A fluoro analogue of the menadione derivative 6-[2′-(3′-methyl) -1′,4′-naphthoquinolyl]hexanoic acid is a suicide substrate of glutathione reductase. Crystal structure of the alkylated human enzyme

Article abstract of DOI:10.1021/ja061155v

Glutathione reductase is an important housekeeping enzyme for redox homeostasis both in human cells and in the causative agent of tropical malaria, Plasmodium falciparum. Glutathione reductase inhibitors were shown to have anticancer and antimalarial acti

Full text of DOI:10.1021/ja061155v

Published on Web 07/28/2006
A Fluoro Analogue of the Menadione Derivative
6-[2′-(3′-Methyl)-1′,4′-naphthoquinolyl]hexanoic Acid Is a
Suicide Substrate of Glutathione Reductase. Crystal Structure
of the Alkylated Human Enzyme†
Holger Bauer,‡,§ Karin Fritz-Wolf,| Andreas Winzer,‡ Sebastian Ku¨hner,‡
Susan Little, Vanessa Yardley, Herve´ Vezin,X Bruce Palfey,O
R. Heiner Schirmer,‡ and Elisabeth Davioud-Charvet*,‡,§
Contribution from the Biochemie-Zentrum der UniVersita¨t Heidelberg, Im Neuenheimer Feld
504, D-69120 Heidelberg, Germany, Centre National de la Recherche Scientifique, 3 rue
Michel-Ange, 75794 Paris cedex 16, France, Max-Planck-Institut fu¨r Medizinische Forschung,
Jahnstrasse 29, D-69120 Heidelberg, Germany, Department of Infections and Tropical Diseases,
London School of Hygiene & Tropical Medicine, Keppel Street, London WC1E 7HT, United
Kingdom, Laboratoire de Chimie Organique et Macromole´culaire, UMR-CNRS 8009, Baˆt. C4,
UniVersite´ des Sciences et Technologies de Lille, 59655 VilleneuVe d’Ascq cedex, France, and
Department of Biological Chemistry, The UniVersity of Michigan,
Ann Arbor, Michigan 48109-0606
Received February 27, 2006; E-mail: elisabeth.davioud@gmx.de
Abstract: Glutathione reductase is an important housekeeping enzyme for redox homeostasis both in human
cells and in the causative agent of tropical malaria, Plasmodium falciparum. Glutathione reductase inhibitors
were shown to have anticancer and antimalarial activity per se and to contribute to the reversal of drug
resistance. The development of menadione chemistry has led to the selection of 6-[2′-(3′-methyl)-1′,4′-
naphthoquinolyl]hexanoic acid, called M5, as a potent reversible and uncompetitive inhibitor of both human
and P. falciparum glutathione reductases. Here we describe the synthesis and kinetic characterization of
a fluoromethyl-M5 analogue that acts as a mechanism-based inhibitor of both enzymes. In the course of
enzymatic catalysis, the suicide substrate is activated by one- or two-electron reduction, and then a highly
reactive quinone methide is generated upon elimination of the fluorine. Accordingly the human enzyme
was found to be irreversibly inactivated with a kinact value of 0.4 ( 0.2 min-1. The crystal structure of the
alkylated enzyme was solved at 1.7 Å resolution. It showed the inhibitor to bind covalently to the active site
Cys58 and to interact noncovalently with His467′, Arg347, Arg37, and Tyr114. On the basis of the crystal
structure of the inactivated human enzyme and stopped-flow kinetic studies with two- and four-electron-
reduced forms of the unreacted P. falciparum enzyme, a mechanism is proposed which explains
naphthoquinone reduction at the flavin of glutathione reductase.
Introduction
organisms. GR reduces glutathione disulfide (GSSG) to its thiol
form GSH:
The homodimeric FAD-containing glutathione reductase (GR;
EC 1.8.1.7) belongs to the family of NADPH-dependent
oxidoreductases and is present in many pro- and eukaryotic
GSSG + NADPH + H+ f 2 GSH + NADP+
The catalytic mechanism of GR from various sources has
been investigated by Williams et al. for many decades.1 The
flow of electrons during GR catalysis is from NADPH to the
re face of the flavin (FAD) and then from the si face of FAD
to the active center disulfide (C58VNVGC63, numbering referring
to human GR) and hence to GSSG. Although the enzyme can
accept four electrons per subunit (two at the flavin and two at
the disulfide), the principal catalytically active forms are the
oxidized Eox and the two-electron-reduced EH2 forms (Scheme
1). In the EH2 state the active site disulfide is reduced while
† Abbreviations: BPO, benzoyl peroxide; BuLi, butyllithium; CAN,
cerium(IV) ammonium nitrate; CI-MS, chemical ionization mass spectrom-
etry; DAST, N,N-diethylaminosulfur trifluoride; DMSO, dimethyl sulfoxide;
Eox, enzyme with the flavin and redox-active disulfide in the oxidized state;
EH2, two-electron-reduced enzyme with the redox-active dithiol; EH4, four-
electron-reduced enzyme with the reduced flavin and the redox-active
dithiol; EI-MS, electron ionization mass spectrometry; FAD, flavin; GR,
glutathione reductase; GSH, reduced glutathione; GSSG, glutathione
disulfide; NBS, N-bromosuccinimide; NQ, naphthoquinone; SCE, standard
calomel electrode; TBAF, tetrabutylammonium fluoride.
‡ Biochemie-Zentrum der Universita¨t Heidelberg.
§ Centre National de la Recherche Scientifique.
| Max-Planck-Institut fu¨r Medizinische Forschung.
London School of Hygiene & Tropical Medicine.
X Universite´ des Sciences et Technologies de Lille.
(1) Williams, C. H., Jr. In Chemistry and Biochemistry of FlaVoenzymes;
Mu¨ller, F., Ed.; CRC Press: Boca Raton, FL, 1992; Vol. 3, pp 121-211.
O The University of Michigan.
9
10784
J. AM. CHEM. SOC. 2006, 128, 10784-10794
10.1021/ja061155v CCC: $33.50 © 2006 American Chemical Society

Fluorine-Based Suicide Substrate of Glutathione Reductase
A R T I C L E S
Scheme 1. Catalytic Cycle of Human Glutathione Reductase
Our present research aims at the design of inhibitors of GR,
which has been identified as a gate keeper protein in the
glutathione redox system. Glutathione reductase inhibitors are
usually considered as drug sensitizers: they do not display high
intrinsic antimalarial or antitumoral activities when used alone,
but they can enhance the effects of chloroquine or cytotoxic
agents. By high-throughput screening, it was found that a
carboxylic acid derivative of menadione, M5 (6-[2′-(3′-methyl)-
1′,4′-naphthoquinolyl]hexanoic acid), is a potent inhibitor of P.
falciparum GR in the low micromolar range.13 Double-headed
prodrugs based on M5 showed potent antimalarial activities both
in vitro and in vivo in the mouse model.13 M5 is uncompetitive
with respect to both NADPH and GSSG.14 The binding site of
the inhibitor could, however, not be elucidated so far, which
suggests that it binds to an enzyme-substrate intermediate that
occurs in the catalytic cycle (Scheme 1).
Naphthoquinone derivatives are also substrates of flavoprotein
disulfide reductases.15,16 This finding was our motivation to
design a suicide substrate, or mechanism-based inhibitor of GR.
With a leaving group present in the position â to the carbonyl
group of the quinone core, the system can be activated by one-
or two-electron reduction, and a highly reactive quinone methide
is formed after release of the leaving group (Scheme 2).17 This
intermediate is expected to react easily with nucleophilic
residues in the vicinity.18 We introduced a fluorine atom â to
the carbonyl group of menadione as an almost isosteric
replacement of hydrogen. Further advantages are that the
carbon-fluorine bond is the most stable of the carbon-halogen
bonds and fluorine is meant to support the postulated mechanism
by formation of an intramolecular hydrogen bond (Scheme 2).17
Two fluoromethyl analogues of M5 were synthesized. Time-
dependent inhibition of both GR from human and P. falciparum
was observed using the GSSG reduction assay and studied. The
alkylated human GR was crystallized, and the 3D structure was
resolved at 1.7 Å resolution. The potential binding and reduction
site(s) of naphthoquinone can be discussed in light of the
observed kinetics using the inactivated human enzyme or two-
and four-electron-reduced forms of the unreacted P. falciparum
enzyme, the electrochemical behavior of the newly synthesized
compounds, and the 3D structure of the human GR alkylated
by 6-[2′-(3′-fluoromethyl)-1′,4′-naphthoquinolyl]hexanoic acid
(fluoro-M5).
the FAD has been oxidized. Catalysis2-4 involves electrons
transferred from NADPH to flavin to form FADH2, and the
reduced FAD very rapidly transfers electrons to the disulfide
center (Scheme 1). In the EH2 enzyme a charge-transfer
interaction is formed between the proximal thiolate Cys63 as
the donor and the flavin as the acceptor. After NADP+ release
GSSG is reduced in the oxidative half-reaction. The solvent-
exposed thiol in EH2, Cys58, initiates the dithiol-disulfide
interchange with GSSG that leads to an S-glutathionylated
enzyme and the generation of the first molecule of glutathione.
