Fragment based discovery of a novel and selective PI3 kinase inhibitor

Article abstract of DOI:10.1016/j.bmcl.2011.07.117

We report the use of fragment screening and fragment based drug design to develop a PI3c kinase fragment hit into a lead. Initial fragment hits were discovered by high concentration biochemical screening, followed by a round of virtual screening to identify additional ligand efficient fragments. These were developed into potent and ligand efficient lead compounds using structure guided fragment growing and merging strategies. This led to a potent, selective, and cell permeable PI3c kinase inhibitor with good metabolic stability that was useful as a preclinical tool compound.

Full text of DOI:10.1016/j.bmcl.2011.07.117

Bioorganic & Medicinal Chemistry Letters 21 (2011) 6586–6590
Contents lists available at SciVerse ScienceDirect
Bioorganic & Medicinal Chemistry Letters
journal homepage: www.elsevier.com/locate/bmcl
Fragment based discovery of a novel and selective PI3 kinase inhibitor
Samantha J. Hughes a, , David S. Millan a, , Iain C. Kilty b, Russell A. Lewthwaite a, John P. Mathias a,
Mark A. O’Reilly b, Andrew Pannifer a, Anne Phelan b, Frank Stühmeier b, Darren A. Baldock a,
David G. Brown a
⇑
⇑
a Worldwide Medicinal Chemistry, Pfizer Global Research and Development, Sandwich Laboratories, Ramsgate Road, Sandwich, Kent CT13 9NJ, UK
b Allergy and Respiratory Research Unit, Pfizer Global Research and Development, Sandwich Laboratories, Ramsgate Road, Sandwich, Kent CT13 9NJ, UK
a r t i c l e i n f o
a b s t r a c t
Article history:
We report the use of fragment screening and fragment based drug design to develop a PI3c kinase frag-
Received 7 July 2011
Revised 29 July 2011
Accepted 31 July 2011
Available online 6 August 2011
ment hit into a lead. Initial fragment hits were discovered by high concentration biochemical screening,
followed by a round of virtual screening to identify additional ligand efficient fragments. These were
developed into potent and ligand efficient lead compounds using structure guided fragment growing
and merging strategies. This led to a potent, selective, and cell permeable PI3
metabolic stability that was useful as a preclinical tool compound.
c kinase inhibitor with good
Keywords:
Fragment
Kinase
Ó 2011 Published by Elsevier Ltd.
Fragment based drug discovery is a powerful method to identify
novel lead molecules. It starts from the identification of one or more
low molecular weight, low complexity molecules called fragments.
Suitable starting fragments are identified through screening a
fragment collection by sensitive biophysical techniques such as
nuclear magnetic resonance, surface plasmon resonance and
X-ray crystallography, or by high concentration biochemical screen-
ing.1,2 Althoughfragments typically have potencies in the millimolar
to micromolar range, they often have high ligand efficiency (LE)3 and
can be developed into leads by growing, linking or merging strate-
gies, usually guided by one or more protein X-ray crystal structures.
Kinases, as drug targets, present numerous challenges including
selectivity, cell penetration and crowded patent space. However,
fragment screening has shown to be a successful means of identify-
ing novel, efficient lead compounds in this area.4
Phosphatidylinositol-3 kinases (PI3Ks) are a family of lipid ki-
nases whose activity is central to a plethora of cellular signalling
pathways, and therefore ultimately to the maintenance of physio-
logical homeostasis. Specifically, PI3Ks belonging to the class 1
group have been linked to a number of patho-physiologies of key
interest to the pharmaceutical industry, such as cancer, rheuma-
toid arthritis, cardiovascular disease and respiratory disease.5,6
activities, including proliferation, differentiation, adhesion, migra-
tion, apoptosis and phagocytosis. Of particular note, the
c and d
isoforms are of significant interest in the pharmaceutical arena
due to the increasing evidence of their role in inflammatory dis-
eases together with their largely haematopoietic restricted tissue
distribution that may ultimately afford a more acceptable safety
toleration profile over the more ubiquitously expressed
a and b
isoforms.5
In an effort to identify novel leads for PI3c, the Pfizer fragment
library collection (5960 fragments) was screened using a high con-
centration biochemical assay.7 An advantage of biochemical
screening is that it provides an initial estimate of the Ki value even
for weak binding compounds. A large number of hits were identi-
fied, 312 of which were selected for IC50 determination, resulting in
150 confirmed hits (or dose response curves). Selected hits were
then further confirmed by orthogonal methods, namely isothermal
denaturation and where possible by protein X-ray crystallography.
