Highly selective and potent thiophenes as PI3K inhibitors with oral antitumor activity

Article abstract of DOI:10.1021/ml200126j

Highly selective PI3K inhibitors with subnanomolar PI3Kα potency and greater than 7000-fold selectivity against mTOR kinase were discovered through structure-based drug design (SBDD). These tetra-substituted thiophenes were also demonstrated to have good

Full text of DOI:10.1021/ml200126j

LETTER
pubs.acs.org/acsmedchemlett
Highly Selective and Potent Thiophenes as PI3K Inhibitors with Oral
Antitumor Activity
Kevin K.-C. Liu,* JinJiang Zhu, Graham L. Smith, Min-Jean Yin, Simon Bailey, Jeffrey H. Chen, Qiyue Hu,
Qinhua Huang, Chunze Li, Qing J. Li, Matthew A. Marx, Genevieve Paderes, Paul F. Richardson,
Neal W. Sach, Marlena Walls, Peter A. Wells, and Aihua Zou
Chemistry Department, Oncology, Pfizer Global Research and Development, 10770 Science Center Drive, La Jolla, California 92120,
United States
S Supporting Information
b
ABSTRACT: Highly selective PI3K inhibitors with subnanomolar
PI3Kα potency and greater than 7000-fold selectivity against mTOR
kinase were discovered through structure-based drug design (SBDD).
These tetra-substituted thiophenes were also demonstrated to have
good in vitro cellular potency and good in vivo oral antitumor activity in
a mouse PI3K driven NCI-H1975 xenograft tumor model. Compounds
with the desired human PK predictions and good in vitro ADMET
properties were also identified. In this communication, we describe the
rationale behind the installation of a critical triazole moiety to maintain
the intricate H-bonding network within the PI3K receptor leading to
both better potency and selectivity. Furthermore, optimization of the
C-4 phenyl group was exploited to maximize the compounds mTOR
selectivity.
KEYWORDS: Kinase inhibitor, PI3K, mTOR, thiophene, triazole, antitumor activity
hosphatidylinositol-3-kinases (PI3Ks) are lipid kinases, which
phosphorylate PIP2 (phosphatidylinositol diphosphate) to
AKT-dependent and AKT-independent mechanisms.5 Agents tar-
geting PI3Ks with additional mTOR kinase inhibitory activity may
carry extra toxicity through disruption of mTOR function in normal
cells. To find an inhibitor to inhibit this critical signaling pathway
with a better therapeutic index, we endeavored to develop a drug
selectively targeting PI3Ks.
P
PIP3 (phosphatidylinositol triphosphate), which recruits PDK
and AKT to the cell membrane. AKT is activated through phos-
phorylation of the membrane-bound PDK leading to a myriad of
downstream processes, including modulation of proteins such as
mTOR (mammalian target of rapamycin), GSK3, forkhead, NF-kB
transcription factors, and eNOS. In this manner, the PI3K/AKT
pathway regulates numerous processes such as metabolism, cell pro-
liferation, growth, survival, angiogenesis, and apoptosis.1 Aberrant
activation of PI3K has been linked to both tumor initiation and
maintenance. During tumorigenesis, activation of the PI3K/AKT/
mTOR pathway can occur through various mechanisms, including
loss of PTEN (the phosphatase that regulates PI3K signaling),
overexpression or activation of certain receptor tyrosine kinases
(e.g., EGFR, HER-2), interaction with activated Ras, overexpression
of the PI3K-α gene (PIKC3A), or mutations in PIKC3A that cause
elevated PI3K kinase activity.2,3 Deregulated PI3K/AKT/mTOR
pathway signaling has been implicated in poor prognosis and low
survival rates in numerous forms of cancers including lymphatic
tumors, glioblastomas, melanomas, breast, prostate, lung, colon, and
ovarian cancers.4 Many preclinical and clinical studies now strongly
indicate that PI3K plays a key role in the biology of human
cancers.2,3 However, most of the inhibitors in this signaling pathway
have inhibitory activity against both PI3K and mTOR kinases. It is
known that the mTOR complexes 1 and 2 both regulate cell
proliferation, apoptosis, angiogenesis, and metabolism through both
Several chemotypes were identified with decent PI3K binding
potency in our HTS assay. To derisk idiosyncratic toxicity
associated with our previous PI3K/mTOR dual inhibitors,6À8
we preferred a novel series with an available relevant cocrystal
showing proteinÀligand binding. This structural information
would allow us to design compounds through SBDD to accel-
erate the lead optimization process. Compound 1, a tetra-sub-
stituted thiophene with moderate PI3Kα Ki, 230 nM (Scheme 1),
was therefore picked for further investigation from other HTS hits
because of its novel structure and available PI3KγÀcompound 1
cocrystal structure (Figure 1). Unlike regular thiophenes that are
prone to oxidative and reactive metabolites, this tetra-substituted lead
compound blocks all potential oxidative sites on the thiophene core, a
fact that was subsequently confirmed by our drug metabolism studies.