Then the second thiol Cys63 attacks the mixed disulfide bridge
of the glutathionylated enzyme, releasing the second molecule
of glutathione and re-forming the oxidized enzyme Eox.
Glutathione is a major player in intracellular redox homeo-
stasis and in detoxication processes both in human cells and in
the malarial parasite Plasmodium falciparum.5-7 Plasmodia are
sensitive to disturbances of the redox metabolism. This fact is
of special importance during the intraerythrocytic life stage
where the parasites are exposed to an increased oxidative stress
level.8 It was observed in parasite cultures with different P.
falciparum strains that the degree of chloroquine resistance is
directly linked to the levels of reduced glutathione.9-11 The
importance of glutathione in this field was further substantiated
by the finding that a decrease of the glutathione level in rodent
malaria leads to a potentiation of the antimalarial action of
chloroquine.12
Experimental Section
Glutathione Disulfide Reductase Assays. The standard assay was
conducted at 25 °C in a 1 mL cuvette. The assay mixture contained
100 µM NADPH and 1 mM GSSG in GR buffer (100 mM potassium
phosphate buffer, 200 mM KCl, 1 mM EDTA, pH 6.9). IC50 values
were evaluated in duplicate in the presence of seven inhibitor
concentrations ranging from 0 to 100 µM. Inhibitor stock solutions
(2) Pai, E. F.; Schulz, G. E. J. Biol. Chem. 1983, 258, 1752-1757.
(3) Krauth-Siegel, R. L.; Arscott, L. D.; Schonleben-Janas, A.; Schirmer, R.
H.; Williams, C. H., Jr. Biochemistry 1998, 37, 13968-13977.
(4) Bo¨hme, C. C.; Arscott, L. D.; Becker, K.; Schirmer, R. H.; Williams, C.
H., Jr. J. Biol. Chem. 2000, 275, 37317-37323.
(12) Deharo, E.; Barkan, D.; Krugliak, M.; Golenser, J.; Ginsburg, H. Biochem.
Pharmacol. 2003, 66, 809-817.
(5) Schirmer, R. H.; Mu¨ller, J. G.; Krauth-Siegel, R. L. Angew. Chem., Int.
Ed. Engl. 1995, 34, 141-154.
(13) Davioud-Charvet, E.; Delarue, S.; Biot, C.; Schwo¨bel, B.; Bo¨hme, C. C.;
Mu¨ssigbrodt, A.; Maes, L.; Sergheraert, C.; Grellier, P.; Schirmer, R. H.;
Becker, K. J. Med. Chem. 2001, 44, 4268-4276.
(6) Krauth-Siegel, R. L.; Bauer, H.; Schirmer, R. H. Angew. Chem., Int. Ed.
2000, 44, 690-715.
(14) Biot, C.; Bauer, H.; Schirmer, R. H.; Davioud-Charvet, E. J. Med. Chem.
2004, 47, 5972-5983.
(7) Kanzok, S. M.; Schirmer, R. H.; Tu¨rbachova, I.; Iozef, R.; Becker, K. J.
Biol. Chem. 2000, 275, 40180-40186.
(15) Misaka, E.; Kawahara, Y.; Nakanishi, K. J. Biochem. (Tokyo) 1965, 58,
436-443.
(8) Atamna, H.; Ginsburg, H. Mol. Biochem. Parasitol. 1993, 61, 231-241.
(9) Ginsburg, H.; Famin, O.; Zhang, J.; Krugliak, M. Biochem. Pharmacol.
1998, 56, 1305-1313.
(16) Salmon-Chemin, L.; Buisine, E.; Yardley, V.; Kohler, S.; Debreu, M. A.;
Landry, V.; Sergheraert, C.; Croft, S. L.; Krauth-Siegel, R. L.; Davioud-
Charvet, E. J. Med. Chem. 2001, 44, 548-565.
(10) Meierjohann, S.; Walter, R. D.; Mu¨ller, S. Biochem. J. 2002, 368, 761-
768.
(17) O’Shea, K. E.; Fox, M. A. J. Am. Chem. Soc. 1991, 113, 611-615.
(18) Modica, E.; Zanaletti, R.; Freccero, M.; Mella, M. J. Org. Chem. 2001,
66, 41-52.
(11) Dubois, V. L.; Platel, D. F.; Pauly, G.; Tribouley-Duret, J. Exp. Parasitol.
1995, 81, 117-124.
9
J. AM. CHEM. SOC. VOL. 128, NO. 33, 2006 10785

A R T I C L E S
Bauer et al.
a
Scheme 2. Proposed Mechanism of the Irreversible Inactivation of Glutathione Reductase by Fluoro-M5
a The naphthoquinone is activated via GR-catalyzed one-electron reduction (two-electron transfer is also possible). Elimination of HF promoted by a
newly formed intramolecular hydrogen bond results in a quinone methide. Nucleophilic attack of the catalytic Cys58 leads to covalent modification of the
enzyme. The radical can further react with oxygen, leading to the formation of superoxide.
were prepared in 100% DMSO. A 1% final DMSO concentration was
kept constant in the assay cuvette. The reaction was started by adding
enzyme (8 mU of human GR, 6.5 mU of P. falciparum GR), and initial
rates of NADPH oxidation were monitored at 340 nm (ꢀ340 nm ) 6.22
mM-1 cm-1).
Time-Dependent Inactivation of Glutathione Reductases by
Fluoro-M5. For determining the rate constants of human and P.
falciparum GR inactivation, the residual GR activity was monitored
over time by following an incubation protocol.19 All reaction mixtures
(final volume of 50 µL) contained 160 µM NADPH, varying inhibitor
concentrations (0-5 µM in the human GR assay, 0-50 µM in the P.
falciparum GR assay), GR (6.85 pmol of human GR, 21 pmol of P.
falciparum GR), and 2% DMSO in GR buffer at 25 °C. At different
time points, 5 µL aliquots of each reaction mixture were removed and
the residual activity was measured in the standard GSSG reduction
assay at 25 °C (1 mM GSSG and 100 µM NADPH). DMSO (2%) was
used in control assays.
Menadione Reductase Activity of Human and P. falciparum
Alkylated Glutathione Reductases. The ability of alkylated GRs to
reduce menadione was assayed by monitoring the oxidation of NADPH
at 340 nm (ꢀ340 nm ) 6.22 mM-1 cm-1). Over a 2 h incubation period
300 µM fluoro-M5 was allowed to react with 86 µM human GR (891
µg) in a 198 µL total volume containing 400 µM NADPH at 25 °C
prior to the naphthoquinone reduction assay. A similar alkylation
experiment was done with 72 µM P. falciparum GR (1.06 mg) over a
15 h long incubation time in a total volume of 257 µL, due to gradual
addition of NADPH (final 780 µM) and the alkylating agent (final 690
µM fluoro-M5). At different time periods, residual GSSG reduction
ability was tested in the standard GSSG reduction assay to check
alkylation completion. When total enzyme alkylation was reached, the
enzymes were tested in the menadione reductase assays. The reduction
assay mixture consisted of 100 mM potassium phosphate buffer, pH
6.9, 200 mM KCl, 1 mM EDTA, and 100 µM NADPH in a total volume
of 1 mL. Menadione reductase activity was determined by recording
the initial velocities in the presence of increasing naphthoquinone
concentrations (0-300 µM) after addition of aliquots of the prereacted
enzymes. Menadione was first dissolved in DMSO, and a final 1%
DMSO concentration was kept constant in the menadione reductase
assays. To measure the intrinsic NADPH oxidase activity in the absence
of the naphthoquinone, the reaction was started by adding GR as the
last component in the cuvette containing 1% DMSO. For the determi-
nation of Km and Vmax values, the steady-state rates were fitted by using
nonlinear regression analysis software (Kaleidagraph) to the Michaelis-
Menten equation.20 From these values, the turnover number kcat and
the catalytic efficiency kcat/Km were calculated. The initial rate for
intrinsic NADPH oxidation activity was compared with the rates
measured in the presence of the naphthoquinone for unreacted and
alkylated enzymes.
