The merits of bipartite transition-state mimics for inhibition of uracil DNA glycosylase

Article abstract of DOI:10.1016/j.bioorg.2004.03.001

The glycosidic bond hydrolysis reaction of the enzyme uracil DNA glycosylase (UDG) occurs by a two-step mechanism involving complete bond breakage to the uracil anion leaving group in the first step, formation of a discrete glycosyl cation-uracil anion in

Full text of DOI:10.1016/j.bioorg.2004.03.001

BIOORGANIC
CHEMISTRY
Bioorganic Chemistry 32 (2004) 244–262
www.elsevier.com/locate/bioorg
The merits of bipartite transition-state mimics
for inhibition of uracil DNA glycosylase
Yu Lin Jiang,a Chunyang Cao,a James T. Stivers,a,*
Fenhong Song,b and Yoshi Ichikawac
a
Department of Pharmacology and Molecular Sciences, The Johns Hopkins University School of Medicine,
725 North Wolfe Street, Baltimore, MD 21205-2185, USA
b
Center for Advanced Research in Biotechnology of the National Institute of Standards and Technology,
University of Maryland Biotechnology Institute, Rockville, MD 20850, USA
c
Optimer Pharmaceuticals, Inc., 10130 Sorrento Valley Road, Suite D, San Diego, CA 92121, USA
Received 9 January 2004
Available online 20 March 2004
Abstract
The glycosidic bond hydrolysis reaction of the enzyme uracil DNA glycosylase (UDG) oc-
curs by a two-step mechanism involving complete bond breakage to the uracil anion leaving
group in the first step, formation of a discrete glycosyl cation-uracil anion intermediate, fol-
lowed by water attack in a second transition-state leading to the enzyme-bound products of
uracil and abasic DNA. We have synthesized and determined the binding affinities of unimo-
lecular mimics of the substrate and first transition-state (TS1) in which the uracil base is co-
valently attached to the sugar, and in addition, bimolecular mimics of the second addition
transition state (TS2) in which the base and sugar are detached. We find that the bipartite
mimics of TS2 are superior to the TS1 mimics. These results indicate that bipartite TS2 inhib-
itors could be useful for inhibition of glycosylases that proceed by stepwise reaction mecha-
nisms.
Ó 2004 Elsevier Inc. All rights reserved.
Keywords: Uracil DNA glycosylase; 1-Azadeoxyribose; Transition-state mimics; Enzyme inhibition
* Corresponding author. Fax: 1-410-955-3023.
E-mail address: jstivers@jhmi.edu (J.T. Stivers).
0045-2068/$ - see front matter Ó 2004 Elsevier Inc. All rights reserved.
doi:10.1016/j.bioorg.2004.03.001

Y.L. Jiang et al. / Bioorganic Chemistry 32 (2004) 244–262
245
1. Introduction
As a DNA repair enzyme, uracil DNA glycosylase catalyzes the removal of
uracil bases from DNA that may arise from the deamination of the normal base
cytosine, or by misincorporation of dUTP into DNA during DNA replication
[1]. Although the genome protective role of this enzyme is well-established, re-
moval of uracil is also important or essential for the life cycle of several viruses,
including pox-, herpes- and cytomegalo-, and thus inhibitors of the enzyme could
serve as clinically useful antiviral agents [2–4]. In addition, inhibition of UDG
could enhance the effectiveness of current anticancer therapies such as 5-fluoro-
uracil and methotrexate that lead to increased accumulation of uracil in DNA
[5,6]. Indeed, in yeast UDG knockout strains, high levels of uracil in DNA leads
to cell cycle arrest at a G2 checkpoint [6], supporting the validity of this ap-
proach.
The mechanistic basis for the extraordinary catalytic power of UDG has been
extensively investigated for the enzyme from Escherichia coli and humans [1]. A
key insight from this work is that the enzyme facilitates a stepwise mechanism
(Fig. 1A), that involves complete breakage of the N-glycosidic bond in the first
transition state (TS1), the generation of a discrete oxacarbenium ion-uracil anion
intermediate [7], followed by attack of the nucleophilic water at C10 of the inter-
mediate in the second transition state (TS2). Crystal structures have been ob-
tained of a reactant analogue complex (UW, Fig. 1B) [8], a bimolecular mimic
of TS2 consisting of a cationic 1-aza-20-deoxyribose (1-aza-dR) sugar and the ura-
cil anion (I þ Uꢀ, Fig. 1B), and the reaction products of abasic DNA and uracil
[9]. The cationic 1-aza-dR component of the TS2 mimic has been previously char-
acterized as a tight binding ligand for the UDG–uracil anion binary complex
(KD ¼ 0:5 nM) [10]. The high affinity of this glycosyl cation mimic for the binary
complex arises in large part from favorable electrostatic interactions with a con-
served aspartate [11], the uracil anion [11], and anionic DNA phosphodiester
groups [12]. In contrast, the very similar (but neutral) tetrahydrofuran abasic site
product mimic binds weakly to the enzyme–uracil anion complex (Fig. 1B,
KD > 15 lM) [10], establishing the importance of the sugar cation in promoting
tight binding.
Although the previously characterized bipartite TS2 mimic is a potent inhibitor of
UDG at pH values in which uracil component is anionic [10,11], at neutral pH values
the affinity diminishes greatly, due to the unfavorable equilibrium for deprotonation
of the base (pKaN1 ¼ 9:8) [11,13]. To address this shortcoming we have now synthe-
sized monopartite mimics of TS1 in which the uracil base is covalently attached to
the 1-azasugar in two ways (Fig. 1B), and we have also explored several low pKa ura-
cil analogues as improved coinhibitors with 1-aza-dR. We find that one of the new
TS1 mimics has significantly greater affinity than two substrate analogues, and that
the highest affinity TS2 mimic, consisting of 1-aza-dR and urazole (pKa ¼ 5:8), is su-
perior at neutral pH to the original bipartite construct using uracil (Fig. 1B). The
general merits and limitations of targeting TS1 and TS2 in stepwise glycosylase re-
actions are discussed.