A total of five X-ray structures of the fragment complexes were ob-
tained. Hits that made it through the screening sequence to this
point, with a protein co-crystal structure, were then considered
on the basis of other properties, such as LE, potential for kinase
selectivity and synthetic tractability. Compound 1, an amino pyraz-
olopyrimidine, was identified as an inhibitor with good LE. An X-
This class 1 group of PI3Ks contain four isoforms (a, b, c and d) that
are activated upon ligation of particular cell surface receptors
ray crystal structure of 1 in complex with PI3c was obtained which
which can subsequently generate a diverse range of cellular
demonstrates that 1 binds at the ATP site with the amino group
acting as a hydrogen bond donor to the main chain carbonyl of
Val882, and the pyrazole N3 accepting a hydrogen bond from
Val882 main chain NH. The 5-methyl group is buried in a hydro-
phobic pocket, packing against the gatekeeper residue Tyr867 as
⇑
Corresponding authors. Tel.: +44 1304 644965 (S.J.H); +44 1304 643497
(D.S.M.).
E-mail addresses: samantha.hughes@pfizer.com (S.J. Hughes), david.millan@
pfizer.com (D.S. Millan).
0960-894X/$ - see front matter Ó 2011 Published by Elsevier Ltd.
doi:10.1016/j.bmcl.2011.07.117

S. J. Hughes et al. / Bioorg. Med. Chem. Lett. 21 (2011) 6586–6590
6587
heavy atom count but it should be noted that the LE of bromine
containing compounds does tend to be higher than methyl ana-
logues, as bromine is more lipophilic and capable of achieving a
greater binding affinity than a carbon atom. The amino benzoxaz-
ole 3 is another similarity search hit with improved potency and LE
relative to the initial hit 1. Compound 3 demonstrates that remov-
ing the ring junction nitrogen is therefore not deleterious to the
potency, whereas addition of a nitrogen atom to form a triazolo-
pyrimidine as in 4 and 5, reduces activity. Interestingly, the pyraz-
olopyrimidine substructure hits 6 and 7 retain activity, despite
replacement of the amino substituent in 1. The most potent frag-
ment from the close-in-mining 8 is also the most ligand efficient.
The 7-chloro substituent in 8 is likely to offer a different vector
for growth compared to the other cores substituents. However, it
was not possible to obtain an X-ray crystal structure of the com-
plex with 8, and therefore this was not pursued.
N
N
H2N
N
1
IC50 915µM
LE 0.35
Figure 1. Structure of compound 1 a fragment screening hit, and X-ray crystal
structure of compound 1 in PI3K
c
.
well as Val882, Phe 961 and Glu880 (Fig. 1), and may potentially
offer wider kinase selectivity.8
As the data above suggests, the close in virtual screening pro-
cess revealed several useful SAR data points and identified struc-
turally differentiated fragments with improved potency and
increased LE. At this point we began exploration around hit 1,
and in particular to investigate binding interactions of the inhibitor
with the hinge region of the kinase. Addition of just three heavy
atoms to compound 1, to produce the acetamide 9 led to a signif-
icant improvement in both potency and LE. As was noted for the
close-in-mining results, it is worth exploring small changes to
the initial fragment ahead of larger changes. The identification of
good initial starting points through fragment screening is clearly
a useful strategy but to leverage the power of this technique the
fragment hit must be synthetically tractable such that further
SAR around the fragment can be explored. Moreover, being able
to explore several different vectors from a single fragment is also
important. Fragment hit 1 was unfortunately not synthetically
tractable to the point where we could rapidly explore the SAR
around it but fortunately hit 2 had greater potential. We adopted
fragment growing and merging strategies on hit 2 to further opti-
mise for potency whilst attempting to maintain the excellent LE of
the starting point.