Rapid-Reaction Studies with the EH2 and EH4 Species of P.
falciparum Glutathione Reductase. EH2 and EH4 were prepared by
titrating anaerobic enzyme with NADPH in a tonometer equipped with
a side arm cuvette and an attached gastight syringe containing the
NADPH solution. Enzyme solutions for rapid-reaction studies were
made anaerobic in glass tonometers by repeated cycles of evacuation
and equilibration over an atmosphere of purified argon. Reduction of
24.7 µM P. falciparum GR with 1 equiv of NADPH led to the thiolate-
flavin charge-transfer complex EH2(FAD)(SH)2 with its characteristic
absorbance at 540 nm. The preparation of the EH4 form was performed
separately by using an NADPH-regenerating system consisting of
glucose-6-phosphate dehydrogenase and glucose 6-phosphate as previ-
ously described.4 In the titration experiment, the anaerobic cuvette
contained 28 µM GR (4 mL), 1 mM glucose 6-phosphate and 1 unit of
glucose-6-phosphate dehydrogenase (at 25 °C), and 7.7 µM NADPH
and the reaction was allowed to proceed for 3.5 h. Formation of EH2
and EH4 was assessed spectrophotometrically using a Hewlett-Packard
8452A diode array spectrophotometer. Stopped-flow experiments were
performed on a Hi-Tech SF-61 DX2 instrument at 4 °C. Solutions of
menadione for rapid reaction studies were made anaerobic by bubbling
them with purified argon within the syringes that were to be loaded
onto the stopped-flow instrument. Time courses were fitted to the sum
of two or more exponential functions (eq 1, where At is the absorbance
at time t, A1 to A3 are the amplitudes of the phases of reaction described
by the observed pseudo-first-order rate constants, k1 to k3, and c is a
constant).
At ) A1e-k t + A2e - k t + A3e-k t + c
(1)
1
2
3
Protein Modification and Crystallization. Recombinant human GR
purified according to ref 21 had a specific activity of 182 U/mg (9500
units/µmol of subunit). For modifying the enzyme with fluoro-M5, 36
µg/mL GR subunits were added to 100 mL of GR assay buffer, 5 µM
fluoro-M5 (7.3 equiv/enzyme subunit, 0.05% DMSO final), and 100
µM NADPH at room temperature. GR activity was measured every
15 min using the GR assay. After 90 min, inhibition of 96% was
observed. Subsequently, the incubation mixture was concentrated by
centrifugation through a Centriprep YM-30 filter device (Amicon) at
3000g to give a final volume of 300 µL. The protein was then dialyzed
overnight at 4 °C against the crystallization buffer (100 mM potassium
phosphate, pH 7.5, 120 mM ammonium sulfate). Crystals were grown
at 15 °C using the hanging-drop technique in 24-well plates with 100
mM potassium phosphate, pH 8.0, and 640 mM ammonium sulfate as
(19) Gromer, S.; Merkle, H.; Schirmer, R. H.; Becker, K. Methods Enzymol.
2002, 347, 382-394.
(20) Segel, I. H. Enzyme Kinetics. BehaVior and analysis of rapid equilibrium
and steady-state enzyme systems; John Wiley and Sons: New York, 1975.
(21) Nordhoff, A.; Bu¨cheler, U. S.; Werner, D.; Schirmer, R. H. Biochemistry
1993, 32, 4060-4066.
9
10786 J. AM. CHEM. SOC. VOL. 128, NO. 33, 2006

Fluorine-Based Suicide Substrate of Glutathione Reductase
A R T I C L E S
Scheme 3. Synthesis Fluoro-M5 and Analoguesa
a Conditions: (a) (1) 3.0 equiv of SnCl2/HCl, EtOH, rt; (2) 3.0 equiv of (CH3O)2SO2, 4 equiv of KOH, acetone, 60 °C; (b) (1) 1.05 equiv of NBS, CCl4,
2 h, reflux; (2) 1.05 equiv of NBS, 0.1 equiv of BPO, 2 h, reflux; (c) 2 equiv of TBAF‚3H2O, CH3CN, 2 h, rt; (d) 1.1 equiv of BuLi, 15 min, -78 °C, then
quenching; (e) 3 equiv of CAN, CH3CN/H2O, 15 min, rt; (f) 3.0 equiv of pimelic acid or 1,4-phenylenediacetic acid, 0.1 equiv of AgNO3, 1.3 equiv of
NH4S2O8, 2.5 h, 70 °C.
the reservoir buffer. In the drop, the initial protein concentration was
∼10 mg/mL, and the concentration of fluoro-M5 was 100 µM. Prior to
the X-ray diffraction experiment, crystals were soaked in buffer (10
mM potassium phosphate, pH 7.5, 1.5 M ammonium sulfate) with an
initial concentration of 0% glycerol and transferred in 10 steps to buffers
with increasing glycerol content to give a final concentration of 20%
glycerol.
Results
Chemistry. Initial attempts to synthesize 2-(fluoromethyl)-
1,4-naphthoquinone (5) by bromination of menadione and
subsequent nucleophilic substitution by fluoride using tetrabu-
tylammonium fluoride trihydrate (TBAF) were not successful.
Neither the variation of the conditions with respect to solvent
and pH nor the use of inorganic fluorides gave the desired
product. This is most probably due to the basicity of fluoride
that promotes elimination processes, resulting in complete
degradation. We have therefore followed a route via reduced
and O-methylated menadiol 1 (Scheme 3, entry 1). The
subsequent step involves the bromination of the aromatic core
and the side chain to yield 2-bromo-3-(bromomethyl)-1,4-
dimethoxynaphthalene (2).26 The benzylic bromine can be
quantitatively substituted by fluorine by incubation with hy-
drated tetrabutylammonium fluoride in acetonitrile.27 This
method is straightforward and more convenient than using alkali-
metal or silver fluoride.28,29 Debromination of the fluoro
intermediate 3 was performed by BuLi treatment followed by
quenching with AcOH. The protonated product 4 was subse-
quently oxidized to the respective naphthoquinone 5 by cerium
ammonium nitrate (CAN), which is a well-established oxidation
reagent in naphthoquinone chemistry.26,30 The introduction of
the carboxyalkyl chain was performed using a silver(II)-
mediated radical decarboxylation reaction previously reported
Data Collection and Structure Determination. All diffraction data
were recorded by the rotation method and processed by XDS, which
includes routines for space group determination, scaling, and conversion
to reduced structure factor amplitudes.22 The crystals obey P21 space
group symmetry and contain two monomers in the asymmetric unit.
The data set (Table S1, Supporting Information) was collected at 100
K using synchrotron radiation at the ESRF (Grenoble, France, beamline
ID14-4, rotation/image 1.0°, wavelength 0.93927 Å, Quantum4 CCD
detector (ADSC)). Reflection phasing and structure refinements were
carried out using the CNS program of Bru¨nger et al.,23 and model
correction was done with the graphic program O.24 The structure was
solved by molecular replacement, using unmodified human glutathione
reductase (1gra, 1) as the search model. The initial difference map (|Fo|
- |Fc|) revealed the presence of electron density showing M5 linked
via a thioether bond at the active sites of the two monomers in the
asymmetric unit. The model displays good stereochemistry as verified
by Procheck.25 The geometry of 6-[2′-(3′-thiomethyl)-1′,4′-naphtho-
quinolyl]hexanoic acid was built with GaussView and refined with
molecular mechanics methods using the UFF force field (Gaussian,
Inc.).
(22) Kabsch, W. J. Appl. Crystallogr. 1993, 26, 795-800.
(23) Bru¨nger, A. T.; Adams, P. D.; Clore, G. M.; DeLano, W. L.; Gros, P.;
Grosse-Kunstleve, R. W.; Jiang, J. S.; Kuszewski, J.; Nilges, M.; Pannu,
N. S.; Read, R. J.; Rice, L. M.; Simonson, T.; Warren, G. L. Acta
Crystallogr., D 1998, 54 (Part 5), 905-921.
(24) Jones, T. A.; Zou, J. Y.; Cowan, S. W.; Kjeldgaard. Acta Crystallogr., A
1991, 47 (Part 2), 110-119.
(25) Laskowski, R. A.; MacArthur, M. W.; Moss, D. S.; Thornton, J. M. J.
Appl. Crystallogr. 1993, 26, 283-291.
(26) Van Tuyen, N.; Kesteleyn, B.; De Kimpe, N. Tetrahedron 2002, 58, 121-
127.
(27) Albanese, D.; Landini, D.; Penso, M. J. Org. Chem. 1998, 63, 9587-9589.
(28) Silverman, R. B.; Oliver, J. S. J. Med. Chem. 1989, 32, 2138-2141.
(29) Fritz-Langhals, E. Tetrahedron: Asymmetry 1994, 5, 981-986.
(30) Terada, A.; Tanoue, Y.; Hatada, A.; Sakamoto, H. Bull. Chem. Soc. Jpn.
1987, 60, 205-213.