246
Y.L. Jiang et al. / Bioorganic Chemistry 32 (2004) 244–262
Fig. 1. The reaction coordinate of uracil DNA glycosylase (UDG) and reaction coordinate mimics. (A)
UDG uses a stepwise mechanism to hydrolyze the glycosidic bond of deoxyuridine in DNA. The forma-
tion and decay of the oxacarbenium ion intermediate involves two high energy transition states (TS1 and
TS2) that exhibit unique geometric and electronic features that may be mimicked by stable chemical con-
structs. (B) Stable chemical mimics of UDG reaction coordinate species. The 20-fluoro-20-deoxyuridine
substrate analogue and the tetrahydrofuran abasic product analogues have been previously studied
[11,15]. The other reactant and TS analogues are investigated in this work.
2. Materials and methods
2.1. Materials
As previously described, the recombinant UDG from E. coli strain B was purified
to >99% homogeneity using a T7 polymerase-based over expression system [14]. The

Y.L. Jiang et al. / Bioorganic Chemistry 32 (2004) 244–262
247
concentration of the enzyme was determined using an extinction coefficient of
38.5 mMꢀ1 cmꢀ1
.
2.2. Phosphoramidites
All nucleoside phosphoramidites were purchased from Applied Biosystems or
Glen Research (Sterling, VA), except for the 20-b-fluoro-20-deoxyuridine phospho-
ramidite which was synthesized as previously described [15], and the 1-aza-1, 2-dide-
oxy-4a-carba-D
-ribitol 50-trityl-30-phosphoramidite, which was synthesized as
described below.
2.3. 1-Aza-1,2-dideoxy-4a-carba-
-ribitol 50-trityl nucleoside (2)
D
To a solution of 1 (0.6 g, 1.67 mmol) in CH2Cl2 (10 ml) [16], was added NEt3
(0.47 ml, 3.4 mmol) and Fmoc-Cl (0.52 g, 2.0 mmol). The reaction mixture was
stirred under nitrogen for 2 h, and then purified by chromatography on silica with
1
ethyl acetate–hexanes (1:1, v/v) to give product 2 (0.78 g) in 80% yeild. H NMR
(CDCl3, ppm) d 7.77 (m, 2H); 7.58 (m, 2H); 7.43 (m, 19H); 4.37 (m, 3H); 4.22 (m,
1H); 3.62 (m, 2H); 3.25 (m, 2H); 3.10 (m, 2H); and 2.42 (m, 1H).
2.4. 1-Aza-1,2-dideoxy-4a-carba-
-ribitol 50-trityl-30-phosphoramidite (3)
D
To a solution of 2 (0.226 g, 0.39 mmol) in CH2Cl2 (10 ml), was added diisopropyl-
ethylamine (0.2 ml) 2-cyanoethyl diisopropylchlorophosphoramidite (130 ll,
0.57 mmol). The reaction mixture was stirred under nitrogen for 0.5 h, which was pu-
rified by chromatography on silica with ethyl acetate–hexanes–NEt3 (1:1:0.01, v/v/v)
to give product 3 (0.22 g) in 71% yield. 1H NMR (CDCl3, ppm) d 7.78 (m, 2H); 7.60
(m, 2H); 7.42 (m, 19H); 4.37 (m, 3H); 4.24 (m, 1H); 3.78 (m, 3H); 3.51 (m, 3H); 3.29
(m, 1H); 3.10 (m, 3H); 2.60 (m, 3H); 1.14 (m, 12H). 31P NMR (CDCl3, ppm) d 150.2
(s). ESI calc for C48H52N3NaO5P (M + Na) 804, found 804. This amidite is fairy un-
stable in trace amounts of NEt3, which likely catalyzes the cleavage of Fmoc group,
resulting in polymerization of the amidite. Therefore, the amidite should be freshly
made for DNA synthesis.
2.5. Oligonucleotide synthesis
The 4 mer oligonucleotides, UF, UW, I, and / (see Fig. 1B), were synthesized using
standard phosphoramidite chemistry with an Applied Biosystems 390 synthesizer. In
these sequences, UF, 20-b-fluoro-20-deoxyuridine nucleotide; UW, 20-deoxypseudouri-
dine nucleotide; I, 1-aza-2-dideoxy-4a-carba- -deoxyribonucleotide; and /, tetrahy-
D
drofuran abasic site analogue. During synthesis of oligonucleotide containing I, the
coupling time was increased to 10 min. In addition, the time for the trityl cleavage
step was increased to 180 s instead of standard 90 s. These modifications were found
to increase the incorporation efficiency at this step from 30 to 80 %. After synthesis
and deprotection, the oligonucleotides were purified by anion exchange HPLC and

248
Y.L. Jiang et al. / Bioorganic Chemistry 32 (2004) 244–262
desalted by C-18 reversed phase HPLC (Phenomenex Aqua column). The correct
size, purity, and nucleotide compositions of the final products were assessed by
analytical reversed phase HPLC [Phenomenex Aqua column (250 mm ꢁ 10 mm)],
MALDI mass spectrometry, and by enzymatic digestion to the constituent nucleo-
sides (see below). The concentrations of the 4 mer oligonucleotides were determined
by UV absorption measurements at 260 nm, using the pair wise extinction coeffi-
cients for the constituent nucleotides [17].
2.6. Synthesis of TS1 and TS2 mimics
As outlined in Fig. 2B, the TS1 analogues with the sequences AU6AA and
AU5AA were synthesized by derivatizing the 1-nitrogen of AIAA [12], using the ap-
propriate uracil derivative [18].
AU6AA was prepared as follows: to dry AIAA (0.9 lmol), were added H2O
(10 ll), CH3CN (30 ll), diisopropylethylamine (10 ll), and 6-chloromethyluracil
(7 ll of a 29 mg/ml solution in 90:10 MeOH/diisopropylethylamine). The clear solu-
tion was stirred for 3 days at room temperature and dried in vacuo. The final product
was obtained in 50% yield from the residue by HPLC using a C-18 reverse phase col-
umn. Mass (ESI) calc for MW 1181, found 1181.
AU5AA was prepared as follows: to AIAA (0.2 lmol in 20 ll H2O), were added
uracil (7 ll of a 1.1 mg/ml solution H2O) and HCHO (9 ll of a 0.28 mg/ml solution
Fig. 2. Synthesis of 1-aza-deoxyribose 50-trityl-30-phosphoramidite (3) and the synthesis of TS1 mimics.
(A) 3 was obtained through the sequential reaction of 1 with Fmoc-Cl and cyanoethyl-diisopropyl-chlo-
rophosphoramidite. (B) 3 was incorporated into the DNA sequence A3AA (abbreviated as I throughout
the text) using standard solid phase chemistry. I was used as a common synthon for the synthesis of the
two TS1 analogues U5 and U6.

