Discovery of 7-aminofuro[2,3-c]pyridine inhibitors of TAK1: Optimization of kinase selectivity and pharmacokinetics

Article abstract of DOI:10.1016/j.bmcl.2013.06.054

The kinase selectivity and pharmacokinetic optimization of a series of 7-aminofuro[2,3-c]pyridine inhibitors of TAK1 is described. The intersection of insights from molecular modeling, computational prediction of metabolic sites, and in vitro metabolite i

Full text of DOI:10.1016/j.bmcl.2013.06.054

Bioorganic & Medicinal Chemistry Letters 23 (2013) 4511–4516
Contents lists available at SciVerse ScienceDirect
Bioorganic & Medicinal Chemistry Letters
journal homepage: www.elsevier.com/locate/bmcl
Discovery of 7-aminofuro[2,3-c]pyridine inhibitors of TAK1:
Optimization of kinase selectivity and pharmacokinetics
⇑
Keith R. Hornberger , Xin Chen, Andrew P. Crew, Andrew Kleinberg, Lifu Ma, Mark J. Mulvihill, Jing Wang,
Victoria L. Wilde, Mark Albertella, Mark Bittner, Andrew Cooke, Salam Kadhim, Jennifer Kahler,
Paul Maresca, Earl May, Peter Meyn, Darlene Romashko, Brianna Tokar, Roy Turton
OSI Pharmaceuticals LLC, 1 Bioscience Park Drive, Farmingdale, NY 11735, USA
a r t i c l e i n f o
a b s t r a c t
Article history:
Available online 27 June 2013
The kinase selectivity and pharmacokinetic optimization of a series of 7-aminofuro[2,3-c]pyridine inhib-
itors of TAK1 is described. The intersection of insights from molecular modeling, computational predic-
tion of metabolic sites, and in vitro metabolite identification studies resulted in a simple and unique
solution to both of these problems. These efforts culminated in the discovery of compound 13a, a potent,
relatively selective inhibitor of TAK1 with good pharmacokinetic properties in mice, which was active in
an in vivo model of ovarian cancer.
Keywords:
TAK1
Inhibitors
7-Amino-furo[2,3-c]pyridine
Cancer
Ó 2013 Elsevier Ltd. All rights reserved.
Inflammation
In the preceding article,1 we reported on the discovery and ini-
tial optimization of a potent series of 7-aminofuro[2,3-c]pyridines
as inhibitors of transforming growth factor b receptor-associated
kinase 1 (TAK1 or MAP3K7), resulting in identification of the po-
tent lead compounds 1a, 1b, and 1c (Fig. 1). These early tool com-
pounds, however, were relatively multi-targeted kinase inhibitors
and also displayed sub-optimal pharmacokinetic properties in
mice. We thus felt they were unsuitable for validating the potential
for TAK1 inhibition to be useful in the treatment of cancer or
inflammatory diseases. Herein we report a concise solution to both
of these problems, culminating in the discovery of tool compound
13a, which showed activity in an in vivo model of ovarian cancer.
Optimization of the kinase inhibitory profile of a given chemical
series from a multi-targeted profile toward a more selective profile
is by its nature a highly multivariate problem. There are over 500
kinases in the human kinome,2 and it is generally not feasible or
cost-effective to screen the entire kinome with full dose–response
curves against even a single compound. Furthermore, there is typ-
ically an upward shift when kinase inhibitory activity is evaluated
in cells, reflective of both the more complicated cellular milieu and
the much higher ATP concentration in cells relative to the Km for
ATP for most kinases.3,4 Thus, biochemical inhibition does not al-
ways linearly translate into cellular inhibition and true conse-
quences in vivo. At the outset, we screened compounds at a
single concentration in a broad panel of 192 kinases in a mobility
shift assay and then conducted follow-up cellular mechanistic
screening on kinases known to have toxicity concerns and/or the
potential to confound efficacy readouts. Medicinal chemistry ef-
forts were then focused on removing activity against kinases of
concern, with the aspiration that the overall selectivity profile
would be improved in the process.
