Bivalent Transition-State Analogue Inhibitors of Human Glyoxalase I

Article abstract of DOI:10.1021/ol035917s

(Equation Presented). A new class of competitive inhibitors of homodimeric human glyoxalase I has been created by cross-linking two molecules of the transition-state analogue S-(N-4-chlorophenyl-N-hydroxycarbamoyl)glutathione (CHG) through their γ-glutamy

Full text of DOI:10.1021/ol035917s

ORGANIC
LETTERS
2003
Vol. 5, No. 25
4855-4858
Bivalent Transition-State Analogue
Inhibitors of Human Glyoxalase I
Zhe-Bin Zheng* and Donald J. Creighton*
Department of Chemistry and Biochemistry, UniVersity of Maryland,
Baltimore County, Baltimore, Maryland 21250
creighto@umbc.edu
Received September 30, 2003
ABSTRACT
A new class of competitive inhibitors of homodimeric human glyoxalase I has been created by cross-linking two molecules of the transition-
state analogue S-(N-4-chlorophenyl-N-hydroxycarbamoyl)glutathione (CHG) through their γ-glutamyl-NH groups with poly-â-alanyl tethers of
2
differing length: [CHG(â-ala)n]2 suberate diamide (n ) 1−7). The strongest inhibitors of this antitumor target enzyme likely bind simultaneously
to the active site on each subunit to give K values as small as 0.96 nM (n ) 6).
i
Human glyoxalase I (hGlxI) is a homodimeric Zn2+ metal-
loisomerase with one active site per monomer.1 The apparent
physiological function of this ubiquitous enzyme is to
catalyze the conversion of cytotoxic methylglyoxal (as the
GSH thiohemiacetal) to S-D-lactoylglutathione2,3 via a
proton-transfer mechanism involving a Zn2+-bound ene-
diolate intermediate GSC(OH)dC(O-)CH3, where GS equals
glutathionyl.4 Methylglyoxal arises as a normal byproduct
of carbohydrate metabolism5 and is capable of covalently
modifying proteins and polynucleic acids critical to cell
viability.6,7 Inhibitors of the enzyme have long been sought
as possible anticancer agents, because of their potential ability
to induce elevated concentrations of methylglyoxal in tumor
cells8 and the observation that rapidly dividing tumor cells
are exceptionally sensitive to the cytotoxic effects of exog-
enous methylglyoxal, reviewed elsewhere.9
In support of this hypothesis, the dialkyl ester prodrugs
of the hGlxI inhibitors S-(N-4-chlorophenyl-N-hydroxy-
carbamoyl)glutathione (CHG, Ki 46 nM)10 and S-4-bromo-
benzylglutathione (Ki 170 nM)11 inhibit the growth of murine
and human tumors in laboratory mice at dosing levels of 80
and 100 mg/kg, respectively.
The high affinity of CHG for the enzyme appears to reflect
the fact that it is a hydrophobic transition-state analogue of
the GlxI reaction.4,12 Compounds of this type generally bind
in the mid-nanomolar concentration range. Nevertheless,
more potent inhibitors with higher specificities are needed
in order to decrease the dosing levels of inhibitor required
(1) Aronsson, A.-C.; Marmstål, E.; Mannervik, B. Biochem. Biophys.
Res. Commun. 1978, 81, 1235-1240.
(2) Creighton, D. J.; Pourmotabbed, T. In Molecular Structure and
Energetics: Principles of Enzyme ActiVity; Liebman, J. F., Greenberg, A.,
Eds.; VCH Publishers: 1988; Vol. 9, pp 353-386.
(3) Vander Jagt, D. L. In Coenzymes and Cofactors: Glutathione;
Dolphin, D., Poulson, P. R., Avramovic, O., Eds.; John Wiley and Sons:
New York, 1989; Vol. 3, Part A, pp 597-641.
(8) Vince, R.; Daluge, S. J. Med. Chem. 1971, 14, 35-37.
(9) Creighton, D. J.; Hamilton, D. S.; Kavarana, M. J.; Sharkey, E. M.;
Eiseman, J. L. Drugs Future 2000, 25, 385-392.