Y.L. Jiang et al. / Bioorganic Chemistry 32 (2004) 244–262
249
in 48:48:4 EtOH/CH3CN/NEt3). The clear solution was then dried in vacuo. To this
mixture, CH3OH (5 ll) and CH3CN (5 ll) were added. The resulting mixture was
sealed and incubated at 80–85 °C for 24 h and dried in vacuo. The final product
was obtained in 12% yield from the residue by HPLC using a C-18 reverse phase col-
umn. Mass (ESI) calc for MW 1181, found 1181.
2.7. Nucleotide composition analysis
The nucleotide compositions of AU5AA and AU6AA were confirmed by digestion
with P1 nuclease and alkaline phosphatase (both obtained from Roche Diagnostics)
followed by separation of the constituent nucleosides using reversed phase HPLC
(Phenomenex C-18 Aqua column, 5 mm ꢁ 250 mm) with monitoring at 260 nm and
isocratic elution (7% CH3CN, 0.1 M TEAA, pH 7.0). The identity of the peaks
was confirmed by comparison with the retention times of authentic nucleoside
standards and the ratio of the peak areas were consistent with the expected stoichi-
ometries and extinction coefficients of these tetramers. The standards for the
20-deoxynucleoside forms of U5 and U6 were synthesized as previously described
[18]. The structures of the deoxynucleoside standards were ascertained by
1H-NMR spectroscopy and ESI–MS analyses.
2.8. Competitive inhibition measurements
For measuring the binding of the single stranded 4 mer UF substrate and / prod-
uct analogue DNA to the free enzyme, competitive kinetic inhibition measurements
were performed using the substrate ApUpAp [19]. Conditions were chosen whereby
[UDG]tot ꢂ [inhibitor] or [ApUpAp], and [ApUpAp] ꢂ Km. Accordingly, Ki could be
obtained directly from a plot of k=k0 against [inhibitor] as shown in Eq. (1), where k
is the observed rate constant (v=½UDGꢃ ) at a given [inhibitor], and k0 is the ob-
tot
served rate constant in the absence of the inhibitor:
k=k0 ¼ 1=ð1 þ ½Xꢃ=KiÞ:
ð1Þ
For these measurements, a sensitive HPLC kinetic assay for monitoring the formation
of the abasic product was employed [20]. All experiments were performed at 25 °C
using TMN buffer at pH 8 (10 mM Tris–HCl, 2.5 mM MgCl2, and 25 mM NaCl). For
determination of the dissociation constants of the uracil analogues in Table 1, an al-
ternative fluorescence-based competitive inhibition kinetic assay was used [10].
2.9. Binding of 1-aza-dR to the UDG binary complex
The dissociation constants for binding of I to the UDG–uracil or UDG–urazole
complex were determined by competition binding measurements in which a 2-amino-
purine (2AP) labeled abasic analogue DNA (/19) was displaced from the EU or
EUz binary complex as previously described [10]. The sequence of /19 has been pre-
viously reported [21], and this DNA construct shows a strong fluorescence decrease
when it binds to the EU binary complex that can be used as a spectroscopic signal in

250
Y.L. Jiang et al. / Bioorganic Chemistry 32 (2004) 244–262
competition binding measurements. Measurements were performed at 25 °C in TMN
buffer at pH 7 and 8. Before performing the competition binding experiments, the KD
values of /19 for the EU and EUz complexes were determined at pH 8 and 7 using
direct binding measurements. In these measurements, the decrease in 2AP fluores-
cence was followed upon titrating a solution containing 200 nM /19 and 1 mM ura-
cil (or urazole) with increasing concentrations of UDG. This concentration of uracil
(or urazole) is over 10-fold greater than the apparent KD of uracil or urazole for the
enzyme. This insures that UDG is saturated with either base, and that the measure-
ments reflect binding of /19 to the enzyme–base binary complex. Excitation was at
320 nm and emission spectra from 340 to 450 nm were collected using a Spex Fluoro-
max 3 spectrofluorimeter. The fluorescence intensity (F) at 370 nm was plotted
against [EU]tot, or [EUz]tot to obtain the KD from Eq. (2), where [X]tot represents
either [EU]tot or [EUz]tot.
1=2
F ¼ F0 ꢀ fðF0 ꢀ Ff Þ½/19ꢃ =2gfb ꢀ ðb2 ꢀ 4½Xꢃ ½/19ꢃ Þ g;
ð2Þ
tot
tot
tot
b ¼ KD þ ½Xꢃ þ ½/19ꢃ :
tot
tot
To determine the affinity of I for the EU and EUz binary complexes, titrations
included a saturating concentration of uracil (1 mM) so that at the beginning of
the titration UDG was completely bound as E ꢄ U or E ꢄ U ꢄ /19. The concentrations
of /19 and UDG in the individual experiments are reported in the legends to Figs. 4
and 8. The dissociation constants of I for the EU or EUz complex (KDI;EU) were then
determined using the computer program Dynafit [22] and the equilibria shown in
Eqs. (3) and (4), employing the known dissociation constants of /19 for the EU
and EUz complexes as determined from Eq. (2) above:
KD/;EU
E ꢄ U ꢄ /19 ꢀ E ꢄ U þ /19
ð3Þ
ð4Þ
KDI;EU
E ꢄ U ꢄ I ꢀ E ꢄ U þ I
1
2.10. H NMR spectroscopy
Samples (0.5 ml in 90% H2O and 10% D2O for frequency lock) contained 0.3 mM
UDG, 2 mM uracil, 5-azauracil or urazole, 10 mM NaH2PO4 (pH 7.5 or 9), and
150 mM NaCl. The samples were placed in 5 mm NMR tubes (Wilmad 535-PP, Bue-
na, NJ) and sealed with parafilm. The experiments were collected at 25 °C on a Var-
ian INOVA 500 MHz spectrometer using a binomial 1-5-10-5-1 pulse sequence that
minimizes excitation of water [23]. Acquisition and processing parameters were: 2k
complex points, 68 ms acquisition time, and 15 Hz line broadening.
2.11. Computational modeling
The structural models and electrostatic potential surfaces shown in Fig. 9 were
obtained with the program Spartan Pro (Wavefunction) using semi-empirical meth-

