Design of novel N-(2,4-dioxo-1,2,3,4-tetrahydro-thieno[3,2-d]pyrimidin-7-yl)-guanidines as thymidine phosphorylase inhibitors, and flexible docking to a homology model.

Article abstract of DOI:10.1016/S0960-894X(02)00828-4

A novel class of thymidine phosphorylase (TP) inhibitors has been designed based on analogy to the enzyme substrate as well as known inhibitors. Flexible docking studies, using a homology model of human TP, of the designed N-(2,4-dioxo-1,2,3,4-tetrahydro-thieno[3,2-d]pyrimidin-7-yl)-guanidines as well as their synthetic precursors provide insight into the observed experimental trends in binding affinity.

Full text of DOI:10.1016/S0960-894X(02)00828-4

Bioorganic & Medicinal Chemistry Letters 13 (2003) 107–110
Design of Novel N-(2,4-Dioxo-1,2,3,4-tetrahydro-thieno[3,2-
d]pyrimidin-7-yl)-guanidines as Thymidine Phosphorylase
Inhibitors, and Flexible Docking to a Homology Model
Melissa L. P. Price,a,* Wayne C. Guida,b,c Tara E. Jackson,b Jason A. Nydick,b
Patricia L. Gladstone,a Jose C. Juarez,a Fernando Donatea and Robert J. Ternanskya
aAttenuon, L.L.C., San Diego, CA 92121, USA
bDepartment of Chemistry, Eckerd College, St. Petersburg, FL 33711, USA
cDrug Discovery Program, H. Lee Moffitt Cancer Center and Research Institute at the University of South Florida,
Tampa, FL 33612, USA
Received 22 May 2002; accepted 24 September 2002
Abstract—A novel class of thymidine phosphorylase (TP) inhibitors has been designed based on analogy to the enzyme substrate as
well as known inhibitors. Flexible docking studies, using a homology model of human TP, of the designed N-(2,4-dioxo-1,2,3,4-
tetrahydro-thieno[3,2-d]pyrimidin-7-yl)-guanidines as well as their synthetic precursors provide insight into the observed experi-
mental trends in binding affinity.
# 2002 Published by Elsevier Science Ltd.
Thymidine phosphorylase (TP) catalyzes the reversible
phosphorolysis of thymidine (1) to thymine and
2-deoxyribose-1-phosphate (Scheme 1).1,2 By analogy to
purine nucleoside phosphorylase,3 the reaction likely
proceeds through an oxocarbenium ion-like transition
state. The phosphorylated sugar product is further con-
verted to 2-deoxy-d-ribose which has been shown to
have chemotactic activity in vitro and angiogenic activ-
ity in vivo.4 Therefore, inhibitors of TP may find utility
as suppressors of tumor growth.5
Taiho Pharmaceutical Co. has reported the most potent
TP inhibitor to date, 5-chloro-6-(2-iminopyrrolidin-1-
yl)methyl-2,4(1H,3H)-pyrimidinedione hydrochloride
(2) which inhibits human TP with an IC50 value of 35
nM.6 Our designed thienopyrimidinyl guanidines, 3 and
4, preserve the interactions of the pyrimidine ring in 1
and 2, substitute an isosteric sulfur atom for the chlor-
ine atom in 2, and maintain the positive-charge charac-
ter of the iminopyrrolidine in 2, which like 2, may mimic
the oxocarbenium ion-like transition state in thymidine
phosphorolysis.
Pyrimidine nucleoside phosphorylase (PyNP) (42%
sequence identity using Clustalw7 for alignment against
the human sequence) catalyzes the same reaction as
TP, as well as the conversion of uridine to uracil.
Unlike other structures of TP,8,9 the high-resolution
Scheme 1.
*Corresponding author. Fax: +1-858-622-0517; e-mail: price@
attenuon.com
0960-894X/03/$ - see front matter # 2002 Published by Elsevier Science Ltd.
PII: S0960-894X(02)00828-4

