Analysis of complexes of inhibitors with Cryptosporidium hominis DHFR leads to a new trimethoprim derivative

Article abstract of DOI:10.1016/j.bmcl.2006.05.047

Cryptosporidiosis, an opportunistic infection affecting immunocompromised patients, the elderly, and children, is still an untreatable disease since the causative agent, Cryptosporidium hominis, is essentially resistant to all clinically used antimicrobia

Full text of DOI:10.1016/j.bmcl.2006.05.047

Bioorganic & Medicinal Chemistry Letters 16 (2006) 4366–4370
Analysis of complexes of inhibitors with Cryptosporidium hominis
DHFR leads to a new trimethoprim derivative
Veljko M. Popov,a David C. M. Chan,a Yale A. Fillingham,a W. Atom Yee,a,
Dennis L. Wrightb and Amy C. Andersonb,*
aDepartment of Chemistry, Dartmouth College, Hanover, NH 03755, USA
bDepartment of Pharmaceutical Sciences, University of Connecticut, Storrs, CT 06269, USA
Received 3 April 2006; revised 16 May 2006; accepted 16 May 2006
Available online 5 June 2006
Abstract—Cryptosporidiosis, an opportunistic infection affecting immunocompromised patients, the elderly, and children, is still an
untreatable disease since the causative agent, Cryptosporidium hominis, is essentially resistant to all clinically used antimicrobial
agents. In order to accelerate the design of new potent and selective inhibitors targeting dihydrofolate reductase of C. hominis
(ChDHFR), we determined the structural basis for the potency of existing DHFR inhibitors using superpositions of the structure
of ChDHFR with other species and analysis of active site complexes of ChDHFR bound to ligands exhibiting a wide range of IC50
values. This information was used to develop an accurate docking model capable of identifying potent inhibitors in silico. A series of
C7-trimethoprim derivatives, designed to exploit a unique pocket in ChDHFR, was synthesized and evaluated; 7-ethyl TMP has
four times higher activity than TMP against ChDHFR.
ꢀ 2006 Elsevier Ltd. All rights reserved.
Cryptosporidiosis is an opportunistic infection caused
by the parasitic protozoan, Cryptosporidium hominis,
also classified as a class B biodefense pathogen. Crypto-
sporidiosis, for which there are no effective therapies,
causes wasting disease in immune-compromised pa-
tients, the elderly, and children. Dihydrofolate reductase
(DHFR) is a major drug target for this family of para-
sitic protozoa,1 which also includes Plasmodium, the
causative agent of malaria. DHFR is an important com-
ponent of the folate biosynthetic pathway, responsible
for providing the DNA base, dTMP. Alarmingly, none
of the clinically used DHFR inhibitors including tri-
methoprim (TMP) and pyrimethamine (PYR) show
any efficacy against C. hominis DHFR (ChDHFR), pre-
sumably due to the structural differences in the enzyme.
To address this problem, several compounds were test-
ed2 with in vitro inhibition assays and although some
of them were potent, none have been introduced to clin-
ical use.
In order to design potent and selective ChDHFR inhib-
itors, we first superposed ChDHFR with other DHFR
enzymes bound to TMP and PYR, and determined the
structural features of ChDHFR that explain the poor
binding of TMP and PYR. We then developed a struc-
ture–activity relationship (SAR) for ChDHFR using a
docking method based on ensembles of ChDHFR struc-
tures and complexes of 30 ligands with a wide range of
IC50 values. Based on the results of these studies, we
proposed several TMP derivatives predicted to have
increased activity toward ChDHFR. An efficient
synthetic route to the TMP derivatives and their evalu-
ation resulted in a better inhibitor that occupies a un-
ique pocket as predicted by the structural analysis.
Sequence and structural alignments of ChDHFR (PDB
ID: 1QZF3) with DHFR of other species revealed why
several clinically used DHFR inhibitors fail to inhibit
ChDHFR. Both pyrimethamine (Fig. 1), which has high
binding affinity for Plasmodium falciparum DHFR
(PfDHFR), and a dibenzazepine inhibitor, compound
4,4 which has potent activity against Pneumocystis cari-
nii DHFR (PcDHFR), exhibit very low potency against
ChDHFR (Table 1).