Y.L. Jiang et al. / Bioorganic Chemistry 32 (2004) 244–262
251
ods (HF/AM1). The truncated models included all atoms of the base and sugar, but
the 50 and 30 phosphodiester groups were omitted and replaced with hydrogen atoms.
The reactant structure was obtained directly from the coordinates of pdb deposition
1EMH without any further optimization before calculating the electrostatic potential
surface. The TS1 model was obtained by constraining the sugar in a 30-exo confor-
mation calculated from KIE measurements [7], and orienting the base in the position
observed in the structure of uracil and 1-aza-dR (1QF3), except that the N1 nitrogen
ꢀ
was moved to a distance of 2.75 A from the anomeric carbon, which corresponds to a
dissociative transition state with less than 0.01 bond order to the leaving group [24].
The model for TS2 was obtained directly from the structure of uracil and 1-aza-dR
(1QF3) by substituting a carbon atom for the 1-NHþ2 of the sugar. The geometry op-
timized structures and electrostatic potential surfaces for U5 and U6 in Fig. 9B were
calculated using semi-empirical methods (HF/AM1), and were manually superim-
posed with the structural models for TS1. The model for bound urazole and 1-
aza-dR in Fig. 9C was obtained from 1QF3 by substituting the appropriate atoms
of the uracil base, followed by constrained energy minimization.
3. Results
3.1. Synthesis of novel TS1 analogues
The characteristics of TS1 are an elongated glycosidic bond, a significant positive
charge development on the sugar, and negative charge development on the uracil
leaving group (Fig. 1A). To imitate these attributes, we synthesized the two TS1 mim-
ics shown in Fig. 1B. The first mimic (U6), has the 1-nitrogen linked to the uracil base
at the 6-position through a methylene bridge. This analogue was anticipated to imi-
tate the elongated glycosidic bond and the developing positive charge on the sugar in
a dissociative transition state. In addition, since the glycosidic nitrogen of the uracil
base of U6 is not involved in a covalent bond to the sugar, it is free to lose a proton
and generate an anion in the active site, as previously observed for uracil [13,25]. This
anion would be anticipated to provide stabilization to the glycosyl cation [11]. The
second TS1 mimic (U5), is similar to U6 in that a methylene bridge connects the sugar
and base (Fig. 1B). However, the bridge connects to the 5-position of the uracil, and
therefore, the orientation of the base heteroatoms differs from U6. In addition, the
N1-nitrogen of U5 is not positioned as close to the sugar as U6, and therefore this po-
sition may not ionize and provide stabilization to the glycosyl cation.
The TS1 analogues shown in Fig. 1B were all synthesized from a common 4 mer
oligonucleotide precursor (I) that contained the 1-azadeoxyribose moiety (Figs. 1A
and B). Although UDG does bind more tightly to longer oligonucleotides [10,20],
the 4 mer has been shown to possess all of the binding determinants required for
tight binding of the 1-aza-dR group [10,12,19]. The 4 mer was prepared using stan-
dard solid phase phosphoramidite DNA chemistry using the commercially available
adenosine nucleotide 30-phosphoramidite, and the custom made 1-aza-1,2-dideoxy-
4a-carba-D
-ribitol 50-trityl-30-phosphoramidite (3, Fig. 1A) [10,16].

252
Y.L. Jiang et al. / Bioorganic Chemistry 32 (2004) 244–262
Once the I synthon was in hand (Fig. 1B), it was fairly straight forward to deriv-
atize its 1-nitrogen with 6-chloromethyluracil or uracil in the presence of formalde-
hyde to yield the U6 and U5 TS1 analogues shown in Fig. 2. These syntheses are
similar to those used previously to obtain analogous 1-aza-uridine nucleosides
[26], but this report is the first to generate these nucleotides in a DNA scaffold. Al-
though the yields are only in the range 12–30%, and the crude reaction products re-
quired purification by high-performance liquid chromatography, sufficient material
was obtained to perform a large number of biochemical experiments.
3.2. Relative binding affinities of substrate, TS1, and bipartite TS2 analogues
Two substrate analogues, UW and UF, were constructed for binding affinity com-
parisons with the TS1 and TS2 analogues described below (Fig. 1B). Dissociation
constants for the substrate and TS1 mimics were determined by a competitive
inhibition kinetic assay as previously described (Fig. 3) [12]. UW and UF were found
to bind with similar affinities of 5.5 and 4.8 lM, respectively, while the TS1 mimic U5
bound 10 to 12-fold more tightly (KD ¼ 0:5 ꢅ 0:04 lM). The other TS1 mimic, U6,
was found to bind with similar affinity as the two substrate mimics (KD ¼ 4:8 ꢅ
0:5 lM), indicating that linking uracil via a methylene bridge to the 6-position is less
effective than to the 5-carbon (i.e., U5). These results suggest that U5 captures some
of the electronic and geometric features of TS1.
An assumption in ascribing the enhanced binding affinity of U5 to mimicry of TS1
is that the 1-nitrogen of the sugar is protonated. Previous NMR and pH studies have
established that the pKa for free 1-aza-dR is 9.5, and that its nitrogen is protonated
when bound to the EUꢀ complex in the pH range 7–9 [11]. Because direct measure-
ment of the 1-nitrogen pKa values in the context of U5 by NMR spectroscopy is not
trivial, and we examined binding of U5 at pH 6.5 to assess whether the measurements
at pH 8 involved the neutral 1-azasugar. In contrast with the anticipated increase in
binding affinity if protonation of the sugar occurred as the pH was lowered from 8.0
Fig. 3. Inhibition of UDG by the substrate analogues UW, UF, and the TS1 analogue U5. A competitive
kinetic inhibition assay was used. The curves are nonlinear best fits to Eq. (1).

