Structure-based design, synthesis, and SAR evaluation of a new series of 8-hydroxyquinolines as HIF-1α prolyl hydroxylase inhibitors

Article abstract of DOI:10.1016/j.bmcl.2006.08.040

A new series of potent 8-hydroxyquinolines was designed based on the newly resolved X-ray crystal structure of EGLN-1. Both alkyl and aryl 8-hydroxyquinoline-7-carboxyamides were good HIF-1α prolyl hydroxylase (EGLN) inhibitors. In subsequent VEGF inducti

Full text of DOI:10.1016/j.bmcl.2006.08.040

Bioorganic & Medicinal Chemistry Letters 16 (2006) 5517–5522
Structure-based design, synthesis, and SAR evaluation
of a new series of 8-hydroxyquinolines as HIF-1a prolyl
hydroxylase inhibitors
Namal C. Warshakoon,* Shengde Wu, Angelique Boyer, Richard Kawamoto,
Justin Sheville, Sean Renock, Kevin Xu, Matthew Pokross, Songtao Zhou,
Carol Winter, Richard Walter, Marlene Mekel and Artem G. Evdokimov
Drug Discovery Division, Procter and Gamble Pharmaceuticals Inc., 8700 Mason-Montgomery Road, Mason, OH 45040, USA
Received 11 July 2006; revised 7 August 2006; accepted 8 August 2006
Available online 22 August 2006
Abstract—A new series of potent 8-hydroxyquinolines was designed based on the newly resolved X-ray crystal structure of EGLN-1.
Both alkyl and aryl 8-hydroxyquinoline-7-carboxyamides were good HIF-1a prolyl hydroxylase (EGLN) inhibitors. In subsequent
VEGF induction assays, these exhibited potent VEGF activity. In addition, this class of compounds did show the ability to stabilize
HIF-1a.
Ó 2006 Elsevier Ltd. All rights reserved.
Hypoxia inducible factor (HIF) is a transcriptional com-
plex that plays a key role in mammalian oxygen homeo-
stasis and activates a host of hypoxic response genes that
regulate angiogenic, glycolytic, and erythropoietic pro-
cesses.1–3 The subunit components, HIF-1a and HIF-
1b (ARNT), are constitutively expressed and regulation
is achieved by the selective destruction of HIF-1a. HIF-
1a is a major control element of the cellular response to
hypoxia (low partial pressure of oxygen).
Since this hydroxylation of HIF-1a is strictly dependent
on oxygen concentration; under hypoxia this reaction is
inhibited, and the surviving HIF-1a binds to ARNT to
form a functional transcriptional activator that turns
on genes with hypoxic response elements (e.g., VEGF,
EPO, and glycolytic enzymes).7 Therefore, inhibition
of HIF-1a prolyl hydroxylases, resulting in HIF-1a sta-
bilization, has the potential to be a viable therapeutic
approach for ischemic diseases including myocardial
infarction, stroke, peripheral arterial disease, heart fail-
ure, diabetes, and anemia.
A family of non-heme iron-containing prolyl hydroxy-
lases (EGLN-1, EGLN-2, EGLN-3) effects post-transla-
tional modification of HIF-1a. In a reaction requiring
Fe2+, oxygen, and ascorbate, these prolyl hydroxylases
catalyze the oxidative decarboxylation of 2-oxoglutarate
(2-OG) and the hydroxylation of proline564 and pro-
line402 in the oxygen-dependent degradation domain
(ODDD) of HIF-1a.4–6 Proline hydroxylation then tar-
gets HIF-1a for proteasomal degradation via binding to
the VHL (Von Hippel Lindau tumor suppressor pro-
tein), elongin C/B, Cul2, Rbx1 ubiquitin ligase complex.
In the recent years, some small molecule HIF-1a inhib-
itors have emerged as possible therapeutics for ischemic
diseases.8 These are small ‘peptide-like’ molecules (e.g.,
1 and 2) that resemble 2-oxoglutarate (2-OG). One
example, isoquinoline 3, acts as a 2-oxoglutarate mimet-
ic and thus exerts prolyl hydroxylase inhibition (IC50:
1.4 lM in our EGLN-1 assay) (Fig. 1).9
Our goal was to design novel HIF-1a prolyl hydroxylase
inhibitors based on the recently solved X-ray crystal
structure of the complex between catalytic domain of
EGLN-1 and the isoquinoline 3.10 The isoquinoline
nitrogen and the amide carbonyl interact with iron,
while the phenolic OH and the amide NH are involved
in hydrogen bonding interactions that impose structural
rigidity to the molecule. The side-chain carboxylate
Keywords: Prolyl hydroxylase inhibitors; HIF-1a; Hypoxia; Ischemia;
Peripheral arterial disease (PAD); Anemia; 8-Hydroxyquin-
olinecarboxyamides.
*
Corresponding author. Tel.: +1 513 622 2284; fax: +1 513 622
0523; e-mail: Warshakoon.nc@pg.com
0960-894X/$ - see front matter Ó 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmcl.2006.08.040