Keywords: Cryptosporidium hominis; Dihydrofolate reductase; Diami-
nopyrimidine; Docking.
*
Corresponding author. Tel.: +1 860 486 6145; fax: +1 860 486
6857; e-mail: Amy.Anderson@uconn.edu
Superposition
ChDHFR:TMP, ChDHFR:4, or ChDHFR:PYR with
of
modeled
complexes
of
Present address: Department of Chemistry, Santa Clara University,
Santa Clara, CA 95053, USA.
0960-894X/$ - see front matter ꢀ 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmcl.2006.05.047

V. M. Popov et al. / Bioorg. Med. Chem. Lett. 16 (2006) 4366–4370
4367
Cl
MeO
NH2
NH2 Cl
N
N
N
N
H
OMe
H2N
N
H2N
H2N
N
2
1
NH2
NH2
OMe
OMe
N
N
N
N
H2N
N
N
OMe
4
3
Figure 1. The most potent compound (1) docked to ChDHFR
(IC50 = 6.5 nM), pyrimethamine (PYR) (2), trimethoprim (TMP) (3),
and a dibenzazepine inhibitor (4).
Figure 2. (a) Connolly surface of PfDHFR (gold) and pyrimethamine;
(b) Modeled Connolly surface of ChDHFR (purple) with pyrimeth-
amine. Note the ethyl group of pyrimethamine protruding into the
surface of Leu 33 and the chlorine atom touching the surface of Ile 62.
Table 1. IC50 values
IC50a(lM)
ChDHFR
PfDHFR
PcDHFR
PYR (2)
TMP (3)
4
ꢀ1000
14
119
0.08
0.01
N/D
N/D
12
smaller substituents on the phenyl ring. Compounds
with halogens at the 30 and 40 positions and those with
methoxy groups at the 20 and 50 positions (e.g., 1) (see
III-5, III-6 and V-16 in reference2) are more potent than
compounds with methoxy groups at the 30, 40, and 50
positions (see V-6, V-7, and V-9 in reference2).
0.21
N/D, not determined.
a IC50 values for ChDHFR were determined in our laboratory using
standard spectrophotometric assays;5 values for PfDHFR:PYR6 and
PcDHFR4 as well as a Ki value for PfDHFR:TMP7 were determined
as described.
Due to the difference in their volumes and conforma-
tions, the substitution of Ile 125 of PcDHFR (or Ile
164 of PfDHFR) by Cys 113 (in ChDHFR) widens
the latter’s linker-binding region (Fig. 4). This pocket
can be occupied by the substituent on C5 of a quinazo-
line ring (R1 in Scheme 1) (e.g., 1) or by the substituent
on a linker region (R2) or R4 in the TMP example. It is
worth noting that the higher activity of the compounds
with methylated linkers such as R2@CH3 or R3@CH3
may be due to the improved van der Waals interactions
or to the restriction of the ligands’ conformational
freedom.
crystal structures of PcDHFR:TMP (1DYR8),
PcDHFR:4
(1KLK9),
and
PfDHFR
(C59R/
S108N):PYR (1J3J6) confirmed the high structural
homology of the residues that bind the 2,4-diaminopyr-
imidine ring and the linker regions. However, there are
significant variations in the residues involved in binding
the substituted phenyl rings that may relate to differenc-
es in ligand binding (Table 2); these differences will be
discussed in more detail.
Leu 33 in ChDHFR narrows the pocket and restricts the
size of a substituent that can be placed at the C6 posi-
tion of the pyrimidine ring. Rigid superposition of
ChDHFR and PfDHFR bound to pyrimethamine re-
veals the steric hindrance of pyrimethamine’s ethyl and
chlorine substituents with Leu 33 and Ile 62 of
ChDHFR, respectively (Fig. 2). In PfDHFR, Met 55
adopts a rotamer conformation that allows it to avoid
steric hindrance with the ethyl group.