(4) Creighton, D. J.; Hamilton, D. S. Arch. Biochem. Biophys. 2001, 387,
1-10.
(5) Richard, J. P. Biochemistry 1991, 30, 4581-4585.
(6) White, J. S.; Rees, K. R. Chem. Biol. Interact. 1982, 38, 339-347.
(7) Papoulis, A.; Al-Abed, Y.; Bucala, R. Biochemistry 1995, 34, 648-
655.
(10) Sharkey, E. M.; O’Neill, H. B.; Kovaleva, E. G.; Kavarana, M. J.;
Wang, H.; Creighton, D. J.; Sentz, D. L.; Eiseman, J. L. Cancer Chemother.
Pharmacol. 2000, 46, 156-166.
(11) Sakamoto, H.; Mashima, T.; Sato, S.; Hashimoto, Y.; Yamori, T.;
Tsuruo, T. Clin. Cancer Res. 2001, 7, 2513-2518.
10.1021/ol035917s CCC: $25.00 © 2003 American Chemical Society
Published on Web 11/11/2003

to inhibit tumor growth in vivo and to reduce the possibility
of side effects.9
Scheme 1
Here, we report for the first time that chemically cross-
linking two CHGs with a linker (L) long enough to allow
simultaneous binding to each active site of hGlxI dramatically
increases binding affinity. This general strategy has been used
in the past to increase binding affinities, as the binding
energies at each site can potentially contribute to the overall
free energy of binding.13 The X-ray crystal structure of hGlxI
in complex with one CHG at each active site (Figure 1, PDB
The bivalent transition-state analogues shown in Scheme
1 proved to be competitive inhibitors of glyoxalase I from
human erythrocytes (hGlxI)12b and from yeast (yGlxI); e.g.,
Figure 2.
Figure 1. hGlxI in complex with CHG (ball and stick models).
code 1QIN)14 shows that the γ-glutamyl-NH2 groups of the
bound CHGs are exposed to bulk solvent, are about 30 Å
apart, and therefore could be chemically cross-linked.
In Figure 1, green spheres represent catalytically essential
Zn2+ ions at the active sites, which directly coordinate the
cisoid oxygens of the bound CHGs. The dimer interface
extends vertically through the center of the figure.
A family of potential bivalent inhibitors was prepared by
the successive introduction of n â-alanyl residues at the
γ-glutamyl-NH2 function of CHG, followed by conversion
to a suberate diamide to complete cross-linking (Scheme 1).
Synthetic methods were adapted from published proce-
dures,15 and chemical identities were confirmed by 1H NMR
(500 MHz) and HRMS (Supporting Information).
Figure 2. Reciprocal plot of the initial rate of the hGlxI reaction
versus concentration of GSH-methylglyoxal thiohemiacetal substrate
(S) at different fixed concentrations of [CHG(â-ala)6]2 suberate
diamide [I]. Conditions: 50 mM sodium phosphate buffer, pH 7.0,
25 °C.
Model building of the inhibitors into the X-ray structure
of hGlxI indicated that analogues with n g 4 could bridge
the active sites, Figure 1. Indeed, a comparison of the
competitive inhibition constants with linker length shows a
dramatic increase in binding affinity when n ) 6 or 7,
corresponding to linker lengths of 70 and 80 Å, respectively;
Table 1, Figure 3.
(12) (a) Hamilton, D. S.; Creighton, D. J. J. Biol. Chem. 1992, 267,
24933-24936. (b) Murthy, N. S. R. K.; Bakeris, T.; Kavarana, M. J.;
Hamilton, D. S.; Lan, Y.; Creighton, D. J. J. Med. Chem. 1994, 37, 2161-
2166. (c) Kalsi, A.; Kavarana, M. J.; Lu, T.; Whalen, D. L.; Hamilton, D.
S.; Creighton, D. J. J. Med. Chem. 2000, 43, 3981-3986.
These results suggest that the longest bivalent inhibitors
(n ) 6, 7) span the active sites on each enzyme molecule.