The initial fragment hit 1 was followed up by virtual screening
of the Pfizer solid compound collection (MW <300 Da) to identify
similar compounds, using a custom written, proprietory, close-in-
mining tool that uses both substructure and nearest neighbour
(ECFP4 fingerprints) searches.9 The searches were restricted to
small changes in the heavy atom count in order to explore the
chemical space around the fragment hit. Virtual screening identi-
fied a further round of compounds for biochemical dose–response
screening (ꢀ20 compounds), resulting in a hit rate of ꢀ50%. Several
structurally differentiated fragments were identified as a result
(Table 1).
Pleasingly compound 2, an amino imidazopyridine, proved to
be both more potent and had greater LE relative to compound 1.
The potency increase was achieved without any change in the
Table 1
Close-in-mining hits around compound 1
Compound
Structure
Kinase-Glo™ IC50
156
(
lM)
LE
N
Fragment merging is an alternative fragment elaboration strat-
egy, where a fragment is merged with another compound that has
a binding mode that overlaps with that of the fragment. X-ray crys-
tallography showed that compound 1 and a literature compound
H2N
H2N
2
0.47
N
Br
Cl
N
O
1010 overlayed well in the ATP-binding site of PI3
c (Fig. 2a). There-
3
4
380
0.42
fore hybridisation of 1 with the methylene thiazolidinedione (TZD)
moiety of compound 10 was hypothesised to improve potency, but
as mentioned above 1 was synthetically difficult to manipulate.
The more LE and synthetically amenable hit 2 is predicted to retain
the same binding mode as 1 and therefore addition of the methy-
lene TZD moiety onto 2 led to compound 11. Compound 11 was
significantly more potent in the biochemical assay, relative to 2,
and even more interestingly was improved in LE, which is not usu-
ally seen when fragments are significantly increased in molecular
weight.11 Gratifyingly, incorporation of an acetamide onto the pyr-
azolopyridine of 11 led to a sevenfold increase in potency in 12
with a similar LE. An X-ray crystal structure of the complex of hy-
N
N
>1600
<0.32
H2N
N
N
Cl
N
N
N
5
6
>1670
395
<0.34
0.33
N
N
N
N
brid 12 with PI3c was obtained and the compound was observed to
bind as anticipated from the component fragments (Fig. 2b).8
Compound 9 was prepared by a simple acetylation procedure
from the fragment hit 1, in good yield, which was originally pre-
pared by the procedure of Ried et al. (Scheme 1).12
N
N
N
N
7
8
773
53
0.35
0.49
The imidazopyridine 11 was prepared by first reacting commer-
cially available methyl 6-aminonicotinate with chloroacetalde-
hyde, followed by reduction of the methyl ester 13 with lithium
aluminium hydride to afford the alcohol 14. The alcohol 14 was
then oxidised to the aldehyde 15 with manganese dioxide, fol-
lowed by condensation with commercially available 1,3-thiazoli-
dine-2,4-dione to afford the desired product 11 (Scheme 2).
N
N
N
N
H2
N
Cl

6588
S. J. Hughes et al. / Bioorg. Med. Chem. Lett. 21 (2011) 6586–6590
OH
N
N
H
N
F
O
N
O
O
S
N
H
O
9
10
HTRF IC50 19uM
HTRF IC50 30nM
LE 0.46
N
N
H
N
N
N
O
O
S
N
O
S
H
O
N
H
O
12
11
HTRF IC50 34nM
HTRF IC50 230nM
LE 0.48
LE 0.53
a
b
Figure 2. X-ray structures of (a) compound 1 (pink) and 10 (green), and (b) the hybrid compound 12 in PI3c.
O
O
N
N
N
a
N
N
H
N
N
N
H2N
b
a
HO
N
c
O
N
O
N
N
14
1
9
O
13
NH2
Scheme 1. Reagents and conditions: (a) acetic anhydride, reflux, 2 h, 100%.