In the in-house 192 kinase panel, compound 1a inhibited 42/
192 kinases >50% at 250ꢀ TAK1 Ki, indicating a relatively multi-
targeted profile. Other compounds closely related to 1b and 1c
showed comparable results, indicating a more widespread kinase
selectivity issue within the chemotype. Among the 42 kinases reg-
istering as hits in the panel were two of particular concern for cel-
lular follow-up: Aurora B and KDR/VEGFR2. Aurora B is a cell cycle
kinase which has critical functions in mitosis and has been the
subject of intense preclinical and clinical investigation for the
treatment of cancer.5 Aurora family inhibitors can, however,
R
2
N
N
O
N
O
4
7
H2N
1a: R =
N
S
R =
1b: X = CH
1c: X = N
X
N
S
⇑
N
Corresponding author at present address: Boehringer Ingelheim Pharmaceuti-
cals, 900 Ridgebury Road, Ridgefield, CT 06877, USA. Tel.: +1 203 798 5137; fax: +1
203 791 6072.
Figure 1. Representative 7-aminofuro[2,3-c]pyridine leads from early optimization
efforts.
E-mail address: keith.hornberger@boehringer-ingelheim.com (K.R. Hornberger).
0960-894X/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.bmcl.2013.06.054

4512
K. R. Hornberger et al. / Bioorg. Med. Chem. Lett. 23 (2013) 4511–4516
be thought of as ‘targeted cytotoxics’ and such activity has the po-
tential for both dose-limiting neutropenia5 and contributing to
non-TAK1-mediated efficacy. KDR is a clinically validated target
for oncology, with multiple inhibitors approved for treatment of
renal cell carcinoma.6 However, KDR inhibition also is associated
with hypertension clinically7 and could further confound efficacy
readouts by inhibition of tumor angiogenesis. Both activities were
deemed of enough concern that the medicinal chemistry campaign
focused on the removal of these activities in the cellular setting.
Examination of the previously acquired X-ray crystal structure of
1c (Fig. 2) in the TAK1–TAB1 fusion protein highlighted two poten-
tial areas where selectivity inroads could be made. S111 in the ribose
pocket of TAK1 corresponds to E161 of Aurora B, suggesting that a
substituent reaching into the ribose pocket and clashing with the
larger E161 would improve selectivity. V42 of TAK1, at the beginning
of the P loop, corresponds to L838 in KDR and forms the top of a rel-
atively narrow channel that the 4-pyrazole substituent fits through.
Although this Val versus Leu difference was relatively small, we felt
that introduction of some torsion to the 4-substituent, which was
nearly coplanar with the furopyridine core in 1c (dihedral angle
ꢁ4°), could improve KDR selectivity. Conceptually, replacing the
pyrazole with a larger ring or introducing a substitution at the pyr-
azole 3- or 5-position should (a) cause increased steric repulsion
with the furopyridine core C3 proton to introduce torsion and (b)
reach further into the ribose binding pocket, potentially affording
selectivity for both Aurora B and KDR simultaneously.
Compounds exploring this concept were synthesized as out-
lined in Scheme 1. Beginning from the previously described1 com-
pound 2, Suzuki coupling with the appropriate boronic ester (3)
installed the 2-substituent as in 4, which was then iodinated (5)
and transformed to final analogs (6) via a second Suzuki coupling.
For a few compounds where the coupling partners were not readily
accessible, it was necessary to reverse the nucleophile/electrophile
partners in the coupling event. Compound 5 was thus Boc pro-
tected and converted to the corresponding stannane (7), which al-
lowed some additional analogs to then be prepared by successive
Stille coupling and Boc deprotection. One acetylenic analog (6m)
was accessed by Sonogashira coupling from 5.