108
M. L. P. Price et al. / Bioorg. Med. Chem. Lett. 13 (2003) 107–110
X-ray structure of PyNP, co-crystallized with pseu-
douridine (5), shows both an open (inactive) conforma-
tion and a closed (presumed active) conformation of the
enzyme.10 Using these coordinates, we have built a
homology model of human TP (HuTP). While others
have built a homology model of HuTP in the open
conformation based on the X-ray structure of Escher-
ichia coli TP,11 to our knowledge this is the first model
that depicts the enzyme in the active conformation.
Flexible docking of the compounds 1–5 and the syn-
thetic precursors 12–14 using the homology model pro-
vide insight into the observed experimental trends in
binding affinity.
iron in refluxing acetic acid/ethanol gave amines 10 and
11,20 which cyclized with sodium methoxide in metha-
nol to provide thieno-pyrimidine diones 12 and 13.21
The desired guanidine salts 3 and 14 were obtained by
treating 12 and 13, respectively, with cyanamide in
acetic acid at 110 ꢀC.
Benzyl guanidine 4 was prepared starting from benzyl
amine, which was converted to the corresponding cyan-
amide 15 with cyanogen bromide and sodium bicarbo-
nate in methanol. Thienopyrimidine dione 12 and 15 in
hexafluoroisopropanol were heated at 110 ꢀC in a sealed
tube for 65 h,22 and guanidine 4 was separated from the
crude reaction mixture by preparative HPLC (Scheme 3).
Structural coordinates for PyNP co-crystallized with 5
(pdb entry: 1brw, chain A)10 were obtained from the
RCSB Protein Data Bank.12 The sequence of HuTP was
obtained from the SwissProt database13 and Swiss-
Model was used to construct the homology model.14
Crystallographic waters were not included in the
homology model. The model was refined via restrained
energy minimization using the MacroModel/BatchMin
software.15 The AMBER* force field was employed
using a distance-dependent dielectric of 4r. Successive,
restrained, energy minimizations were performed using
a harmonic restraining potential on each atom starting
with a restraining force constant of 500 kJ/mol A2. This
force constant was decreased by a factor of ca. 2 until
the final energy minimization was unconstrained.
The quality of our HuTP homology model was exam-
ined with ProCheck.23 For the 20 A shell of residues
surrounding the binding site, 89.0% of the f and c
angles fell into the most favorable region of the Rama-
chandran map and 10.4% were in the additionally-
allowed region. Furthermore, w1 and w2 angles were
found to have fewer eclipsed conformations than a
typical 2 A resolution structure. Finally, the analysis
provided an overall g-factor of 0.1 for our model, where
a g-factor is a measure of the conformational violations
The same force field and dielectric was also used in all of
the flexible docking simulations with the exception that
ligand charges were fit to the electrostatic potential from
ab initio calculations at the HF/6-31G* level using
Jaguar.16 The protein was truncated to a ca. 15 A shell
from any ligand atom in 4. The protein side chains for
twenty residues were fully flexible; the backbone atoms
of these residues were constrained with a force constant
of 100 kJ/mol A2. All other residues were frozen at their
initial positions and thus only their nonbonded poten-
tial energy contributed to the overall potential.
Restricting the conformational freedom of residues
beyond the first shell of interaction with the ligand pro-
vides a necessary balance between docking accuracy and
simulation time. Each flexible docking simulation con-
sisted of 20,000 steps of the LMOD docking search
procedure,17,18 which performs orientational as well as
conformational sampling, using Batchmin.15 In order to
validate the LMOD method, 5 was docked into PyNP
(closed conformation) starting with an arbitrary orien-
tation and conformation for the ligand. The global
minimum had an rmsd of 1.09 A for the uracil ring
atoms. Comparison of the ribose was not possible since
this part of the electron density was ill-defined in the
crystal structure.
Scheme 2. (a) HCl (concd), MeOH, Á; (b) ethyl isocyanatoformate,
CHCl3, Á; (c) Fe(s), AcOH, EtOH, Á; (d) NaOMe, MeOH, Á; (e)
cyanamide, AcOH, Á.
The unsubstituted guanidines 3 and 14 were synthesized
as shown in Scheme 2. The nitro thiophene regioisomers
6 and 7 were prepared and separated according to the
procedure described by Elliott et al.19 The acetyl groups
were removed with HCl in methanol, and the resulting
free amines were treated with ethyl isocyanatoformate
to afford 8 and 9. Reduction of the nitro groups with
Scheme 3. (a) Cyanogen bromide, NaHCO3, MeOH, 0 ꢀC; (b) 12,
HFIP, 100 ꢀC.