The strongest bivalent inhibitors have Ki values ∼500-fold
smaller than the monovalent inhibitors CHG(â-ala)1, CHG-
(â-ala)5, and CHG(â-ala)6. This corresponds to a ∼250-fold
difference in affinity per CHG group, given that the bivalent
(13) (a) Slon-Usakiewicz, J. J.; Purisima, E.; Tsuda, Y.; Sulea, T.;
Pedyczak, A.; Fethuere, M. C.; Yonishi, Y. Biochemistry 1997, 36, 13494-
13502 and references therein. (b) Lyon, R. P.; Hill, J. J.; Atkins, W. M.
Biochemistry 2003, 42, 10418-10428.
(14) Cameron, A. D.; Ridderstro¨m, M.; Olin, B.; Kavarana, M. J.;
Creighton, D. J.; Mannervik, B. Biochemistry 1999, 38, 13480-13490.
(15) Zheng, Z.-B.; Nagai, S.; Iwanami, N.; Kobayashi, A.; Natori, S.;
Sankawa, U. Chem. Pharm. Bull. 1999, 47, 777-782.
4856
Org. Lett., Vol. 5, No. 25, 2003

as evidenced by the Ki value of 583 nM for CHG(â-ala)6
versus that for CHG.
Table 1. Competitive Inhibition Constants (Ki) of CHG,
CHG(â-ala)n, and [CHG(â-ala)n]2 Suberate Diamide with hGlxI,
Yeast GlxI (yGlxI), and Bovine Liver GlxII (bGlxII)a
Tight binding by the bivalent inhibitors (n ) 6, 7) appears
to reflect the fact that (a) once the first CHG group binds to
the enzyme there is a dramatic increase in the probability of
binding of the remaining CHG group and (b) the observed
affinity constant for the bivalent inhibitor with the enzyme
is the product of the binding constants for each CHG group.
This can be understood on the basis of the following system
of equilibria, Scheme 2.
linker
length
(Å)
Ki, hGlxI
(nM)b
Ki, yGlxI Ki, bGlxII
(µM)c (nM)c
compound
CHG
46 ( 4d
3.6 ( 0.3d 1700 ( 17
CHG(â-Ala)n
n )
1
5
6
4
24
30
330 ( 6
744 ( 39
583 ( 33
1200 ( 12
870 ( 60
Scheme 2
[CHG(â-Ala)n]2sub.
n )
1
2
3
4
5
6
7
21
31
41
50
60
70
80
70.1 ( 1.3
75.9 ( 7.4
84.1 ( 2.4
59.6 ( 2.0
7.5 ( 0.5
3.0 ( 0.2
9.0 ( 0.3
35 ( 4
50 ( 5
0.96 ( 0.06 7.2 ( 0.1
0.97 ( 0.02 4.1 ( 0.5
142 ( 1
79 ( 8
a Conditions: sodium phosphate buffer, 50 mM, pH 7, 25 °C. b Mean
(( SD) for triplicate determinations. c Mean (( SD) for duplicate deter-
minations. d Taken from ref 12b.
In Scheme 2, E ) enzyme, X-L-X is the bivalent
inhibitor with identical binding groups X, connected by the
linker L, and the Ks are defined as binding constants. Under
dilute conditions, where K23 . K24[X-L-X], only the
equilibria associated with K12 and K23 must be considered.
Thus, the binding constant of the bivalent inhibitor, where
X ) CHG (n ) 6), for the enzyme (Ki-1) is directly
proportional to the product of the two microscopic binding
constants (eq 1).