S
N
Cl
a
O
B
O
O
+
H2N
N
2
3
S
S
N
N
b
O
O
I
H2N
4
N
N
H2N
N
5
S
c
6a-6b
6d-6g
6i-6p
O
R
H2N
N
S
N
d, e
f, g
6c, 6h,
6q, 6r
5
O
SnMe3
Boc2N
N
7
Scheme 1. Preparation of 4-substituted analogs 6. Reagents and conditions: (a)
PdCl2dppf, K2CO3, 1,4-dioxane/H2O, reflux (79%); (b) NIS, DMF, rt (77%); (c) R-
B(OH)2 or R-B(pin), PdCl2dppf, K2CO3, 1,4-dioxane/H2O, 80–100 °C,
Pd(PPh3)4, CuI, Et3N, THF/DMF, 40 °C (for 6m); (d) Boc2O, DMAP, toluene/DMF,
50 °C; (e) Me3SnSnMe3, Pd(PPh3)4, toluene, 120 °C (89%, 2 steps); (f) R-I or R-Br,
Pd2dba3, P(o-tolyl)3, Et3N, DMF, 70 °C; (g) 4 N HCl, 1,4-dioxane, rt.
lW or
high-throughput screening hits was the optimal platform to
conduct further selectivity explorations from, with most other
five-membered and all six-membered cyclic substituents showing
significantly reduced TAK1 activity. The linear linker typified by
acetylene 6m also lost substantial TAK1 activity. Although a
thiazole 4-substituent (6i) was also relatively potent against
TAK1, it offered no advantage in overall physicochemical proper-
ties or lipophilic ligand efficiency8 (LLE = 6.2 for 6a vs LLE = 5.7
for 6i) and was deprioritized for further follow-up.
A survey of the initial 4-position SAR (Table 1) showed that the
4-pyrazole substituent (6a) analogous to that from the early
Focusing attention more closely on the furopyridine 4-pyrazole
substituent, we briefly explored a palette of 3- and 5-substitutions
on the pyrazole ring itself, with the anticipation that 3-substitu-
tion, facing into the ribose binding pocket, would deliver enhanced
selectivity for Aurora B and KDR. Cellular inhibition of Aurora B
was tracked by measuring inhibition of histone H3 S10 phosphor-
ylation (pHH3) in HT-29 cells, as this histone site is characterized
to be phosphorylated by Aurora B.5 Cellular inhibition of KDR
was tracked by measuring inhibition of KDR autophosphorylation
in VEGF-stimulated HEK293 cells overexpressing KDR. Unfortu-
nately, we observed a substantial TAK1 potency loss from both
simple 3-methyl (6n) and 5-methyl (6o) pyrazole substitutions, ef-
fects which were additive in combination (6p). The poor tolerance
of the 5-methyl was anticipated due to a steric clash with the prox-
imal Y106 of TAK1, immediately adjacent to the hinge binding res-
idues. The loss of activity with 3-substitution is less easily
rationalized; presumably the introduced torsion precluded an ade-
quate fit to the narrow channel the pyrazole occupies in TAK1. A 3-
methoxy substituent (6q) was better tolerated, but there was still a
four-fold loss in activity versus the unsubstituted compound. A
Figure 2. Potential selectivity sites for Aurora family kinases and KDR/VEGFR2.

K. R. Hornberger et al. / Bioorg. Med. Chem. Lett. 23 (2013) 4511–4516
4513
Table 1
Table 2
SAR of 4-substitutions
SAR of 4-pyrazole substitutions
S
S
N
N
R3
N
N
O
O
R
R5
H2N
N
H2N
N
a
a
a
Compd
R
IC50
(
lM) Compd
R
IC50
(
lM)
Compd
R3
R5
IC50
(lM)
TAK1b
pJNKc
pHH3d
pKDRe
N
N
N
S
6a
0.15
4.9
6h
1.4
N
N
6a
6n
6o
6p
6q
6r
H
Me
H
Me
OMe
CN
H
H
Me
Me
H
0.15
3.5
4.1
23.9
0.60
22.8
0.19
1.0
3.0
4.4
1.7
3.0
3.4
>20
17.4
NT
>20
NT
>20
19.9
NT
>20
NT
N
6b
6c
6i
6j
0.23
S
H
>10
a
12.7
3.2
Values are the mean of P2 experiments; NT = not tested.
b
N
Biochemical ALPHAScreen assay with TAK1–TAB1 fusion protein in the pres-
M ATP.
