Discovery of selective 4-amino-pyridopyrimidine inhibitors of MAP4K4 using fragment-based lead identification and optimization

Article abstract of DOI:10.1021/jm500155b

Mitogen-activated protein kinase kinase kinase kinase 4 (MAP4K4) is a serine/threonine kinase implicated in the regulation of many biological processes. A fragment-based lead discovery approach was used to generate potent and selective MAP4K4 inhibitors. The fragment hit pursued in this article had excellent ligand efficiency (LE), an important attribute for subsequent successful optimization into drug-like lead compounds. The optimization efforts eventually led us to focus on the pyridopyrimidine series, from which 6-(2-fluoropyridin-4-yl)pyrido[3,2-d]pyrimidin-4-amine (29) was identified. This compound had low nanomolar potency, excellent kinase selectivity, and good in vivo exposure, and demonstrated in vivo pharmacodynamic effects in a human tumor xenograft model.

Full text of DOI:10.1021/jm500155b

Article
pubs.acs.org/jmc
Discovery of Selective 4‑Amino-pyridopyrimidine Inhibitors of
MAP4K4 Using Fragment-Based Lead Identification and Optimization
Terry D. Crawford,* Chudi O. Ndubaku, Huifen Chen, Jason W. Boggs, Brandon J. Bravo,
Kelly DeLaTorre, Anthony M. Giannetti, Stephen E. Gould, Seth F. Harris, Steven R. Magnuson,
Erin McNamara, Lesley J. Murray, Jim Nonomiya, Amy Sambrone, Stephen Schmidt, Tanya Smyczek,
Mark Stanley, Philip Vitorino, Lan Wang, Kristina West, Ping Wu, and Weilan Ye
Genentech Inc., 1 DNA Way, South San Francisco, California 94080, United States
S
* Supporting Information
ABSTRACT: Mitogen-activated protein kinase kinase kinase kinase 4 (MAP4K4) is a
serine/threonine kinase implicated in the regulation of many biological processes. A
fragment-based lead discovery approach was used to generate potent and selective
MAP4K4 inhibitors. The fragment hit pursued in this article had excellent ligand
efficiency (LE), an important attribute for subsequent successful optimization into
drug-like lead compounds. The optimization efforts eventually led us to focus on the
pyridopyrimidine series, from which 6-(2-fluoropyridin-4-yl)pyrido[3,2-d]pyrimidin-4-
amine (29) was identified. This compound had low nanomolar potency, excellent
kinase selectivity, and good in vivo exposure, and demonstrated in vivo
pharmacodynamic effects in a human tumor xenograft model.
INTRODUCTION
■
Mitogen-activated protein kinase kinase kinase kinase 4
(MAP4K4, also known as HGK), a member of the sterile 20
(STE20) family of kinases,1 is involved in a variety of signaling
pathways. In particular, the expression and function of Map4k4
are linked to many physiological processes as well as pathological
conditions, including metabolism,2−7 inflammation,8 neural
degeneration,9 and cancer,10−15 suggesting that inhibiting
Figure 1. (A) Compound 1. The red area of the molecule represents
possible points of electrostatic contact with MAP4K4. (B) Predicted
binding mode of compound 1 in the ATP binding site of MAP4K4
kinase domain crystal structure. Green dashed lines indicate potential
intermolecular hydrogen bonds.
MAP4K4 function may be beneficial in treating diseases
associated with these processes. In order to further investigate
MAP4K4 as a potential therapeutic target, identification of a
potent and selective inhibitor will be necessary. However, there
has been limited precedence for a selective small molecule
MAP4K4 inhibitor in the literature.16 In the absence of suitable
literature tool compounds, we conducted a screen of a diverse
fragment library using surface plasmon resonance (SPR) with the
aim of identifying a suitable hit for further development.17
oxazole heterocycle while directing the aromatic group into a
hydrophobic pocket adjacent to the methionine gatekeeper
(Met105) and catalytic lysine (Lys54) (Figure 1B).22
Our strategy for prosecuting this fragment was to pursue both
close-in analogues of the oxazole as well as larger groups that
would extend the fragment deeper into the hydrophobic pocket.
In order to ensure that the optimization efforts did not deviate
too far from ideal drug-like space, initial fragment follow-up
compounds around 1 were restricted to a molecular weight less
than 350 and cLogP less than 3.5.23,24 We hypothesized that
replacing the oxazole with readily available hinge binding motifs
with the ability to form two conventional hydrogen bonds could
provide more efficient interactions and thereby enhance potency.
As confirmation of this hypothesis, simple biaryl compounds
RESULTS AND DISCUSSION
■
The fragment screen provided an array of hits with reasonable
binding affinity (Kd ≈ 10−3400 μM) and ligand efficiencies (225
hits with LE > 0.35).18 Several fragment hits were actively
investigated, and this article will focus on the progression of the
oxazole fragment (1, Figure 1A).19
On the basis of the chemical structure of 1 and comparisons
with other known kinase inhibitors, we were confident that it was
ATP-competitive and bound to the hinge region of the kinase
utilizing the motif colored in red.20 The docking model of 1 in the
ATP-binding site of MAP4K4 (Figure 1B) was generated using
the Glide program21 indicating that it formed a conventional as
well as an atypical hydrogen bond to the hinge through the
Received: January 28, 2014
Published: March 27, 2014
© 2014 American Chemical Society
3484
dx.doi.org/10.1021/jm500155b | J. Med. Chem. 2014, 57, 3484−3493

Journal of Medicinal Chemistry
Article
bearing complementary hydrogen bond donor/acceptor pairs
(Table 1, 2−7) showed modest benefit in potency of up to 8-fold
pocket within the protein as suggested by the docking model.