For compounds to get good cellular potency, permeability, and
absorption, our initial priority was to replace the free carboxylic acid
Received: June 13, 2011
Accepted: September 19, 2011
Published: September 19, 2011
r
2011 American Chemical Society
809
dx.doi.org/10.1021/ml200126j ACS Med. Chem. Lett. 2011, 2, 809–813
|

ACS Medicinal Chemistry Letters
LETTER
Scheme 1. Progression from Initial Thiophene HTS Hit 1 to
Lead 4a
Scheme 2. Proposed Binding Tautomer of Triazole, T2, Is
Also the Calculated Global Minimum Energy (The Exception
Being the OPLS2005 Calculation)
a Compound 3 is a dual PI3K/mTOR inhibitor (ref 6).
The corresponding carboxylic amide 2 was thus made, and the
compound displayed a greater than 10-fold improvement on the
PI3Kα binding and less than 1 uM cellular potency in our cellular
S473 AKT phosphorylation assay. In addition, compound 2 also
provided good selectivity over mTOR kinase (Scheme 1). En-
couraged by this result, we carefully evaluated the cocrystal
structures of both compound 1 and our previously reported
potent PI3K/mTOR dual inhibitors from the 4-methyl pyrido-
pyrimidinone series.6À8 We wanted to exploit structureÀactivity
relationship (SAR) learnings from our dual inhibitor series and
apply these to improve both potency and selectivity for our novel
thiophene series. The carboxylic acid moiety of compound 1
corresponds to the methoxy pyrimidine moiety in the 4-methyl
pyridopyrimidinone series as shown in the overlapped cocrystal
structures in Figure 1. For clarity, compound 1 is colored in
brown, and compound 3 is colored in purple. Of particular note,
there is a water molecule serving as a bridge between the pyridyl
nitrogen of compound 3 and the PI3Kγ enzyme. In the cocrystal
structure of compound 3 shown in Figure 2 (structure on the
left), there is a water molecule that sits in between the conserved
Asp841 and Tyr867 residues of PI3Kγ, which correspond to the
Asp810 and Tyr836 residues of PI3Kα, respectively. This water
molecule is making H-bonds with both the pyridyl nitrogen of
compound 3 as previously mentioned and the Asp and Tyr
residues in the PI3Ks, thus forming a H-bond network (Figure 2,
left). To keep this intact within PI3K enzymes, a potent inhibitor
needs to be designed not to disrupt this intricate H-bond
network, or the inhibitor will end up losing potency because of
the energy penalty paid. This observation had important rami-
fications on our compound designs. With this observation in
mind, different amide bioisosteres were modeled and synthesized
to test our hypothesis. Among these analogues, compound 4 with
a 1,2,4-triazole group stands out, as this has excellent PI3Kα Ki
(1.7 nM), good cellular potency (S473 IC50 = 73 nM, Scheme 1),
and good selectivity against mTOR (mTOR Ki = 434 nM; 255Â
selectivity).
Figure 1. Cocrystal structures of compounds 1 (brown color) and 3
(purple color; PDB code: 3ML9) in PI3Kγ. The nitrogen of the
morpholine ring of compound 1 represents the kinase hinge binder.
Figure 2. Left: compound 3 in PI3Kγ (PDB code: 3ML9). There is a
water between Asp 841 and Tyr 867 to maintain a H-bonding network,
and the nitrogen on the pyridine ring is making contact with this water
molecule. Right: Proposed binding model for compound 4 in PI3Kα
(see the text for details).