N
O
N
N
d
+
N
S
O
N
O
The imidazopyridine core in 12 was prepared according to the
procedure of Hamdouchi et al. by first reacting the commercially
available 2-amino-5-iodopyridine with tosyl chloride, affording
16, followed by alkylation with iodoacetamide to give 17.13 Ring
closure with trifluoroacetic anhydride afforded the imidazopyri-
dine 18. The trifluoroacetyl group in 18 was swapped for an acetyl
group to afford 19, followed by metallation and quenching with
N,N-dimethyl formamide to afford 20. The aldehyde 20 was then
condensed with 1,3-thiazolidine-2,4-dione, as before, to afford
the desired product 12 (Scheme 3).14
S
11
O
15
N
H
O
Scheme 2. Reagents and conditions: (a) chloroacetaldehyde, NaHCO3, MeOH,
reflux, 8 h, 80%; (b) LiAlH4, THF, 0 °C, 2 h, 32%; (c) MnO2, CH2Cl2, reflux, 5 h, 66%;
(d) b-alanine, AcOH, 100 °C, 2 h, 33%.
inhibition at 1
l
M. Pan-kinase selectivity in this panel was impres-
sive with only 3 kinases exhibiting >50% inhibition at 1
l
M,
As compound 12 proved to be a potent and LE lead we were
keen to assess its general kinase selectivity, and therefore potential
utility as a tool compound. Kinase selectivity of 12 was assessed in
an in-house panel of 43 kinases by measuring single point
namely CLK1 (94%), Caesin kinase II alpha chain (CKIIa) (95%)
and Tao kinase 3 (TAO3) (57%) (Fig. 3). This selectivity is notable
for a compound that binds in the ATP site through a standard

S. J. Hughes et al. / Bioorg. Med. Chem. Lett. 21 (2011) 6586–6590
6589
I
I
O
O
O O
I
a
b
S
S
N
N
N
N
H
N
NH2
H2N
16
17
O
c
N
H
N
H
N
N
N
e
H
N
d
N
F
F
N
N
I
F
O
I
O
O
19
O
20
18
f
N
H
N
N
O
O
S
12
N
H
O
Scheme 3. Reagents and conditions: (a) TsCl, pyridine, rt, 16 h, 92%; (b) iodoacetamide, di-iso-propylethylamine, DMF, rt, 16 h, 53%; (c) TFAA, CH2Cl2, reflux, 16 h, 44%; (d) (i)
NaOH, rt, 16 h, 100%, (ii) acetic anhydride, AcOH, 16 h, rt, 77%; (e) iPrMgCl, DMF, THF, ꢂ40 °C, 2 h, 68%; (f) b-alanine, 1,2-thiazolidinedione, AcOH, 100 °C, 3.5 h, 18%.
Figure 3. Heatmap showing single point activity of compound 12 against 43 kinases at 1 lM.
kinase donor/acceptor interaction at the hinge. However, the com-
pound did not display any selectivity versus the PI3 kinase iso-
form (HTRF IC50 25 nM). The hybrid 12 displayed good metabolic
stability in both human and rat liver microsomes (10 l/min/mg
and 15 L/min/mg, respectively), and crucially demonstrated good
Acknowledgments
a
We thank Sara Chunn and Sandra Newman for synthesis of
compounds.
l
l
membrane permeability in a parallel articifical membrane perme-
References and notes
ability assay (PAMPA) (13.0 ꢁ 10ꢂ6 cm/s). In line with the good
1. Rees, D. C.; Congreve, M.; Murray, C. W.; Carr, R. Nat. Rev. Drug Disc. 2004, 3,
660.
PAMPA data, the compound has shown activity (IC50 0.760 lM)
in a primary whole-cell assay, where compounds were evaluated
for their ability to inhibit fMLP stimulated superoxide production
by human isolated neutrophils.
In conclusion, screening of the Pfizer fragment library against
PI3c by a high concentration biochemical assay led to the identifi-
cation of a weakly potent, but LE amino pyrazolopyrimidine frag-
ment hit 1. Close in virtual screening allowed exploration of
chemical space around this initial hit, generating several structur-
ally differentiated fragment cores with improved potencies and LE.