Y.L. Jiang et al. / Bioorganic Chemistry 32 (2004) 244–262
253
to 6.5, U5 bound more weakly at pH 6.5, with KD ¼ 8:3 lM. The weaker binding at
pH 6.5 may be partially attributed to protonation of the catalytic aspartate (pKa 6.5)
that is positioned below the a face of the sugar (Fig. 1A) [9,27].
The salient features of TS2 are the lack of covalent bonding between the sugar
and the anionic uracil, and significant positive charge at C10 (Fig. 1). Thus a good
mimic of TS2 is the bipartite combination of uracil and I (Fig. 1B). Binding of I
to EUꢀ complex can be followed by competitively displacing a fluorescent abasic
DNA analogue (/, Fig. 4) [10], and is pH dependent because interaction of this
glycosyl cation mimic is enhanced by the negative charge on the bound uracil base
(pKaEU ¼ 7:5) [25]. As previously shown, I binds very tightly at pH 8.0 with a
KD ¼ ð5 ꢅ 0:4Þ ꢁ10ꢀ4 lM (solid curve, Fig. 4) [10]. However at pH 7, where the
bound uracil is 75% neutral, I binds 300-fold more weakly (KD ¼ 0:14 ꢅ 0:03 lM).
A bar chart is shown in Fig. 5 that compares the relative affinities of the substrate
and TS1 mimics at pH 8.0, and the TS2 mimic at pH 8.0 and 7.0. Although this com-
parison clearly reveals that the bipartite TS2 mimic (I þ Uꢀ) is superior at pH 8.0,
the AchillesÕ heel of this mimic is that the pKa values for the free and bound uracil
are too high to generate the required anionic form at neutral pH, leading to weak
binding of both U and I under physiological conditions. A further discussion of
the merits and limitations of TS2 mimicry is presented later.
3.3. Screening low pKa uracil analogues for improved coinhibitors
One strategy to improve the potency of the TS2 mimic at a physiological pH value
is to find uracil analogues that have decreased pKa values as compared to uracil itself
Fig. 4. pH dependence of I binding to the EU complex. A competitive displacement assay was used in
which a fluorescent abasic site analogue (/) was dissociated from the EU complex upon binding of I
[10]. The solid curve shows the previously obtained fitted curve for binding of I to the EU complex at
pH 8 [10]. The measurements at pH 7 used 1 mM uracil, 330 nM UDG, and 200 nM /. The curve was cal-
culated using the program Dynafit [22].

254
Y.L. Jiang et al. / Bioorganic Chemistry 32 (2004) 244–262
Fig. 5. Relative KD values (lM) of substrate and TS1 analogues at pH 8.0 (left graph axis), and the bipar-
tite TS2 mimic at pH 8.0 and 7.0 (right graph axis).
free
bound
(pKaðN1Þ ¼ 9:8, pKaðN1Þ
¼ 6:4) [13]. With this aim, we screened six uracil an-
alogues with pKa values ranging from 5.8 to 8 for their ability to competitively in-
hibit the reaction of UDG at pH 8 and 7 (Table 1). As a point of reference, uracil
binds 8.2-fold more weakly as the pH is lowered in this range, due to protonation
free
of the bound uracil [25]. Most of the uracil analogues in Table 1 with pKaðN1Þ
values of 6.7–8.0 show 1.5- to 5.9-fold weaker binding as the pH was lowered from
8 to 7, which arises from protonation of both the bound and free uracil. The mag-
nitude of the binding decrement parallels the proton affinities of these analogues (Ta-
ble 1), but it is impossible to be more quantitative about the trends because the pKa
values for the bound uracil analogues are not known. One analogue, urazole
(pKa ¼ 5:8), showed pH independent binding over the pH range investigated
(Fig. 6), and furthermore, bound 14-fold more tightly than uracil at pH 7
(KD ¼ 98 lM) (Fig. 6B, Table 1).
Table 1
pKa values and dissociation constants for uracil analoguesa
pKa
Kd (mM)
K8=K7b
pH 8
pH 7
U
9.8
8.0
7.6
7.4
7.2
6.7
5.8
0.17 ꢅ 0.02
6.2 ꢅ 1.2
1.4 ꢅ 0.1
0.12
0.65
0.27
0.37
0.17
0.44
0.94
5-FU
6-AU
6-CF3-U
CA
9.6 ꢅ 1.6
1.4 ꢅ 0.1
0.39 ꢅ 0.01
0.11 ꢅ 0.01
0.098 ꢅ 0.016
0.43 ꢅ 0.04
0.092 ꢅ 0.008
0.30 ꢅ 0.03
0.58 ꢅ 0.11
0.98 ꢅ 0.02
0.098 ꢅ 0.004
5-AU
Uz
a The pKa values for U, 5-FU, 6-AU, 6-CF3-U, CA, 5-AU, and Uz were obtained from [46–52],
respectively.
b The ratio of the dissociation constants at pH 8 and 7.

Y.L. Jiang et al. / Bioorganic Chemistry 32 (2004) 244–262
255
Because urazole showed favorable binding properties to the free enzyme, we in-
vestigated its binding interactions in more detail using NMR spectroscopy. Previous
NMR studies of uracil binding to UDG have revealed the presence of a very down-
field shifted proton resonance that arises from a hydrogen bond between uracil O2
and the imidazole NH of His187 in the active site (Fig. 1A) [11,13,28]. This hydrogen
bond stabilizes the uracil anion in the active site by 5 kcal/mol and plays an indirect
but essential role in binding of I by facilitating formation of the uracil anion compo-
nent of the electrostatic sandwich that cradles the sugar cation [11]. To directly test
whether the urazole anion persists at neutral pH in the UDG active site, we collected
NMR spectra of UDG in the presence of uracil, 5-azauracil, and urazole at pH 9.0
and 7.5 (Figs. 7A and B). At pH 9.0, uracil and urazole both show the expected res-
onance at 14.6 and 14.8 ppm, but 5-azauracil does not. The absence of the proton
resonance for 5-azauracil is consistent with its weaker binding to the enzyme, and
suggests that this may be due to its improper positioning in the active site. When
the pH is lowered to 7.5, only the downfield resonance of urazole remains, while that
for uracil is broadened beyond the limits of detection. This NMR result strongly sup-
ports the proposal that the pH independent binding of urazole, and its enhanced
Fig. 6. Competitive inhibition of the reaction of UDG with the substrate AUAp (1 lM) by uracil and
urazole at (A) pH 8 and (B) pH 7.