It is unfortunate that we could not obtain further cocrystal
structures in this thiophene series to prove our model. However,
we believe the excellent PI3Kα potency of compound 4 is due to
the 1,2,4-triazole group replacing the water molecule in the ATP-
binding site, while the N1 nitrogen in the triazole functions as a
H-bond acceptor to interact with Tyr836, and the N2 nitrogen
moiety, which is known in general to be associated with both poor
permeability and absorption.9
810
dx.doi.org/10.1021/ml200126j |ACS Med. Chem. Lett. 2011, 2, 809–813

ACS Medicinal Chemistry Letters
LETTER
Table 1. Examples of Compounds from C-4 Phenyl Library:
Compounds 9 and 10 with Biological and in Vitro ADME
Data
Figure 3. X-ray structure of compound 4, which is consistent with both
our hypothesis and computational calculations.
Scheme 3. Amide Bioisosteric Analogues 4À8
the C-4 position was designed and synthesized after careful
filtration of a series of virtual compounds to optimize a series of
calculated physicochemical properties, such as clogP, molecular
weight, and polar surface area.
Indeed, this C-4 aryl library generated a number of very potent
compounds with excellent selectivity, that is, compounds 9 and
10 (Table 1). These two extremely potent compounds have
PI3Kα Ki less than 1 nM and mTOR Ki more than 1 μM. It is
worth noting that compound 9 has more than 7000-fold
selectivity over mTOR kinase. In addition, these two compounds
also have excellent selectivity over 40 other kinases, and no major
CYP inhibitions were observed. Less than 30% inhibition was
observed in 1A2, 2C9, 2D6, and 3A4 CYP enzymes at 3 μM.
According to the occupancy maps of PI3Kα and mTOR shown
in Figure 4, the out-of-plane 2,4-disubstituted C-4 phenyl group
bumps against a region of the mTOR protein that is defined by
the Ile-78, Leu-26, Lys-28, and Pro-10 residues. However, there
are no such unfavorable steric interactions in the corresponding
region in PI3K enzymes (Figure 4, smaller structure shown on
the right). For different PI3K isoforms, these compounds have
about equal potency against PI3Kδ and about 10À40Â less
potent against PI3Kγ.
The general synthesis of thiophenes and in particular com-
pound 4 is outlined in Scheme 4 below. The synthesis started
with the known compound A, which was prepared according to a
patent procedure.11 The thio-methyl ether A was subjected to
m-CPBA oxidation to give a mixture of the sulfone and sulfoxide B,
which was subsequently treated with morpholine to give com-
pound C in 77% yield. Compound C was coupled with 4-chlor-
ophenylboronic acid under optimized Suzuki conditions to give
compound D, which was hydrolyzed under basic conditions to
yield free carboxylic acid E in high yield. Compound E was then
converted to the amide 2 via the acid chloride under standard
conditions. Finally, compound 2 was treated with DMF and
DMA at elevated temperature to give the amide ylide intermedi-
ate, which was reacted directly with hydrazine to the desired
triazole 4 in 70% yield over the final two steps.12
acts as a H-bond donor interacting with Asp810, thus maintain-
ing and stabilizing the crucial PI3Kα protein structure (Figure 2,
structure on the right). The 1,2,4-triazole can exist in three
different tautomeric forms as indicated in Scheme 2. Different
computational calculation tools were employed to predict the
preferred global minimum energy species. Three out of four
calculations utilized predict T2 as the preferred tautomeric form,
which is consistent with our binding hypothesis for compound 4.
Further supporting evidence is provided by a single crystal X-ray
structure of compound 4, which also exists as the predicted T2
tautomer (Figure 3).
The potencies of other amide bioisosteres are shown in
Scheme 3. Compounds 4À6 display excellent potencies, whereas
compounds 7 and 8 do not, which is again consistent with our
H-bond donor/acceptor binding model. Compounds 4À6 are
capable of making the H-bond acceptor/donor interactions with
the aforementioned Tyr and Asp enzyme residues, respectively,
but compounds 7 and 8 are not. Compound 4 further differ-
entiates itself from compounds 5 and 6 due to its increased
metabolic stability with a lower extraction ratio10 of 0.3 vs 0.7 for
compounds 5 and 6 (Scheme 3).
With these potent and stable inhibitors in hand, we then
further focused to improve compound selectivity against mTOR.
