Fragment-based design, synthesis, biological evaluation, and SAR of 1H-benzo[d]imidazol-2-yl)-1H-indazol derivatives as potent PDK1 inhibitors

Article abstract of DOI:10.1016/j.bmcl.2017.10.041

In this work, we describe the use of the rule of 3 fragment-based strategies from biochemical screening data of 1100 in-house, small, low molecular weight fragments. The sequential combination of in silico fragment hopping and fragment linking based on S1

Full text of DOI:10.1016/j.bmcl.2017.10.041

Accepted Manuscript
Fragment-Based Design, Synthesis, Biological Evaluation, and SAR of 1H-
benzo[d]imidazol-2-yl)-1H-indazol Derivatives as Potent PDK1 Inhibitors
Ting Chen, Venkataswamy Sorna, Susie Choi, Lee Call, Jared Bearss, Kent
Carpenter, Steven L. Warner, Sunil Sharma, David J. Bearss, Hariprasad
Vankayalapati
PII:
S0960-894X(17)31035-1
https://doi.org/10.1016/j.bmcl.2017.10.041
BMCL 25370
DOI:
Reference:
To appear in:
Bioorganic & Medicinal Chemistry Letters
Received Date:
Revised Date:
Accepted Date:
13 August 2017
3 October 2017
19 October 2017
Please cite this article as: Chen, T., Sorna, V., Choi, S., Call, L., Bearss, J., Carpenter, K., Warner, S.L., Sharma,
S., Bearss, D.J., Vankayalapati, H., Fragment-Based Design, Synthesis, Biological Evaluation, and SAR of 1H-
benzo[d]imidazol-2-yl)-1H-indazol Derivatives as Potent PDK1 Inhibitors, Bioorganic & Medicinal Chemistry
Letters (2017), doi: https://doi.org/10.1016/j.bmcl.2017.10.041
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Bioorganic & Medicinal Chemistry Letters
Fragment-Based Design, Synthesis, Biological Evaluation, and SAR of 1H-
benzo[d]imidazol-2-yl)-1H-indazol Derivatives as Potent PDK1 Inhibitors
a
c
c
c
b
b
Ting Chen , Venkataswamy Sorna , Susie Choi , Lee Call , Jared Bearss , Kent Carpenter , Steven L.
b
c
b *
c *
Warner , Sunil Sharma , David J. Bearss , Hariprasad Vankayalapati
a
Department of Physiology & Developmental Biology, Brigham Young University, Provo, UT 84602 U.S.A.
Tolero Pharmaceuticals, Inc., 2975 Executive Parkway, Suite 320, Lehi, UT 84043, U.S.A.
Division of Oncology of School of Medicine and Center for Investigational Therapeutics at Huntsman Cancer Institute, University of Utah, 2000 Circle of
b
c
Hope, Salt Lake City, UT 84112 U.S.A.
ARTICLE INFO
ABSTRACT
Article history:
Received
Revised
Accepted
Available online
In this work, we describe the use of the rule of 3 fragment-based strategies from biochemical
screening data of 1100 in-house, small, low molecular weight fragments. The sequential
combination of in silico fragment hopping and fragment linking based on S160/Y161/A162
hinge residues hydrogen bonding interactions leads to the identification of novel 1H-
benzo[d]imidazol-2-yl)-1H-indazol class of Phosphoinositide-Dependent Kinase-1 (PDK1)
inhibitors. Consequent SAR and follow-up screening data led to the discovery of two potent
PDK1 inhibitors: compound 32 and 35, with an IC50 of 80 nM and 94 nM, respectively. Further
biological evaluation showed that, at the low nanomolar concentration, the drug had potent
ability to inhibit phosphorylation of AKT and p70S6, and selectively kill the cancer cells with
mutations in both PTEN and PI3K. The microarray data showed that DUSP6, DUSP4, and
FOSL1 were down-regulated in the sensitive cell lines with the compound treatment. The in vivo
test showed that 35 can significantly inhibit tumor growth without influencing body weight
growth. Our results suggest that these compounds, especially 35, merit further pre-clinical
evaluation.
Keywords:
PDK1 inhibitor
Fragment-based design
1
H-benzo[d]imidazol-2-yl)-1H-indazoles
AN3-CA
DUSP6
DUSP4
FOSL1
KATO-III
MV4-11
2
009 Elsevier Ltd. All rights reserved.
*
Corresponding author’s e-mail: hari@hci.utah.edu, dbearss@toleropharma.com

The Phosphoinositide-Dependent Kinase-1 (PDK1), by passing the signal down to the AKT, plays the central role in
PI3K/AKT/mTOR pathway. Several AGC super family kinases including PKC, SGK, PKB/AKT, p70S6K, and PDK1 itself can be
phosphorylated by PDK1. Mutations in genes (PI3K and PTEN) that regulate PIP3 production exist in variety of tumors, such as breast,
ovarian, prostate, gastric, lung, and hematological cancers. These mutations usually lead to elevated levels of PIP3, which can enhance
the activation of PKB/AKT, p70S6K, and SGK. Inhibitors of PDK1 present a viable therapeutic option for the treatment of PDK1
1
elevated cancer types . If the selective PDK1 inhibitor that was also orally bioavailable small molecules can be discovered, it would
show the most desirable therapeutic effect in these cancers, particularly ovarian cancers where the physiological role of the PDK1
2
mediated pathway is well defined .