Para substitutions also significantly decreased potency, indicating
that the limit of the hydrophobic pocket had been reached
(compounds 12, 15, 16, and 21). Halogens were well tolerated at
the meta position, with the m-Cl (24) essentially equipotent and
the m-F (19) being 2-fold more potent relative to the
unsubstituted phenyl 9.
Table 1. MAP4K4 SPR Binding Affinity of Oxazole
Replacements with Hinge Binding Motifs
An X-ray cocrystal structure of compound 22 with the
recombinant MAP4K4 kinase domain was solved (Figure 2). It
revealed a “glove-like” fit for the ligand contained in the ATP-
binding pocket and offered structure-based explanations for the
SAR reported in Table 2. Importantly, the structure was
consistent with the high ligand efficiency observed for the
compound. The aminoquinazoline core forms two conventional
hydrogen bonds and a weaker C−H type hydrogen bond with
the kinase hinge residues Glu106 and Cys108. The substituted
phenyl reaches into a hydrophobic pocket that is blocked off by a
salt bridge between Lys54 and Asp171, which prevented further
expansion at the para position while allowing small substitutions
at either meta positions. Of particular note was the fact that the P-
loop tyrosine side chain (Tyr36) swept down to form an edge to
face interaction with the quinazoline core. Contemporaneous
with this observation, a group at Pfizer published work suggesting
that this folded P-loop conformation in MAP4K4 and a few other
kinases may provide a potential opportunity for kinase
selectivity.16 Since the halogen was involved in a favorable van
der Waals interaction with the gatekeeper Met105, we also
probed disubstituted aromatic substitutions off the 5 position,
with 3-fluoro-5-aminophenyl 18 being equipotent to the 3-
fluorophenyl.
While this fragment progression showed a remarkable increase
in potency through just two design iterations, the molecular
properties were beginning to trend outside of the ideal drug-like
space, particularly with respect to cLogP. This was of concern
both with regard to instability in the in vitro liver microsome
experiments, as well as the ability to achieve a desired PK profile
in vivo. Evaluation of the ligand-lipophilicity efficiency (LLE)26
of compounds in Table 2 validated this concern, as potency
improvements from quinazoline 8 (LLE = 4.25) to 18, 19, and 22
did not improve their LLEs (4.48, 3.94, and 3.03, respectively).
Further fragment optimization therefore focused on reducing
lipophilicity through optimization of the core and also placing
polar substituents on the periphery. To increase the polarity of
the core, a nitrogen walk around the quinazoline core was
undertaken (Table 3, 24−26).
Incorporation of nitrogen at the 7- or 8-positions (Table 3, 24
and 25) resulted in 100-fold and 35-fold losses in potency,
respectively. This data highlights the importance of the P-loop
Tyr36 edge to face interaction with the quinazoline core, which
was weakened by the repulsive interactions between the lone pair
of aromatic nitrogen and π-electrons of the Tyr36 side chain.
Conversely, placement of a nitrogen at the 5-position
(compound 26) resulted in a nearly 4-fold increase in potency,
while maintaining an excellent LE (0.59) and also significantly
improving the LLE relative to 19 (5.33 from 3.94). Possible
reasons for the potency relative to lipophilicity improvement
(ΔLLE = 1.40) could be that (1) the intramolecular hydrogen
bond between the 4-amino and 5-aza would make the NH2 group
in 26 more planar, providing better conjugation of the amino
group and also a better vector to the hinge, which will result in a
stronger hydrogen bond between 4-amino and Glu106 backbone
carbonyl and that (2) the addition of 5-aza in the core allows the
a
See Experimental Procedures for LE and LLE calculations.
improvement over the oxazole (1), while generally retaining high
ligand efficiencies. Gratifyingly, extending the molecule with the
quinazoline (8) showed a significant (55-fold) increase in
potency, achieving a single digit micromolar dissociation
constant as measured by SPR, while also maintaining ligand
efficiency and lipophilic ligand efficiency levels comparable to
those of 1.
Evaluation of compound 8 in a biochemical assay using a
synthetic peptide substrate (Z′-LYTE)25 demonstrated an IC50
of 0.189 μM (Table 2). With submicromolar biochemical
potency achieved with our prototype compound 8, we advanced
with biochemical activity as the primary assay from that point
forward.
The structure−activity relationship (SAR) of the 6-aryl-1-
aminoquinazoline scaffold was then explored further. Unsub-
stituted phenyl compound 9 (Table 2) was roughly 2-fold more
potent than 8, suggesting some negative effects from inserting a
meta-NH2. Ortho-substitution on the phenyl group (e.g., 10)
was not well tolerated presumably due to the large unfavorable
torsional strain that results from binding into a nearly planar
3485
dx.doi.org/10.1021/jm500155b | J. Med. Chem. 2014, 57, 3484−3493

Journal of Medicinal Chemistry
Article
Table 2. MAP4K4 Affinities for 6-Aryl Derivatives of Quinazoline 8
a
IC50 data are an average of at least 3 independent experiments. Each value has a standard deviation of less than one geometric mean (biochemical
b
data with standard deviations can be found in the Supporting Information). Compounds were incubated with mouse liver microsomes for 1 h, and
the percent remaining was determined by LC-MS/MS analysis.
phenyl to adopt a smaller torsion angle (global minimum at 20
degrees) that is required for binding.