Both PI3Kα and mTOR occupancy maps were generated, and
these mappings suggest that the C-4 phenyl moiety partially con-
tributes to mTOR selectivity for this series of compounds. There-
fore, a library with a diverse set of aryl and heteroaryl groups on
From this series, compound 9 was selected for further studies
due to the favorable balance of it is potency, selectivity, and
physicochemical properties. Drug metabolic studies suggested
that this compound would not have any issues with conjugated or
811
dx.doi.org/10.1021/ml200126j |ACS Med. Chem. Lett. 2011, 2, 809–813

ACS Medicinal Chemistry Letters
LETTER
Table 3. Selective PI3K Inhibitor, Compound 9, Is More
Active in NSCLCs Harboring the PI3KCA Mutation than
Other Cancer Cell Lines
Figure 4. Left: mTOR protein labeled in orange and PI3Kα labeled in
light blue with surface area shown. Out-of-plane 2,4-dichlorophenyl of
compound 9 impinges onto the mTOR occupancy map region defined
by Ile-78, Leu-26, Lys-28, and Pro-10. Right: compound 9 shown sitting
inside the PI3Kα occupancy map.
Scheme 4. Synthesis of Compound 4
Figure 5. DoseÀresponse curve of compound 10 in H1975 TGI
studies.
Table 2. Human PK Predictions of Compound 9 Based on
HLM
compd 9
in vitro HLM and Vdss from rats (SSS)
CLp
mL/min/kg
0.52
1.65
37
Vdss
L/kg
h
t1/2 effective
F
%
97
Figure 6. Free concentrations of compound 10 in H1975 TGI studies.
active metabolites. In rats with bile duct cannulation, less than 1%
parent drug was excreted unchanged into urine and bile. From
this data, we decided to use HLM for human PK predictions as
the major metabolic pathways are clearly through oxidative
transformations. Compound 9 has favorable human PK predic-
tions with low clearance, long half-lives, and excellent bioavail-
ability (Table 2).
Compound 9 was tested in different cell lines for its sensitivity
in the MTT cell viability assay. Interestingly, this selective PI3K
inhibitor is more sensitive with the NSCLC (nonsmall cell lung
cancer) cell lines that harbor PI3KCA mutations than cell lines
with the PTEN deletion as shown in Table 3.
812
dx.doi.org/10.1021/ml200126j |ACS Med. Chem. Lett. 2011, 2, 809–813

ACS Medicinal Chemistry Letters
LETTER
Markowitz, S.; Kinzler, K. W.; Vogelstein, B.; Velculescu, V. E. High
frequency of mutations of the PIK3CA gene in human cancers. Science
2004, 304, 554.
(5) Sarbassov, D. D.; Ali, S. M.; Sabatini, D. M. Growing roles for the
mTOR pathway. Curr. Opin. Cell Biol. 2005, 17, 596.
(6) Cheng, H.; Bagrodia, S.; Bailey, S.; Edwards, M.; Hoffman, J.;
Hu, Q.; Kania, R.; Knighton, D. R.; Marx, M. A.; Ninkovic, S.; Sun, S.;
Zhang, E. Med. Chem. Comm. 2010, 1 (2), 139.
(7) Liu, K. K.-C.; Huang, X.; Bagrodia, S.; Chen, J. H.; Greasley, S.;
Cheng, H.; Sun, S.; Knighton, D.; Rodgers, C.; Rafidi, K.; Zou, A.; Xiao,
J.; Yan, S. Quinazolines with intra-molecular hydrogen bonding scaffold
(iMHBS) as PI3K/mTOR dual inhibitors. Bioorg. Med. Chem. Lett.
2011, 21 (4), 1270–1274.
(8) Liu, K. K.-C.; Bagrodia, S.; Bailey, S.; Cheng, H.; Chen, H.; Gao,
L.; Greasley, S.; Hoffman, J. E.; Hu, Q.; Johnson, T. O.; Knighton, D.;
Liu, Z.; Marx, M. A.; Nambu, M. D.; Ninkovic, S.; Pascual, B.; Rafidi, K.;
Rodgers, C. M.-L.; Smith, G. L.; Sun, S.; Wang, H.; Yang, A.; Yuan, J.;
Zou, A. 4-Methylpteridinones as orally active and selective PI3K/mTOR
dual inhibitors. Bioorg. Med. Chem. Lett. 2010, 20 (20), 6096–6099.
(9) Kalgutkar, A. S.; Daniels, J. S. 1(Metabolism, Pharmacokinetics
and Toxicity of Functional Groups), Carboxylic acids and their bioisos-
teres. RSC Drug Discovery Ser. 2010, 99–167.