inhibitors contain two CHG groups and the monovalent
inhibitors contain one CHG group. Since the three monova-
lent derivatives bind to the enzyme with about the same
affinity, the linker groups appear not to affect binding affinity
significantly. The 4-fold higher affinities per CHG group for
the shorter bivalent inhibitors (n ) 1-4) versus the control
compounds such as CHG(â-ala)6 might be indicative of a
peripheral site near the active site that can weakly bind a
CHG group, or that the average solution conformation of
the bivalent inhibitors is nearer to that of the bound species,
reducing the unfavorable entropy of binding. The bivalent
inhibitor [CHG(â-ala)6]2 suberate diamide (Ki ) 0.96 nM)
binds about 50-fold more tightly than CHG (Ki ) 46 nM)
to hGlxI. Binding would be tighter, except that amidation
of the γ-glutamyl-NH2 decreases binding affinity by 13-fold,
(Ki)-1 ) K12K23 ) 1 × 109 M-1
(1)
To a good first approximation,
K12 ≈ 2(Kii)-1 ) 3.4 × 106 M-1
(2)
where Kii is the observed inhibition constant for CHG(â-
ala)n, n ) 6. The coefficient of 2 in eq 2 corrects for the
fact that the bivalent inhibitor contains two CHG groups
available for binding while the monovalent inhibitor contains
one available CHG group. Thus, binding of CHG in the
intermediate complex 2 is a highly favorable process, on the
basis of the calculated value of K23, eq 3.
K23 ) 1 × 109 M-1/3.4 × 106 M-1 ) 294
(3)
The concept of effective concentration (EC)16 can also be
used to access the thermodynamic advantage of placing two
interacting groups in the same “molecular entity” versus
Figure 3. Log plot of Ki versus linker length for the bivalent
inhibitors of Table 1 for hGlxI (b) and yGlxII (2). Open symbols
are for CHG alone.
(16) McNaught, A. D.; Wilkinson, A. IUPAC Compendium of Chemical
Terminology, 2nd ed.; Blackwell Science, 1997.
Org. Lett., Vol. 5, No. 25, 2003
4857

having the reacting groups in separate molecules.17 By
replacing the term “molecular entity“ with “complex 2”,
shown in Scheme 2, the EC for unbound CHG in this
complex is obtained by dividing the unimolecular equilibrium
constant by the bimolecular equilibrium constant for binding
of the monovalent inhibitor to the enzyme (eq 4).
residues indicative of two active sites per monomer.20 Indeed,
topological mapping of the yGlxI polypeptide on the crystal
structure of hGlxI indicates the presence of two active sites
that are analogous in structure to that observed in the X-ray
crystal structure of the human enzyme.21 Mutagenesis studies
indicate that both active sites are functional. Nevertheless,
there must be subtle differences between the active sites on
the two enzymes, as CHG binds 78-fold less tightly to yGlxI
than to hGlxI (Table 1). Perhaps there are differences in the
surface topology of yGlxI not found in hGlxI that preclude
simultaneous binding to each active site using the bivalent
inhibitors described in this study. Whatever the explanation,
cross-linking has increased inhibitor selectivity by almost
100-fold, as CHG binds 78-fold more tightly to hGlxI than
to yGlxI, while [CHG(â-ala)6]2 suberate diamide binds about
7500-fold more tightly (Table 1).
EC ) K23/(Kii)-1 ) 174 µM
(4)
The EC significantly exceeds the dissociation constant of
CHG(â-ala)6 (Ki ) 0.58 µM) with the enzyme. Therefore,
binding is a cooperative phenomenon in which the initial
binding of the first tethered CHG group to the first active
site dramatically increases the probability that the second
tethered CHG group will bind to the remaining active site.
For the thioester hydrolase bGlxII, the CHG(â-ala)n
monomers inhibit the enzyme as well as CHG (Table 1),
indicating that the γ-glutamyl-NH2 group of the bound
inhibitor probably extends into bulk solvent and that sub-
stituents at this position do not interfere with binding. In
addition, there is no dramatic increase in binding affinity
with increasing linker length, fully consistent with a single
active site for GlxII. As observed with hGlxI, the affinity of
the bivalent inhibitors for bGlxII is somewhat greater per
CHG group (7-fold) than for CHG(â-ala)6 monomer. Com-
paring the inhibition constants of CHG and [CHG(â-ala)6]2
suberate diamide for hGlxI versus bGlxII, cross-linking
increases binding selectivity from 37-fold to 148-fold, a
4-fold increase. Finally, CHG and [CHG-(â-ala)6]2 suberate
diamide show no inhibitory activity with GSH peroxidase
or human placental GSH transferase up to an inhibitor
concentration of 20 µM.