Inhibition of JNK phosphorylation in TNF
Inhibition of histone H3 (HH3) S10 phosphorylation in HT-29 cells.
Inhibition of KDR autophosphorylation in HEK293/KDR cells.
ence of 100
l
N
c
a
-stimulated HCT-116 cells.
O
S
6d
6e
0.56
2.1
6k
6l
3.8
d
e
NH
OMe
23.6
O
6f
2.2
4.6
6m
2.9
Table 3
Mouse pharmacokinetic parameters for selected compounds
6g
Ar
S
1a: Ar = A, R = D
N
N
a
Biochemical ALPHAScreen assay with TAK1–TAB1 fusion protein in the pres-
ence of 100 M ATP; values are the mean of P2 experiments.
1b: Ar = B, R = D
1d: Ar = B, R = E
1e: Ar = C, R = E
R
O
l
H2N
N
S
more rigid 3-cyano substituent (6r) was not tolerated. Compound
6n offered some signs of improved cellular selectivity for Aurora
B and KDR, validating the initial design concept. In the final analy-
sis, however, the loss of TAK1 activity suffered by these compounds
was deemed too great for use as potential tools and further evolu-
tion of this SAR was also deprioritized (Table 2).
N
N
N
N
O
S
S
A
B
C
N
OH
In parallel with this initial exploration of the kinase selectivity
SAR, we also evaluated the pharmacokinetic properties of our
lead compounds and attempted to tune physicochemical proper-
ties by modulation of the solvent-exposed pyrazole N-substitu-
tion. Although we conducted an extensive medicinal chemistry
effort around pyrazole N-substitution, potency and selectivity
SAR were relatively flat and therefore only key pharmacokinetic
observations for representative compounds are detailed here (Ta-
ble 3). First, we noted that the thienopyridine 2-substituent had
inferior bioavailability relative to other lead structures (compare
1a and 1b). Second, replacement of the pyrazole N-acetyl piperi-
D
E
Route
iva
Parameters
Cl (mL/min/kg)
(h)
1a
1b
1d
1e
178
0.1
187
2.5
0.03
21c
1
128
0.5
261
2.5
1.6
285
11
111
3.3
299
12
1.1
821
27
70
1.6
477
4.1
7.8
t
½
AUC0–1 (ng h/mL)
Vss (L/kg)
Cmax
p.o.b
(l
M)
AUC0–1 (ng h/mL)
16,003
335
F (%)
a
b
c
iv PK dosed at 2 mg/kg.
p.o. PK dosed at 20 mg/kg.
AUC0–last; compound was below the limit of quantitation at 8 h after oral
dine substitution with
a cyclohexanol resulted in further
improvement in oral bioavailability (compare 1b and 1d). Finally,
the benzothiadiazole 2-substitution had the best overall oral
exposure (compare 1d and 1e). Compound 1e had a late oral
tmax = 8 h with a second, lower peak concentration at 2 h, together
indicative of potential enterohepatic recycling.9 This observation
may afford a rationale for the apparent >100% bioavailability.
Based on these pharmacokinetic results, we therefore shifted
our future SAR focus to the benzothiadiazole 2-substituent
(chemotype C) and cyclohexanol pyrazole N-substituent (chemo-
type E).
administration and AUC0–1 was not calculated.
the curve (AUC) and concentration–time profile were still sub-
optimal for all of these compounds. A major factor hampering
these chemotypes was that all had high systemic clearance, and
we anticipated that little further progress would be made on
improving pharmacokinetics without understanding and reducing
compound metabolism. A computational analysis of compounds 1a
and 1e for sites of CYP metabolism using the StarDrop program10
identified few clear leads. The thienopyridine sulfur of 1a was
predicted to be highly labile, but this potential metabolic site
was already corrected for in 1e. Compound 1e itself had no highly
While we had made gains in oral bioavailability as a result of
this exploration, setting aside the substantial potential contribu-
tions from enterohepatic recycling, the absolute oral area under

4514
K. R. Hornberger et al. / Bioorg. Med. Chem. Lett. 23 (2013) 4511–4516
labile sites identified that would alone explain its high clearance.