(10) μL/min/mg protein is the unit for CLint, which can be directly
obtained from the in vitro HLM experiments. We have scaled up the
HLM CLint with known liver physiology parameters, such as mg HLM/g
liver and g liver/kg body weight. Predicted hepatic clearance (CLH, mL/
min/kg) was calculated with the scaled up CLint (mL/min/kg) using the
well-stirred model. ER is the ratio of predicted CLH and hepatic blood
flow (QH). As a result, the predicted ER is a parameter that directly
reflects the HLM stability (CLint in HLM).
(11) Castano-Mansanet, A. M.;Dominguez-Manzanares, E.; Escribano,
A. M.; Fernandez, M. C.; Hornback, W. J; Jimenez-Aguado, A. M.;
Tromiczak, E. G.; Wu, Z.; Zarrinmayeh, H.; Zimmerman, D. M. Pre-
paration of thiophene and furan compounds for potentiating glutamate
receptor function. WO2005070916, 2005.
(12) For a detailed discussion of the synthetic chemistry, see Huang,
Q; Richardson, P. F.; Sach, N. W.; Zhu, J; Liu, K. K.-C.; Smith, G. L.;
Bowles, D. M. Development of Scalable Syntheses of Selective PI3K
inhibitors. Org. Process Res. Dev. 2011, 15 (3), 556–564.
Figure 7. Body weight change of compound 10 in H1975 TGI studies.
Because of a better rodent PK, compound 10 was dosed orally
in our in vivo antitumor model, PI3K driven NCI-H1975 xeno-
graft tumors.13 As shown in Figure 5, the compound demon-
strated dose responsive tumor growth inhibitory activity from 25
to 200 mg/kg in QD oral dosing. In Figure 6, free concentrations
of compound 10 at 50 and 200 mg have good exposures to cover
the compound cell IC50 (80 nM). There was no body weight
reduction in these in vivo studies as shown in Figure 7.
In summary, structure-guided lead optimization was applied
to a HTS hit to allow us to quickly discover extremely potent and
selective PI3K inhibitors with novel thiophene structures. Orally
active compounds with favorable human PK predictions were
also identified in this series. More details regarding the optimiza-
tion and development of these compounds will be disclosed in
due course.
’ ASSOCIATED CONTENT
(13) For animal studies, 6À8 week old nu/nu athymic female mice
were obtained from Jackson Laboratories; the mice were maintained in
pressurized ventilated caging at the Pfizer La Jolla animal facility. All
studies were done in compliance with Institutional Animal Care and Use
Committee guidelines. Tumors were established by injecting 2 Â 106
cells suspended 1:1 (v/v) with reconstituted basement membrane
(Matrigel, BD Biosciences). For tumor growth inhibition studies, mice
with established tumors of ∼150 mm3 were randomized and orally
treated with vehicle or PI3K inhibitor daily. Tumor dimensions were
measured with vernier calipers, and tumor volumes were calculated
using the formula: π/6(larger diameter) Â (smaller diameter)2. Tumor
growth inhibition percentage (TGI %) was calculated as 100(1 À ΔT/ΔC).
S
Supporting Information. Representative synthetic schemes
b
and procedures and NMR and MS data on new compounds. This
material is available free of charge via the Internet at http://pubs.acs.
org.
’ AUTHOR INFORMATION
Corresponding Author
*E-mail: kevin.k.liu@pfizer.com.
’ ACKNOWLEDGMENT
We thank Dr. Deepak Dalvie for generating the metabolite
data and Dr. Ted O. Johnson for useful discussions.
’ REFERENCES
(1) Engelman, J. A.; Luo, J.; Cantley, L. C. The evolution of
phosphatidylinositol 3-kinases as regulators of growth and metabolism.
Nat. Rev. Genet. 2006, 7, 606–619.
(2) Engelman, J. A. Targeting PI3K signalling in cancer: Opportu-
nities, challenges and limitations. Nat. Rev. Genet. 2009, 9, 550–62.
(3) Nuss, J. M; Tsuhako, A. L; Anand, N. K. Emerging Therapies
based on inhibitors of phosphatidyl-inositol-3-kinases. Annu. Rep. Med.
Chem. 2009, 44, 339–356.
(4) Samuels, Y.; Wang, Z.; Bardelli, A.; Silliman, N.; Ptak, J.; Szabo,
S.; Yan, H.; Gasdar, A.; Powell, S. M.; Riggins, G. J.; Willson, J. K.;
813
dx.doi.org/10.1021/ml200126j |ACS Med. Chem. Lett. 2011, 2, 809–813