3,4
Fragment-based lead discovery has become a powerful tool for the discovery and optimization of potent leads . As a substitute to
more conventional high-throughput screening methods of hit identification process, the screening of low molecular weight (MW <300
Da) and less lipophilic (clog P < 3) fragments can often serve as key moieties for the identification of novel, tractable hits with a high
5
hit rate . These fragments exhibit superior ligand efficiency and can be improved into potent leads with promising “rule of five”
properties. Thus, the design and optimization on the final lead compound performed by identifying and optimizing individual, high
ligand efficient fragments. This is followed by the synthetic linkage of two or more fragments to produce the required affinity for the
target protein with ideal lead-like properties. The main advantage of fragment-based design techniques is that it requires the synthesis
and testing of fewer compounds and significantly increases the amount of chemical space tested by using a comparatively small library
of compounds.
6
In the present study, structural analysis of a previously reported crystal structure of PDK1 kinase bound to BX-320 facilitated the
7,8,9
primary design and subsequent discovery of novel 1H-benzo[d]imidazol-2-yl)-1H-indazoles PDK1 inhibitors . Two (32 and 35 in
Table 1) of the compounds we discovered showed low nanomolar activity in 10 dose-response curve cell viability assays, particularly in
7
cell lines with PTEN and PI3K mutations . When compound 32 screened against 21 kinases, it showed good potency against PDK1.
Compound 32 also effectively lowered phospho-AKT (Thr308) levels in a dose-dependent manner.
Design and Synthesis of PDK1 Inhibitors: We employed fragment-based virtual screening, using varied selection of shape
similarity methods for searching commercially available compound databases for molecules with similar configurations, but altered
10
chemical connectivity when compared to the known PDK1 inhibitors BX-320, and BX-517 . Additionally, the application of fragment-
based design strategies coupled with knowledge-based scaffold-hopping led to the collection of approximately 1100 low molecular
11,12,13
14
weight fragments
that were physically acquired and screened in a luciferase-based PDK1 biochemical kinase assay . The
combination of in silico virtual screening based on hydrogen bonding to amino acid residues (Ser160/Ala162) in the hinge region of
PDK1 and a biochemical kinase assay led to the identification of multiple PDK1 inhibitor scaffolds that were all within the “rule of
15, 16
three” criteria of chemical space, suggesting they were suitable for further optimization
.
Active fragments and their binding modes were validated using protein-ligand computational methods and the PDK1 crystal
6
structure (PDB ID: 1Z5M). IC50 values for the five most active fragments were between 29-122 μM of which 4-iodo-1H-indazol-3-
amine (3) and 6-bromo-1H-indazol-3-amine (4) (Figure 1A and 1B) showed the most significant promise based on the results of the
biochemical assay and information obtained from computational efforts.
Figure 1A. Both the panels depict the binding mode of fragment 3 within the PDK1 ATP binding site. The dashed line in blue/red shows hing hydrogen bonding
interactions. Critical residues labeled and depicted in surface representation.
Figure 1B. Chemical structure of fragment hits 1-5 to leads and its PDK1 activities.
Compared to the other hits, these two fragments demonstrated sufficient potency and ligand efficiency that allowed for further
optimization to more complex lead candidates possessing the required potency and selectivity toward PDK1. Furthermore, these

fragments appeared to be excellent starting points for optimizing cellular activity and the physicochemical properties required to
investigate candidate compounds in pre-clinical studies and, ultimately, in clinical trials. The ICM and GOLD docking experimental
results on these fragments provided a clear rational on the less potent activities of the 5- or 6-bromo-1H-indazol-3-amines (fragments 1
and 5) based on its binding mode when compared to the corresponding, more potent activity of 4-iodo-1H-indazol-3-amine (3) fragment
th
series. The 6 position of bromine atom in fragment 1 is involved in direct steric clash with Leu111 residue which is located in the
hydrophobic site of PDK1. In the case of 5-Br-1H-indazol-3-amine fragment (1), 5-Br substitution positioned towards Thr122 residue
of PDK1 pocket, whereas the indazole involved in interactions with Tyr122 is retained, albeit with weak binding energy compared to
6
the active hit fragment 4-iodo-1H-indazol-3-amine (3). Utilizing 3-D crystal structures , computational screening, and standard
medicinal chemistry efforts, fragments 3 and 4 systematically optimized to yield small molecule inhibitors (Figure 1B) with a
15, 16
significant improvement in potency toward PDK1 and possessing favorable physicochemical properties
for drug development.