We also conducted an in vitro metabolite identification study on
1a in mouse and human liver microsomes in the presence of both
NADPH and UDPGA cofactors. Although several minor oxidative
metabolites were detected, the common major metabolite was
an M+176 metabolite corresponding to a single glucuronidation.
Suspecting this glucuronidation to be occurring on the furopyri-
dine 7-amino substituent, we also ran metabolite identification
on the 7-des-amino counterpart of 1a and observed that the glucu-
ronide metabolite had completely disappeared. A glucuronide
metabolite also aligned with the apparent >100% bioavailability
of 1e; depot of glucuronides in bile followed by cleavage and recir-
culation of the parent compound in the intestine is a known mech-
anism for enterohepatic recycling.9
Although we now had a clear sense of the major site of metab-
olism within the chemical series, this was a particularly vexing
medicinal chemistry problem due to the essential function of the
7-amino group in anchoring binding to the hinge of TAK1. Accept-
ing that we could not remove the 7-amino group altogether, we in-
stead wondered whether it would be possible to modulate its
nucleophilicity, for example by attachment of an electron-with-
drawing group to an open site on the furopyridine core. These
hypotheses around metabolism were coming to the fore just as
the kinase selectivity exploration around the pyrazole 4-substitu-
ent was concluding and we were concomitantly seeking alternate
ways to achieve Aurora B and KDR selectivity. We realized that
the furopyridine 3-position offered an opportunity to solve both
the metabolism and kinase selectivity problems in a single stroke.
In addition to providing a venue for anchoring an electron-
withdrawing group for metabolic stability, the 3-position also
directly faces the ribose binding pocket of TAK1. Such a 3-substitu-
tion could thus provide the desired steric clash with E161 of Aurora
B and introduce torsion to the pyrazole 4-substituent for KDR
selectivity. The only remaining obstacle in reduction of this strat-
egy to practice was the synthesis of the 3-substituted compounds.
At this point we also realized that the means to achieve such a
synthesis were already within reach (Scheme 2). Previously1 we re-
ported the synthesis of compound 8 as a by-product of the desired
C2 mono-chlorination of the furopyridine core en route to com-
pound family 1 and 6. Ironically, this by-product, initially per-
and then transformed to 11 and 12 by a series of sequential Suzuki
couplings analogous to that used for compound series 1. Global re-
moval of all protecting groups afforded the 3-Cl substituted com-
pound 13a, and coupling of this chloride with zinc cyanide in
turn gave 13b. For the preparation of all other compounds, it was
necessary to Boc protect the 7-amino group to give 14. Various het-
erocyclic substituents (13c, 13d) could be accessed by Suzuki cou-
pling, although substantial optimization of conditions was needed
to affect coupling of this sterically congested aryl chloride.12 Some
fluorinated analogs (13e, 13f) were accessed by Suzuki coupling of
a vinyl group, followed by ozonolysis with either PPh3 or NaBH4
work-up, and then subsequent fluorination with DeoxoFluor.
The SAR of the furopyridine 3-substituted compounds 13 is de-
scribed in Table 4. In general, most compounds demonstrated good
maintenance of TAK1 biochemical and cellular potency, although a
larger 6-membered heterocyclic substituent (13c) displayed some
loss of activity. All of the 3-substituted compounds traded back
some lipophilic ligand efficiency, but the loss was deemed neces-
sary to address the other deficits of the chemotype. Gratifyingly,
most compounds demonstrated enhanced cellular selectivity for
Aurora B and especially KDR compared to 1e. In particular, com-
pound 13a showed an excellent balance of TAK1 activity and ki-
nase cellular selectivity (>300ꢀ for Aurora B and >1000ꢀ for
KDR). We also evaluated compound 13f against the broader bio-
chemical panel of 192 kinases and confirmed that as selectivity
for Aurora B and KDR improved, so too did the overall selectivity
profile. At ꢁ50ꢀ TAK1 Ki, 13f showed >50% inhibition for only 4/
192 kinases in the panel: FLT3D835Y, HGK, MAP4K5, and MINK1.