In summary, the longest bivalent transition state analogues
described here are the strongest competitive inhibitors of
hGlxI yet reported. Moreover, they exhibit improved selec-
tivity over several other GSH-dependent enzymes tested. This
lays the experimental and conceptual foundation for the
development of a new class of powerful inhibitors for this
important antitumor target enzyme. The alkylester prodrug
strategy10 could be used to deliver the bivalent inhibitors into
tumor cells, as the inhibitors can be readily converted to the
tetra-O-ethyl esters by incubation in ethanolic/HCl (Zheng
and Creighton, unpublished).
The maximum EC depends to a significant extent on the
entropy lost on binding.18 In the present case, the magnitude
of the EC is determined not only by the increase in the
average local concentration of the unbound CHG group with
respect to the unoccupied active site in complex 2 over that
of the unbound monovalent inhibitor but also by an “orienta-
tion effect”. This reflects the effect of the linker on the
orientation of the unbound CHG group with respect to the
unoccupied active site in complex 2, which either makes
binding more favorable by reducing unfavorable degrees of
translational and/or rotational freedom or makes binding less
favorable by reducing favorable degrees of translational and/
or rotational freedom. Therefore, the EC value reported here
for n ) 6 is primarily a property of the linker with hGlxI.
This EC can be used to calculate expected Ki values for
bivalent inhibitors in which (â-ala)n suberate diamide chemi-
cally cross-links any one of a number of different GSH
derivatives reported to be either reversible or irreversible
inhibitors of hGlxI.19
To assess the specificity of the bivalent inhibitors for hGlxI
versus other GSH-dependent enzymes, the inhibition studies
were extended to include yGlxI and bGlxII, a thioester
hydrolase for S-D-lactoylglutathione. These enzymes were
of interest because they are both inhibited by CHG12b and
both are monomeric enzymes, although yGlxI has been
argued to contain two active sites.
Yeast GlxI does not show evidence of a dramatic increase
in binding affinity with increasing linker length (Table 1,
Figure 3). Thus, there is no evidence that binding of the
longest bivalent inhibitor to yGlxI involves simultaneous
binding to two active sites. This is somewhat surprising in
view of sequence comparisons indicating that yGlxI is the
result of a gene duplication event, which retained amino acid
Acknowledgment. This work was supported by a grant
from the NIH (CA 59612). We thank Diana S. Hamilton for
helpful discussions.
Supporting Information Available: Kinetic and syn-
thetic methods and analytical data on synthetic products. This
material is available free of charge via the Internet at
http://pubs.acs.org.
(17) (a) Kirby, A. J. AdV. Phys. Org. Chem. 1980, 17, 183-278. (b)
Morawetz, H.; Goodman, N. Macromolecules 1970, 3, 699-700.
(18) (a) Page, M. I.; Jencks, W. P. Proc. Natl. Acad. Sci. U.S.A. 1971,
68, 1678-1683. (b) Jencks, W. P. AdV. Enzymol. 1975, 43, 219-410. (c)
Page, M. I. AdV. Chem., Int. Ed. Engl. 1977, 16, 449-459.
(19) Creighton, D. J.; Zheng, Z.-B.; Holewinski, R. J.; Hamilton, D. S.;
Eiseman, J. L. Biochem. Trans. 2003, 31, 1378-1382.
OL035917S
(20) (a) Ridderstro¨m, M.; Mannervik, B. Biochem. J. 1996, 316, 1005-
1006. (b) Clugston, S. L.; Daub, E.; Kinach, R.; Miedema, D.; Barnard, J.
F. J.; Honek, J. F. Gene (Amsterdam) 1997, 186, 103-111.
(21) Frickel, E.-M.; Jemth, P.; Widersten, M.; Mannervik, B. J. Biol.
Chem. 2001, 276, 1845-1849.
4858
Org. Lett., Vol. 5, No. 25, 2003