256
Y.L. Jiang et al. / Bioorganic Chemistry 32 (2004) 244–262
Fig. 7. 1H NMR spectra of uracil (U), 5-azauracil (AzU), and urazole (Uz) bound to UDG at (A) pH 9.0
and (B) pH 7.5. The downfield resonance arises from the interaction of the NeH of His187 with the O2
anion of U and Uz (see text).
Fig. 8. Binding of I to the EUz complex at pH 8 and 7. The measurements at pH 8 used 1.4 lM UDG,
200 nM /, and 1 mM urazole (Uz). The measurements at pH 7 used 330 nM UDG, 200 nM /, and 1 mM
urazole (Uz). The curves were calculated with the program Dynafit [22].
affinity as compared to uracil, arises from its reduced pKa, allowing it to be nega-
tively charged at neutral pH.
Because of the enhanced binding of urazole at neutral pH, we investigated this
analogue as a coinhibitor with I (Fig. 8). Using the competitive displacement
fluorescence assay, I was found to bind to the EUzꢀ complex with an affinity of

Y.L. Jiang et al. / Bioorganic Chemistry 32 (2004) 244–262
257
14 ꢅ 4 and 49 ꢅ 5 nM at pH 8 and 7, respectively. Thus at a physiological pH 7.4, I
binds to the EUzꢀ complex with an affinity of about 30 nM as estimated from inter-
polation between these measured values. This affinity is 170-fold greater than the
substrate analogues bind to the free enzyme.
4. Discussion
4.1. Targeting TS1 in stepwise glycosylase reactions
Glycosylase reactions have long been amenable to the design of inhibitors that
mimic features of high energy structures along the reaction coordinate, thereby cap-
turing a portion of the enzymeÕs strong binding energy for these species [29–35].
In the case of a stepwise glycosylase reaction, where a discrete oxacarbenium ion
Fig. 9. Structural and electrostatic potential models for reactant, TS1, TS2, and chemical mimics. (A)
Structural models for the bound substrate analogue UW and the two transition states of the stepwise re-
action catalyzed by UDG. (B) Geometry optimized structural models for U5 and U6 superimposed on
the model for TS1. The electrostatic potentials are plotted on the van der Waals surfaces, which are shown
to the right of each model. (C) Geometry optimized structural model for the bipartite TS2 mimic of uraz-
ole and I superimposed on the model for TS2. The electrostatic potential surface of Uz þ I is shown to the
right. The details of how these models were calculated are described in Section 2.

258
Y.L. Jiang et al. / Bioorganic Chemistry 32 (2004) 244–262
intermediate is formed, it is possible to envision the design of molecules that mimic
TS1, the intermediate, or TS2. In general, which of these mimics becomes the most
useful inhibitor will depend on how closely the unique charge and geometric prop-
erties of these reaction coordinate species are reproduced by the inhibitor. In addi-
tion, to become useful inhibitors, bipartite intermediate and TS2 mimics must also
possess sufficient binding energy to overcome the unfavorable entropic penalties as-
sociated with binding of the two molecular components from solution. A transition
state mimicry strategy employing both 10 and 40-azasugars has been used previously
to potently inhibit several DNA glycosylases [36,37].
Mimics of TS1 in which the sugar and base are covalently tethered must reproduce
the elongated bond lengths of this dissociative transition state [24,38], as well as the
charge distributions on the sugar and base. In addition, the linker must allow favorable
positioning of the leaving group and sugar relative to enzyme groups that form stabi-
lizing interactions with these parts of the substrate. Using the classic pseudo-thermo-
dynamic framework developed by Radzicka and Wolfenden [39], the maximum
binding affinity for a perfect TS1 mimic would be equal to KD ¼ knon=ðkcat=KmÞ, where
k
non is the rate of the uncatalyzed glycosidic bond cleavage reaction. In the case of UDG
and its substrate AUAA, knon=ðkcat=KmÞ ¼ 10ꢀ10 sꢀ1/(3 ꢁ 106 Mꢀ1 sꢀ1) ꢆ 10ꢀ17 M [1,10].
Accordingly, a perfect TS1 mimic would be expected to bind almost twelve-orders of
magnitude more tightly than the UW and UF substrate analogues (Fig. 1B). Although
the TS1 mimics studied here bind as much as 12-fold more tightly than the substrate
analogues, they capture only a tiny fraction of the theoretical interaction energy ex-
pected from a perfect TS1 mimic.
Previous structural and mechanistic studies provide an informative basis for un-
derstanding the comparatively weak binding affinity of the TS1 mimics. In the crystal
structure of UDG bound to DNA containing the substrate analogue deoxypseudo-
uridine (UW), the nucleotide base and sugar are oriented in a highly distorted confor-
mation in which the base is nearly coplanar with the sugar ring, and the C–C
ꢀ
glycosidic bond is elongated (1.55 A, Fig. 9A) [8]. This ground-state conformation
appears to be significantly strained, and likely represents a high energy conformation
approaching that of TS1, which is highly dissociative [7]. A reasonable structural
model for TS1 is shown in Fig. 9A, which was obtained from three pieces of infor-
mation (see Section 2): (i) the crystallographic coordinates of UW [8], (ii) the results
from KIE studies that provide the sugar pucker and C10–N1 bond order in TS1
(<0.01) [7], and the position of the uracil base in the crystal structure of the bipartite
TS2 mimic of uracil and 1-aza-dR [40]. A model of TS1 obtained using this informa-
tion is superimposed with energy minimized models of U6 and U5 in Fig. 9B. As
compared to the glycosidic bond of the substrate analogue UW, the methylene
bridges of U6 and U5 increase the linear distance between the base and sugar from
ꢀ
1.55 to about 2.6 A. However as shown in Fig. 9B, the constraints imposed by the
methylene bridge groups do not allow an orientation of the base and sugar that pre-
cisely mimics that of TS1.
In addition to the relatively poor structural mimicry by U6 and U5, these ana-
logues also poorly match the charge distribution on the sugar and base in TS1.
This can be appreciated from comparison of the electrostatic potential surfaces of