9
J. AM. CHEM. SOC. VOL. 128, NO. 33, 2006 10787

A R T I C L E S
Bauer et al.
Table 1. IC50 Values for M5 Derivatives as Inhibitors of P.
falciparum and Human Glutathione Reductases
e10 µM. Under these specific conditions M5 is the most potent
known inhibitor of both P. falciparum GR and human GR,
displaying IC50 values of 4.5 and 3.2 µM, respectively.14 The
bioisosteric replacement of three methylene groups by a phenyl
group in 7 improved the inhibitory potency of the parent M5
with respect to the human enzyme and in the fluoro analogue
6 with respect to the Plasmodium enzyme (Table 1).
IC50
(µ
M)
IC50 (µM)
P. falciparum
GR assaya
human
GR assaya
P. falciparum
GR assaya
human
compd
compd
GR assaya
menadione
M5
fluoro-M5
42.0
4.5b
6.4
27.5
3.2b
4.1
6
7
2.0
7.7
3.5
1.4
To investigate the inhibition of human GR at varying GSSG
concentrations (40-1000 µM), kinetic studies were first per-
formed with saturating NADPH concentration (100 µM) in the
presence of fluoro-M5 (0-10 µM). A Ki value of 5.4 ( 0.4
µM for fluoro-M5 was determined as shown in panel C, Figure
S1 (Supporting Information). The Cornish-Bowden plot (panel
A, Figure S1) as well as the Lineweaver-Burk plot (panel B,
Figure S1) for fluoro-M5 was consistent with noncompetitive
inhibition, whereas equations for competitive and uncompetitive
inhibition could not accurately describe the observed relationship
between V and GSSG concentration.
Glutathione Reductase-Catalyzed Naphthoquinone Re-
ductase Activity. The ability of P. falciparum GR to reduce
the naphthoquinone moiety was studied by following the
oxidation of NADPH in the presence of 7. The naphthoquinone
reductase activity catalyzed by P. falciparum GR was compared
to the intrinsic NADPH oxidation activity of the enzyme in the
absence of naphthoquinone (NQ) (eq 2). P. falciparum GR
displayed a 7.4-fold higher NADPH oxidation activity in the
presence of 200 µM 7 (eq 3 or 4),
a The values were determined at pH 6.9 and 25 °C in the presence of 1
mM GSSG. b Data from ref 14.
for M5 (Scheme 3, entry 2) and applied to combinatorial
synthesis of naphthoquinones.16,31 The reaction leads to a very
selective alkylation at the free position of the quinone moi-
ety.16,32 This reaction was applied for the preparation of the new
carboxylic acids fluoro-M5, 6, and 7 (Scheme 3, entries 1 and
2) starting from pimelic acid and 1,4-phenylenediacetic acid
and the menadione derivatives, menadione itself, and the
(fluoromethyl)menadione 5.
Electrochemical Reduction of Substituted Menadione
Derivatives. The redox potentials of the non-fluoro- and the
fluoronaphthoquinones were determined by cyclic voltammetry
to study the influence of the fluorine atom on the π-π electron
acceptor properties (Table S2, Supporting Information). For
comparison with menadione, the introduction of the hexanoic
acid chain at the C-3 of menadione in M5 produced a shift to
more negative potentials on both waves showing one-electron-
and two-electron-transfer reactions (Table S2, Supporting
Information). Thus, as we previously reported,14 M5 is a poorer
substrate of both GRs than menadione because it is less readily
reduced. Two interesting conclusions can be drawn from the
electrochemical behavior of the new compounds. First the
substitution of the benzyl chain at C-3 of menadione leads to a
more positive E1° value (-0.69 V) for 7 compared to the E1°
value (-0.75 V) for M5. These values might suggest stronger
electron-donating properties of the hexane chain at C-3 in M5
versus the benzyl chain in 7. The presence of one fluorine atom
at the methyl side chain of both menadione derivatives fluoro-
M5 and 6 shifts both one-electron- and two-electron-reduction
waves to more positive potentials, rendering these (fluorom-
ethyl)menadione derivatives more reducible than the non-fluoro
analogues. This behavior was favorable for the design of fluoro-
based suicide substrates for both GRs. The combined effects
due to the presence of both the fluorine atom at the methyl
position and the benzyl chain at the C-3 of the menadione core
are additive in the resulting GR inhibition, as indicated by the
low IC50 values determined for 6 versus fluoro-M5 in both GR
assays and versus M5 in P. falciparum GR assays (Table 1).
Inhibition of Human Glutathione Reductase under Steady-
State Conditions. Determination of the IC50 values of the
compounds fluoro-M5, 6, and 7 in comparison with M5 under
steady-state conditions was performed in the assay using 1 mM
GSSG. As previously discussed, this high GSSG concentration
does not mimic physiological conditions but cell-pathological
conditions where the GSSG level starts to be toxic for the
parasite. Very few inhibitors described earlier in the literature
were found to be potent inhibitors in GR assays: methylene
2O2 + NADPH + H+ f 2O2•- + NADP+ + 2H+ (2)
2NQ + NADPH + H+ f 2NQ•- + NADP+ + 2H+
NQ + NADPH + H+ f NQH2 + NADP+
(3)
(4)
the NADPH oxidation rates being 0.012 U/mg of protein in the
absence of naphthoquinone and 0.089 U/mg of protein in the
presence of 200 µM 7. The Km and kcat values, 115 ( 23 µM
and 0.24 ( 0.02 s-1 corresponding to Vmax ) 0.25 U/mg of
protein, were derived from measurements at six different
substrate concentrations. When NADPH consumption was
followed, 7 was reduced by P. falciparum GR with a catalytic
efficiency kcat/Km of 2.1 × 103 M-1 s-1, a value 1.1-fold and
2.3-fold higher than the values determined for menadione and
M5,14 respectively. When 7 was studied as a substrate of human
GR, the Km and kcat values were determined as 355 ( 91 µM
and 0.34 ( 0.06 s-1, respectively. This results in a catalytic
competence kcat/Km of 0.9 × 103 M-1 s-1, a value 5.6-fold lower
than that determined for menadione. For comparison, M5 when
measured up to 200 µM was found not to be a substrate of the
human GR.
Reaction of Fluoro-M5 with Glutathione Reductases from
Man and P. falciparum. The binding locus of the lead structure
M5 has remained unknown. On the basis of the kinetic data, it
is clear that the parent compound M5sfrom which our new
mechanism-based inhibitor fluoro-M5 derivessis an uncom-
petitive inhibitor with respect to both substrates of glutathione
reductase, NADPH and GSSG.14 This was found to be true for
both glutathione reductases from man and from P. falciparum.
Since cocrystallization of M5 and M5-derived reversible inhibi-
tors with the respective enzymes was not successful, we
followed the approach of labeling the binding site by an
14
blue, some arylisoalloxazines, and M5 displayed IC50 values
(31) Salmon-Chemin, L.; Lemaire, A.; De Freitas, S.; Deprez, B.; Sergheraert,
C.; Davioud-Charvet, E. Bioorg. Med. Chem. Lett. 2000, 10, 631-635.
(32) Anderson, J. M.; Kochi, J. K. J. Am. Chem. Soc. 1970, 92, 1651-1659.
9
10788 J. AM. CHEM. SOC. VOL. 128, NO. 33, 2006

Fluorine-Based Suicide Substrate of Glutathione Reductase
A R T I C L E S
Figure 1. Time-dependent inactivation of human and P. falciparum glutathione reductases by fluoro-M5. (A) Human GR (6.85 pmol in 50 µL) was incubated
in the presence of 160 µM NADPH and inhibitor for 0, 5, 10, 15, 30, and 60 min incubation periods. Fluoro-M5 concentrations used were 0 (b), 0.5 (0),
1.25 ([), 2.5 (O), and 5.0 (2) µM. A 2% final concentration of DMSO was present in all incubation mixtures. (B) The kobs data versus [I] were fitted to
eq 5 (see the text), which resulted in the hyperbolic curve (dashed line and zoom in the inset). At low inhibitor concentration the dissociation constant for
fluoro-M5 and the first-order rate constant for irreversible inactivation were determined as KI ) 7 ( 5 µM and ki ) 0.4 ( 0.2 min-1, respectively. (C) The
time dependency of inactivation of P. falciparum GR was revealed by determining the residual activity of different aliquots of incubated enzyme (21 pmol
in 50 µL) for different time periods: 0.25, 5, 10, and 15 min. Fluoro-M5 concentrations used were 0 (b), 10 (0), 25 ([), and 50 (O) µM. A 2% concentration
of DMSO was present in all cases.
