Discovery of a potent and highly selective PDK1 inhibitor via fragment-based drug discovery

Article abstract of DOI:10.1016/j.bmcl.2011.03.032

We report the use of a fragment-based lead discovery method, Tethering with extenders, to discover a pyridinone fragment that binds in an adaptive site of the protein PDK1. With subsequent medicinal chemistry, this led to the discovery of a potent and hig

Full text of DOI:10.1016/j.bmcl.2011.03.032

Bioorganic & Medicinal Chemistry Letters 21 (2011) 3078–3083
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry Letters
journal homepage: www.elsevier.com/locate/bmcl
Discovery of a potent and highly selective PDK1 inhibitor via fragment-based
drug discovery
Daniel A. Erlanson a, , Joseph W. Arndt b, Mark T. Cancilla a, Kathy Cao a, Robert A. Elling a, Nicki English b,
⇑
Jessica Friedman b, Stig K. Hansen a, Cathy Hession b, Ingrid Joseph b, Gnanasambandam Kumaravel b,
Wen-Cherng Lee b, Ken E. Lind a, Robert S. McDowell a, Konrad Miatkowski b, Christine Nguyen a,
Thinh B. Nguyen a, Sophia Park a, Nuzhat Pathan b, David M. Penny a, Michael J. Romanowski a, Daniel Scott b,
Laura Silvian b, Robert L. Simmons a, Bradley T. Tangonan a, Wenjin Yang a, Lihong Sun b,
⇑
a Sunesis Pharmaceuticals, Inc., 395 Oyster Point Blvd., South San Francisco, CA 94080, USA
b Biogen Idec, Inc., 14 Cambridge Center, Cambridge, MA 02142, USA
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 a fragment-based lead discovery method, Tethering with extenders, to discover a
pyridinone fragment that binds in an adaptive site of the protein PDK1. With subsequent medicinal
chemistry, this led to the discovery of a potent and highly selective inhibitor of PDK1, which binds in
the ‘DFG-out’ conformation.
Received 9 December 2010
Revised 8 March 2011
Accepted 9 March 2011
Available online 17 March 2011
Ó 2011 Elsevier Ltd. All rights reserved.
Keywords:
PDK1
Selective
Tethering
Pyridinone
DFG-out
Fragment-based drug discovery
The phosphatidylinositol-3 kinase (PI3K) signalling pathway
could provide a compelling intervention point to treat cancer. When
the pathway is activated, PI3K generates phosphatidylinositol 3,4,5-
triphosphate, which causes protein 3-phosphoinositide-dependent
protein kinase-1 (PDK1) to localize to the plasma membrane, where
it first autophosphorylates and subsequently phosphorylates the
protein Akt. This activates Akt, which in turn phosphorylates a num-
ber of substrates ultimately resulting in increased cell survival and
resistance to apoptosis. Because the pathway is activated in most tu-
mors, blocking the pathway may sensitize tumors to other antican-
cer agents.1
The importance of the PI3K signalling pathway is reflected in
the number of programs targeting pathway members. However,
although there have been considerable drug discovery efforts on
PI3K and Akt, PDK1 has been relatively ignored. Indeed, when we
first started working on the target, there were almost no inhibitors
reported in the literature. Since then there have been some re-
ports,2–6 but most of these compounds are not specific for PDK1
over other kinases. In terms of validating the target, a potent and
highly specific inhibitor would be a useful tool compound.
We used fragment-based drug discovery to search for such
compounds. Fragment-based drug discovery builds drug leads
from smaller component molecules, or ‘fragments’.7–12 Screening
fragments rather than fully elaborated molecules can increase
the efficiency of sampling chemical space, and can potentially in-
crease the novelty of resulting molecules.
A previously reported fragment-based discovery technology
called Tethering13 uses reversible disulfide bonds between a cys-
teine residue in a protein and a thiol-containing fragment to enable
the capture and identification of weak binding fragments by mass
spectrometry.14 A later version of the technology, Tethering with
extenders, identifies companion fragments in the presence of a
known binding moiety, or ‘extender’.15,16 An extender can both
irreversibly modify a target cysteine residue and reversibly capture
companion fragments that bind to an adjacent site. We have also
reported a variant of this technology, Tethering with dynamic
extenders, and used it to discover Aurora A inhibitors.17 Here, we
report the use of Tethering with extenders to discover highly
selective PDK1 inhibitors.