Next, we wanted to examine whether ligands with higher
potency, but unavailable in the clinic, avoided interac-
tions with Leu 33 and created additional van der Waals
interactions in the area of the absent loop or with Cys
113. We developed an accurate and reliable method to
dock inhibitors with a wide range of IC50 values in the
ChDHFR active site using the crystal structure of the
enzyme (PDB ID:1QZF3), a flexible protein model avail-
able in the program, Flo9810, and ensemble averaging.11
To ensure that docking results accurately predict the
affinities of novel compounds, our docking method
was refined against a set of 30 known ChDHFR inhibi-
tors, all of which had a general scaffold containing a 2,4-
diaminopyrimidine, a linker region, and a substituted
phenyl group (representative compounds are shown in
Fig. 1). Using three independent energy terms: electro-
static (Eest), van der Waals (Evdw), and structural com-
plementarity (Ecnt), from the docking scores to
estimate the logIC50 according to the equation:
logIC50 = 18.379 + 0.296 · Evdw + 0.281 · Eest + 0.206 ·
Ecnt, we achieved a reasonable correlation (R2 = 72.9%)
between the scores of the docked protein:ligand com-
plexes and the measured IC50 values, thereby validating
the docked poses for these inhibitors.
The phenyl-binding pocket in ChDHFR is more super-
ficial (Fig. 3) since ChDHFR lacks a loop analogous to
Pro 66-Phe 69 of PcDHFR that makes many van der
Waals and p-stacking interactions. In ChDHFR, the
shallower phenyl-binding pocket appears to require
Table 2. Comparison of key residue differences in DHFR active sites
ChDHFR
PfDHFR
PcDHFR
Flag
Leu 33
Absent
Absent
Cys 113
Met 55
Pro 113
Phe 116
Ile 164
Ile 33
Leu 33—PYR
Pro 66
Phe 69
Ile 123
Absent—TMP, 4
Absent—TMP, 4
Cys 113—Reduced van
der Waals
Flag column indicates a compound has a lower affinity for ChDHFR
due to the combination of amino acids.

4368
V. M. Popov et al. / Bioorg. Med. Chem. Lett. 16 (2006) 4366–4370
Figure 3. (a) Compound 4 in PcDHFR; (b) Compound 4 in ChDHFR. The buried surface of the ligand is colored green and the non-buried surface is
colored yellow.
10, and Leu 33. The phenyl ring of most of the ligands
was positioned in one of two major conformations in
the active site, in either an ‘up’ conformation, defined
by residues Leu 25, Thr 58, Ser 61, and Ile 62, or a
‘down’ conformation, defined by residues Leu 33 and
Leu 67 (Fig. 5).
The effects of the three unique features observed in the
superposition of ChDHFR with other species were obvi-
ous in the docking study. Potent compounds often have
a bicyclic diaminopyrimidine ring, which avoids steric
conflict with Leu 33, yet maximizes van der Waals con-
tacts. Potent compounds were also usually larger, form-
ing compensating van der Waals contacts for those lost
Figure 4. The pocket created by Cys 113. TMP (yellow) and NADPH
(purple) are also shown.
by the absence of the loop at the active site. Finally, po-
tent compounds often contained a substituent at C5 of
the quinazoline that interacted with Cys 113.
NH2 R4
NH2 R1 R2
OMe
OMe
N
X
7
N
Y
R3
5
H2N
N
H2N
N
OMe
Scheme 1. A general scaffold for the linker region of quinazoline
ligands. Position 5 is noted; X or Y can be C or N; positions where
substitutions occur are also noted. Also shown is a scaffold for TMP
ligands with C7 noted.
For all 30 ligands, we found that Eest is the single best
predictor of ligand activity (R2 = 0.4 between Eest and
logIC50). A salt bridge between the 2,4-diaminopyrimi-
dine ring and Asp 32 is probably the major contributor
of the electrostatic energy; achieving the proper orienta-
tion of the ring to form these bonds is critical. Structural
complementarity between the ligand and the active site
is also important, as ligands that either have steric inter-
actions with the active site or do not adequately occupy
the site have increased Ecnt values.
Analysis of the docked poses of potent inhibitors reveals
that the 2,4-diaminopyrimidine ring assumes the same
orientation in the ChDHFR active site as observed in
corresponding crystal structures, with good geometry
between the ring and Asp 32 as well as hydrophobic
interactions between the ring and Phe 36, Val 9, Val
Figure 5. Interactions of compound 1 and TMP (3) with ChDHFR
(cyan). TMP is shown in blue, the ‘up’ conformation of 1 is shown in
green, and the ‘down’ conformation of 1 is shown in purple. For
clarity, only one conformation of TMP is shown.