Several of the compounds demonstrated mid-nM inhibitory activities against PDK1 and several rounds of optimization afforded two
compounds, namely compounds 32 and 35. These lead compounds exhibited potent PDK1 activity with IC50 values of 80 and 90 nM,
respectively (Table 1). The known PDK1 inhibitor BX-517 was used as positive control exhibited an IC50 of 9 nM. These compounds
1
2
discovered through the exploration of SAR at the R and R positions of the core scaffold (Scheme 1). Initial studies were aimed at the
substitutions of the directly attached 5-position of indazole ring (19-21, Table 1), which were prepared from commercially available
phenylboronic acids. Evaluation of compounds 19-23 indicated that either 3-substituted methylsulfone (19), hydrophobic, electron
withdrawing groups (20), or a cyclopentylurea (23) show poor PDK1 kinase inhibitory activity. However, the introduction of a 2-
methyl-2-phenylpropanenitrile at the fourth position (21) showed a modest improvement in activity (IC50 = 5.2 µM) and provided a path
1
7
for further optimization .
Table 1. Inhibition of PDK1 by 1H-benzo[d]imidazol-2-yl)-1H-indazol derivatives.
a
b
Supporting information for assay details, NA = Not Active
The introduction of an amide with a small cycloalkyl ring substituent, such as cyclopropyl (30, IC50 = 1.23 µM), or adding electron-
withdrawing groups, such as fluorine atoms, was well tolerated and significantly improved the PDK1 activity to 280 and 410 nM (27
and 28 respectively). Introduction of an amide substituted heteroaryl group, such as in compound 25, was not tolerated. Extending
morpholine amide moiety with greater flexible spacer and basic benzyl morpholine that is slightly basic moiety resulted in a significant
loss of its activity, as seen with compound 31. We further explored the effects of moving the amide position by making compound 37,
which exhibited only mildly potent activity against PDK1 (IC50 = 385 nM). Similarly, moving the substituent to the sixth position of the
1H-indazole ring completely abolished PDK1 activity (38). In comparison, the insertion of a 2, 2-difluoro cyclopropyl substituent and a
directly substituted morpholin provided potent PDK1 inhibitory activity (32) with a IC50 value of 80 nM.
18
The predicted binding modes of these fragments and actual compounds we synthesized were confirmed by ICM docking and
molecular dynamic simulations within the active site of PDK1 kinase domain when in complex with compound 32 (Figure 1C). These
results revealed that the 1H-indazoles of 1H-benzo[d]imidazol-2-yl)-1H-indazole scaffold and its -N atoms involved in a critical
interaction with the hinge region of the PDK1 through a potential hydrogen bond with the backbone of Ser160. Besides that, another
hydrogen bond forms between the backbone -N atom of Tyr161 and the second -N atom in the indazole scaffold of compound 32.
Whereas the benzo[d] imidazol ring involved in pi-pi stacking interactions with the aryl ring side chain of Tyr161. The fused aryl ring
of the 1H-indazole associated with non-bonded interaction with side chains of the Leu159 residue located at the gate-keeper site, and
had a favorable hydrophobic contact between aryl ring of 1H-indazole, side chain of Leu88 and the backbone atoms of Gly89. The 2, 2-
difluoro-N-phenylcyclopropanecarboxamide group participates in multiple interactions with Glu90, Gly91 and Lys111 residues.

Figure 1C. Compound 32 in complex with PDK1. Docking model of the complex formed between compound 32 and the PDK1 crystal structure model (ID:
Z5M) of the ATP binding site. The critical residues involved in H-bonding interaction shown solid dashed lines. The gatekeeper L159 and DFG sites depicted
1
with CPK and dotted surfaces.
The amide carbonyl group oriented within hydrogen-bonding distance of Lys111. The 2, 2-difluorocyclopropyl rings occupy a
hydrophobic portion of the cavity surrounded by the DFG loop and the side chain of Thr122. To explore the solvent binding site,
th
several compounds with a morpholine ring substitution at the 6 position of 1H-benzo[d]imidazol improved the physicochemical
properties of compounds and improved the cell-based activity as seen with compound 35 against AN3-CA cells. The morpholine group
positioned at the solvent binding site is also involved in interactions with Lys165 (Figure 1C). The introduction of 2-methyl morpholine
did not significantly alter the PDK1 kinase activity, whereas the introduction of a one-carbon spacer (31) abolished the PDK1 activity
(1.22 mM). The addition of this flexible spacer caused a conformational shift of the molecules losing the hydrogen bond interactions
with the hinge region. The in vivo pharmacokinetic profile of compound 32 is in evaluation, but based upon the in vitro profile,
compound 32 and 35 represent attractive starting points for a more extensive medicinal chemistry program to optimize the molecules.
The entire series of compounds synthesized during this optimization effort is shown in Table 1.
To synthesize 1H-benzo[d]imidazol-2-yl)-1H-indazoles series, we followed the synthesis outlined in Scheme 1, starting with the
THP protected substituted aldehyde 7 and treated it with aryl-1,2-diamine in concentrated HCl and NaHSO
3
, which provided the (1H-
benzo[d]imidazol-2-yl)-1-(tetrahydro-2H-pyran-2-yl)-1H-indazole (9). In subsequent steps for the synthesis of compounds 19-21,
variously substituted aryl boronic acid groups were added under cross-coupling Pd(OAc)
2
, PPh
3
conditions in presence of Na
2
CO
3
in
THF followed by subsequent deprotection in presence of TFA and DCM, which afforded compound 10 and 11, respectively.
a
a
Scheme 1 . Synthesis of 1H-benzo[d]imidazol-2-yl)-1H-indazoles series. Reagents and conditions: (a) TsOH (0.1 eq), THF:DCM (1:1), RT, 4-6h; (b) (i)
o
con.HCl, THF, 65 C, 30 min (ii) NaHSO
3
, Reflux, 12h; (c) Pd(OAc)
2
, PPh
3
, Na
2
CO
3
, THF, Reflux, 12h; (d) TFA, DCM, RT, 12h; (e) 10% Pd/C, H2, Ethanol,
RT; (f) Pyridine, DCM, RT; (g) HOBt, EDC, NaHCO3, DMF, RT, 24h; (h) TFA, DCM, RT,12h.