⇑
Corresponding authors. Tel.: +1 415 978 2159 (D.A.E.); tel.: +1 617 679 2063;
fax: +1 617 679 3635 (L.S.).
E-mail addresses: derlanson@carmot.us (D.A. Erlanson), lihong.sun@biogenidec.
com (L. Sun).
0960-894X/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmcl.2011.03.032

D. A. Erlanson et al. / Bioorg. Med. Chem. Lett. 21 (2011) 3078–3083
3079
SH
PDK1
(1) Extender
O
H
N
H
N
X
S
R
S
S
(1)
N
H
S
NH2
N
N
(2) Extender-modified protein
S
SH
(2)
O
H
H
N
N
Protein
S
N
SH
PDK1
H
N
N
(3) Selected fragment
(3)
S
R
S
O
O
H2N
S
Cl
(4)
R
S
S
N
H
N
S
(4) Example of an unselected fragment
O
S
O
O S
N
S
S
(5)
H2N
S
S
N
H
PDK1
(5) Covalent complex
O
O
O
H
N
H
N
S
Cl
Protein
S
N
H
S
N
H
N
N
N
(6) Stable inhibitor
(6)
O
O
H
N
H
N
Cl
N
H
N
PDK1
O
N
N
Scheme 1. Tethering with extenders applied to PDK1.
Our approach is shown schematically in Scheme 1. We first
introduced a cysteine residue into PDK1 at position 166 (E166C).
Next, we reacted the protein with an extender based on diamino-
pyrimidine (DAP), a moiety known to bind in the purine binding
site, or hinge region.17,18 The extender contains an acrylamide moi-
ety that reacts with the introduced cysteine residue, as well as a
disulfide linker. Under the labeling conditions, this disulfide is re-
duced to reveal a free thiol positioned for probing the adaptive
binding site. We were able to obtain a crystal structure that reveals
the extender binding as expected (Fig. 1), with the kinase in the ac-
tive ‘DFG-in’ conformation.
peated the screen with phosphorylated (active) PDK1, selection of
the pyridinone was much weaker.
Subsequently, we replaced the disulfide linkage between the
DAP moiety and the pyridinone with a simple alkyl carbon chain.
We were pleased to find that the resulting compound 6 inhibited
a cascade PDK1-Akt assay with an IC50 of 0.2 lM. Under the condi-
tions of the assay, PDK1 autophosphorylates; when we repeated
the assay, this time with pre-phosphorylated PDK1, we found that
compound 6 was less potent, with an IC50 of 2.9 lM. This is consis-
tent with the Tethering results, and both suggest that the inhibitor
binds to the inactive form of the protein. Control experiments re-
vealed that the compounds were completely inactive against Akt
(data not shown).
We were interested to probe the linker-length dependence of
this chemotype. We hypothesized that if the pyridinone is binding
specifically to the protein, a linker, that is, either too long or too
short would cause a loss in activity. Table 1 shows that there is
in fact a pronounced linker length dependence, with 5 or 6 meth-
ylene units being optimal (compounds 10 and 6), which is consis-
tent with the linker length expected from Tethering. Furthermore,
aware of the potential for artifacts among low-affinity binders, we
tested some of our compounds at a high ATP concentration to en-
sure they were ATP-competitive. All the inhibitors are ATP-com-
petitive, suggesting that the DAP element is contributing to the
affinity.
We next performed a Tethering screen against a library of
roughly 3000 compounds. Of these, one of the strongest hits was
the pyridinone shown in Scheme 1. We used an unphosphorylated
form of the protein for the initial screen; interestingly, when we re-
Having established the optimal linker length, we next sought to
improve the potency of our molecules by exploring replacements
for the DAP. We made a small library of roughly 50 different purine
mimetics, many of which had previously been reported as frag-
ments of kinase inhibitors (Table 2).