5518
N. C. Warshakoon et al. / Bioorg. Med. Chem. Lett. 16 (2006) 5517–5522
O
OH
O
Fe
O
O
OH
O
O
N
N
CO2H
H
O
O
H
H
N
N
N
HO
OH
O
HO
O
O
Arg383
O
N
H
Cl
N
H
1
2
3
2-Oxoglutarate(2-OG)
Figure 1. Small molecule HIF-1a prolyl hydroxylase inhibitors.
reaches toward Arg383 to form a salt-bridge interaction
(Fig. 2).
8-Hydroxyquinolines can exert their physiological prop-
erties through bidentate chelation of metal ions.11 Since
metal ions are important cofactors of many physiologi-
cally active enzymes, 8-hydroxyquinolines, have been
explored as a viable drug discovery platform in many
instances.12 We envisioned the same premise in this par-
ticular instance since iron is essential for the activity of
EGLN-1.
According to our modeling experiments (see the refer-
ences and notes section), the 8-hydroxyquinoline 4 can
coordinate iron via the amide carbonyl and the 8-OH,
while the carboxylate interacts with Arg383 in the active
site (Fig. 3). The quinoline ring likely stacks with Tyr310
at the top of the active site.
Figure 3. Predicted binding of compound 4 to the active site of
EGLN-1. Only the most important interactions are shown.
Quinoline analogs were synthesized via a reaction be-
tween corresponding amines and the acid chloride of
the 8-hydroxyquinoline-7-carboxylic acid. Good yields
were obtained via the coupling between the acid chloride
and the corresponding amine/aniline, while other cou-
pling reagents (CDI, HBTU, HATU, and EDAC) gave
poor yields (Scheme 1).
R
N
1. SOCl2/CH2Cl2
N
R1
2. iPr2EtNH/THF/NHRR1/50oC
75-80%
N
CO2H
OH
O
OH
6
5
Scheme 1.
The biological activity of the 8-hydroxyquinolines de-
rived against EGLN-1 enzyme and subsequent VEGF
induction is described in the tables below. The 8-
hydroxyquinoline-7-carboxylic acid was completely
inactive (Table 1, Compound 1), while introduction of
the side chains bearing carboxylic acid (Table 1, Com-
pound 1) delivered analog with good activity against
EGLN-1. Interestingly, the corresponding naphthalene
derivative had marginal activity (26.8 lM) showing the
importance of quinoline nitrogen. Next, we examined
the possibility of replacing the carboxylic acid moiety
with a suitable functional group that is smaller in size.
N-(cyanomethyl)-8-hydroxy-7-carboxyamide (Table 1,
Compound 3) showed good activity comparable to its
carboxylic acid counterpart. This was significant in
improving the pharmacokinetic profile, particularly
regarding membrane permeation in subsequent cellular
assays.
The next direction of our SAR studies was focused on
the analogs derived from the coupling between the 8-
hydroxyquinoline-7-carboxylic acid and various substi-
tuted benzylamines. Upon examining the binding mode
as predicted by the FlexX program, the potential cat-
ion–p interaction between the distal phenyl group and
Arg322 at the entrance of the active site (Fig. 4) struck
us as interesting. The p–cation interaction which is the
electrostatic interaction between a cation and the nega-
tive electrostatic potential associated with the face of a
Figure 2. Binding of compound 3 to the active site of EGLN-1. Only
the most important interactions are shown.