M. L. P. Price et al. / Bioorg. Med. Chem. Lett. 13 (2003) 107–110
109
present. For reference, a g-factor of ꢁ0.4 is expected for
a crystal structure of 2 A resolution, and more positive
g-factors would be expected for higher resolution mod-
els. Figure 1 shows a ribbon-type diagram of the Calpha
superimposition of the HuTP model and PyNP.
hydrogen bonds in PyNP are maintained in the HuTP
model as well.
The global minimum conformations from docking of 1
versus 2 (Fig. 3a) and 2 versus 3 (Fig. 3b) are illustrated
below. The proximity of the endocyclic nitrogen atom in
the iminopyrrolidine of 2 to the C40 ribose oxygen atom
in 1 supports the hypothesis that 2 confers inhibition by
transition-state mimicry (Fig. 3a). Furthermore, the
similar binding orientation of 3 suggests that this com-
pound may also resemble the transition state electro-
nically (Fig. 3b and c). In fact, an additional low-energy
docked conformation of 3, which is ca. 1.5 kcal/mol
higher in energy than the global minimum, shows
excellent overlay with 2 (Fig. 3c). It is unclear to what
extent these two conformations are actually populated,
considering that 1.5 kcal/mol is probably within the
inherent error in using a homology model and a force
field that did not include these ligands in the
parameterization set.
In addition to having comparable backbone geometries,
we expect that the arrangement of the conserved resi-
dues in the binding site should be maintained between
the two proteins. To ensure that the model-building
process did not distort this expected arrangement, we
docked 5 into both PyNP and HuTP and compared the
relative positioning of these residues. Figure 2 shows the
overlay of the sidechains; all of the intermolecular
As shown in Table 1, compound 3 has a significantly
lower binding affinity compared to 2. While other active
‘purine-like’ TP inhibitors have been previously repor-
Table 1. Experimental binding affinitiesa
Compd
KI (mM)b
1
2
3
4
—
0.005 (ꢂ0.003)
76.0 (ꢂ6.4)
64.3 (ꢂ1.0)
nac
Figure 1. Structural comparison of PyNP10 (blue) and HuTP (yellow)
homology model.
5
12
13
14
>600d
>300d
>2400d
aThe enzyme assay was performed with recombinant TP as reported
by Wataya24 with the exception that arsenate was substituted for
phosphate to reduce the reversibility of the reaction.
bValues represent the mean of at least three experiments (standard
deviations in parentheses).
cna, not available.
dLow solubility prevented accurate determination of the binding affinity.
Figure 2. Comparison of the PyNP (cyan) and HuTP (red) binding
sites when complexed with 5.
Figure 3. Docked conformations of (a) 1 (green) and 2 (purple) and
(b)–(c) 2 (purple) and 3 (cyan). (a) and (b) show global minimum
conformations; (c) illustrates another low-energy structure for 3. The
protein has been removed for clarity.
Figure 4. Electrostatic potential surface of the HuTP binding pocket
occupied by the benzyl subtituent in 4 (green). Phosphate is shown in
purple.