Pyridopyrimidine 26 had a cLogP value nearly a log unit lower
than quinazoline 19 (cLogP = 2.6 vs 3.4), which returned the
3486
dx.doi.org/10.1021/jm500155b | J. Med. Chem. 2014, 57, 3484−3493

Journal of Medicinal Chemistry
Article
with the Lys54 side chain (Figure 3). Compound 29 also showed
a significantly improved stability profile compared to that of 26−
28, which established it as a viable lead molecule with the
potential to have good in vivo stability.
Acetamide 30 showed a significant decrease in potency relative
to that of parent aniline 18. This potency drop was expected
based on the narrowness of the channel and the polar nature of
the solvent exposed region as shown in Figure 2. Incorporation of
a basic center into the amide tail with 2-pyrrolidinyl acetamide 31
resulted in potency equivalent to that of 18. The crystal structure
of 31 in complex with MAP4K4 showed that the basic center was
critical for regaining potency lost from acetamide (30) due to its
ability to disrupt the Lys54−Asp171 salt bridge (Figure 4). Lys54
was pushed out, while Asp171 formed a charge−charge
interaction with the basic amine from the bound ligand. The
carbonyl and amide nitrogen formed water-mediated hydrogen
bonds to the protein. Incorporation of nitrogen at the 5 position
of the quinazoline core (see Table 3 for numbering system)
yielded amide 32 with 10 nM potency and moderate stability in
human microsomes, although still poor in mouse microsomes.
Attempts to reverse the amide (e.g., 34) also resulted in a large
drop in potency, while removing the carbonyl altogether to
provide amine 33 resulted in a 15-fold potency decrease.
Kinase Selectivity. On the basis of the folded P-loop
conformation that MAP4K4 adopts with the quinazoline and
pyridopyrimidine-type ligands, we expected that they would
exhibit significant selectivity over other kinases. An earlier lead,
quinazoline 22 (Activity assay IC50 = 77 nM), was screened in a
kinase selectivity panel available from Invitrogen, and out of 61
kinases, only 2 showed >50% inhibition at 1 μM (Abl, Inh = 66%;
Mink1, Inh = 84%) (Figure 5; the full selectivity table can be
found in the Supporting Information). Notably, Abl is one of a
few kinases capable of adopting the unusual folded P-loop
conformation;16 therefore, activity of 22 against Abl was
anticipated. Significant inhibition of kinase Mink1 was also
expected, given that the ATP-binding pocket of Mink1 had
∼100% sequence identity with that of MAP4K4 and the fact that
they are closely related family members of MAP4K4. Compound
26 provided an 8-fold improvement in MAP4K4 potency while
maintaining good kinase selectivity. Abl affinity was mitigated
(11% at 1 μM), but modest PI4Kb activity was now observed
(67% at 1 μM). Compound 32 showed the best selectivity profile
over the 67 kinases screened, with only the inhibition of Mink1
showing activity >35% at 1 μM. Compound 29 showed modest
TSSK1 (57% at 1 μM) and MARK1 (54% at 1 μM) affinity but
was otherwise also highly selective considering that the 1 μM test
concentration was ∼60-fold higher than the MAP4K4 IC50 (17
nM).
Figure 2. Crystal structure of 22 (cyan) in complex with MAP4K4.
(PDB ID: 4OBO). (A) Binding details viewed from the opening of the
ATP binding site with the hinge on the left and the N-terminal domain
on top. The P-loop is folded down to enclose the compound with the
side chain of Tyr36 (yellow sticks). Hydrogen bond interactions to the
hinge are shown in green dashed lines, and distances are shown in
Angstroms. (B) Binding details viewed from the top of the N-terminal
domain. Solvent accessible surface of the binding site shows a snug fit of
22 in a hydrophobic channel formed by the kinase hinge, N- and C-
terminal domains, and the folded-in P-loop.
Table 3. MAP4K4 Affinities for Pyridopyrimidine Positional
Isomers Relating to 19 and 22
a
IC50 data are an average of at least 3 independent experiments. Each
value has a standard deviation of less than one geometric mean
(biochemical data with standard deviations can be found in the
Supporting Information).
Pharmacokinetics/Pharmacodynamics. Because of their
favorable in vitro profiles, the in vivo clearance of lead
compounds 29 and 32 was assessed in female CD-1 mice
(Table 5). Compound 32 showed a high clearance rate that was
under-predicted by in vitro liver microsomes (Table 4). The
rapid clearance was potentially due to anilide cleavage by
amidases or esterases. However, there was reasonable correlation
between in vitro and in vivo clearance rates for 29.
Oral exposure of 29 at a 5 mg/kg dose was a modest F = 8.5%
(Table 5B). However, when a dose escalation study was
conducted the AUClast was nearly dose-proportional up to 150
mg/kg, and the free fraction exposure levels at this high dose
suggested that 29 would have sufficient pathway engagement in
vivo to elicit biomarker response. We have recently uncovered
that inhibition of MAP4K4 by ATP-competitive small molecular
series back into the desired molecular property space and
permitted further growth of the molecule to increase potency.
Close-in analogues of 26 were then synthesized (Table 4, 27−
34) to explore the size of the hydrophobic pocket at the meta-
position (27) as well as the potential favorable interactions with
Asp171 (28) and the catalytic lysine (Lys54) (29).
The activities for 27−29 were essentially equipotent to 26,
with the additional H-bond interactions gained by 28 and 29
possibly offset by desolvation penalties. Notably, 29 demon-
strated another significant improvement in LLE (6.34),
indicating that we were driving this series into an excellent
property space.