Initial fragment growing efforts were restricted to small changes,
and afforded significant gains in potency. Subsequent merging of
the fragment with a literature compound 10 led to the identifica-
2. Alex, A. A.; Flocco, M. M. Curr. Top. Med. Chem. 2007, 7, 1544.
3. Hopkins, A. L.; Groom, C. R.; Alex, A. A. Drug Discovery Today 2004, 9, 430.
4. Erlanson, D. A.; Arndt, J. W.; Cancilla, M. T.; Cao, K.; Elling, R. A.; English, N.;
Friedman, J.; Hansen, S. K.; Hession, C.; Joseph, I.; Kumaravel, G.; Lee, W.-C.;
Lind, K. E.; McDowell, R. S.; Miatkowski, K.; Nguyen, C.; Nguyen, T. B.; Park, S.;
Pathan, N.; Penny, D. M.; Romanowski, M. J.; Scott, D.; Silvian, L.; Simmons, R.
L.; Tangonan, B. T.; Yang, W.; Sun, L. Bioorg. Med. Chem. Lett. 2011, 21, 3078.
5. Rückle, T.; Schwarz, M. K.; Rommel, C. Nat. Rev. Drug Disc. 2006, 5, 903.
6. Kong, D.; Yamori, T. Cancer Sci. 2008, 99, 1734.
7. The initial primary screen of the fragment library against recombinant human
His-PI3c was performed by quantifying the depletion of ATP using the Kinase-
Glo™ luciferase assay format. The assay tolerated DMSO concentrations of up
to 15%, thereby enabling primary screening at compound loads of 0.5–1.5 mM.
The fragment library was screened in duplicate, with an excellent correlation
between the first and the second data set and typical Z0 values of greater than
0.6. The assay, which was used for the profiling of the primary hits in a
concentration response format, allowed the quantification of Ki values in a
tion of a novel, potent and ligand efficient PI3c inhibitor 12 which
shows activity in a cell based assay, and with promising general ki-
nase selectivity and metabolic stability. The strategies outlined in
this paper have allowed the rapid identification of a tool compound
for further preclinical investigation.
range from 500 nM to 500 lM. During later stages of hit to lead optimisation
the cost efficient and robust Kinase-Glo™ luciferase assay was replaced by a
HTRF based assay allowing the quantitative determination of Ki values down to
1 nM, with typical Z0 values of greater than 0.8.

6590
S. J. Hughes et al. / Bioorg. Med. Chem. Lett. 21 (2011) 6586–6590
8. The coordinates of compounds 1 and 12 bound to PI3
c
have been deposited
11. Reynolds, C. H.; Bembenekb, S. D.; Tounge, B. A. Bioorg. Med. Chem. Lett. 2007,
17, 4258.
with the Protein Data Bank and assigned PDB ID codes 3zvv (2.5 Å) and 3zw3
(2.8 Å), respectively.
9. Rogers, D.; Hahn, M. J. Chem. Inf. Model. 2010, 50, 742.
10. Pomel, V.; Klicic, J.; Covini, D.; Church, D. D.; Shaw, J. P.; Roulin, K.; Burgat-
Charvillon, F.; Valognes, D.; Camps, M.; Chabert, C.; Gillieron, C.; Francüon, B.;
Perrin, D.; Leroy, D.; Gretener, D.; Nichols, A.; Vitte, P. A.; Carboni, S.; Rommel,
C.; Schwarz, M. K.; Rückle, T. J. Med. Chem. 2006, 49, 3857.
12. Ried, W.; Peuchert, K. P. Jus. Liebigs Ann. der Chem. 1965, 682, 136.
13. Hamdouchi, C.; Zhong, B.; Mendoza, J.; Collins, E.; Jaramillo, C.; Eugenio De
Diego, J.; Robertson, D.; Spencer, C. D.; Anderson, B. D.; Watkins, S. A.; Zhanga,
F.; Brooks, H. B. Bioorg. Med. Chem. Lett. 2005, 15, 1943.
14. All compounds showed 1H NMR and mass spectra consistent with their
structure and were >95% purity by HPLC.