110
M. L. P. Price et al. / Bioorg. Med. Chem. Lett. 13 (2003) 107–110
ted,25,26 the bulkier thiophene ring warrants considera-
tion that sterics may play a role in the reduced affinity.
However, calculations of the van der Waals contribu-
tions to the intermolecular interaction energies of the
docked complexes (data not shown) refute this sugges-
tion. Estimates of the relative solvation free energies for
binding (ÁÁGsolvation) of 2 and 3 were then obtained by
reducing the solvation energy of the complex by that of
the unbound protein and ligand (single-point calculations
using the GB/SA solvation model);27 the ÁÁGsolvation
for 3 is larger than 2 by ca. 9 kcal/mol. While the mag-
nitude of these relative solvation free energies may be
overestimated, these results indicate that there is an
enhanced desolvation penalty for the more hydrophilic
guanidine which is likely a dominant contributor to the
reduced affinity of 3 compared to 2. Although the
amino compounds 12 and 13 suffer less from this sol-
vation penalty than 3, there is a >10 kcal/mol loss of
favorable intermolecular interactions in the binding site.
References and Notes
1. Iltzsch, M. H.; el Kouni, M. H.; Cha, S. Biochemistry 1985,
24, 6799.
2. Friedkin, M.; Roberts, D. J. Biol. Chem. 1954, 207, 245.
3. Kline, P. C.; Schramm, V. L. Biochemistry 1995, 34, 1153.
4. Haraguchi, M.; Miyadera, K.; Uemura, K.; Sumizawa, T.;
Furukawa, T.; Yamada, Y.; Akiyama, S. Nature 1994, 368, 198.
5. Matsushita, S.; Nitanda, T.; Furukawa, T.; Sumizawa, T.;
Tani, A.; Nishimoto, K.; Akiba, S.; Miyadera, K.; Fukush-
ima, M.; Yamada, Y.; Yoshida, H.; Kanzaki, T.; Akiyama, S.
Cancer Res. 1999, 59, 1911.
6. Fukushima, M.; Suzuki, N.; Emura, T.; Yano, S.; Kazuno,
H.; Tada, Y.; Yamada, Y.; Asao, T. Biochem. Pharmacol.
2000, 59, 1227.
7. Thompson, J. D.; Higgins, D. G.; Gibson, T. J. Nucleic
Acids Res. 1994, 22, 4673.
8. Spraggon, G.; Stuart, D.; Ponting, C.; Finnis, C.; Sleep, D.;
Jones, Y. J. Mol. Biol. 1993, 234, 879.
9. Walter, M. R.; Cook, W. J.; Cole, L. B.; Short, S. A.;
Koszalka, G. W.; Krenitsky, T. A.; Ealick, S. E. J. Biol. Chem.
1990, 265, 14016.
It is worth noting that the bulkier 4 maintains a similar
level of activity to the unsubstituted 3. Our docking
studies indicate that space is available in the binding site
for the extra aromatic ring, and that the electrostatic
potential in this region is positively charged, as shown in
Figure 4. Attempting to modulate the activity of the
compound, by functionalizing the benzene ring or
exploring other aromatic substituents, is a promising
direction for future work. Another attractive approach
for enhancing binding affinity is functionalization at the
6-position on the thiophene ring which should pre-
organize the conformation that best mimics the transition
state.
10. Pugmire, M. J.; Ealick, S. E. Structure 1998, 6, 1467.
11. Cole, C.; Marks, D. S.; Jaffar, M.; Stratford, I. J.; Dou-
glas, K. T.; Freeman, S. Anti-Cancer Drug Des. 1999, 14, 411.
12. Berman, H. M.; Westbrook, J.; Feng, Z.; Gilliland, G.;
Bhat, T. N.; Weissig, H.; Shindyalov, I. N.; Bourne, P. E.
Nucleic Acids Res. 2000, 28, 235.
13. Bairoch, A.; Apweiler, R. Nucleic Acids Res. 2000, 28, 45.
14. SwissModel, 2002. (http://www.expasy.org/swissmod/
SWISS-MODEL.html).
15. Mohamadi, F.; Richards, N. G. J.; Guida, W. C.; Lis-
kamp, R.; Lipton, M.; Caufield, C.; Chang, G.; Hendrickson,
T.; Still, W. C. J. Comput. Chem. 1990, 11, 440.
16. Jaguar, Ver. 4.0. Schrodinger: Portland, OR, 2000.
17. Kolossvary, I.; Guida, W. C. J. Am. Chem. Soc. 1996, 118,
5011.
18. Kolossvary, I.; Guida, W. C. J. Comput. Chem. 1999, 20,
1671.
19. Elliott, R. L.; O’Hanlon, P. J.; Rogers, N. H. Tetrahedron
1987, 43, 3295.
20. Jarvest, R. L.; Connor, S. C.; Gorniak, J. G.; Jennings,
L. J.; Serafinowska, H. T.; West, A. Bioorg. Med. Chem. Lett.
1997, 7, 1733.
21. Stanovnik, B.; Tisler, M. Synthesis 1972, 308.
22. Snider, B. B.; O’Hare, S. M. Tetrahedron Lett. 2001, 42,
2455.
23. Laskowski, R. A.; MacArthur, M. W.; Moss, D. S.;
Thornton, J. M. J. Appl. Cryst. 1993, 26, 283.
24. Wataya, Y.; Santi, D. W. Anal. Biochem. 1981, 112, 96.
25. Hirota, K.; Sawada, M.; Sajiki, H.; Sako, M. Nucleic
Acids Symposium Series 1997, 37, 59.
In summary, we have designed a novel class of TP
inhibitors based on the structural features of known
inhibitors and on the mechanism of catalysis. We have
also built the first homology model of human TP in the
active conformation, and have performed flexible dock-
ing studies using this model in order to analyze the
details of ligand binding. In addition, these results help
rationalize the experimental binding affinities and will
assist further lead optimization studies. Finally, this
novel class of ‘purine-like’ compounds may find utility
as nucleoside mimics with inhibitory activity against
other therapeutic targets.
26. Balzarini, J.; Gamboa, A. E.; Esnouf, R.; Liekens, S.;
Neyts, J.; De Clercq, E.; Camarasa, M. J.; Perez-Perez, M. J.
FEBS Lett. 1998, 438, 91.
Acknowledgements
We thank Professor Steve Ealick of Cornell University
for numerous helpful discussions.
27. Still, W. C.; Tempczyk, A.; Hawley, R. C.; Hendrickson,
T. J. Am. Chem. Soc. 1990, 112, 6127.