Scheme 4. Oxidoreduction of Naphthoquinonesa
irreversible mechanism-based inhibitor closely related to M5 in
structure. This lead compound is a weak substrate of glutathione
reductase from P. falciparum, the kcat and Km values being 0.10
s-1 and 109 µM.14 The specific activity as low as 0.11 U/mg
for P. falciparum GR-catalyzed reduction of M5 does not
interfere with our approach to a mechanism-based inhibitor that
is activated in a reduction process directly catalyzed by the GR
enzyme (Scheme 2). Also the presence of one fluorine atom
may have an influence on the oxidant character of the parent
core M5. The electrochemical properties of the newly synthe-
sized compounds were also evaluated in parallel.
Inactivation of human GR by fluoro-M5 followed pseudo-
first-order reaction kinetics. The experimental data allowed
application of the derivation by Kitz and Wilson33 for irrevers-
ible inactivation. A semilogarithmic plot (panel A, Figure 1) of
the fraction of noninhibited enzyme activity ln(Vi/V0) versus
incubation time yielded linear curves with increasing slopes,
equivalent to the apparent rate constant of irreversible inhibition
(kobs). The kobs values for fluoro-M5 and human enzyme ranged
between 22.7 × 10-3 and 443 × 10-3 min-1 within the log-
linear range of the inhibition curve (panel B, Figure 1). The
secondary plot expressing kobs as a function of inhibitor
concentration followed eq 5, where KI represents the dissociation
Abbreviations: OX, oxidized form; RED, reduced form; SQ rad,
semiquinone radical; QM rad, quinone methide radical; QM, quinone
methide.
suggesting that more than one binding site of the naphtho-
quinone is involved in the presence of different enzyme species
of reacted and unreacted enzymes in mixture (Scheme 5).
Secondary binding sites with lower affinity are well-known for
the related enzyme trypanothione reductase.34 Similar to the
recent observation for a nitrofuran compound and the human
thioredoxin reductase,35 increased naphthoquinone reductase and
NADPH oxidase activities of enzyme alkylated by fluoro-M5
obscure the inactivation of the unreacted enzyme, rendering the
measurement of the residual activity more complex (see the
highest inhibitor concentration and the longest time in panel A
of Figure 1). The results are consistent with our concept of
mechanism-based inhibition by (monofluoromethyl)menadione
derivatives. In additional kinetic experiments (panel C, Figure
ki[I]
kobs
)
(5)
KI + [I]
constant of the inhibitor and ki is the first-order rate constant
for irreversible inactivation. It was illustrated for fluoro-M5 and
the human GR, by a hyperbolic curve only at low inhibitor
concentration, allowing estimation of ki as 0.4 ( 0.2 min-1, KI
) 7 ( 5 µM, and second-order rate constant ki/KI as 1 × 103
M-1 s-1 (inset, panel B, Figure 1). The resulting half-time value
(t1/2) of inactivation of the human GR was determined as 1.7
min. A double-reciprocal replot of kobs versus [I] fitted to a linear
relationship (data not shown). At higher inhibitor concentration,
the curve was not hyperbolic anymore but became polynomial,
(34) Inhoff, O.; Richards, J. M.; Briet, J. W.; Lowe, G.; Krauth-Siegel, R. L. J.
Med. Chem. 2002, 45, 4524-4530.
(35) Millet, R.; Urig, S.; Jacob, J.; Amtmann, E.; Moulinoux, J. P.; Gromer, S.;
Becker, K.; Davioud-Charvet, E. J. Med. Chem. 2005, 48, 7024-7039.
(33) Kitz, R.; Wilson, I. B. J. Biol. Chem. 1962, 237, 3245-3249.
9
J. AM. CHEM. SOC. VOL. 128, NO. 33, 2006 10789

A R T I C L E S
Bauer et al.
zene.38 Although covalent inactivation of both enzymes was
clearly observed, the kinetics of inactivation indicate that fluoro-
M5 interacts in a more complex way with P. falciparum GR
than with human GR (data not shown).
Scheme 5. Sketch of the Postulated Alkylation Mechanism of GR
by Fluoro-M5 and of Oxidation of Reduced GR by Menadionea
Dialysis of the alkylated enzyme solution and subsequent
enzymatic assays clearly showed that the type of inhibition was
indeed irreversible since GSSG reduction was completely
suppressed. Incubation of the enzymes with inhibitor in the
absence of NADPH did not result in any inactivation, indicating
thatsas predicted from our modelsreduction of the enzyme is
a prerequisite for this mechanism-based inactivation.
Naphthoquinone Reductase Activity of Human and P.
falciparum Alkylated Glutathione Reductases. The ability of
human and P. falciparum fluoro-M5-alkylated GRs to reduce
menadione was studied by following NADPH oxidation.
Complete alkylation for both enzymes was revealed by full
inactivation of the glutathione reductase activity as described
in the Experimental Section. The naphthoquinone reductase
activity (eqs 3 and 4) of both alkylated GRs was first compared
to the intrinsic NADPH oxidase activities in the absence of
naphthoquinone (eq 2). Human and P. falciparum alkylated GRs
displayed a 3-fold higher NADPH oxidase activity in the
presence of oxygen (eq 2) than unreacted enzymes; the observed
NADPH oxidation rates in the absence of menadione were 0.036
U/mg for human alkylated GR and 0.031 U/mg for parasitic
alkylated GR. Then the menadione reductase activities of human
and P. falciparum alkylated GRs were compared to those of
the unalkylated enzymes. Human and P. falciparum alkylated
GRs showed increased rates in the reduction of menadione
(Figure 2) compared to their unmodified forms (kcat of 0.16 s-1
for both unreacted GRs).14 The kcat/Km values for human and
P. falciparum alkylated GRs, 1.23 and 3.55 mM-1 s-1 (corre-
sponding to Vmax ) 0.41 U/mg human GR and 0.81 U/mg
parasitic GR) were derived from measurements at five different
substrate concentrations. For the unmodified enzymes from man
and P. falciparum kcat/Km values of 1.50 and 1.94 mM-1 s-1
were determined under the same conditions.
Comparison of the Oxidation of EH2 and EH4 Forms of
P. falciparum Glutathione Reductase by Menadione. We
generated the EH2 and EH4 forms of P. falciparum glutathione
reductase and measured the rates of oxidation of these two
enzyme forms by menadione in anaerobic stopped-flow experi-
ments. The two-electron-reduced enzyme [(EH2)(FAD)(SH)2]
has thiolate-flavin charge-transfer absorbance at 540 nm, while
the four-electron-reduced enzyme (EH4), where both the cys-
teines and the flavin are reduced, has no absorbance from the
charge-transfer complex or oxidized flavin. The reactions of
EH2 and EH4 with menadione were studied in anaerobic
stopped-flow experiments (Figure 2B,C). EH2 reacted very
slowly with 150 µM menadione. After a long lag, oxidized
enzyme, monitored by the increase in absorbance at 485 nm,
appeared after several hundred seconds (Figure 2B). This
corresponded to the disappearance of charge-transfer absorbance
at 540 nm. Fitting the 485 nm trace to sums of exponentials
(eq 1) gave an apparent rate constant for the formation of Eox
of 0.0089 s-1 (a half-life of 77 s); the same value was obtained
a The dashed arrow means that EH4 was formed in an independent
experiment using the NADPH-regenerating system and might be generated
from EH2 or another enzyme intermediate described in Scheme 1.
Naphthoquinone (fluoro-M5 or menadione) reduction by EH4 generates EH2.
Upon reduction by the EH4 species, the reduced naphthoquinone (NQ)
undergoes fluorine elimination. The released highly reactive quinone methide
(QM*) is attacked by the solvent-exposed thiol Cys58, in accordance with
the crystal structure of the human alkylated GR active site. After alkylation
of Cys58, NADPH oxidase and naphthoquinone reductase activities can be
continuously maintained at the expense of NADPH.
1), we observed that Plasmodium GR was also quickly
inactivated upon incubation of the enzymes with NADPH and
fluoro-M5. The first-order rate constant ki for irreversible
inactivation could not be determined because of the concomitant
inactivation of the parasitic enzyme by NADPH itself.36 In
addition, the resulting parasitic GR inhibition is very complex
because the irreversible component resulting from incubation
of enzyme, NADPH, and fluoro-M5, is partly masked by both
the redox activity of the free compound and the increased
NADPH oxidase activity of the alkylated enzyme. This last
activity was also reported for the related P. falciparum thiore-
doxin reductase inactivated by 6,7-dinitroquinoxaline37 and the
human thioredoxin reductase alkylated by chlorodinitroben-
(37) Andricopulo, A. D.; Akoachere, M. B.; Krogh, R.; Nickel, C.; McLeish,
M. J.; Kenyon, G. L.; Arscott, D. L.; Williams, C. H., Jr.; Davioud-Charvet,
E.; Becker. Bioorg. Med. Chem. Lett. 2006, 16, 2283-2292.