Analysis of the cocrystal structure of 29 in MAP4K4 indicated
that the pyridine nitrogen does indeed form a hydrogen bond
3487
dx.doi.org/10.1021/jm500155b | J. Med. Chem. 2014, 57, 3484−3493

Journal of Medicinal Chemistry
Article
Table 4. MAP4K4 Affinities for Pyridopyrimidine Analogues Relating to 19 and 26
a
IC50 data are an average of at least 3 independent experiments. Each value has a standard deviation of less than one geometric mean (biochemical
b
data with standard deviations can be found in the Supporting Information). Compounds were incubated with mouse liver microsomes for 1 h, and
the percent remaining was determined by LC-MS/MS analysis.
Figure 4. Crystal structure of compound 31 bound to the ATP binding
site of MAP4K4 (PDB ID: 4OBQ). Intermolecular hydrogen bonds are
shown in green dashed lines, and distances are shown in Angstroms.
Figure 3. Crystal structure of compound 29 bound to the ATP binding
site of MAP4K4 (PDB ID: 4OBP). Intermolecular hydrogen bonds are
shown in green dashed lines, and distances are shown in Angstroms.
inhibitor causes an elevation in phosphorylated (T181)
MAP4K4 levels in vitro.27 This finding indicated that ATP-
3488
dx.doi.org/10.1021/jm500155b | J. Med. Chem. 2014, 57, 3484−3493

Journal of Medicinal Chemistry
Article
Figure 5. Heat map of Invitrogen kinase selectivity for 22, 26, 32, and 29 at 1 μM. Bands are colored according to a progressive color gradient, with green
indicating 0% inhibition and red as 100% inhibition (white indicates no data available for that kinase). The complete table of kinase inhibition data can be
found in the Supporting Information.
Table 5. Summary of PK Parameters Following IV Bolus (A) and PO (B) Administration in Female CD-1 Mice
(A)
analyte
dose (mg/kg)
AUC0‑∞ (h·μM)
CL (mL/min/kg)
T1/2 (h)
Vss (L/kg)
32
29
1
1
0.25 0.01
5.99 0.62
174.0 3.8
11.6 1.3
0.34 0.23
0.56 0.05
3.9 1.7
0.6 0.03
(B)
analyte
dose (mg/kg)
AUClast (h·μM)
Cmax (μM)
Tmax (h)
F (%)
32
29
29
29
29
5
0.35 0.04
2.55 0.21
30.3 9.12
41.2 26.2
107.0 22.5
0.15 0.04
1.33 0.28
6.34 0.84
7.65 4.62
14.5 3.1
0.33 0.14
1.00 0.00
0.83 0.29
2.67 2.89
6.0 3.0
5.7 0.5
5
8.5 0.7
50
100
150
23.6 7.1
16.0 10.2
27.8 5.8
competitive MAP4K4 inhibitor-induced elevation of MAP4K4-
pT181 constitutes a pharmacodynamic (PD) response. There-
fore, treatment of mice with a potent and selective MAP4K4
inhibitor such as 29 would be expected to elicit similar PD
response in vivo provided there is sufficient exposure.
Compound 29 was evaluated for its ability to induce a PD
response in HT-1080 human fibrosarcoma xenograft tumors.
After a single oral dose of compound 29, plasma and tumor
samples were isolated at 4 and 24 h time points. PD response was
assessed through Western blotting of MAP4K4-pT181 normal-
ized to total MAP4K4. Compound 29 caused a dose-dependent
increase in pMAP4K4 (T181) at both time points above the
baseline of the vehicle-treated mice (Figure 6).
Chemistry. Fragment analogues (2−8) were synthesized by
Suzuki coupling of commercially available 3-aminophenylbor-
onic acid with appropriate aryl bromides. Fragments derived
from 6-aryl quinazoline (9−23) were prepared in a similar
fashion with commercial 4-amino-6-bromoquinazoline and
phenylboronic acids (Scheme 1).
Pyridopyrimidine analogues 24−27 were synthesized as
described in Scheme 2. Cyclization of 3-amino-6-chloropicoli-
namide (37) with triethylorthoformate yielded pyridopyrimidi-
none 38.28 Chlorination with refluxing thionyl chloride afforded
chloropyrimidine 39, followed by amination with ammonia to
give amine 40. Subsequent Suzuki coupling reactions then
provided analogue 26. The regioisomeric pyridopyrimidinone
intermediate 42 was derived by cyclization of amino acid 41 with
formamide.29 Analogue 27 was synthesized via the same route as
26.
CONCLUSIONS
■
Herein we described the discovery of a potent, ligand efficient,
and highly selective series for inhibition of MAP4K4. By taking a
different approach than most conventional FBDD efforts, we
incorporated a broad scope of structural diversity relative to the
initial hit to quickly obtain compounds that possessed nanomolar
biochemical activity. Subsequently, we focused on fine-tuning the
scaffold to optimize the lipophilic ligand efficiency of the
compounds. The series benefitted by the folded P-loop
conformation, which helped to impart high ligand efficiency
and superb kinase selectivity. These efforts culminated in the
identification of compound 29, a low molecular weight MAP4K4
inhibitor that had favorable in vivo PK properties. Compound 29
was evaluated in a PD experiment where it demonstrated
pathway functional response. The discovery of this compound
paved the way for further lead optimization efforts, which will be
the subject of future communications.
EXPERIMENTAL PROCEDURES
■
General Methods. All solvents and reagents were used as obtained.