N. C. Warshakoon et al. / Bioorg. Med. Chem. Lett. 16 (2006) 5517–5522
Table 1. SAR of 8-hydroxyquinoline derivatives (95% Cl)
5519
N
R
OH
Compound
R
EGLN IC50 (lM)
>100
Compound
R
EGLN IC50 (lM)
O
O
OH
1
11
9.7
3.9
6.1
4.1
N
H
CF3
CN
O
H
N
2
3
4
5
5.0
12
13
14
15
CO2H
CN
N
H
O
O
O
CN
H
N
O
O
4.4
4.2
4.6
N
H
H
N
N
H
N
H
N
O
O
10
OMe
Cl
N
H
CO2Me
O
O
H
N
OMe
OMe
6
7
4.8
7.4
16
17
3.2
7.2
N
H
OMe
O
O
H
N
N
H
O
O
OMe
F
H
N
8
9
4.4
28.7
6.0
18
19
20
5.3
6.9
5.8
N
H
Cl
F
O
O
O
O
O
N
H
OMe
10
N
H
N
H
simple p system is a fairly new hypothesis and is argued
as an important non-covalent interaction in biological
recognition.13 In contrast to the binding mode of 4
(Fig. 3), in this instance, the quinoline nitrogen coordi-
nates the iron. It is likely that the 8-hydroxyl is also in-
volved in coordination, even though the present model
positions it too far from the metal ion.
Earlier, we have regarded the interaction between the li-
gands and the active site as mostly lipophilic in nature
(the active site consists of more lipophilic residues than
hydrophilic residues). The unsubstituted benzamide
analog (Table 1, Compound 4) showed good activity
while substitutions at the 4-position of the phenyl ring
with both electron-donating (Table 1, Compound 5)
and electron-withdrawing groups (Table 1, Compound
6) generated compounds with comparable activity. The
benzyl ester derivative (Table 1, Compound 11) showed
Figure 4. Predicted binding of 8-hydroxyquinoline derivative (Table 1,
Compound 4) to the active site of EGLN-1. Only the most important
interactions are shown.