Y.L. Jiang et al. / Bioorganic Chemistry 32 (2004) 244–262
259
these TS1 mimics with that of the TS1 model (Figs. 9A and B). For instance, U5 and
U6 both mimic the positive charge character at C10, but they poorly match the neg-
ative potential on the uracil leaving group, because the electrons of their intact gly-
cosidic bonds are not able to delocalize onto the O2 atom of the base. A potent
purine based TS1 inhibitor (4) analogous to U5 and U6 has been synthesized for
the enzyme purine nucleoside phosphorylase (PNP) from bovine and Mycobacterium
tuberculosis (Mt) [34,41].
The MtPNP follows a dissociative but concerted mechanism and the methylene
bridge analogue (4) presumably mimics the features of this transition-state
(KD ¼ 24 pM) [34]. In contrast, the bovine PNP shows an earlier more associative
transition-state, and accordingly, (4) bound much more weakly [34,42]. For UDG,
neither type of TS1 inhibitor is very effective because of the highly unusual orienta-
tion of the sugar and base, and the optimization of the active site towards stabiliza-
tion of the electronic features of the oxacarbenium ion intermediate and TS2.
4.2. The merits of targeting TS2
The properties of TS2 are a fully dissociated uracil base, which has accumulated a
full negative charge, and a glycosyl cation that will have developed some bonding to
the incoming water nucleophile. The recent crystal structure of I and the uracil anion
bound to human UDG shows that I assumes a distorted conformation in which the
imino nitrogen is displaced from the plane of the sugar ring towards the attacking
water [40]. An implication of this structure is that the planar conformation of the
oxacarbenium ion intermediate is broken in TS2, and that the electrophilic anomeric
carbon migrates to meet the water nucleophile. Thus, although I was originally
thought to mimic the planar oxacarbenium ion intermediate, its nonplanar confor-
mation, and its charge localization on the 10 position, makes it a much better mimic
of TS2. The structure and corresponding electrostatic model of the bipartite TS2 mi-
mic consisting of urazole and I is shown in Fig. 9 for comparison with a model of
TS2 (Fig. 9A), which was constructed from the crystallographic coordinates of
1Q3F. The excellent geometric and electrostatic match between the bipartite inhibi-
tor and TS2 is much better than that observed for any of the TS1 mimics (see above).
The development of bipartite TS2 inhibitors for glycosylase reactions is a nontra-
ditional approach, because a single substrate molecule is deconstructed into two

260
Y.L. Jiang et al. / Bioorganic Chemistry 32 (2004) 244–262
parts, each of which mimics the presumed features of a bimolecular transition state.
This contrasts with the bisubstrate analogue strategy, where two reactants are cova-
lently tethered to capture the binding energy of the two halves, thereby bypassing the
entropic penalty for binding two molecules from solution [43,44]. The inherent dif-
ficulty with this bimolecular approach is that the binding of each component is com-
paratively weak in the absence of the other component. Thus in general, the
individual binding energy of at least one of the parts must be sufficiently large to sig-
nificantly populate the enzyme under physiological conditions. Under such condi-
tions, binding of the first component creates the binding environment that allows
tight binding of the second component, and creation of the high affinity bimolecular
inhibitor complex. In practical terms, this requires that the intracellular concentra-
tion of at least one of the component parts be high relative to its KD. Of course, a
monopartite TS2 mimic could provide a significant entropic advantage over the bi-
partite mimic if the correct structural features could be captured.
In principle, the above thermodynamic problem for a bipartite inhibitor may be
overcome if one of the components is also the last reaction product to be released
(such as the uracil anion in this case) [10]. For such an ordered product release mech-
anism, the second component (1-aza-dR in this case) will bind avidly to the enzyme–
product complex, and the inhibition mechanism will be noncompetitive with respect
to substrate. For UDG, in vitro studies at pH 8.0 have shown that 80 nM I provided
95% inhibition of UDG in the absence of any added uracil coinhibitor, and that this
strong inhibition arose from I binding tightly to the EU product complex [10]. Of
course, both U and I also bind to the free enzyme, but these inhibition pathways will
not be significant given the weak affinities of the individual parts for the free enzyme
at physiological pH. A noncompetitive mode of inhibition is especially advantageous
for targeting a DNA repair enzyme that not only binds to damaged target sites in
DNA, but also binds nonspecifically to nontarget DNA, which is present at high
concentrations in cells.
5. Conclusion
We have explored chemical analogues that mimic the features of several species
along the stepwise enzymatic reaction coordinate of UDG. In general, these findings
show that bipartite TS2 mimics can serve as effective inhibitors for glycosylase reac-
tions that proceed by stepwise mechanisms. In favorable cases, the bipartite ap-
proach can offer advantages over unimolecular TS1 inhibitors because stable
chemical analogues of TS2 may be better able to match the geometries and charge
distributions of the cationic sugar and leaving group. Furthermore, we suggest that
this strategy is likely to be most successful when the following conditions are met: (i)
when one of the inhibitor components is the anionic product leaving group of the
reaction, (ii) when the free and bound forms of the leaving group product have
the correct anionic ionization state at physiological pH, and (iii) when the intracel-
lular concentration of the leaving group product is comparable to or greater than its
KD. In addition, the inhibition is noncompetitive with respect to substrate, which is