We further obtained an X-ray crystal structure of 13a bound to
TAK1–TAB1 fusion protein (Fig. 3) and found that it possessed ex-
actly the features hypothesized to enhance selectivity. The single
largest change in this structure compared to the earlier structure
of 1c was the increased dihedral angle between the furopyridine
core and the 4-pyrazole substituent (ꢁ25° vs ꢁ4° in 1c), bringing
the pyrazole C5 ꢁ1 Å closer to the selectivity residue V42 (L838
in KDR).
Equally gratifying was that the majority of the 3-substituted
analogs had somewhat reduced extraction ratios in mouse and hu-
man liver microsomes compared to 1e. However, our microsomal
preparations were not run in the presence of the cofactors needed
for Phase II metabolism, for example UDPGA, and thus may not
fully capture the impact of the 3-substitutions on glucuronidation.
Indeed, the mouse pharmacokinetics of 13a and 13f (Table 5) were
dramatically improved compared to 1e. Both compounds had
ceived to be
a
synthetic nuisance requiring painstaking
chromatographic separation, now became the linchpin of the syn-
thetic campaign for the core 3-substitution.11 Compound 8 was
deprotected under acidic conditions to 9, iodinated (10) with NIS,
Cl
Cl
Cl
R2
Cl
Cl
Cl
Cl
b
d
c
R4
R4
O
O
O
O
I
R2N
N
H2N
N
H2N
N
H2N
N
8: R = Boc
a
10
11
12
g
9: R = H
R2
e
R2 =
N
Cl
h,i
13c-d
R4
f
O
j, k, l
j, m, l
N
13b
13a
S
13e
13f
R4 =
N
Boc2N
N
N
OTBS
14
Scheme 2. Preparation of 3-substituted analogs 13. Reagents and conditions: (a) 4 N HCl in 1,4-dioxane, 55 °C (68%); (b) NIS, MeCN, 60 °C (41%); (c) R4-B(pin), PdCl2dppf,
K2CO3, 1,4-dioxane/H2O, 75 °C (51%); (d) R2-B(pin), Pd(PPh3)4, Na2CO3, 1,4-dioxane/H2O, 120 °C,
W; (e) concd HCl, 40 °C (50%, 2 steps); (f) Zn(CN)2, Pd(PPh3)4, 1,4-dioxane/
DMF, 140 °C, W (12%); (g) Boc2O, DMAP, DCM, rt (79%); (h) Ar-B(pin), Pd2dba3, XPhos, K3PO4, 1,4-dioxane/H2O, 100 °C, W; (i) 4 N HCl in 1,4-dioxane, MeOH, rt; (j) vinyl-
W (69%); (k) (i) O3, DCM, (ii) NaBH4 (50%); (l) (i) DeoxoFluor, DCM, 35 °C, (ii) 4 N HCl in 1,4-dioxane, 70 °C; (m) (i) O3,
l
l
l
B(pin), Pd2dba3, XPhos, K3PO4, 1,4-dioxane/H2O, 100 °C,
l
DCM, (ii) PPh3 (69%).