5520
N. C. Warshakoon et al. / Bioorg. Med. Chem. Lett. 16 (2006) 5517–5522
to the enzyme in a mode similar to that of the benzyl-
amine analogs (Fig. 4). A similar p–cation interaction
with Arg322 was predicted for the distal phenyl group
extending from the quinoline nucleus (Fig. 6).
Most of these 8-hydroxyquinoline analogs are potent
inducers of VEGF (a downstream biomarker for HIF-
1 stabilization). The most potent compound (Table 2,
Compound 14) has an EC50 of 140 nM. In addition, sev-
eral other analogs exhibited potent activity in the nano-
molar range. In summary, we were able to design and
synthesize novel HIF-1a prolyl hydroxylase inhibitors
with the aid of the X-ray structure of EGLN-1 catalytic
domain in complex with a potent inhibitor. Our com-
pounds were shown to be potent VEGF inducers in a
cell-based assay. Currently, these 8-hydroxyquinoline
carboxyamides are undergoing selectivity and PK
studies.
Figure 5. Predicted binding of 8-hydroxyquinoline derivative (Table 1,
Compound 12) to the active site of EGLN-1. Only the most important
interactions are shown.
Modeling studies. The 3D coordinates of receptor were
directly adapted from the in-house X-ray structure of
human EGLN-1. The 3D structures of compounds to
be docked were generated and minimized by Sybyl 7.9
(Sybyl; Tripos Inc., 1699 South Hanley Rd, St. Louis,
MO 63144, USA). The protonation state of each com-
pound was assigned based on physiologic condition.
For instance, all carboxylic acids were deprotonated.
All docking calculations were performed by using FlexX
(FlexX; Tripos Inc., 1699 South Hanley Rd, St. Louis,
MO 63144, USA) with standard docking procedure
and parameters. The active site is defined as the collec-
reduced activity thus underlining the importance of
amide functionality at this position Figure 5.
Next, we investigated the aniline-derived analogs: mod-
eling studies revealed that these compounds may bind so
that the distal phenyl ring of the side-chain points to-
ward the active-site Arg383. The 8-hydroxyl and the
amide carbonyl groups probably coordinate iron in a
manner similar to that of compound 4 (Fig. 3). Hetero-
aromatics may also be used instead of the phenyl group
(Table 1, Compound 14). Interestingly, N-(3,5-dimeth-
oxyphenyl)-8-hydroxyquinoline-7-carboxamide (Table
1, Compound 16) is twofold more active than the corre-
sponding N-(3,5-dimethoxybenzyl)-8-hydroxyquinoline-
7-carboxamide (Table 1, Compound 7) suggesting that
the compound extension is limited by the depth of the
binding site.
˚
tion of amino acids lying within 6.5 A of atoms of com-
pound 3 determined by the in-house X-ray structure.
FlexX is a fragment-based docking program. The dock-
ing conformations were ranked by using FlexX scores.
The final protein–ligand complexes were not minimized
after the docking. With the same docking setup, we were
able to successfully reproduce the binding conformation
of compound 3 to EGLN-1. In the predicted protein–li-
gand complex with the best FlexX score, the RMSD
(Root Mean Square Deviation) between the coordinates
of the predicted ligand and those of the cocrystal ligand
The analogs derived from phenethyl amines (Table 1,
Compounds 19–22) also exhibited good activity
(Fig. 6). Modeling suggested that these compounds bind
˚
is less than 1.5 A.
EGLN-1 activity assay. The EGLN-1 enzyme activity
was determined using mass spectrometry (matrix-assist-
ed laser desorption ionization, time-of-flight MS, MAL-
DI-TOF-MS).14 The HIF-1a peptide corresponding to
residues 556–574 (DLDLEALAPYIPADDDFQL) was
used as substrate. The reaction was conducted in a total
volume of 50 lL containing Tris–Cl (5 mM, pH 7.5),
ascorbate (120 lM), 2-oxoglutarate (3.2 lM), HIF-1a
(8.6 lM), and bovine serum albumin (0.01%). EGLN-
1, quantity predetermined to hydroxylate 20% of sub-
strate in 20 min, was added to start the reaction. Where
inhibitors were used, compounds were prepared in
dimethylsulfoxide at 10-fold final assay concentration.
After 20 min at room temperature, the reaction was
stopped by transferring 10 lL of reaction mixture to
50 lL of a mass spectrometry matrix solution (a-cya-
no-4-hydroxycinnamic acid, 5 mg/ml in 50% acetoni-
trile/0.1% TFA, 5 mM NH4PO4). Two microliters of
the mixture was spotted onto a MALDI-TOF-MS target
Figure 6. Predicted binding of 8-hydroxyquinoline derivative (Table 1,
Compound 19) to the active site of EGLN-1. Only the most important
interactions are shown.

N. C. Warshakoon et al. / Bioorg. Med. Chem. Lett. 16 (2006) 5517–5522
Table 2. VEGF activity of 8-hydroxyquinoline derivatives (95% Cl)
5521
N
R
OH
Compound
R
VEGF EC50 (lM)
Compound
R
VEGF EC50 (lM)
H
N
O
O
1
>100
11
>100
CO2H
CN
N
H
CF3
CN
O
H
N
2
3
4.1
12
13
0.14
N
H
O
O
CN
H
N
O
O
0.89
0.62
3.1
0.42
1.9
N
H
H
N
4
5
14
15
OMe
Cl
N
H
O
O
N
H
N
O
O
>100
N
H
CO2Me
OMe
OMe
OMe
H
N
6
—
16
3.2
N
H
O
O
OMe
O
O
H
N
8
9
0.89
17
18
0.58
N
H
F
O
—
—
O
O
N
H
Cl
O
O
10
0.63
19
20
—
—
N
H
N
H
OMe
N
H
plate for analysis with an Applied Biosystems (Foster
City, CA) 4700 Proteomics Analyzer MALDI-TOF-
MS equipped with a Nd:YAG laser (355 nm, 3 ns pulse
width, 200 Hz repetition rate). Hydroxylated peptide
product was identified from substrate by the gain of
16 Da. Data defined as percent conversion of substrate
to product were analyzed in GraphPad Prism 4 to calcu-
late IC50 values.
Following overnight incubation, the cells were washed
with 100 lL of Opti-MEM (Gibco, Carlsbad, CA) to
remove serum. Compound in DMSO was serially
diluted (beginning with 100 lM) in Opti-MEM and
added to the cells. The conditioned media were ana-
lyzed for VEGF with a Quantikine human VEGF
immunoassay kit (R&D Systems, Minneapolis, MN).
Optical density measurements at 450 nm were record-
ed using the Spectra Max 250 (Molecular Devices,
Sunnyvale, CA). Data defined as % of DFO stimula-
tion were used to calculate EC50 values with Graph-
Pad Prism 4 software.
VEGF ELISA. HEK293 cells were seeded in 96-well
poly-lysine coated plates at 20,000 cells per well in
DMEM (10% FBS, 1% NEAA, and 0.1% glutamine).