We chose the five-methylene linker rather than the slightly
more potent six-methylene linker to try to keep the molecules as
compact as possible. Removing the DAP moiety (compound 12)
led to a complete loss of activity, and most amide (compound
13) or sulfonamide (compound 14) linked compounds were also
inactive or only weakly active. On the other hand, a number of ami-
no-heterocycles were tolerated, with compound 18 being slightly
Figure 1. Crystal structure of complex 2, (orange) extender-modified PDK1 (gray).
Co-ordinates have been deposited with the protein data bank (PDB code: 3PWY).

3080
D. A. Erlanson et al. / Bioorg. Med. Chem. Lett. 21 (2011) 3078–3083
Table 1
Linker length-dependence of initial hit from Tethering with extenders
Cl
O
O
N
H
H
H
N
N
N
n
O
N
N
Compound
n
Unphosphorylated
Phosphorylated
Phosphorylated PDK-AKT,
PDK-AKT IC50
(
lM)
PDK-AKT IC50
(
lM)
20ꢀ Km IC50 (lM)
7
8
9
10
6
11
2
3
4
5
6
7
ND
ND
ND
0.64
0.2
ND
>50
>50
57
10.8
2.9
22
ND
ND
>200
111
28
63
ND = Not determined.
Table 2
SAR around hinge-binding region
Cl
O
N
H
H
N
R
N
O
Compound
R
Unphosphorylated
Phosphorylated
Phosphorylated
PDK-AKT IC50
(
lM)
PDK-AKT IC50
(l
M)
PDK-AKT, 20ꢀ Km IC50 (lM)
O
N
10
0.64
10.8
111
N
H
N
O
12
13
>100
25
>200
80
>200
ND
O
NH2
N
N
O
O
O
S
14
15
ND
1.2
>50
18
ND
O
O
N
>200
N
N
N
16
17
18
2.0
31
23
5.3
>100
ND
O
N
3.4
N
N
N
0.78
45
O
more active than compound 10 as well as easier to work with syn-
thetically.20 Finally, the SAR suggested that a hydrogen-bond
acceptor ortho to the exocyclic nitrogen was important for activity.
Having settled on the methoxyaminopyrimidine as a hinge-bin-
der, we next explored changes to the pyridinone (Table 3). We
found that the chlorine contributed modestly to the potency
(compare compound 19 to compound 18), and that a chlorine
was also tolerated at the para position (compound 20), but not at
the ortho position (compound 21). Heterocycles such as compound
22 were generally not tolerated, nor were alkyl substituents (com-

D. A. Erlanson et al. / Bioorg. Med. Chem. Lett. 21 (2011) 3078–3083
3081
Table 3
SAR around pyridinone
R
O
O
N
H
H
O
N
N
N
N
Compound
R
Unphosphorylated
Phosphorylated
PDK-AKT IC50
(lM)
PDK-AKT IC50
(lM)
18
19
0.78
5.3
Cl
Cl
1.3
12
20
21
0.24
1.6
14
70
Cl
Cl
22
23
24
>100
ND
>50
>50
2.3
N
F
F
0.21
pound 23). Compound 24, with two fluorine atoms, was one of the
more potent compounds obtained. As also seen above, all of these
compounds show greater potency for unphosphorylated than pre-
phosphorylated PDK1, consistent with binding to the inactive form
of the protein.
the nitrogen, with a 30-fold difference in potency between the
methyl esters 27 and 28. Hydrolyzing the methyl ester 27 to the
free acid 29 led to a loss in potency, and substitution with either
branched (compound 30), unbranched (compound 31), or benzyl
substituents also reduced potency compared to compound 27.
However, compound 33, with a phenyl substituent directly off
the linker, proved to be a highly potent molecule with an IC50 of
2 nM in the pre-phosphorylated assay, and an EC50 of 400 nM in
the cell assay. Synthesis of compound 33 is shown in Scheme 3.