V. M. Popov et al. / Bioorg. Med. Chem. Lett. 16 (2006) 4366–4370
4369
Conversely, a comparison of the most potent compound
(1) with TMP (compound 3; 1000-fold less potent)
(Fig. 5) shows that TMP has much less favorable con-
tact energies, most likely due to a lack of optimization
of the interactions of the trimethoxyphenyl ring and
the linker region. Although it adopts a similar binding
orientation, the single pyrimidine ring and shorter linker
region do not extend the trimethoxyphenyl ring to fully
occupy the hydrophobic pockets visualized with com-
plexes of compound 1.
but we would expect that the enantiomerically pure
compound will have even higher activity. These results
validated our docking methodology and identified a
key design element that can be incorporated in future
designs to increase potency. Substitution at this position
may also increase selectivity since human DHFR has a
bulkier valine at this position.
In conclusion, structural information and SAR data can
be used to expedite the drug discovery process. We have
superposed ChDHFR with several structures of DHFR
from other species and identified three features that are
unique in ChDHFR: Leu 33 restricts the active site and
van der Waals interactions are reduced through the sub-
stitution of Cys 113 as well as the absence of a loop at
the active site. Docking a broad range of inhibitors into
the active site of ChDHFR allowed the development of
an SAR model. The primary interactions that occur be-
tween ChDHFR and potent inhibitors include the
canonical 2,4-diaminopyrimidine interactions, a substi-
tuted two-atom linker that interacts with Cys 113 and
extends the phenyl group into two different hydrophobic
pockets, and the interactions of the substituted phenyl
group with those pockets. In order to experimentally
test whether interactions with Cys 113 will improve the
potency of a lower molecular weight compound, we
docked several TMP analogs with substitutions at C7,
predicted to interact with Cys 113. The synthesis and
evaluation of these C7-TMP analogs has led to a deriv-
ative that is four times more potent than the parent
compound.
In order to validate hypotheses presented by our struc-
tural analysis, we docked, synthesized, and evaluated a
series of compounds that should utilize the unique pock-
et created by the substitution of Ile 164 (PfDHFR) by
Cys 113. These compounds served also as a test set to
validate the predictions made from the docking study.
A small library of C7-TMP analogs, with both R and
S stereochemistry, was constructed in silico and docked
into the ChDHFR active site. 7-Ethyl TMP (R) was pre-
dicted to be the most potent, with the sum of Evdw, Eest,
and Ecnt equal to ꢁ58.46 kJ/mol. In comparison, the
energies of 7-methyl TMP and 7-isopropyl TMP were
ꢁ56.85 and ꢁ56.98, respectively.
A versatile synthetic route, presented here, provided ac-
cess to a variety of C7-derivatized TMP analogs which
could, in principle, be used for the design of larger li-
braries of compounds that share the TMP scaffold
(Scheme 2).12–14 7-Ethyl TMP was synthesized (as a
racemic mixture) along with two related analogs, 7-Me
TMP15 and 7-isopropyl TMP, both of which had higher
docking energies and served as negative controls.
Acknowledgment
Benzylic oxidation of commercially available TMP was
performed at elevated temperatures with activated man-
ganese dioxide to give the ketone 5. Addition of an ex-
cess of the Grignard reagent led to the tertiary
carbinol that was subsequently dehydrated to give the
alkene 6–8 in good yield. Catalytic hydrogenation over
palladium on carbon produced the final analogs 9–11
for evaluation.
Financial support by the National Institutes of Health
(GM67542) is gratefully acknowledged.
References and notes
1. Anderson, A. Drug Discovery Today 2005, 10, 121.
2. Nelson, R.; Rosowsky, A. Antimicrob. Agents Chemother.
2001, 45, 3293.
3. O’Neil, R.; Lilien, R.; Donald, B.; Stroud, R.; Anderson,
A. J. Biol. Chem. 2003, 278, 52980.
4. Rosowsky, A.; Cody, V.; Galitsky, N.; Fu, H.; Papoulis,
A.; Queener, S. J. Med. Chem. 1999, 42, 4853.