Reactions involving air or moisture sensitive reagents were carried out
under nitrogen atmosphere. Microwave reactions were performed using
CEM Discover and Biotage Initiator reactors. NMR spectra were
recorded in a deuterated solvent with Varian Avance 300-MHz or
Bruker Avance 400- or 500-MHz NMR spectrometers and referenced to
3489
dx.doi.org/10.1021/jm500155b | J. Med. Chem. 2014, 57, 3484−3493

Journal of Medicinal Chemistry
Article
Figure 6. (A) Pharmacodynamic evaluation of compound 29 in HT-1080 human fibrosarcoma xenografts. (B) Quantitation of the p-MAP4K4 levels
normalized by total MAP4K4 over a 24 h period. Error bars correspond to the standard error of the means.
a
phase indicated and operating at a 35 mL/min flow rate. Chemical
Scheme 1. Suzuki Couplings to Afford Compounds 2−23
purities were >95% for all final compounds as assessed by LCMS
analysis. The following analytical method was used to determine
chemical purity of final compounds: HPLC-Agilent 1200, water with
0.05% TFA, acetonitrile with 0.05% TFA, Agilent SB-C18, 1.8 μM, 2.1 ×
30 mm, 25 °C, 3−95% B in 8.5 min, 95% in 2.5 min, 400 μL/min, 220
and 254 nm, equipped with Agilent quadrupole 6140, ESI positive, and
90−1300 amu.
MAP4K4 SPR-Based Fragment Screen. MAP4K4 (residues 3−
328) containing an N-temrinal His-tag and a C-terminal Avi-tag
(Avidity, LLC) was biotinylated during baculovirus expression and
purified to homogeneity by nickel-affinity and size-exclusion chroma-
tography. Mass analysis indicated that there was no phosphorylation of
the purified material. For SPR experiments, the biotinylated protein was
captured to CM5 chips (GE Healthcare) manually coupled with ∼8,000
RUs of neutravidin (Pierce) inside a Biacore T100 (GE Healtchare) as
a
previously described.30 After priming into the experimental running
Reagents and conditions: (i) bis(di-tert-butyl(4-
dimethylaminophenyl)phosphine)dichloropalladium(II), potassium
acetate, acetonitrile/water, 80 °C; (ii) bis(di-tert-butyl(4-
dimethylaminophenyl)phosphine)dichloropalladium(II), potassium
phosphate tribasic, dioxane/water, 100 °C.
buffer (50 mM HEPES, pH 7.2, 150 mM NaCl, 2 mM MgCl2, 1 mM
TCEP, 0.01% Tween-20, 1% PEG-3350, and 5% DMSO), 4,000−8,000
RUs of M4K4 was generally captured. Good long-term binding stability
of the surface was observed during pilot fragment screening that
included the use of a control compound with a KD of 1 μM and a
dissociation half-life of ∼3 s. A screen of the Genentech edition one
fragment library (∼2,500 compounds) was executed using a single-point
screen at 100 μM compound. Raw sensorgram data were reduced,
solvent corrected, double reference inspected for poor sensorgrams and
aggregating compounds,31 and fit using Scrubber II (BioLogic Software,
Campbell, Australia; http://www.biologic.com.au). Primary data were
additionally corrected for drift, daily solvent mismatches, and scaled to
fractional binding occupancy using BioBook (IDBS) as previously
discussed.30 Approximately 300 hits were further characterized in a
dose−response format by SPR at 200 μM in a 2-fold dilution series to
yield KD and LE information. A total of 225 hits were characterized and
confirmed with a KD range of 10 to 2000 μM and LEs ranging from 0.24
to 0.59 with a median of 0.33.
trimethylsilane (TMS). Chemical shifts are expressed as δ units using
TMS as the external standard (in NMR descriptions, s = singlet, d =
doublet, t = triplet, q = quartet, m = multiplet, and br = broad peak). All
coupling constants (J) are reported in Hertz. Mass spectra were
measured with a Finnigan SSQ710C spectrometer using an ESI source
coupled to a Waters 600MS HPLC system operating in reverse mode
with an X-bridge Phenyl column of dimensions 150 mm × 2.6 mm, with
5 μm-sized particles. Preparatory-scale silica gel chromatography was
performed using medium-pressure liquid chromatography (MPLC) on
a CombiFlash Companion (Teledyne ISCO) with RediSep normal
phase silica gel (35−60 μm) columns and UV detection at 254 nm.
Reverse-phase high performance liquid chromatography (HPLC) was
used to purify compounds where indicated by elution on a Waters 3100
mass-directed prep instrument using a Phenomenex Gemini-NX C18
column (20.2 × 50 mm, 5 μm) as stationary phase using the mobile
MAP4K4 Inhibition Biochemical Assay Protocol. The kinase
activity of purified human MAP4K4 kinase domain on a peptide
3490
dx.doi.org/10.1021/jm500155b | J. Med. Chem. 2014, 57, 3484−3493

Journal of Medicinal Chemistry
Article
a
Scheme 2. Preparation of Pyridopyrimidine
a
Reagents and conditions: (i) (EtO)3CH, reflux, 3 h; (ii) thionyl chloride, cat. DMF, 85 °C, 3 h; (iii) 7 N NH3/MeOH, 50 °C, 2 h; (iv) bis(di-tert-
butyl(4-dimethylaminophenyl)phosphine) dichloropalladium(II), potassium phosphate tribasic, dioxane/water, 100 °C; (v) formamide, 130 °C, 12
h.
substrate (5-FAM-LGRDKYKTLRQIRQ-COOH) was monitored
using Z′-LYTE technology according to the manufacturer’s protocol.
Ligand Efficiency (LE) and Ligand-Lipophilicity Efficiency
(LLE) calculations. SPR LE and LLE calculated as LE = −RT ln(Kd)/
Nheavy atoms. LLE = pKd − cLogP. Biochemical assay LE = −RT ln(IC50)/
Nheavy atoms. LLE = pIC50 − cLogP.