An X-ray crystal structure of compound 33 in complex with the
PDK1 kinase domain is shown in Figure 2. As expected, the cyclic
urea moiety lies in the ATP-binding pocket while the fragment
identified through Tethering binds in the allosteric DFG-out pock-
et. The hinge interactions consist of a pair of hydrogen bonds be-
tween the amino group and the carbonyl of the cyclic urea that
interact with, respectively, the carbonyl oxygen of Ser 160 and
At this point we were somewhat frustrated, as none of the sub-
stitutions had significantly improved potency over the initial com-
pound 6. We attempted to replace the flexible linker with a variety
of aromatic and heteroaromatic spacers as we had done in previous
programs,15 but none of these improved activity, and indeed most
of them decreased it (data not shown). Although we did not yet
have a crystal structure of any of our compounds bound to PDK1,
model-building suggested that the linker was buried far back in
the active site, in a sterically congested region, and thus anything
larger than a simple alkyl chain would not be accommodated. With
this in mind, we designed a new molecule that would incorporate
the hydrogen bond donor and acceptor that had been found to be
important in the hinge-binding analysis. We mapped the six atoms
between the hydrogen-bond donor and pyridinone amide of com-
pound 25 onto compound 26. Although the aromatic bond lengths
in the latter compound would be shorter than their counterparts in
compound 25, we reasoned that since the five-atom linker of com-
pound 24 was equipotent, the binding mode should be similar.
Much to our delight, we found that compound 26 was more potent
by roughly a factor of 50 in both the unphosphorylated and prep-
hosphorylated assays (Scheme 2).
F
F
unphosphorylated phosphorylated
O
O
N
N
N
PDK-AKT
PDK-AKT
H
N
µ
µ
IC50[ M]
IC50[ M]
N
O
N
H
O
2
4
6
0.32
2.6
25
26
F
F
We examined the cellular activity of compounds in PC-3 cells.
As mentioned above, Akt is a substrate for PDK1, and we used a cel-
lular assay from Meso Scale Discovery (MSD) to measure the phos-
phorylation of residue Thr 308. We found that compound 26 was a
<0.013
0.045
H
N
O
O
H
N
N
H
O
4
2
6
weak but reproducible inhibitor, with an EC50 of 5.1
The ethanolamine linker of compound 26 is readily amenable to
SAR studies using available chiral -amino alcohols (Table 4). We
found a strong stereochemical preference for substitution next to
lM.
a
Scheme 2. Redesign of linker-hinge region. Note that the lower limit of detection
for the unphosphorylated assay is 13 nM.

3082
D. A. Erlanson et al. / Bioorg. Med. Chem. Lett. 21 (2011) 3078–3083
Table 4
SAR around linker region
F
F
O
O
N
H
H
N
O
N
L
O
N
H
Compound
L
Unphosphorylated
Phosphorylated
Cellular assay p308
PDK-AKT IC50
(l
M)
PDK-AKT IC50
(l
M)
(AKT) EC50
(lM)
26
27
<0.013
0.045
5.1
<0.013
0.045
0.029
0.006
0.18
0.26
2.1
>10
CO2Me
CO2Me
CO2H
28
29
ND
30
31
0.077
0.024
0.5
2.6
2.7
0.083
32
33
0.042
0.46
5.4
<0.013
0.002
0.40
the backbone amide of Ala 162 of PDK1, which is reminiscent of
the binding mode of the indole based inhibitors of PDK1 from Ber-
lex.4 The phenyl substitutent at the linker of the inhibitor engages
phobic contacts with gatekeeper residue Leu 159 and Val 143.
Finally, the terminal difluorobenzyl ring resides in the allosteric
DFG-out pocket, that is, formed by mostly hydrophobic residues
including Met 134, Leu 137, Phe 142, Leu 196 and Ile 221.
in a key cation–p stacking interaction with the catalytic Lys 111.