2
18
The R substitutions were instituted using the substituted 1, 2-dinitrobenzene (15) as starting material provided in Scheme 2 .
Additional analogues outlined in Table 1 were prepared from compound 9 by reduction under 10% Pd/C in EtOH in presence of H
2
leading to 12. Compound 12 was subsequently transformed to urea or amide (13) using base or HOBt/EDC coupling conditions
followed by a deprotection step leading to compound 14 (see Supporting Information for details) and variously substituted compounds
22-34 given in Table 1.
b
b
Scheme 2 . Synthesis intermediates. Reagents and conditions: (a) SOCl
2
, Et
3
N, DMF, THF; (b) NaBH
4
, BF
3
.OEt
2
, THF; (c) 10% Pd/C, H
2
, Ethanol, RT.
To sum up, only compounds 32 and 35 can inhibit PDK1 at the low nanomolar concentration, so the following part will focus on
these two compounds.
Kinase Selectivity Profiling: To explore kinase specificity toward PDK1, 32 was screened at a single concentration (500 nM)
against a panel of relevant kinases for selectivity profiling (Figure. 2). In addition to PDK1, 32 showed appreciable activity against
p70S6K, which is downstream of the PI3K/AKT pathway. Compound 32 also inhibited VEGFR1/2 and PDGFRA with reasonable
potency. Therefore, not only the kinases in PI3K/AKT pathway, but also the ones in the MAPK pathway were inhibited by 32. More
efforts were still needed in order to make 32 a pure PDK1 inhibitor. The kinase selectivity profile of 35 has not been determined,
although it is expected to be similar to 32.

Figure 2. Kinase selectivity profile for compound 32.
Cell-based Activity: To assess the activities of 32 in relevant cell-based assays, the compounds were screened in cell viability
assays against three PI3K/AKT pathway mutation cancer cell lines: AN3-CA, RL-95-A, and HEC-1A (Figure 3).AN3-CA cells have
mutations in both PTEN and PIK3R1 (regulatory subunit of PI3K). RL-95-A cells have mutations in PTEN and in H-RAS (can regulate
PI3K). HEC-1A cells are mutated in PIK3CA (catalytic subunit of PI3K). Interestingly, 32 efficiently killed AN3-CA cells at the lower
concentration with an IC50 of 39.0 nM, but the IC50 for RL-95-A and HEC-1A cell lines were at 449.1 nM and 3.4 μM. (Figure 3). The
activity for 35 was similar to 32. The IC50 for 35 treated AN3-CA cells was 0.02 μM, and for 35 treated RL-95-A was 0.7 μM. (Data not
shown)
Figure 3. Inhibition of cell viability in the selected endometrial cancer cell lines (AN3-CA, RL95-2, and HEC-1A) using compound 32.
These preliminary results from a small panel of cell lines suggested that the cancer cell lines with PTEN and PI3K mutations, like
AN3-CA cells, were more sensitive to 32 and 35. It seems like that lost PTEN function and PI3K activation are important for the
sensitivity to 32 and 35. To confirm this importance, a special cell line, HS578T, that has mutated PIK3R1 and wild type PTEN was
used. We knocked-down PTEN using shRNA and see if modified HS578T can be sensitive to 32 and 35. A cell viability assay showed
that loss of PTEN sensitized this cell line to both 32 and 35, lowering the IC50 of 32 from 3.1 to 1.2 mM and that of 35 from 2.2 to 0.80
mM (Figure. 4).
Figure 4. Toxicity of compound 32 and 35 on HS578T transduced with PTEN shRNA or control shRNA. Toxicity represents cell death percentage.
To further explore how PI3K/AKT/mTOR pathway was affected by 32 in AN3-CA cells. Phospho-AKT (Thr308) was checked
using the Meso Scale Discovery technology platform after 24 hours treatment (Figure. 5A). 32 showed good ability of down-regulating
pAKT (about 50%) with lower concentration (100nM). Furthermore, phosphorylation of ribosomal protein s6, the downstream of
PI3K/AKT/mTOR pathway, were also checked using an in-cell western, (Figure. 5B). Both 32 and 35 showed great ability of
decreasing phosphorylation of s6 with EC50 values of 75.6 nM and 76.7 nM, respectively.
Figure 5. A. Phosphorylated AKT (T308) levels in AN3-CA treated with different concentrations of compound 32 for 24h. B. Visual overlay of total ribosomal
protein s6 (red) and phospho-ribosomal protein s6 (green). Image appears yellow where both channels overlap. Quantification of signal intensity and EC50 values
for the two drugs used shown on the right.