6-Chloropyrido[3,2-d]pyrimidin-4(3H)-one (38). To a round-
bottomed flask was added 3-amino-6-chloro-pyridine-2-carboxamide
32, (9.0 g, 52.5 mmol) and 100 mL of triethoxymethane. The reaction
was stirred at reflux for 3 h. The reaction was then cooled to room
temperature and the white solid collected by filtration. The precipitate
was then purified by flash column chromatography (3−10% MeOH/
DCM) yielding 4.6 g of product as a white solid (48%).1H NMR (400
MHz, DMSO) δ 12.69 (s, 1H), 8.19 (s, 1H), 8.14 (d, J = 8.6 Hz, 1H),
7.88 (d, J = 8.6 Hz, 1H).
4,6-Dichloropyrido[3,2-d]pyrimidine (39). To a round-bottomed
flask was added 6-chloro-3H-pyrido[3,2-d]pyrimidin-4-one (4.6 g, 25.33
mmol), followed by thionyl chloride (80 mL) and DMF (0.05 mL). The
reaction was stirred at 85 °C for 3 h, at which point the reaction was a
clear solution. The reaction was then concentrated via rotovap and
azeotroped 3× with chloroform. The crude product was dried overnight
on high vacuum and then carried forward to the amination step without
purification.
6-Chloropyrido[3,2-d]pyrimidin-4-amine (40). To 4,6-
dichloropyrido[3,2-d]pyrimidine (5.0 g, 25 mmol) was added 30 mL
of 7 M ammonia in methanol (210 mmol), and then it was stirred at 50
°C for 2 h. The reaction was then concentrated via rotovap and
triturated with dichloromethane. The filter cake was then washed with
water 3× to yield a clean product; 2.83 g as a yellow solid (62% over two
steps). 1H NMR (400 MHz, DMSO) δ 8.44 (s, 1H), 8.13 (d, J = 8.8 Hz,
1H), 8.02 (bs, J = 14.8 Hz, 1H), 7.89 (bs, 1H), 7.86 (d, J = 8.8 Hz, 1H).
6-(3-Fluorophenyl)pyrido[3,2-d]pyrimidin-4-amine (26). To an 8
mL screw-cap vial was added 6-chloropyrido[3,2-d]pyrimidin-4-amine
(75 mg, 0.42 mmol), (3-fluorophenyl)boronic acid (87 mg 0.62 mmol),
bis(di-tert-butyl(4-dimethylaminophenyl)phosphine)-
dichloropalladium(II) (15 mg, 0.021 mmol), potassium phosphate
tribasic (0.63 mL, 2.00 mol/L in water 1.25 mmol), and 1,4-dioxane
(0.75 mL). The mixture was capped and shaken at 100 °C for 2 h. The
reaction was cooled to room temperature, diluted with 4 mL of DCM,
and washed with 2 mL of water. Combined organics were concentrated
under reduced pressure. The residue was purified by prep HPLC (5% to
85% acetonitrile with 0.1% ammomium hydroxide in 8 min gradients)
yielded of 6-(3-fluorophenyl)pyrido[3,2-d]pyrimidin-4-amine (51 mg,
51%). 1H NMR (400 MHz, DMSO) δ 8.48 (d, J = 8.9 Hz, 1H), 8.44−
8.38 (m, 2H), 8.30−8.18 (m, 2H), 8.15 (d, J = 8.8 Hz, 1H), 7.99 (bs,
1H), 7.62−7.53 (m, 1H), 7.36−7.29 (m, 1H). LCMS m/z (M + H) 241.
6-(2-Fluoropyridin-4-yl)pyrido[3,2-d]pyrimidin-4-amine (29). To
6-chloropyrido[3,2-d]pyrimidin-4-amine (4.0 g, 22.2 mmol) in 100 mL
of acetonitrile was added 2-fluoro-4-(4,4,5,5-tetramethyl-1,3,2-dioxa-
borolan-2-yl)pyridine (5.93 g, 26.6 mmol), [1,1′-bis-
(diphenylphosphino)ferrocene]dichloropalladium(II), complex with
dichloromethane (1.62 g, 2.2 mmol), and 66 mL of 1 M potassium
carbonate in water (66 mmol). The reaction was heated to 80 °C for 1 h.
The crude was filtered through a thin layer of Celite, and the solid was
then treated with 250 mL of ethanol and heated to reflux. The solid was
filtered, then triturated with ethyl acetate and methanol to afford 6-(2-
fluoropyridin-4-yl)pyrido[3,2-d]pyrimidin-4-amine (5.3 g, 99%). 1H
NMR (400 MHz, DMSO) δ 8.61(d, J = 8.8 Hz, 1H), 8.46 (s, 2H), 8.42
(d, J = 5.3 Hz, 1H), 8.37 (d, J = 5.3 Hz, 1H), 8.34 (s, 1H), 8.22 (d, J = 8.8
Hz, 1H), 8.12 (s, 1H). LCMS m/z (M + H) 242.
Modeling Methods. The molecular modeling software available
through the Maestro interface (Schrodinger, Inc.)21 were used for ligand
preparation and docking. Starting ligand conformations for all
compounds were prepared by LigPrep at pH 7.0 1 using default
parameters. The prepared ligand structures were then docked into the
ATP-binding site of a MAP4K4 crystal structure (2.35 Å) in complex
with an HTS hit (not published) using Glide SP. The docking poses
were evaluated based on a combination of criteria including the Glide
docking score, the formation of hydrogen bond(s) to the hinge residues
Glu106 and Cys108, other favorable intermolecular interactions, low
strain energy of the bound ligand (Estrain <2 kcal/mol), etc. All of the
docking model and X-ray structure figures were produced using the
software PyMOL (Schrodinger, Inc.).21
ASSOCIATED CONTENT
■
S
* Supporting Information
PK study protocols and analysis; PD in vitro and in vivo analysis;
single point inhibition Invitrogen kinase selectivity data for 22,
26, 32, and 29; protein purification and crystallization;
preparation and characterization of 2−25, 27−28, and 30−34;
and biochemical assay data with standard deviations. This
material is available free of charge via the Internet at http://pubs.
acs.org.