(38) Arne´r, E. S.; Bjornstedt, M.; Holmgren, A. J. Biol. Chem. 1995, 270, 3479-
3482.
(36) Schirmer, M.; Scheiwein, M.; Gromer, S.; Becker, K.; Schirmer, R. H. In
FlaVins and FlaVoproteins 1999; Guisla, S., Kroneck, P., Macheroux, P.,
Sund, H., Eds.; Rudolf Weber, Agency for Scientific Publications: Berlin,
Germany, 1999; pp 857-862.
9
10790 J. AM. CHEM. SOC. VOL. 128, NO. 33, 2006

Fluorine-Based Suicide Substrate of Glutathione Reductase
A R T I C L E S
Figure 2. Menadione reductase activity of alkylated glutathione reductases and of EH2 and EH4 forms of P. falciparum glutathione reductase. (A) The
menadione reductase activity of human (O) and P. falciparum (b) GR alkylated by fluoro-M5 was tested by monitoring the oxidation of NADPH at 340 nm,
25 °C, and pH 6.9 under steady-state conditions. Menadione was dissolved in DMSO, and the NADPH oxidation activity was measured at five different
concentrations in duplicate (0-300 µM) in the presence of 1% DMSO. For the determination of Km and Vmax values, the steady-state rates were fitted by
using nonlinear regression analysis software (Kaleidagraph) to the Michaelis-Menten equation,20 and the turnover number kcat and the catalytic efficiency
kcat/Km were calculated. The initial rate for aerobic NADPH oxidation activity of modified P. falciparum GR was not subtracted from the rates measured in
the presence of the naphthoquinone because it proved negligible in comparison to that of GR-catalyzed menadione reductase activity. (B, C) Stopped-flow
absorbance traces after mixing 24.7 µM P. falciparum EH2 (B) or 28 µM P. falciparum EH4 (C) with 150 µM menadione at 25 °C and pH 6.9 and monitored
at the indicated wavelengths. The bold points are the experimental data, and the solid lines represent data fitted according to eq 1. Note the logarithmic time
scales.
for the disappearance of charge-transfer absorbance at 540 nm.
Interestingly, EH4 reacted much faster with menadione, forming
EH2 within ∼1 s. After this prominent initial phase, smaller,
slower phases occurred in the 100 s reactions. Traces were fitted
to sums of exponentials, giving an apparent rate constant for
the formation of EH2 from EH4 of 6.0 s-1 (a half-life of 0.12 s)
at both 450 and 540 nm (Figure 2C). The other phases, which
might be due to traces of EH2 in the EH4 preparation, were
much slower and were not considered catalytically significant
(data not shown). Thus, reduction of 150 µM menadione by
EH4 was about 660-fold faster than by EH2, suggesting that
EH4 is the species primarily responsible for menadione reduc-
tion. In contrast, the reduction of GSSG occurs exclusively by
EH2.
Figure 3. Close-up of the human GR active site showing the alkylation of
Cys58 by the naphthoquinone methide. The superimposed difference
electron density omit map is contoured at 3σ. Relevant residues in the
surroundings of the inhibitor are shown. The FAD molecule is drawn with
lemon-yellow sticks. Inhibitor and residues are depicted with gray (C atoms),
blue (N atoms), red (O atoms), and gold-yellow (Sγ atoms) sticks. The
structure, solved at 1.7 Å resolution, supports our proposed mechanism:
after reduction of fluoro-M5 the reactive intermediate undergoes fluorine
elimination to yield the quinone methide, which is a highly reactive Michael
acceptor. The nucleophile is the most solvent exposed cysteine (Cys58) of
the catalytic site, which is also the residue responsible for the thiol-disulfide
exchange with the substrate GSSG. The crystal structure shows a stoichi-
ometry of one covalently bound inhibitor molecule per enzyme subunit.
Furthermore, defined noncovalent interactions were observed. The car-
boxylate function of the M5 moiety is in ionic interaction with the two
active site residues Arg37 and Arg347 and is stabilized by a hydrogen bond
with Tyr114. The figure was drawn with bobscript and Raster3D.50-52
Crystal Structure of Human Glutathione Reductase
Inactivated by Fluoro-M5. For crystallographic analysis of the
alkylated enzyme we selected the human enzyme because it
crystallizes readily even in the presence of inhibitors. The 3D
structures of unliganded human GR and of various ligand-
protein complexes are known.39,40 Also, it is a good model for
the parasitic enzyme since the crystal structure of unliganded
P. falciparum GR has been solved39 and shows a scaffold very
similar to that of human GR.
We obtained monoclinic crystals (space group P21, Table S1,
Supporting Information) of human GR, alkylated by the
postulated quinone methide generated from fluoro-M5. The
refined structure revealed the biological dimer of human GR in
the asymmetric unit of the crystal. The two catalytic sites are
occupied by M5 linked via a thioether bound. Except for the
first 17 N-terminal residues and the side chains of Phe78 and
Phe78′, the structure is well represented by the electron density.
The side chains of nine residues in the asymmetric unit assume
alternative conformations. Our structure was refined at 1.7 Å
resolution with an Rfree of 25.6% and with an overall temperature
factor of 30.3 Å2 (Table S1, Supporting Information). All
backbone torsion angles are within the allowed region of the
Ramachandran plot. The structural data indicate a clear stoi-
chiometry with one inhibitor molecule bound per enzyme
subunit of the homodimeric GR. The binding site of the inhibitor
is characterized by covalent and noncovalent interactions. The
compound is covalently linked to Cys58 of the active site via
a sulfur-carbon bond at the position previously occupied by
the fluorine atom (Figure 3). This is in full agreement with the
prediction that a nucleophilic residue in the enzyme should easily
add to the highly reactive quinone methide generated as the
inhibitor intermediate formed after reduction and elimination
of the leaving fluorine atom (Scheme 2). The inhibitor is
covalently bound, but it interacts with the enzyme by additional
(39) Sarma, G. N.; Savvides, S. N.; Becker, K.; Schirmer, M.; Schirmer, R. H.;
Karplus, P. A. J. Mol. Biol. 2003, 328, 893-907.
(40) Schulz, G. E.; Schirmer, R. H.; Sachsenheimer, W.; Pai, E. F. Nature 1978,
273, 120-124.
9
J. AM. CHEM. SOC. VOL. 128, NO. 33, 2006 10791

A R T I C L E S
Bauer et al.
noncovalent interactions. One keto group of the naphthoquinone
is bound via a hydrogen bridge to His467′ of the other subunit.
The conformation of the histidine is further stabilized by the
interaction with Glu472′ (Figure 3). The carboxylate function
of the inhibitor side chain is in ionic contact with the active
site residues Arg37 and Arg347 and is hydrogen-bonded to
Tyr114. Arg37 is known to be important for binding the
negatively charged glycine carboxylate in natural substrate
glutathione disulfide.41 This finding is consistent with our
previous finding that the carboxylate function of M5 seems to
be essential for targeting the inhibitor to the active site. On the
basis of the structural information, we can further conclude that
the binding site of the inhibitor is specific and well defined
because no further electron densities could be detected at other
sites.
led to decomposition of the naphthoquinone in enzyme assays,
and the chemical preparation yielded only 2% of the desired
epoxide.28 In our hands, the reported route28 based on the
coupling of a lithium dialkyl cuprate with a bromide failed.
Instead a new route (Scheme 3) was successfully achieved by
improvement of two key steps: (i) Finkelstein exchange of
bromine by fluorine using tetrabutylammonium fluoride, (ii)
alkylation of 2-(fluoromethyl)menadione using the radical
decarboxylation of carboxylic acids by silver nitrate and
ammonium peroxodisulfate. The developed route proved to be
convenient for synthesizing a large array of diversely 3-substi-
tuted 2-(fluoromethyl)menadione derivatives, as illustrated by
the synthesis of fluoro-M5 and its analogue 6. Furthermore, the
rationale given for the GR inhibition mechanism appears to be
valid since inactivation of the enzyme occurred at a high rate
and led to the alkylated protein as hypothesized in Scheme 2
and proven by the crystal structure of human alkylated GR. In
particular, the parasitic alkylated enzyme, inactivated in its
physiological antioxidant function, retained increased NADPH
oxidase activity that might contribute to high oxidative stress
in the parasites along with GSSG accumulation and continuous
waste of the NADPH level.