Cell-line screen: To find out more sensitive cell lines for 32 and 35, a large cell line panel of over 100 cancer cell lines was screened
with these two compounds. Both 32 and 35 demonstrated remarkable selectivity for cell killing in several cell lines (KATO3, KG-1,
MV4-11, Kasumi-1, MFE296 and AN3-CA) in the low nanomolar range compared to all of the remaining cell lines in which the
compounds showed low micromolar activity (Table 2 in supporting information). The cell lines that turned out to be highly sensitive
had PTEN deletions or silencing as well as activation of PIK3CA through activating mutations in PIK3CA or RAS proteins.

Microarray check: Then we ran the whole genome microarray for the ten different cell lines (six primary sensitive cell lines:
KATO3, KG-1, MV4-11, Kasumi-1, MFE296, AN3-CA; and four insensitive cell lines: J82, Molt4, HEC1A, MFE280) and each cell
line treated with either 32, 35, or DMSO control. Single replicates of each treatment performed. RNA degradation plot (Figure 6a in
supporting information) and the density plot (Figure 6b in supporting information) show no degradation of the samples and the density
plots of all the arrays were quite similar. We tried applying linear models to the data, to find cell-line-specific and drug-specific effects,
but it is impossible to identify any cell line/drug interaction effects. Then we attempted to simplify the patterns of expression displayed
in Figure 6c (Supporting information) by applying k-means clustering across the expression profiles of the 5144 genes showing changes
>
2-fold. This technique groups the expression profiles into k different clusters, and we hope the k is a small number, the variance
within each of the k clusters is small, and the variance between the clusters is large. Even at values of k as large as 40, however, the
variance within the clusters is quite high: 40% of the total. This implies that the expression profile landscape is very complex, due to
many different patterns of expression across the cell lines. However, all hierarchical cluster results showed that the samples clustered
primarily by cell line, that the response to either compound was relatively subtle compared to the differences between cell lines, and
that the response to the two compounds was similar within each cell line, but was different between the cell lines. Therefore, we looked
at the similarity of response to the two treatments within each cell line by generating scatter plots comparing 32/DMSO expression and
3
5/DMSO expression and calculated the Pearson’s correlation between the 32/DMSO and 35/DMSO ratios for each cell line. We
noticed that the correlation was highest among four of the sensitive cell lines (AN3-CA: 0.900, KATO3: 0.886, KG-1: 0.886, and MV4-
1: 0.890), and that each of those cell lines showed a handful of genes with large decreases of expression (> 4-fold down) in both 32
1
and 35 treated cells (Figure 6d in supporting information). Figure 6d (Supporting information) shows that, although each cell line has its
own group of highly down-regulated genes, the 4 cell lines share several genes that are highly down-regulated: DUSP6, FOSL1, and to
a lesser degree DUSP4.
RT-PCR and Western: To further confirm the results from microarray analysis, 7 cell lines (3 sensitive cell lines, AN3-CA,
KATO3, MV4-11 and 4 insensitive cell lines, HCT-116, CAPAN-1, SW480, Molt4) were treated with 35 at three different
concentrations (1μM, 0.1μM and 0.01μM), and then checked the mRNA expression level for those three shared genes (DUSP6, FOSL1
and DUSP4). 35 can down-regulate the transcription of those 3 genes in the three sensitive cell lines, but the regulate directions are
quite different in those four insensitive cell lines (Figure 7 in supporting information). Then, Western blot was done to confirm the
change in the protein level, and the results showed the same changing direction (Figure 8). For the sensitive cell lines, the protein level
for DUSP6, FOSL1 and DUSP4 all decreased with 35 treatment, while this is not true for the insensitive cell lines. PDK1 mRNA
(Figure 7 in supporting information) and protein (Figure 8) expression level had also been checked, but didn’t show any consistent
change in either sensitive or insensitive cell lines.
Figure 8. Western Blot for PDK1, DUSP6, FOSL1, DUSP4, and β-actin in three sensitive cell lines (AN3-CA, KATO-3 and MV4-11) and four insensitive cell
lines (CAPAN-1, HCT-116, SW480 and Molt4) after 35 treatment.
ShRNA: To see if down-regulating DUSP6, FOSL1 or DUSP4 are the key factors that contribute to the 35-induced sensitive cancer
cell lines death; si-RNA (all purchased from Thermo Scientific) experiments had been done to the three sensitive cell lines. ShRNA had
also been used to knockdown PDK1 to imitate the 35’s PDK1 inhibitory effect. After transfection at the optimal condition, ATPLite
assay has been used to reflect the amount of alive cells. The results showed that knocking-down DUSP6, FOSL1 or DUSP4
individually can kill the sensitive cell lines, but not as well as knocking-down PDK1 (Figure 9a in supporting information), while
knocking-down PDK1 and 35 treatment shown the really similar killing effect (Figure 9b in supporting information). For MV4-11 cell
line, just knocking-down DUSP6 shown the same effect as knocking down PDK1. For KATO3 cell line, knocking-down DUSP4 seems
more important for killing the cell. For AN3-CA, the important role seems to switch to FOSL1.