Y.L. Jiang et al. / Bioorganic Chemistry 32 (2004) 244–262
261
desirable when high concentrations of substrate are present in the cellular environ-
ment. Several reactions that potentially meet these criteria are those catalyzed by su-
gar-NDP hydrolases or transferases [32,45], nucleoside phosphorylases [42], and
phosphoribosyl transferases [42].
Acknowledgments
This work was supported by National Institutes of Health Grant GM46835
(J.T.S.), and by the DOD Breast Cancer Research Program DAMD17-03-1-1251.
References
[1] J.T. Stivers, Y.L. Jiang, Chem. Rev. 103 (2003) 2729–2759.
[2] R. Chen, H. Wang, L.M. Mansky, J. Gen. Virol. 83 (2002) 2339–2345.
[3] A.K. Millns, M.S. Carpenter, A.M. DeLange, Virology 198 (1994) 504–513.
[4] R.B. Pyles, R.L. Thompson, J. Virol. 68 (1994) 4963–4972.
[5] R.D. Ladner, Curr. Protein Pept. Sci. 2 (2001) 361–370.
[6] B.A. Tinkelenberg, M.J. Hansbury, R.D. Ladner, Cancer Res. 62 (2002) 4909–4915.
[7] R.M. Werner, J.T. Stivers, Biochemistry 39 (2000) 14054–14064.
[8] S.S. Parikh, G. Walcher, G.D. Jones, G. Slupphaug, H.E. Krokan, G.M. Blackburn, J.A. Tainer,
Proc. Natl. Acad. Sci. USA 97 (2000) 5083–5088.
[9] S.S. Parikh, C.D. Mol, G. Slupphaug, S. Bharati, H.E. Krokan, J.A. Tainer, EMBO J. 17 (1998)
5214–5226.
[10] Y.L. Jiang, Y. Ichikawa, J.T. Stivers, Biochemistry 41 (2002) 7116–7124.
[11] Y.L. Jiang, A.C. Drohat, Y. Ichikawa, J.T. Stivers, J. Biol. Chem. 277 (2002) 15385–15392.
[12] Y.L. Jiang, Y. Ichikawa, F. Song, J.T. Stivers, Biochemistry 42 (2003) 1922–1929.
[13] A.C. Drohat, J.T. Stivers, J. Am. Chem. Soc. (2000) 1840–1841.
[14] A.C. Drohat, G. Xiao, M. Tordova, J. Jagadeesh, K.W. Pankiewicz, K.A. Watanabe, G.L. Gilliland,
J.T. Stivers, Biochemistry 38 (1999) 11876–11886.
[15] J.T. Stivers, K.W. Pankiewicz, K.A. Watanabe, Biochemistry 38 (1999) 952–963.
[16] K. Makino, Y. Ichikawa, Tetrahedron Lett. 39 (1998) 8245–8248.
[17] G.D. Fasman, Handbook of Biochemistry and Molecular Biology: Nucleic Acids, third ed., vol. 1,
CRC Press, Boca Raton, FL, 1975.
[18] V.V. Filichev, E.B. Pedersen, Tetrahedron 57 (2001) 9163–9168.
[19] Y.L. Jiang, J.T. Stivers, Biochemistry 40 (2001) 7710–7719.
[20] Y.L. Jiang, K. Kwon, J.T. Stivers, J. Biol. Chem. 276 (2001) 42347–42354.
[21] J.T. Stivers, Nucleic Acids Res. 26 (1998) 3837–3844.
[22] P. Kuzmic, Anal. Biochem. 237 (1996) 260–273.
[23] P.J. Hore, J. Magn. Reson. 61 (1983) 567–570.
[24] V.L. Schramm, Curr. Opin. Chem. Biol. 5 (2001) 556–564.
[25] A.C. Drohat, J.T. Stivers, Biochemistry 39 (2000) 11865–11875.
[26] S.U. Hansen, M. Bols, Acta Chem. Scand. 52 (1998) 1214–1222.
[27] A.C. Drohat, J. Jagadeesh, E. Ferguson, J.T. Stivers, Biochemistry 38 (1999) 11866–11875.
[28] J. Dong, A.C. Drohat, J.T. Stivers, K.W. Pankiewicz, P.R. Carey, Biochemistry 39 (2000).
[29] A. Bulow, I.W. Plesner, M. Bols, Biochim. Biophys. Acta 1545 (2001) 207–215.
[30] H. Liu, X. Liang, H. Sohoel, A. Bulow, M. Bols, J. Am. Chem. Soc. 123 (2001) 5116–5117.
[31] B.L. Mark, D.J. Vocadlo, D. Zhao, S. Knapp, S.G. Withers, M.N. James, J. Biol. Chem. 276 (2001)
42131–42137.
[32] L. Qiao, B.W. Murray, M. Shimazaki, J. Schultz, C.-H. Wong, J. Am. Chem. Soc. 118 (1996) 7653–
7662.

262
Y.L. Jiang et al. / Bioorganic Chemistry 32 (2004) 244–262
[33] G.A. Kicska, L. Long, H. Horig, C. Fairchild, P.C. Tyler, R.H. Furneaux, V.L. Schramm, H.L.
Kaufman, Proc. Natl. Acad. Sci. USA 98 (2001) 4593–4598.
[34] A. Lewandowicz, W. Shi, G.B. Evans, P.C. Tyler, R.H. Furneaux, L.A. Basso, D.S. Santos, S.C.
Almo, V.L. Schramm, Biochemistry 42 (2003) 6057–6066.
[35] A. Lewandowicz, P.C. Tyler, G.B. Evans, R.H. Furneaux, V.L. Schramm, J. Biol. Chem., 2003.
[36] L. Deng, O.D. Schaerer, G.L. Verdine, J. Am. Chem. Soc. 119 (1997) 7865–7866.
[37] O.D. Scharer, H.M. Nash, J. Jiricny, J. Laval, G.L. Verdine, J. Biol. Chem. 273 (1998) 8592–8597.
[38] V.L. Schramm, W. Shi, Curr. Opin. Struct. Biol. 11 (2001) 657–665.
[39] A. Radzicka, R. Wolfenden, Science 267 (1995) 90–93.
[40] M.A. Bianchet, L.A. Seiple, Y.L. Jiang, Y. Ichikawa, L.M. Amzel, J.T. Stivers, Biochemistry 42
(2003) 12455–12460.
[41] L.A. Basso, D.S. Santos, W. Shi, R.H. Furneaux, P.C. Tyler, V.L. Schramm, J.S. Blanchard,
Biochemistry 40 (2001) 8196–8203.
[42] A. Fedorov, W. Shi, G. Kicska, E. Fedorov, P.C. Tyler, R.H. Furneaux, J.C. Hanson, G.J.
Gainsford, J.Z. Larese, V.L. Schramm, S.C. Almo, Biochemistry 40 (2001) 853–860.
[43] W.P. Jencks, Adv. Enzymol. Relat. Areas Mol. Biol. 43 (1975) 219–410.
[44] M.I. Page, W.P. Jencks, Proc. Natl. Acad. Sci. USA 68 (1971) 1678–1683.
[45] P.M. Legler, M.A. Massiah, A.S. Mildvan, Biochemistry 41 (2002) 10834–10848.
[46] E. Kimura, H. Kitamura, T. Koike, M. Shiro, J. Am. Chem. Soc. 119 (1997) 10909–10919.
[47] C. Thibaudeau, J. Plavec, J. Chattopadhyaya, J. Org. Chem. 61 (1996) 266–286.
[48] N.D. Goldberg, J.L. Dahl, R.E. Parks Jr., J. Biol. Chem. 238 (1963) 3109–3114.
[49] C. Heidelberger, D. Parsons, D.C. Remy, J. Am. Chem. Soc. 84 (1962) 3697–3698.
[50] R.C. Hirt, R.G. Schmitt, H.L. Strauss, J.G. Koren, J. Chem. Eng. Data 6 (1961) 610–612.
[51] A. Piskala, J. Gut, Coll. Czechoslovak Chem. Commun. 26 (1961) 2519–2529.
[52] M.J. Bausch, B. David, P. Dobrowolski, C. Guadalupe-Fasano, R. Gostowski, D. Selmarten, V.
Prasad, A. Vaughn, L.H. Wang, J. Org. Chem. 56 (1991) 5643–5651.