K. R. Hornberger et al. / Bioorg. Med. Chem. Lett. 23 (2013) 4511–4516
4515
Table 4
SAR of 3-substitutions
N
N
R
S
N
N
OH
O
H2N
N
b
Compd
R
LLEa
IC50
(
lM)
ERg
TAK1c
pJNKd
pHH3e
pKDRf
Mouse
Human
1e
13a
13b
H
Cl
CN
6.3
5.1
5.1
0.009
0.028
0.16
0.015
0.018
0.017
0.16
6.5
>10
0.65
>20
>20
0.63
0.39
0.43
0.48
0.40
0.38
N
13c
13d
4.8
5.5
0.11
0.13
6.9
1.7
8.2
0.48
0.53
0.41
0.55
N
0.031
0.043
0.76
S
13e
13f
CH2F
CHF2
5.7
5.6
0.017
0.014
0.045
0.020
1.5
1.7
18.3
>20
0.43
0.39
0.49
0.48
a
LLE = TAK1 pKi–clogD7.4. See Ref. 1 for details.
b
c
d
e
f
Values are the mean of P2 experiments; NT = not tested.
Biochemical ALPHAScreen assay with TAK1–TAB1 fusion protein in the presence of 100
Inhibition of JNK phosphorylation in TNF -stimulated HCT-116 cells.
lM ATP.
a
Inhibition of histone H3 (HH3) S10 phosphorylation in HT-29 cells.
Inhibition of KDR autophosphorylation in HEK293/KDR cells.
Extraction ratio (ER) in mouse and human liver microsomes.
g
Table 5
Mouse pharmacokinetic parameters for selected compounds
Route
iva
Parameters
13a
13f
Cl (mL/min/kg)
20
10
t
(h)
2.1
1.8
½
AUC0–1 (ng h/mL)
Vss (L/kg)
1698
2.1
3440
1.0
p.o.b
p.o.c
Cmax
(
l
M)
3.8
8774
52
18.6
33,710
98
AUC0–1 (ng h/mL)
F (%)
Cmax
(l
M)
7.1
NT
AUC0–1 (ng h/mL)
F (%)
26780
32
NT
NT
a
b
c
iv PK dosed at 2 mg/kg.
p.o. PK dosed at 20 mg/kg.
p.o. PK dosed at 100 mg/kg. NT = not tested.
of tumor xenograft samples showed high in vivo levels of the acti-
vating cytokine TNF in the clear cell ovarian line TOV21G. Cellular
mechanistic inhibition of pJNK by ABC-FP in TOV21G (IC50 = 0.028 -
M) was comparable to that seen in HCT-116. It was thus postu-
a
Figure 3. Single crystal X-ray structure of compound 13a bound to TAK1–TAB1
fusion protein (2.5 Å resolution); PDB entry code 4L53. Key hydrogen bonds,
internal contacts, and dihedral angles are displayed.
l
lated that the TOV21G cell line would be a more robust model
for the in vivo evaluation of TAK1 inhibitors than HCT-116. CD-1
nude mice bearing subcutaneous TOV21G tumor xenografts were
treated with 100 mg/kg of ABC-FP daily for 14 days. Over the dos-
ing interval, ABC-FP showed a calculated 62% tumor growth inhibi-
tion on day 15, indicating some potential for a TAK1 inhibitor to be
useful in the treatment of ovarian cancer.
In summary, we have evolved a series of potent but relatively
poorly kinase selective 7-aminofuro[2,3-c]pyridine inhibitors of
TAK1 with poor pharmacokinetics into more selective inhibitors
with excellent oral exposure. These efforts culminated in the dem-
onstration of in vivo efficacy for compound ABC-FP in an ovarian
cancer model. Structure-based fine tuning of steric and conforma-
tional properties allowed for the exploitation of relatively small
substantially reduced iv clearance, with correspondingly excellent
oral exposure and bioavailability, and demonstrated oral PK curves
absent the late concentration increases which would be indicative
of enterohepatic recycling. At a higher 100 mg/kg dose, 13a
showed less than dose-proportional increases in Cmax and
AUC0–1 and decreased bioavailability, indicating solubility- and/
or permeability-limited absorption at the higher dose.