5522
N. C. Warshakoon et al. / Bioorg. Med. Chem. Lett. 16 (2006) 5517–5522
9. Guenzler, -P.; Volkmar, K.; Stephen, J.; Liu, D. Y.; Todd,
References and notes
W. PCT. In. Appl. WO 2005007192, 2005.
10. Evdokimov, A. et. al, Nat. Struc. Biol. 2006, submitted for
publication.
11. Caris, C.; Baret, P.; Pierre, J.-L.; Serratrice, G. Tetrahe-
dron 1996, 52, 4659.
1. Mack, C. A.; Magovern, C. J.; Budenbender, K. T.; Patel,
S. R.; Schwatz, E. A.; Zanzonico, P.; Ferris, B.; Sanborn,
T.; Isom, O. W.; Crystal, R. G.; Rosengart, T. K. J. Vasc.
Surg. 1998, 27, 699.
12. (a) Zhuang, L.; Wai, J. S.; Embrey, M.; Fisher, T. E.;
Egbertson, M. S.; Payne, L. S.; Guare, J. P.; Vacca, J. P.;
Hazuda, D. J.; Felock, P. J.; Wolfe, A. L.; Stillmock, K.
A.; Witmer, M. V.; Moyer, G.; Schleif, W. A.; Gabyelski,
L. J.; Leonard, Y. M.; Lynch, J. J.; Michelson, S. R.;
Young, S. D. J. Med. Chem. 2003, 46, 453; (b) Ujiie, T.
Chem. Pharm. Bull. 1975, 23, 62; (c) Paris, D.; Cottin, M.;
Demonchaux, P.; Augurt, G.; Dupassieux, P.; Lenoir, P.;
Peck, M. J.; Jasserand, D. J. Med. Chem. 1995, 38, 669.
13. Zacharias, N.; Dougherty, D. A. Trends Pharmacol. Sci.
2002, 23, 281.
2. Bruick, R. K.; McKnight, S. L. Genes Dev. 2001, 15, 2497.
3. Safran, M.; Kaelin, W. G., Jr. J. Clin. Invest. 2003, 111, 779.
4. Semenza, G. L. J. Appl. Physiol. 2000, 88, 1474.
5. Semenza, G. L. Cell 2001, 107, 1.
6. Kondo, K.; Kaelin, W. G., Jr. Exp. Cell. Res. 2001, 264, 117.
7. Semenza, G. Biochem. Pharmacol. 2002, 64, 993.
8. (a) McDonough, M. A.; McNeill, L. A.; Tilliet, M.;
Papamicae¨l, C. A.; Chen, Q.- Y.; Benerji, B.; Hewitson, K.
S.; Schofield, C. J. J. Am. Chem. Soc. 2005, 127, 7680; (b)
Mole, D. R.; Schlemminger, I.; McNeill, L. A.; Hewitson,
K. S.; Pugh, C. W.; Ratcliffe, P. J.; Schofield, C. J. Biorg.
Med. Chem. Lett. 2003, 13, 2677; (c) Schlemminger, I.;
Mole, R.; McNeill, L. A.; Dhanda, A.; Hewitson, K. S.;
Tian, Y.-M.; Ratcliffe, P. J.; Pugh, C. W.; Schofield, C. J.
Bioorg. Med. Chem. Lett. 2003, 13, 1451.
14. Greis, K. D.; Zhou, S.; Burt, T. M.; Carr, A. N.; Dolan,
E.; Easwaran, V.; Evodokimov, A.; Kawamoto, R.;
Rosegen, J.; Davis, G. F. J. Am. Soc. Mass. Spectrom.
2006, 815.