Antiparasitic and Cytotoxic Activities in Vitro. Inhibition
of the growth of P. falciparum by our naphthoquinones was
evaluated by determining the inhibitor concentration required
to kill 50% of the parasite (ED50 values). The antimalarial
activities of M5 and its derivatives against the intraerythrocytic
P. falciparum strains 3D7 and K1 are given in Table S3
(Supporting Information). The cytotoxicity of the compounds
for human cells was determined in assays using a human
nasopharyngeal carcinoma (KB) cell line to enable an estimation
of the therapeutic index. This parameter is defined by the ratio
of the ED50 value measuring antimalarial efficacy over the ED50
value measuring toxicity. Interestingly, the weaker cytotoxic
effects against human cells are caused by the reversible GR
inhibitors M5 and 7 while the more pronounced toxic effects
are due to the irreversible enzyme inhibitors fluoro-M5 and 6
and to the nonspecific inhibitor menadione. In particular,
compound 7 showed the highest ED50 value against the human
KB cell line, giving a therapeutic index of ca. 47-52 with
respect to K1 and 3D7 strains, respectively. In addition
compound 7 is as effective against the chloroquine-sensitive
strain 3D7 as against the chloroquine-resistant strain K1. The
therapeutic index of menadione is low, ranging from 2.8 to 3.5,
probably because the compound is known to bind to different
targets. The most potent and irreversible inhibitor 6 displayed
higher cytotoxicity against both human cells and parasites,
resulting in low therapeutic indexes of 1.6-0.4 for 3D7 and
K1 strains, respectively. On the basis of these observations, the
activity profile of the newly prepared naphthoquinones might
be selectively exploited to design antimalarial drug candidates
by optimizing the reversible and specific GR-catalyzed redox
cyclers. In contrast, the irreversible GR inhibitors are more
appropriate as antitumoral agents, as illustrated with the severe
generalized GR deficiency after antitumor chemotherapy with
carmustine, also called BCNU (1,3-bis(chloroethyl)-1-ni-
trosourea).42
Kinetic Studies and Electrochemistry. Human GR was
identified as an attractive target for the development of new
antimalarials and antitumor molecules,5 in particular to circum-
vent drug-resistance mechanisms. Among all types of inhibitors,
the turncoat inhibitorssalso called subversive substrates or redox
cyclerssand the irreversible inhibitors are the most promising
ones. As previously shown, most of the subversive substrates,
including menadione itself and its derivative M5, displayed
noncompetitive or uncompetitive inhibition with respect to both
NADPH and GSSG.6,43,44 Especially the uncompetitive type of
inhibition makes these lead compounds so interesting for a
potential drug use because this inhibitor type cannot be
overcome by physiological counteractions such as increase of
the substrate concentration or increase of the enzyme expres-
sion.45 Uncompetitive inhibitors only bind to enzyme-substrate
complexes, implying that any cellular increase in substrate or
enzyme concentration even reinforces inhibition. This can lead
to dramatic cell responses because of the toxic accumulation
of the substrate.45 The identification of the binding sites of
subversive substratesssuch as naphthoquinones, methylene blue,
and nitrofuran derivativessis still the most challenging issue
in drug discovery applied to disulfide reductases. In agreement
with kinetic data obtained with various un/noncompetitive
naphthoquinones, the crystallographic structures of ligand-
human GR complexes revealed that many ligands, including
menadione,46 3,7-diamino-2,8-dimethyl-5-phenylphenazinium
chloride (safranin),46 6-hydroxy-3-oxo-3H-xanthene-9-propionic
acid,43 a series of 10-arylisoalloxazines,44 S-(2,4-dinitrophenyl)-
glutathione,3 and methylene blue,39 are bound in a large cavity
at the dimer interface. The cavity at the interface of the
homodimer in human GR is linked to the GSSG binding site
via two primary channels.16 For instance, menadione was shown
to bind sandwiched by the Phe78/Phe78′ pair in the cavity at
the interface of the homodimer.
Discussion
Design and Chemistry. 2-(Fluoromethyl)menadione deriva-
tives were first designed as suicide substrates of vitamin K
epoxide reductaseswhich is not a flavoenzymesbut turned out
not to be time-dependent inhibitors.28 In addition, the attempted
synthesis of 2-fluoromethyl vitamin K and of its 2,3-epoxide
(43) Savvides, S. N.; Karplus, P. A. J. Biol. Chem. 1996, 271, 8101-8107.
(44) Scho¨nleben-Janas, A.; Kirsch, P.; Mittl, P. R.; Schirmer, R. H.; Krauth-
Siegel, R. L. J. Med. Chem. 1996, 39, 1549-1554.
(41) Stoll, V. S.; Simpson, S. J.; Krauth-Siegel, R. L.; Walsh, C. T.; Pai, E. F.
Biochemistry 1997, 36, 6437-6447.
(42) Frischer, H.; Ahmad, T. J. Lab. Clin. Med. 1977, 89, 1080-1091.
(45) Cornish-Bowden, A. FEBS Lett. 1986, 203, 3-6.
(46) Karplus, P. A.; Schulz, G. E. J. Mol. Biol. 1989, 210, 163-180.
9
10792 J. AM. CHEM. SOC. VOL. 128, NO. 33, 2006

Fluorine-Based Suicide Substrate of Glutathione Reductase
A R T I C L E S
In this study, new M5 analogues were designed and synthe-
sized for optimizing GR inhibition. The bioisosteric replacement
of three methylene groups by a phenyl ring in compound 7 has
a direct effect on the redox behavior of the menadione moiety
by affecting both one-electron- and two-electron-reduction rates
in agreement with their E1° and E2° values (Table S2, Supporting
Information). However, the redox behavior is not the single
criterion that influences the inhibitory capability of subversive
substrates since menadione and 7 possess similar oxidant
properties, i.e., similar E1° and E2° values, but they displayed
very different IC50 values in both GR assays (Table 1). This
suggests that inhibition depends on both the redox potentials
and on structural features governing recognition by the enzyme.
To get a deeper insight into the binding locus of the inhibitors,
we have designed and synthesized irreversible inhibitors based
on M5 and 7. Introduction of one fluorine atom in a â position
to the carbonyl group of the menadione core was aimed at the
conversion of the reversible inhibitors M5 and 7 into mechanism-
based inhibitors or suicide substrates, fluoro-M5 and 6, to
inactivate both enzymes. First, fluorine significantly affected
both one-electron- and two-electron-reduction potentials of the
fluoro-M5 analogues, i.e., E1° and E2° values (Table S2,
Supporting Information). These effects were desired to ef-
ficiently inactivate the enzyme upon reduction. The combined
effects of both the fluorine and the benzyl chain in 6 lead to
the lowest oxidation potentials upon formation of the semi-
quinonic radical and the dianionic form. The formation of one
of these two reactive forms can be proposed as the first step in
the mechanism of enzyme inactivation, in good agreement with
the IC50 values in both GR assays. The choice of a fluorine
atom as a leaving group was also governed by criteria such as
size, stability, specificity, lipophilicity, and bioavailability.
Irreversible inactivation of both GRs by both fluoro-M5
analogues occurred at a high rate in accordance with the t1/2
and ki values determined for GR inactivation. As predicted, the
elimination of fluorine facilitates Michael addition of a nucleo-
phile to the quinone methide (Scheme 2), in accordance with
the crystal structure of the human alkylated GR active site. The
presence of a nucleophilic residue (Cys) near the region where
the naphthoquinone is reduced agrees with our structural data.
The highly reactive quinone methide (Scheme 2) can in principle
be attacked by N-, O-, or S-nucleophiles.18 It was not surprising
to observe that the nascent thiol Cys58, which is the most
reactive and the most solvent-exposed group, was found to be
alkylated. However, inconsistencies between the structural and
kinetic datasthe binding at the GSSG site and the uncompetitive
inhibition type, respectivelysare most likely due to the presence
of both EH2 and EH4 species in mixture and to the reaction of
the generated quinone methide in the disulfide active site of
EH2 where complete alkylation of the most exposed Cys58
occurred (Scheme 5).
noteworthy to mention also that IC50 values of M5, fluoro-M5,
and compounds 6 and 7 of Table 1 are low but not very different.
This is due to the fact that reduction of naphthoquinones is
thought to occur at the reduced flavin of EH4 and not at the
two-electron-reduced species where either the flavin or the
redox-active Cys pair is reduced. Thus, naphthoquinone reduc-
tion and alkylation of Cys58 do not take place on the same
time scale, in agreement with both the similar IC50 values of
M5, fluoro-M5, and compounds 6 and 7 in Table 1 and half-life
values of EH2 (77 s) and EH4 (∼0.12 s) species determined in
stopped-flow kinetics of menadione reduction. These conclu-
sions are mainly based on detailed kinetic studies of the catalytic
mechanism of the diaphorase reactions by Mycobacterium
tuberculosis lipoamide dehydrogenase, a flavoenzyme closely
related to GR.47 Both naphthoquinone and oxygen reductions
were shown to take place at the EH4 level and not at the (EH2)-
(FADH2)(S-S)‚NADP+ complex or EH2.47 Reduction by EH4
is likely to be responsible for noncompetitive or uncompetitive
inhibition behavior published for a large panel of structurally
diverse GR inhibitors. Our EH4 experiment was done with P.
falciparum GR because this reduced enzyme species was
quantitatively produced under O2-free conditions using an
NADPH-regenerating system while the human EH4 form was
never observed. We can assume that EH4 of human GR does
form but is not visible because of the faster formation of EH2,
which is under kinetic control in the catalytic cycle. The
NADPH-regenerating system supplies the electrons that drive
the otherwise thermodynamically unfavored formation of EH4.