In vivo Efficacy: To demonstrate in vivo activity in a preliminary animal efficacy model, initially the study protocol was submitted
for review and was approved by Huntsman Cancer Institute’s IACUC.We did subcutaneous implantation to put AN3-CA cells into
3
athymic nude mice at the hind flank. Then wait until the tumors reached a size of approximately 200 mm , mice were then stratified into
three groups and intraperitoneally (IP) administered either vehicle, 32, or 35. The compounds were formulated in a solution of 20%
DMSO, 20% cremophor, and 60% water and dosed at 50 mg/kg daily Monday through Friday for two weeks. Tumors and body weights
were measured twice a week and are shown in Figure 10A and Figure 10B, respectively. Compound 32 caused only a minor decrease in
tumor burden compared to the vehicle control that was not significant. However, treatment with 35 led to a significant decrease in
tumor size; at the end of the two-week study, the group treated with 35 had on average tumors about one-third the size of the vehicle
treated group. Furthermore, neither 32 nor 35 caused any body weight loss during the whole study. In fact, the animals continued to
gain weight at a rate that was very similar to the animals in the vehicle group. These results suggest that both compounds, but
particularly 35, has in vivo efficacy at dose levels that are not toxic to the animals.
Phosho-s6 was also checked by western blot in the tumors from the vehicle and 35 groups (Figure 10C). Phospho-s6 level decreased
in the 35 treated group, not vehicle group. Total-s6 level didn’t change. These data suggest that 35 targets the PDK1/ p70S6K pathway
and 35 decreases the p-S6 to inhibit tumor growth.

A
B
Figure 10 A and B. Tumor volumes (A) and body weights (B) of nude mice IP injected with compound 32, 35, or vehicle for two weeks.
Figure 10 C. Total and phosphorylated ribosomal protein s6 levels in tumor samples from mice treated with vehicle or compound 35.
Discussion: Substantial preclinical and even clinical evidence now exists suggesting that targeting the PI3K/AKT pathway will have
therapeutic benefit in some types of cancer. Numerous PI3K inhibitors and AKT inhibitors are at multiple stages of development, along
with several mTOR inhibitors, some of which are now approved anti-cancer agents. Clearly, this pathway represents an exciting area for
drug development and can, hopefully, provide patients and physicians with new options to treat various malignancies. PDK1 is a vital
component to this signaling pathway that, in our view, has not been completely explored as a therapeutic target. PDK1 is required for
full activation of AKT and is frequently constitutively activated due to loss of PTEN and mutations in PI3K. Early knockout animal
studies suggest that PDK1 inhibition may overcome the consequences of these genetic alterations in cancer and this has been
corroborated by more recent target validation studies using small molecules and shRNA approaches.
19, 20, 21
Using a fragment-based, structure-assisted approach
, lead compounds 32 and 35 were discovered as fairly potent PDK1
inhibitors. When profiled against a small panel of kinases, 32 also demonstrated activity against p70S6K, which functions downstream
of AKT and by itself, represents a potential therapeutic target. Even though 32 had some off-target effect (can strongly inhibit
VEGFR1/2 and PDGFRA), it had good ability to inhibit kinases in the PDK1/PI3K/AKT pathways. Considering VEGFR1/2 and
PDGFRA are also really important for tumor growth, 32 may become a good anti-cancer drug instead of a good PDK1 selective
inhibitor.
In cell-based systems, 32 and 35 inhibited the phosphorylation of AKT at the Threonine 308 position, which is a direct
phosphorylation site for PDK1. Furthermore, the compounds inhibited the downstream activation of the S6 ribosomal protein, which
could be an effect of inhibiting PDK1, and also could be the results from the dual inhibition of PDK1 and p70S6K, which functions
directly on S6. In a small panel of endometrial cancer cell lines, compounds 32 and 35 demonstrated preferential activity in AN3-CA
that was both mutated for PTEN (loss-of-function) and PI3K (gain-of-function), whereas they showed less activity in cell lines that only
had a PTEN or PI3K mutation. To test the hypothesis, that 32 and 35 possess preferential activity in cells with both PTEN and PI3K
mutations, the breast cancer cell line, HS578T, which has an activating mutation in PI3K and wild type PTEN, was using that treated
with PTEN shRNA, to artificially harbor both mutations. In this system, 32 and 35 were 3-fold more potent against the cells with PTEN
knockdown compared to the cells that treated with a scrambled shRNA. Although this did not result in a dramatic shift in activity, it
supports the idea that these PDK1 inhibitors may be most suitable for development in patients with tumors that harbor mutations in both
genes. The absence of a larger shift in activity in these experiments may stem from the fact that the shRNA against PTEN does not
completely knockout its expression like a truncating mutation would be in a cancer cell. The knockdown of PTEN as determined by
RT-qPCR was approximately 80% after the shRNA treatment.