Interestingly, the central pyridinone core of 33 forms two hydro-
gen bonds, one being intramolecular, while the other is with
PDK1 from the backbone amide of Asp 223 (DFG motif). In addi-
tion, the pyridinone ring is flanked on the opposite side by hydro-
The compound 33 was profiled at Upstate Scientific at 10 lM
concentration against what was at the time their entire panel of
241 kinases and showed no significant activity against any other
kinases in the PI3K pathway. Moreover, it inhibited only one kinase
F
F
F
F
F
F
O2N
H2N
O
O
N
O
O
N
NH2
HO
b
a
O
O
N
H
N
H
N
+
O
HO
HO
37
36
35
F
F
F
F
34
H
H2N
H2N
O
O
N
O
O
N
N
d
c
H
N
H
N
O
N
H
O
O
33
38
Scheme 3. Synthesis of compound 33. Reagents and conditions: (a) EDC, HOBT, DIEA, DMF, rt, 16 h, 84%; (b) 5-fluoro-2-nitroaniline, NaH, DMF, rt, 16 h, 53%; (c) Zn, HCl,
methanol–isopropanol, rt, 2 h, 100%; (d) 1,10-carbonyl-diimidazole, DMF, 60 °C, 4 h.

D. A. Erlanson et al. / Bioorg. Med. Chem. Lett. 21 (2011) 3078–3083
3083
Supplementary data
Supplementary data associated with this article can be found, in
the online version, at doi:10.1016/j.bmcl.2011.03.032.
References and notes
1. Mora, A.; Komander, D.; van Aalten, D. M.; Alessi, D. R. Semin. Cell Dev. Biol.
2004, 15, 161.
2. Feldman, R. I.; Wu, J. M.; Polokoff, M. A.; Kochanny, M. J.; Dinter, H.; Zhu, D.;
Biroc, S. L.; Alicke, B.; Bryant, J.; Yuan, S.; Buckman, B. O.; Lentz, D.; Ferrer, M.;
Whitlow, M.; Adler, M.; Finster, S.; Chang, Z.; Arnaiz, D. O. J. Biol. Chem. 2005,
280, 19867.
3. Gopalsamy, A.; Shi, M.; Boschelli, D. H.; Williamson, R.; Olland, A.; Hu, Y.;
Krishnamurthy, G.; Han, X.; Arndt, K.; Guo, B. J. Med. Chem. 2007, 50, 5547.
4. Islam, I.; Bryant, J.; Chou, Y. L.; Kochanny, M. J.; Lee, W.; Phillips, G. B.; Yu, H.;
Adler, M.; Whitlow, M.; Ho, E.; Lentz, D.; Polokoff, M. A.; Subramanyam, B.; Wu,
J. M.; Zhu, D.; Feldman, R. I.; Arnaiz, D. O. Bioorg. Med. Chem. Lett. 2007, 17,
3814.
5. Islam, I.; Brown, G.; Bryant, J.; Hrvatin, P.; Kochanny, M. J.; Phillips, G. B.; Yuan,
S.; Adler, M.; Whitlow, M.; Lentz, D.; Polokoff, M. A.; Wu, J.; Shen, J.; Walters, J.;
Ho, E.; Subramanyam, B.; Zhu, D.; Feldman, R. I.; Arnaiz, D. O. Bioorg. Med.
Chem. Lett. 2007, 17, 3819.
6. Medina, J. R.; Blackledge, C. W.; Heerding, D. A.; Campobasso, N.; Ward, P.;
Briand, J.; Wright, L.; Axten, J. A. ACS Med. Chem. Lett. 2010, 1, 439.
7. Erlanson, D. A.; McDowell, R. S.; O’Brien, T. J. Med. Chem. 2004, 47, 3463.
8. Rees, D. C.; Congreve, M.; Murray, C. W.; Carr, R. Nat. Rev. Drug Disc. 2004, 3,
660.
9. Fragment-based Approaches in Drug Discovery; Jahnke, W., Erlanson, D. A., Eds.;
Wiley-VCH: Weinheim, Germany, 2006; 34,.
10. Erlanson, D. A. Curr. Opin. Biotechnol. 2006, 17, 643.
11. Hajduk, P. J.; Greer, J. Nat. Rev. Drug. Disc. 2007, 6, 211.
12. Congreve, M.; Chessari, G.; Tisi, D.; Woodhead, A. J. J. Med. Chem. 2008, 51,
3661.
13. Tethering is a registered service mark of Sunesis Pharmaceuticals, Inc. for its
fragment-based drug discovery.