Now that suitable tool compounds had been identified, we next
turned to an evaluation of compound 13a (hereafter referred to as
ABC-FP13) in an in vivo model of ovarian cancer (Fig. 4). Although
until this point we had been screening compounds in vitro vs the
HCT-116 colorectal carcinoma cell line, a parallel ongoing analysis

4516
K. R. Hornberger et al. / Bioorg. Med. Chem. Lett. 23 (2013) 4511–4516
Acknowledgments
We would like to thank Ms. Viorica Lazarescu and Dr. Yingjie Li
for analytical support, Proteros Biostructures for solving the crystal
structure of 13a in complex with TAK1–TAB1 fusion protein, and
Drs. Arno Steinig, Andrew Garton, and Maryland Franklin for help-
ful discussions. We also thank Janine Schwedes for technical assis-
tance in the manuscript revisions.
References and notes
1. Hornberger, K.R.; Berger, D.M.; Crew, A.P.; Dong, H.; Kleinberg, A.; Li, A.-H.;
Medeiros, M.R.; Mulvihill, M.J.; Siu, K.; Tarrant, J.; Wang, J.; Weng, F.; Wilde,
V.L.; Albertella, M.; Bittner, M.; Cooke, A.; Gray, M.J.; Maresca, P.; May, E.;
Meyn, P.; Peick Jr., W.; Romashko, D.; Tanowitz, M.; Tokar, B. Bioorg. Med. Chem.
Lett. 2013, 23, 4517.
2. Manning, G.; Whyte, D. B.; Martinez, R.; Hunter, T.; Sudarsanam, S. Science
2002, 298, 1912.
3. Knight, Z. A.; Shokat, K. M. Chem. Biol. 2005, 12, 621.
4. Smyth, L. A.; Collins, I. J. Chem. Biol. 2009, 2, 131.
5. Dar, A. A.; Goff, L. W.; Majid, S.; Berlin, J.; El-Rifai, W. Mol. Cancer Ther. 2010, 9,
268.
Figure 4. Tumor growth inhibition of compound ABC-FP in TOV21G xenograft
model.
6. Korpanty, G.; Smyth, E. Curr. Pharm. Des. 2012, 18, 2680.
7. Eskens, F. A. L. M.; Verweij, J. Eur. J. Cancer 2006, 42, 3127.
8. (a) Leeson, P. D.; Springthorpe, B. Nat. Rev. Drug Disc. 2007, 6, 881; (b) Edwards,
M. P.; Price, D. A. Ann. Rep. Med. Chem. 2010, 45, 381. See footnote 15 of ref. 1 for
details of our computational method.
9. Davies, N. M.; Takemoto, J. K.; Brocks, D. R.; Yáñez, J. A. Clin. Pharmacokinet.
2010, 49, 351.
10. (a) http://www.optibrium.com/stardrop (accessed April 2, 2013); For details of
the underlying computational method, see: (b) Jones, J. P.; Mysinger, M.;
Korzekwa, K. R. Drug Metab. Dispos. 2002, 30, 7.
11. Synthesis conditions to increase the yield of 8 from the originally reported 18%
by increasing the equivalents of base and Cl3CCCl3 have been developed and
will be reported in due course.
12. Billingsley, K. L.; Anderson, K. W.; Buchwald, S. L. Angew. Chem., Int. Ed. 2006,
45, 3484.
differences in the kinase active sites to achieve useful levels of Aur-
ora
B and KDR cellular kinase selectivity. Kinase selectivity
improvements ultimately came along a more subtle path after
the more obvious avenues had failed. In conjunction, analysis of
in vitro metabolite profiling data enabled the design of compounds
with lower clearance through the use of the same single substitu-
tion used for kinase selectivity. We were further able to rapidly
execute the synthesis of these compounds by taking advantage of
an observed by-product in the core synthetic route. Although the
final structures resulting from this effort were only slightly re-
moved from the starting points, they were arrived at through a
carefully-reasoned and systematically evaluated approach. Further
profiling of the in vitro and in vivo pharmacology of ABC-FP will be
reported in due course.
13. ABC-FP = trans-4-{4-[7-amino-2-(1,2,3-benzothiadiazol-7-yl)-3-
chlorofuro[2,3-c]pyridin-4-yl]-1H-pyrazol-1-yl}cyclohexanol.