Thus, our stopped-flow kinetic data suggest that the EH4 form
of P. falciparum GR, although thermodynamically unfavored
in the physiological catalytic cycle (Scheme 1), is the enzyme
species responsible for menadione reduction (Scheme 5).
X-ray Diffraction Analysis of Fluoro-M5-Modified Human
Glutathione Reductase. Crystals of human GR, reacted with
fluoro-M5 and dialyzed, were grown as described in the
Experimental Section. The difference Fourier maps of the
crystallized alkylated enzyme showed electron density exclu-
sively at the GSSG binding site with a stoichiometry of one
inhibitor molecule per subunit. In the crystal structure of human
GR with bound fluoro-M5 presented here, the inhibitor is clearly
located in the active site, covalently linked to the catalytic
Cys58, which is responsible for the cleavage of the disulfide in
the physiological substrate GSSG. However, both alkylated
enzyme species are still able to reduce menadione with catalytic
efficiencies as high as those previously described for unreacted
enzymes. An interesting observation was that the electron
density in the region of Phe78 and Phe78′ was not well defined.
As recently analyzed by Wang et al.,48 small-molecule binding
can induce subtle changes in certain side chain atoms, playing
an important role in ligand-protein binding. High flexibility
of the side chains was revealed by increased B factors (>40
Å2), most of which are located at interfaces either to the FAD
or to the other subunit.49 High disorder had also been shown to
be associated with the first 17 residues and with Phe78 and
Besides the locus where naphthoquinone binding and reduc-
tion take place, the identification of the primary inhibitor-
enzyme-substrate complex formed in the catalytic cycle
(Scheme 1) represents a major aim for any inhibitor develop-
ment. Our kinetic and structural data with the irreversible and
mechanism-based inhibitor fluoro-M5 unambiguously showed
that reduced Cys58 from the EH2 species, which reacts directly
with GSSG, is not involved in naphthoquinone reduction
because the alkylated enzyme can still reduce menadione. It is
(47) Argyrou, A.; Sun, G.; Palfey, B. A.; Blanchard, J. S. Biochemistry 2003,
42, 2218-2228.
(48) Yang, C. Y.; Wang, R.; Wang, S. J. Med. Chem. 2005, 48, 5648-5650.
(49) Karplus, P. A.; Schulz, G. E. J. Mol. Biol. 1987, 195, 701-729.
(50) Kraulis, P. J. J. Appl. Crystallogr. 1991, 24, 946-950.
(51) Merritt, E. A.; Bacon, D. J. Methods Enzymol. (Macromol. Crystallogr.,
Part B) 1997, 277, 505-524.
(52) Esnouf, R. M. Acta Crystallogr., D 1999, 55, 938-940.
9
J. AM. CHEM. SOC. VOL. 128, NO. 33, 2006 10793

A R T I C L E S
Bauer et al.
Phe78′ in the 1.54 Å crystal structure of unliganded human
GR.49 This is also the case for human GR upon binding of
fluoro-M5. It is noteworthy that these Phe78/Phe78′ residues
took a new, very defined and rigid orientation with B factors
dropping to near 15 Å2 upon binding of menadione46 and
6-hydroxy-3-oxo-3H-xanthene-9-propionic acid.43 However, just
because menadione was found bound in the cavity at the 2-fold
axis does not justify that naphthoquinone reduction occurs at
that site. The electron-transfer mechanisms and pathways
operating between the flavin and naphthoquinone remain
speculative until structural studies identify the residues involved
in the enzyme-naphthoquinone complex. On the basis of these
observations, it cannot be excluded that the perturbation at
Phe78/Phe78′ represents a structural change induced by the
transient binding of fluoro-M5. This increase in local protein
flexibility might be significant for protein function and may
therefore have important implications for structure-based design
of small-molecule ligands; i.e., the possibility of subtle con-
formational changes at side chains of ternary complexes upon
binding of pyridine nucleotide and naphthoquinone is strongly
suggested. Because the cavity is rather close to the GSSG
binding site, it is tempting to suggest that effector binding in
this cavity might play a role in enzyme regulation in vivo by
affecting the redox properties of the flavoenzyme via the R-helix
63-80.49 Thus, the local high flexibility observed in the region
of Phe78 and Phe78′ suggests a future direction of research since
physiological ligands binding in the cavity at the 2-fold axis of
the homodimer are still unknown.
Antiparasitic and Cytotoxic Activities in Vitro. The M5
analogues showed activity against the malarial parasites with
ED50 values below 7 µM for compound 7 against both
chloroquine-sensitive 3D7 and chloroquine-resistant K1 strains
(Table S3, Supporting Information). Introduction of fluorine at
the methyl group of menadione led to the new compounds
fluoro-M5 and 6 with no increased antimalarial activities but
with an improved cytotoxicity profile against tumor cells. This
observation suggests another partition coefficient for the fluoro
compounds and an irreversible mechanism occurring very
rapidly in the human cells before reaching the parasitic
compartments. The (fluoromethyl)naphthoquinones based on
menadione reported here, e.g., compound 6, may be considered
in light of alternative leads for adjuvants in cancer chemo-
therapy.
mostly as starting blocks to derivatize chloroquine and related
4-aminoquinolines as drugs for use against chloroquine-resistant
parasite strains.13 However, apart from their effects as chloro-
quine-resistance reversers, they have also been reported to have
intrinsic antimalarial activity, which in some cases has been
attributed to effects on both the redox homeostasis and glu-
tathione level.6 Still another conclusion can be drawn from the
data presented here: the GR-catalyzed redox-cycling activity
and not the irreversible GR inactivation is the dominant principle
to obtain more powerful antimalarial effects. Our previous
double prodrugs of M5sdesigned to target two essential
pathways in the parasite, i.e., haem polymerization and redox
equilibriumswere revealed to be active in vitro in the low
nanomolar range and in Plasmodium berghei-infected mice.13
Work on a new series of antimalarial double drugs based on
M5 and 7 with remarkable antimalarial activities both in vitro
and in vivo is now in progress.
Acknowledgment. We are grateful to B. Jannack and I.
Ko¨nig, who contributed by excellent technical assistance to the
chemical preparation of compound 7 in bulk and to crystalliza-
tion of human GR, respectively. The X-ray data set was
collected at the ESRF, beamline ID14-4, Grenoble, France, and
we thank Dr. I. Schlichting for her help with data collection.
1
Dr. G. Schilling is acknowledged for recording H, 13C, 19F,
and NMR spectra (Organisch-Chemisches Institut, Heidelberg
University). Our work is supported by the CNRS-DFG program
of the Centre National de la Recherche Scientifique (E.D.-C.
and H.B.) and by the Deutsche Forschungsgemeinschaft (SFB
544 “Control of Tropical Infectious Diseases”, Projects B2
(H.S.) and B14 (E.D.-C.)). H.B., a CNRS Junior Fellow, is
grateful for a fellowship from the CNRS-DFG program. We
thank Dr. C. Danzin (Marion Merrell Research Institute) for
fruitful discussion on fluorine-based inhibitors. We also ac-
knowledge Professor Charles Williams, Jr. and Dave Arscott,
The University of Michigan, for the use of their laboratory in
the fast kinetic experiments and for many helpful suggestions.
E.D.-C. is a Delegate and H.B. a Fellow of the Centre National
de la Recherche Scientifique, France, in the frame of a French-
German cooperation with Heidelberg University, Germany.
Supporting Information Available: Detailed experimental
procedures, spectrosopic data, and elemental analyses for the
preparation and the characterization of the new compounds 4-7
and fluoro-M5, data collection and structure refinement tables,
protocols and results on cyclic voltammetry, and protocols and
data for measuring enzyme kinetics, antimalarial activities, and
cytotoxicity. This material is available free of charge via the
Internet at http://pubs.acs.org.
Above all, the improvement of menadione led to the car-
boxylic acid 7 retaining an antimalarial activity similar to that
of menadione itself against the chloroquine-sensitive 3D7 and
chloroquine-resistant K1 strains, but showing much lower
cytotoxicity against human cells. This suggests that 7, the most
potent M5 analogue, may serve as a lead for further antimalarial
development. Naphthoquinones have been studied in malaria
JA061155V
9
10794 J. AM. CHEM. SOC. VOL. 128, NO. 33, 2006