14. Erlanson, D.; Braisted, A.; Raphael, D.; Randal, M.; Stroud, R.; Gordon, E.; Wells,
J. Proc. Natl. Acad. Sci. U.S.A. 2000, 97, 9367.
Figure 2. Co-crystal structure of compound 33 bound to PDK1, showing the
interactions between the compound (green) and the protein (yellow). Figure made
using PyMOL.22 Co-ordinates have been deposited with the protein data bank (PDB
code: 3QC4)
(Musk, 98% inhibition) besides PDK1 (92% inhibition) by more than
80%. With its high kinase selectivity and desirable cell potency,
compound 33 was evaluated in a PC-3 prostate xenograft model.
The compound was dosed intraperitoneally at 100 mg/kg to PC-3
tumor-bearing CB17 SCID mice and showed statistically significant
35% inhibition of phosphorylated Akt-T308 at 2 h but less inhibi-
tion at later time points, due to its poor pharmacokinetic proper-
ties, such as a low plasma exposure and high clearance.19
In conclusion, we have used Tethering with extenders to iden-
tify a fragment that binds in the adaptive region of PDK1 and is
selective for the protein’s inactive conformation. We used medici-
nal chemistry to advance this fragment to a low nanomolar inhib-
itor of PDK1 with activity in cells. An X-ray structure of the
complex of this compound with PDK1 confirmed its unique DFG-
out binding mode, which is to our knowledge the first reported
DFG-out PDK1 inhibitor. This compound should prove a useful tool
for probing PDK1 biology. In fact, after seeing the disclosure of this
class of molecules,20 researchers from Merck have re-synthesized
and characterized compound 33, confirming our findings.21 More-
over, the promising in vivo PD activities demonstrated with this
highly selective PDK1 inhibitor have prompted us to search for po-
tent and selective PDK1 inhibitors that will have better pharmaco-
kinetic properties for improved PD and efficacy activity.
15. Erlanson, D. A.; Lam, J. W.; Wiesmann, C.; Luong, T. N.; Simmons, R. L.; DeLano,
W. L.; Choong, I. C.; Burdett, M. T.; Flanagan, W. M.; Lee, D.; Gordon, E. M.;
O’Brien, T. Nat. Biotechnol. 2003, 21, 308.
16. Erlanson, D. A.; Wells, J. A.; Braisted, A. C. Annu. Rev. Biophys. Biomol. Struct.
2004, 33, 199.
17. Cancilla, M. T.; He, M. M.; Viswanathan, N.; Simmons, R. L.; Taylor, M.; Fung, A.
D.; Cao, K.; Erlanson, D. A. Bioorg. Med. Chem. Lett. 2008, 18, 3978.
18. Bridges, A. J. Chem. Rev. 2001, 101, 2541.
19. PK parameters of compound 33 after intravenous injection in male rats at
1 mg/kg
MRTlast = 2.2 h.
dose:
AUClast = 289 h ꢀ ng/mL,
CL = 45.4 mL/min/kg,
and
20. Lind, K. E.; Cao, K.; Lin, E. Y.; Nguyen, T. B.; Tangonan, B. T.; Erlanson, D. A.;
Guckian, K.; Simmons, R. L.; Lee, W.; Sun, L.; Hansen, S.; Pathan, N.; Zhang, L.
WO 2008/005457, 2008.
21. Nagashima, K.; Shumway, S. D.; Sathyanarayanan, S.; Chen, A. H.; Dolinski, B.;
Xu, Y.; Keilhack, H.; Nguyen, T.; Wiznerowicz, M.; Li, L.; Lutterbach, B. A.; Chi,
A.; Paweletz, C.; Allison, T.; Yan, Y.; Munshi, S. K.; Klippel, A.; Kraus, M.;
Bobkova, E. V.; Deshmukh, S.; Xu, Z.; Mueller, U.; Szewczak, A. A.; Pan, B. S.;
Richon, V.; Pollock, R.; Blume-Jensen, P.; Northrup, A.; Andersen, J. N. J. Biol.
Chem. 2011, 286, 6433.
Acknowledgments
We thank Monya L. Baker for helpful suggestions on the manu-
script and Stuart Lam, Jacky Wang, and Liping Gao for preparatory
HPLC and compound handling.
22. DeLano, W. L. http://pymol.sourceforge.net/.