In order to find out more sensitive cell lines and further prove the hypothesis that has been mentioned in last paragraph, a large cell
line panel screen was done with 32 and 35 treatment. The cell lines that turned out to be highly sensitive (KATO3, KG-1, MV4-11,
Kasumi-1, MFE296 and AN3-CA) had PTEN deletions or silencing as well as activation of PIK3CA through activating mutations in
PIK3CA or ras proteins. Then the whole genome microarray tried to find out any common change that happened in the transcription
level among those highly sensitive cell lines. After trying different ways to do the analysis, some common genes were found: DUSP6
and FOSL1 were highly down-regulated, as well as DUSP4, but to a lesser degree. Followed by RT-PCR and Western blot, the changes
were confirmed in both the transcription and translation level. With the shRNA experiment, we know that knocking-down these three
genes also can kill those sensitive cell lines (AN3-CA, KATO3, MV4-11), but in the varying degrees. The interesting part is, only for
MV4-11 cell line, the lethality for knocking-down DUSP6 is almost the same as the drug treatment or knocking-down PDK1. DUSP6
(Dual specificity phosphatase 6) is a protein phosphatase, also known as MKP3. It can dephosphorylate their target kinases in order to
inactivate them, so it can negatively regulate members of the MAPK superfamily, which play a role in cellular proliferation and
24, 25, 26
differentiation.
As we known it’s always hard to cure the cancer by only using one drug, so the combination of DUSP6 inhibitor
and PDK1 inhibitor may have a better effect on MV4-11 or biphenotypic B myelomonocytic leukemia.
22, 23
The developability of these PDK1 compounds
as therapeutic agents were further validated by a xenograft model in mice using
the AN3-CA cell line, an endometrial cancer cell line that was extremely sensitive to 32 and 35. In the biochemical and cell-based
assays, these two compounds did not differentiate themselves from each other in any meaningful way. However, when dosed at the
same level and schedule, 35 was much more active in the in vivo xenograft model. Structurally, 32 and 35 differ only be the addition of
a methyl group on compound 35. This improvement in vivo activity is likely due to the improved physicochemical properties of 35 over
compound 32 with regards to membrane permeability and in vivo drug exposure due to metabolism and drug clearance. Therefore, 35,
with its additional methyl group, is likely the more superior of the two compounds to move forward in more advanced animal studies.
Pharmacokinetic and more advanced toxicology studies are underway to further explore the possibilities of developing 35 as an anti-

cancer agent. Additionally, 35 has shown to effectively knockdown ribosomal protein S6 signaling in the tumors suggesting the
compound has on-target activity.
In conclusion, we have identified compound 35 as a viable PDK1 inhibitor. This compound exhibited good drug-like properties and
has demonstrated excellent in vivo activity in preliminary animal studies. Compound 35 is currently in more advanced animal studies to
characterize pharmacokinetics, pharmacodynamics effects, and limitations due to toxicity. Compound 35 may not be the prefect PDK1
inhibitor, but it has higher possibility to become a novel compound that can be moved forward toward clinical development, particularly
in endometrial cancer and others with frequent mutations in both PTEN and PI3K.
Acknowledgments
This work was supported in part by NCI grant P30 CA042014 awarded to the Center for Investigational Therapeutics (CIT) at
Huntsman Cancer Institute, as well as funds from the Experimental Therapeutics Program at Huntsman Cancer Institute, University of
Utah, 2000 Circle of Hope, Salt Lake City, UT 84112 U.S.A.
References and notes
1.
Raimondi, C.; Falasca, M. Curr. Med. Chem. 2011, 18, 2763-2769.
2
3
4
5
.
.
.
.
Peifer, C.; Alessi, D. R. ChemMedChem. 2008, 3, 1810-1838.
Chessari, G.; Woodhead, A. J. Drug Discovery Today. 2009, 14, 668–675.
Congreve, M.; Chessari, G.; Tisi, D.; Woodhead, A. J. J. Med. Chem. 2008, 51, 3661–3680.
Friesner, R. A.; Banks, J. L.; Murphy, R. B.; Halgren, T. A.; Klicic, J. J.; Mainz, D. T.; Repasky, M. P.; Knoll, E. H.; Shaw, D. E.; Shelley, M.; Perry,
J. K.; Francis, P.; Shenkin, P. S. J. Med. Chem 2004, 47, 1739–1749.
6
7
8
9
1
.
.
.
.
Feldman, R. I.; Wu, J. M.; Polokoff, M. A.; Kochanny, M. J.; Dinter, H.; Zhu, D, Biroc SL, 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 (20), 9867-9874.
McBride, C. M.; Renhowe, P. A.; Heise, C.; Jansen, J. M.; Lapointe, G.; Ma, S.; Piñeda, R.; Vora, J.; Wiesmann, M.; Shafer, C. M. Bioorg. Med.
Chem. Lett. 2006, 16, 3595–3599.
Medina, J. R.; Blackledge, C. W.; Heerding, D. A.; Campobasso, N.; Ward, P. B.; Briand, J.; Wright, J.; Axten, J. M. ACS Med. Chem. Lett. 2010, 1,
4
39-442.
Johnson, M.G.; Hu, Q.; Lingardo, L.; Ferre, R. A.; Greasley, S.; Yan, J.; John, K,; Chen, P.; Jacques, E.; Gordon, A. J. Computer-Aided Mol. Design
011, 25, 689-698.
2
0. a) 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, 3819-3825, b) 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, 3814-3818.
