values of Thiophenols. pKa values were
measured in 10 mM MES buffer containing 0.2 mM NiSO4, 2 mM
EDTA, 10% DMF, and 1% EtOH, at pHs ranging from 3 to 10.
Thiolate absorbance was measured at the appropriate λmax as indicated
above. pKa values were obtained with the ENZfitter software package
assuming a single ionization. The experimentally determined pKa
values were (1), 7.4; (2), 7.2; (3), 7.1; (4), 6.9; (5), 6.4; (6) 4.5.
Synthesis of Product Thiol-CDNB Conjugates. Synthesis of all
thiol-CDNB conjugates was adapted from the method of Koechel and
Cafruny.19 To a solution of NaHCO3 (5 mmol) in 50 mL of H2O, 5
mmol of CDNB was added. The flask was flushed with argon, followed
by dropwise addition of thiol (5 mmol) in 20 mL of EtOH. The reaction
mixture was stirred and purged with argon overnight, during which a
yellow precipitate formed. The mixture was acidified with HCl (final
pH ∼2) and extracted with diethyl ether. The extract was concentrated
in vacuuo to a yellow solid and recrystallized twice from CH2Cl2.
Analyses were as follows.
4-Methoxybenzenethiol-CDNB Conjugate (Product of (1) and
CDNB). 1H NMR (acetone-d6): ∂ 8.99 (1H, d, J ) 2.5 Hz), 8.33 (1H,
dd, Ja ) 2.5 Hz, Jb ) 9.1 Hz), 7.62 (2H, d, J ) 8.7 Hz), 7.14-7.2
(3H, m), 3.92 (3H, s). FAB-MS: 306 (M+, base), 289 (35.5), 273
(10.2), 259 (8.9), 242 (15.4), 227 (23.0), 196 (30.7). ꢀ340 ) 8500 M-1
Acknowledgment. The authors gratefully acknowledge Mike
Fisher and Professor William F. Trager for helpful discussions
and Dr. Eric C. Dietze for experimental determination of the
GSH pKa values complexed with GST mutants. This work was
supported by The National Institutes of Health (GM51210 and
GM7750) and Merck Research Labs, Rahway, NJ.
cm-1
.
4-Methylbenzenethiol-CDNB Conjugate (Product of (2) and
CDNB). 1H NMR (acetone-d6): ∂ 8.99 (1H, dd, Ja ) 2.5 Hz), 7.58
(2H, d, J ) 8.1 Hz), 7.45 (2H, d, J ) 7.8 Hz), 7.15 (1H, d, J ) 9.1),
3.45 (3H, s). EI/MS: (70 eV) 290 (M+, 69.0), 202 (19.6), 197 (27.7),
JA980816O
180 (base), 153 (18.3), 139 (24.8). ꢀ340 ) 9200 M-1 cm-1
.
(20) Lawrence, R. A.; Burke, R. F. Biochem. Biophys. Res. Commun.
1976, 71, 952.
(19) Koechel, D. A.; Cafruny, E. J. J. Med. Chem. 1973, 16, 1147.