Accession Codes
PDB accession codes: 4OBO, 4OBP, and 4OBQ.
AUTHOR INFORMATION
■
Corresponding Author
*E-mail: terrydc@gene.com. Phone: 650-467-8895.
3491
dx.doi.org/10.1021/jm500155b | J. Med. Chem. 2014, 57, 3484−3493

Journal of Medicinal Chemistry
Article
broadly expressed in human tumor cells and can modulate cellular
transformation, invasion, and adhesion. Mol. Cell. Biol. 2003, 23, 2068−
2082.
Notes
The authors declare no competing financial interest.
(11) Collins, C. S.; Hong, J.; Sapinoso, L.; Zhou, Y.; Liu, Z.; Micklash,
K.; Schultz, P. G.; Hampton, G. M. A small interfering RNA screen for
modulators of tumor cell motility identifies MAP4K4 as a promigratory
kinase. Proc. Natl. Acad. Sci. U.S.A. 2006, 103, 3775−3780.
(12) Liu, A. W.; Cai, J.; Zhao, X. L.; Jiang, T. H.; He, T. F.; Fu, H. Q.;
Zhu, M. H.; Zhang, S. H. ShRNA-targeted MAP4K4 inhibits
hepatocellular carcinoma growth. Clin. Cancer Res. 2011, 17, 710−720.
(13) Liang, J. J.; Wang, H.; Rashid, A.; Tan, T.-H.; Hwang, R. F.;
Hamilton, S. R.; Abbruzzese, J. L.; Evans, D. B.; Wang, H. Expression of
MAP4K4 is associated with worse prognosis in patients with stage II
pancreatic ductal adenocarcinoma. Clin. Cancer Res. 2008, 14, 7043−
7049.
ACKNOWLEDGMENTS
■
We thank Mengling Wong, Chris Hamman, Michael Hayes, and
Shi Choong for compound purification. Baiwei Lin, Deven
Wang, and Yutao Jiang are acknowledged for analytical support.
Grady Howes’ and Gigi Yuen’s help with compound manage-
ment and logistics is also recognized, as is the help from Jinhua
Chen and Charles Ding (Wuxi Apptec) with technical support.
Finally, Chris Lewis, Bing-Yan Zhu, Jeff Blaney, and Sarah
Hymowitz are thanked for project support.
ABBREVIATIONS USED
■
(14) Qiu, M.-H.; Qian, Y.-M.; Zhao, X.-L.; Wang, S.-M.; Feng, X.-J.;
Chen, X.-F.; Zhang, S.-H. Expression and prognostic significance of
MAP4K4 in lung adenocarcinoma. Pathol. Res. Pract. 2012, 208, 541−
548.
cLogP, calculated LogP; DCM, dichloromethane; DMF,
dimethylformamide; DMSO-d6, deuterated DMSO; EGTA,
ethylene glycol-bis(2-aminoethyl ether)-N,N,N′,N′-tetraacetic
acid; (EtO)3CH, triethylorthoformate; FA, formic acid;
HATU, 2-(7-aza-1H-benzotriazole-1-yl)-1,1,3,3-tetramethyluro-
nium hexafluorophosphate; HEPES, 2-[4-(2-hydroxyethyl)-
piperazin-1-yl]ethanesulfonic acid; LogD7.2, log of partition
coefficient between octanol and pH7.2 aqueous buffer; MCT,
methyl cellulose/tween; MeOH, methanol; PMSF, phenyl-
methylsulfonyl fluoride; SPR, surface plasmon resonance;
TBST, tris buffered saline with Tween 20; TCEP, 3,3′,3″-
phosphanetriyltripropanoic acid; TEA, triethylamine
(15) Mao, J.; Ligon, K. L.; Rakhlin, E. Y.; Thayer, S. P.; Bronson, R. T.;
Rowitch, D.; McMahon, A. P. A novel somatic mouse model to survey
tumorigenic potential applied to the hedgehog pathway. Cancer Res.
2006, 66, 10171−10178.
(16) Guimaraes, C. R. W.; Rai, B. K.; Munchhof, M. J.; Liu, S.; Wang, J.;
̃
Bhattacharya, S. K.; Buckbinder, L. Understanding the impact of the P-
loop conformation on kinase selectivity. J. Chem. Inf. Model. 2011, 51,
1199−1204.
(17) Erlanson, D. A. Introduction to Fragment-Based Drug Discovery.
In Fragment-Based Drug Discovery and X-Ray Crystallography; Topics in
Current Chemistry; Springer: Berlin, Germany, 2011; Vol. 317, pp 1−
32.
REFERENCES
■
(1) Dan, I.; Watanabe, N. M.; Kusumi, A. The Ste20 group kinases as
regulators of MAP kinase cascades. Trends Cell Biol. 2001, 11, 220−230.
(2) Tang, X. An RNA interference-based screen identifies MAP4K4/
NIK as a negative regulator of PPAR, adipogenesis, and insulin-
responsive hexose transport. Proc. Natl. Acad. Sci. U.S.A. 2006, 103,
2087−2092.
(3) Baumgartner, M.; Sillman, A. L.; Blackwood, E. M.; Srivastava, J.;
Madson, N.; Schilling, J. W.; Wright, J. H.; Barber, D. L. The Nck-
interacting kinase phosphorylates ERM proteins for formation of
lamellipodium by growth factors. Proc. Natl. Acad. Sci. U.S.A. 2006, 103,
13391−13396.