1
1
1. Cully, M.; You, H.; Levine, A. J.; Mak, T. W. Nat Rev Cancer. 2006, 6(3), 184-192.
2. Angiolini, M.; Banfi, P.; Casale, E.; Casuscelli, F.; Fiorelli, C.; Saccardo, M. B.; Silvagni, M.; Zuccotto, F. Bioorg. Med. Chem. Lett 2010, 20, 4095–
4
099.
1
3. Howard, S.; Berdini, V.; Boulstridge, J. A.; Carr, M. G.; Cross, D. M.; Curry, J.; Devine, L. A.; Early, T. R.; Fazal, L.; Gill. A. L.; Heathcote,
M.; Maman, S.; Matthews, J. E.; McMenamin, R. L.; Navarro, E. F.; O'Brien. M. A.; O'Reilly, M.; Rees, D. C.; Reule, M.; Tisi, D.; William,
G.; Vinković, M.; Wyatt, P.G. J. Med. Chem 2009, 52, 379-388.
1
1
1
4. Xu, Z.; Nagashima, K.; Sun, D.; Rush, T.; Northrup, A.; Andersen, J. N.; Kariv, I.; Bobkova, E. V. J. Biomol. Screen 2009, 14, 1257-1262.
5. Suite 2011, QikProp, version 3.4, Schrödinger, LLC, New York, NY, 2011.
6. Calculator Plugins Calculator Plugins were used for structure property prediction and calculation, Marvin Sketch (6.1.2), 201n (2013), ChemAxon
(http://www.chemaxon.com).
1
7. Early, T. R.; Fazal, L.; Gill. A. L.; Heathcote, M.; Maman, S.; Matthews, J. E.; McMenamin, R. L.; Navarro, E. F.; O'Brien. M. A.; O'Reilly,
M.; Rees, D. C.; Reule, M.; Tisi, D.; William, G.; Vinković, M.; Wyatt, P.G. J. Med. Chem 2009, 52, 379-388.
8. Abagyan, R.; Totrov, M.; Kuznetsov, D. J. Comput. Chem 1994, 15, 488-506.
9. Poulsen, A.; Blanchard, S.; Soh, C. K.; Lee, C.; Williams, M.; Wang, H.; Dymock, B. Bioorg. Med. Chem. Lett 2012, 22, 305-307.
0. Humphrey, J.; Kablaoui, N.; Kazmirski, S.; Kraus, M.; Kupchinsky, S.; Li. J.; Lingardo, L.; Marx, M. A.; Richter. D.; Tanis, S. P.; Tran, K.; Vernier,
W.; Xie, Z.; Yin, M. J.; Yu, X. H. J. Med. Chem 2011, 54, 8490-8500.
1
1
2
2
1. Medina, J. R.; Becker, C. J.; Blackledge, C. W.; Duquenne, C.; Feng, Y. Grant, S. W.; Heerding, D.; Li, W. H.; Miller, W. H.; Romeril, S.
P.; Scherzer, D.; Shu, A.; Bobko, M. A.; Chadderton, A. R.; Dumble, M.; Gardiner, C. M.; Gilbert, S.; Liu, Q.; Rabindran, S. K.; Sudakin, V.; Xiang,
H.; Brady, P. G.; Campobasso, N.; Ward, P.; Axten, J. M. J. Med. Chem 2011, 54, 1871-195.
2
2
2
2
2
2. Najafov, A.; Sommer, E. M.; Axten, J. M.; Deyoung, M. P.; Alessi, D. R.; Dario, R. Biochem J 2011, 433, 357-369.
3. Bearss, D. J.; Vankayalapati, H.; Sorna, V.; Warner, S, L.; Sharma, S. PCT Int. WO 2012135799 A1. Appl 2012.
4. Patterson K.I.; Brummer T.; O'Brien P.M.; Daly R.J.. Biochem. J. 2009, 418, 475–489.
5. Caunt C.J.; Keyse S.M.. FEBS J. 2013, 280, 489–504
6. Bermudez O.; Pagès G.; Gimond C.. Am. J. Physiology - Cell Physiology. 2010, 299, C189–C202
Supplementary Material
Supplementary data (experimental information details and supporting tables and figures) associated with this article can be found, in
the online version.
±
Current Address:
th
TGen, 445 N. 5 Street, Phoenix, AZ 85004, USA.
Email: hvankayalapati@tgen.org

Table 1. Inhibition of PDK1 by 1H-benzo[d]imidazol-2-yl)-1H-indazol derivatives.
a
a
Compound
Structure
IC50 µM
Compound
Structure
IC50 µM
1
2
2
2
9
1
3
5
15.3
20
30.3
b
5.2
53.9
8.8
22
24
26
NA
0.9
12.4
2
2
3
7
9
1
0.3
28
30
32
0.4
1.2
b
NA
1.2
0.2
0.08
3
3
3
3
5
7
34
36
38
0.2
0.8
0.09
0.4
b
NA

Graphical Abstract
Fragment-Based Design, Synthesis,
Biological Evaluation, and SAR of 1H-
benzo[d]imidazol-2-yl)-1H-indazol
Derivatives as Potent PDK1 Inhibitors
Ting Chen, Venkataswamy Sorna, Susie Choi, Lee Call, Jared Bearss, Kent Carpenter, Steven L. Warner, Sunil
Sharma, David J. Bearss, Hariprasad Vankayalapati