5. Anderson, A. Acta Crystallogr., Sect. F 2005, F61, 258.
6. Yuvaniyama, J.; Chitnumsub, P.; Kamchonwongpaisan,
S.; Vanichtanankul, J.; Sirawaraporn, W.; Taylor, P.;
Walkinshaw, M.; Yuthavong, Y. Nat. Struct. Biol. 2003,
10, 357.
7. Sirichaiwat, C.; Intaraudom, C.; Kamchonwongpaisan, S.;
Vanichtanankul, J.; Thebtaranonth, Y.; Yuthavong, Y.
J. Med. Chem. 2004, 47, 345.
Each of the analogs was tested in a spectrophotometric
enzyme assay.5 Racemic 7-ethyl TMP has an IC50 value
of 4.2 lM against ChDHFR, 7-Me-TMP has a value of
340 lM, and 7-isopropyl TMP has a value greater than
100 lM. The IC50 for the racemic mixture of 7-ethyl
TMP is four times lower than that for TMP (14 lM)
NH2
N
O
OMe
OMe
MnO2
N
1) RMgX, 80-97%
TMP
120˚ C
65%
2) CF3CO2H, 66-70%
H2N
OMe
5
8. Champness, J.; Achari, A.; Ballantine, S.; Bryant, P.;
Delves, C.; Stammers, D. Structure 1994, 2, 915.
9. Cody, V.; Galitsky, N.; Luft, J.; Pangborn, W.; Rosowsky,
A.; Queener, S. Acta Crystallogr., Sect. D 2002, D58, 946.
10. McMartin, C.; Bohacek, R. J. Comput. Aided Mol. Des.
1997, 11, 333.
R'
R''
NH2
R
H2N
OMe
OMe
OMe
OMe
N
N
H2 (1 atm)
H2N
N
HOAc, Pd/C
60-78%
H2N
N
OMe
6, R'=R''=H
7, R'=Me, R''=H
8, R'=R''=Me
OMe
9 R=Me (C7-Me)
10 R=Et (C7-Et)
11 R=iPr (C7-iPr)
11. Manuscript submitted. Briefly, the active site was prepared
˚
by adding hydrogens to NADPH and residues within 11 A
of the ligand. Limited flexibility (according to Flo98) was
Scheme 2. Synthesis of C7-substituted analogs of trimethoprim.

4370
V. M. Popov et al. / Bioorg. Med. Chem. Lett. 16 (2006) 4366–4370
allowed for these active site residues, the nicotinamide and
accumulating energy penalties. It is assumed that in
solution the shell would accommodate the slight changes
of the residues and therefore the additional penalties are
overestimated and inaccurate. Average values of three
independent energy terms Eest, Evdw, and Ecnt were used to
calculate the theoretical logIC50 using the formula
obtained from the multivariable linear regression line
equation.
ribose of NADPH. The model of ChDHFR was subjected
to a 1000-step Monte-Carlo docking search followed by a
20-step simulated annealing run. The lowest energy
conformer from this search was used in future docking
searches. The 30 ligands were protonated at N1 and
docked in the prepared site using 500 Monte-Carlo steps
followed by energy minimization. Each docking run
yielded the 25 lowest energy conformers described by five
energy values: the internal energy of the ligand (Elig), the
internal energy of the active site residues (Esh) in addition
to electrostatic (Eest), van der Waals (Evdw), and contact
(Ecnt) energies. We did not include Elig or Esh in the
regression line since Elig is calculated by molecular
mechanics in vacuo and has limited accuracy. Esh was
omitted because the limited flexibility allowed during
docking permitted some residues to move toward the shell,
12. Brossi, A.; Grunberg, E.; Hoffer, M.; Teitel, S. J. Med.
Chem. 1971, 14, 58.
13. Hess, S.; Dolker, M.; Haferburg, D.; Kertscher, H.;
Matysik, F.; Ortwein, J.; Teubert, U.; Zimmermann, W.;
Eger, K. Pharmazie 2001, 56, 306.
14. Nordholm, L.; Dalgaard, L. Arch. Pharmacy. Org. Chem.
Sci. Ed. 1983, 11, 1.
15. Gangjee, A.; Lin, X.; Queener, S. E. J. Heterocycl. Chem.
2003, 40, 507.