(18) Abad-Zapatero, C. Ligand efficiency indices for effective drug
discovery. Expert Opin. Drug Discovery 2007, 2, 469−488.
(19) A full description of other fragment expansion campaigns will be
described in short order. Wang, L., et al., manuscript in Preparation.
(20) Traxler, P.; Furet, P. Strategies toward the design of novel and
selective protein tyrosine kinase inhibitors. Pharmacol. Ther. 1999, 82,
195−206.
(21) Maestro/Glide/PyMOL; Schrodinger, Inc.: New York.
̈
(22) At this stage in the hit triage, X-ray crystal structures of fragments
in complex with MAP4K4 were of too low a resolution to be
interpretable.
(4) Bouzakri, K.; Zierath, J. R. MAP4K4 gene silencing in human
skeletal muscle prevents tumor necrosis factor-alpha-induced insulin
resistance. J. Biol. Chem. 2007, 282, 7783−7789.
(23) Mauser, H.; Stahl, M. Chemical fragment spaces for de novo
design. J. Chem. Inf. Model. 2007, 47, 318−324.
(24) Lipinski, C. A.; Lombardo, F.; Dominy, B. W. Experimental and
computational approaches to estimate solubility and permeability in
drug discovery and development settings. Adv. Drug Delivery Rev. 1997,
23, 3−25.
(5) Bouzakri, K.; Ribaux, P.; Halban, P. A. Silencing mitogen-activated
protein 4 kinase 4 (MAP4K4) protects beta cells from tumor necrosis
factor-alpha-induced decrease of IRS-2 and inhibition of glucose-
stimulated insulin secretion. J. Biol. Chem. 2009, 284, 27892−27898.
(6) Tesz, G. J.; Guilherme, A.; Guntur, K. V. P.; Hubbard, A. C.; Tang,
X.; Chawla, A.; Czech, M. P. Tumor necrosis factor alpha (TNFalpha)
stimulates Map4k4 expression through TNFalpha receptor 1 signaling
to c-Jun and activating transcription factor 2. J. Biol. Chem. 2007, 282,
19302−19312.
(25) Rodems, S. M.; Hamman, B. D.; Lin, C.; Zhao, J.; Shah, S.;
Heidary, D.; Makings, L.; Stack, J. H.; Pollok, B. A. A FRET-based assay
platform for ultra-high density drug screening of protein kinases and
phosphatases. Assay Drug Dev. Technol. 2002, 1, 9−19.
(26) Leeson, P. D.; Springthorpe, B. The influence of drug-like
concepts on decision-making in medicinal chemistry. Nat. Rev. Drug.
Discovery 2007, 6, 881−890.
(27) Please refer to Supporting Information for methodology.
(28) Rewcastle, G. W.; Palmer, B. D.; Thompson, A. M.; Bridges, A. J.;
Cody, D. R.; Zhou, H.; Fry, D. W.; McMichael, A.; Denny, W. A.
Tyrosine kinase inhibitors. 10. Isomeric 4-[(3-Bromophenyl)amino]-
pyrido[ d]- pyrimidines are potent ATP binding site inhibitors of the
tyrosine kinase function of the epidermal growth factor receptor. J. Med.
Chem. 1996, 39, 1823−1835.
(29) Fry, D.; Kraker, A.; Meyer, R. F.; Rewcastle, G. W. Bicyclic
Compounds Capable of Inhibiting Tyrosine Kinases of the Epidermal
Growth Factor Receptor Family. US Patent US5654307A1, 1997.
(30) Giannetti, A. M. From experimental design to validated hits a
comprehensive walk-through of fragment lead identification using
surface plasmon resonance. Methods Enzymol. 2011, 493, 169−218.
̈
(7) Zhao, X.; Mohan, R.; Ozcan, S.; Tang, X. MicroRNA-30d induces
insulin transcription factor MafA and insulin production by targeting
mitogen-activated protein 4 kinase 4 (MAP4K4) in pancreatic β-cells. J.
Biol. Chem. 2012, 287, 31155−31164.
(8) Aouadi, M.; Tesz, G. J.; Nicoloro, S. M.; Wang, M.; Chouinard, M.;
Soto, E.; Ostroff, G. R.; Czech, M. P. Orally delivered siRNA targeting
macrophage Map4k4 suppresses systemic inflammation. Nature 2009,
458, 1180−1184.
(9) Loh, S. H. Y.; Francescut, L.; Lingor, P.; Bahr, M.; Nicotera, P.
̈
Identification of new kinase clusters required for neurite outgrowth and
retraction by a loss-of-function RNA interference screen. Cell Death
Differ. 2007, 15, 283−298.
(10) Wright, J. H.; Wang, X.; Manning, G.; LaMere, B. J.; Le, P.; Zhu,
S.; Khatry, D.; Flanagan, P. M.; Buckley, S. D.; Whyte, D. B.; Howlett, A.
R.; Bischoff, J. R.; Lipson, K. E.; Jallal, B. The STE20 kinase HGK is
3492
dx.doi.org/10.1021/jm500155b | J. Med. Chem. 2014, 57, 3484−3493

Journal of Medicinal Chemistry
Article
(31) Giannetti, A. M.; Koch, B. D.; Browner, M. F. Surface plasmon
resonance based assay for the detection and characterization of
promiscuous inhibitors. J. Med. Chem. 2008, 51, 574−580.
3493
dx.doi.org/10.1021/jm500155b | J. Med. Chem. 2014, 57, 3484−3493