Discovery of selective LRRK2 inhibitors guided by computational analysis and molecular modeling

Article abstract of DOI:10.1021/jm300452p

Mutations in the genetic sequence of leucine-rich repeat kinase 2 (LRRK2) have been linked to increased LRRK2 activity and risk for the development of Parkinson's disease (PD). Potent and selective small molecules capable of inhibiting the kinase activity of LRRK2 will be important tools for establishing a link between the kinase activity of LRRK2 and PD. In the absence of LRRK2 kinase domain crystal structures, a LRRK2 homology model was developed that provided robust guidance in the hit-to-lead optimization of small molecule LRRK2 inhibitors. Through a combination of molecular modeling, sequence analysis, and matched molecular pair (MMP) activity cliff analysis, a potent and selective lead inhibitor was discovered. The selectivity of this compound could be understood using the LRRK2 homology model, and application of this learning to a series of 2,4-diaminopyrimidine inhibitors in a scaffold hopping exercise led to the identification of highly potent and selective LRRK2 inhibitors that were also brain penetrable.

Full text of DOI:10.1021/jm300452p

Article
pubs.acs.org/jmc
Discovery of Selective LRRK2 Inhibitors Guided by Computational
Analysis and Molecular Modeling
Huifen Chen,*,† Bryan K. Chan,† Jason Drummond,‡ Anthony A. Estrada,† Janet Gunzner-Toste,†
Xingrong Liu,§ Yichin Liu,‡ John Moffat,‡ Daniel Shore,† Zachary K. Sweeney,†,⊥ Thuy Tran,†
Shumei Wang,† Guiling Zhao,† Haitao Zhu,∥ and Daniel J. Burdick*,†
‡
§
†Discovery Chemistry Department, Biochemical and Cellular Pharmacology Department, Drug Metabolism and Pharmacokinetics
∥
Department, and Neuroscience Department, Genentech, Inc., 1 DNA Way, South San Francisco, California 94080, United States
S
* Supporting Information
ABSTRACT: Mutations in the genetic sequence of leucine-
rich repeat kinase 2 (LRRK2) have been linked to increased
LRRK2 activity and risk for the development of Parkinson’s
disease (PD). Potent and selective small molecules capable of
inhibiting the kinase activity of LRRK2 will be important tools
for establishing a link between the kinase activity of LRRK2
and PD. In the absence of LRRK2 kinase domain crystal
structures, a LRRK2 homology model was developed that
provided robust guidance in the hit-to-lead optimization of small molecule LRRK2 inhibitors. Through a combination of
molecular modeling, sequence analysis, and matched molecular pair (MMP) activity cliff analysis, a potent and selective lead
inhibitor was discovered. The selectivity of this compound could be understood using the LRRK2 homology model, and
application of this learning to a series of 2,4-diaminopyrimidine inhibitors in a scaffold hopping exercise led to the identification
of highly potent and selective LRRK2 inhibitors that were also brain penetrable.
A few selective inhibitors of LRRK2 kinase activity such as
LRRK2-IN-118 and CZC-2514619 have been described recently.
However, they do not appear to have sufficient CNS exposure
to be used in mammalian models of PD. A potent ALK/LRRK2
kinase inhibitor, TAE684, was recently reported to achieve
significant brain exposure in mouse but did not inhibit LRRK2
phosphorylation in the brain.20 The development of selective
and brain penetrable LRRK2 inhibitors therefore remains a
critical need for the LRRK2 field.21 In this contribution, we
describe our initial efforts to use structure-based design and
computational approaches to identify useful LRRK2 chemical
probes starting from a high-throughput screening effort.
INTRODUCTION
■
Parkinson’s disease (PD) is a multisystem neurodegenerative
disorder that is clinically characterized primarily by tremors,
rigidity, and bradykinesia.1 The current standard of care for PD
patients is limited to symptomatic treatment, which only
provides temporary attenuation of motor symptoms and does
not affect the progression of neurodegeneration. There is,
therefore, a strong demand for disease modifying or neuro-
protective therapies. One of the more attractive targets for
disease modification was identified in 2004 when genetic
variations in the LRRK2 gene were linked to familial PD.2,3 In
particular, the specific G2019S mutation of LRRK2 has been
associated with both familial and idiopathic PD.4−6 The LRRK2
gene encodes a large protein with multiple domains, including a
kinase domain.7−9 The detailed physiological function and
effectors of the LRRK2 kinase are largely unknown and remain
to be determined.10 Importantly, the G2019S mutation in the
kinase domain is a dominant mutation that has been shown to
increase LRRK2 kinase activity in vitro, suggesting that the
kinase activity of LRRK2 is involved in Parkinson’s disease
pathophysiogenesis.11−14 Indeed, recent studies with non-
specific LRRK2 small molecule inhibitors have suggested that
inhibition of LRRK2 activity might ameliorate neurodegener-
ative phenotypes in C. elegans and Drosophila Parkinson’s
disease models and mouse models of LRRK2.15,16 However,
because of the lack of general kinase selectivity of compounds
used in the efficacy studies, the biological effects of LRRK2
kinase inhibition remain to be elucidated.17
RESULTS
■
A high-throughput screening campaign using G2019S LRRK2
protein22 yielded a number of interesting small molecule
inhibitor scaffolds, including triazolopyridines and diaminopyr-
imidines represented by compounds 1 and 2 (Figure 1). The
triazolopyridine compounds were highly potent and had
physical properties consistent with CNS penetration,23 while
the aminopyrimidine inhibitors had excellent ligand efficiency
(LE).24,25 It is well-known, however, that the diaminopyr-
imidine motif is particularly well-represented in the kinase
inhibition literature26 and the potential selectivity of these
compounds was a concern. We therefore focused our initial
computer-aided design efforts on the development of models
Received: March 31, 2012
© XXXX American Chemical Society
A
dx.doi.org/10.1021/jm300452p | J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
Article
MLK1, TIE2, and TAK1 protein structures as templates using
MOE (Chemical Computing Group, Inc.).36 The models built
with these templates were useful for elucidating general features
of the ATP binding site, such as hinge residues, gatekeeper, and
other residues that could participate in ATP and/or inhibitor
binding. However, none of them proved robust enough to
provide reasonable docking poses for all of the top priority
chemical classes found in the screening campaign.
The second approach for template selection was to search for
kinases that exhibited similar inhibition profiles as LRRK2 for a
given set of inhibitors. Analysis of the internal biochemical
activity profiles of LRRK2 inhibitors found in the screening
campaign revealed a surprising correlation between LRRK2 and
JAK3 and a weaker correlation with JAK2. These data
suggested that JAK3 crystal structures could be good structural
templates for building LRRK2 homology models. However,
because of the paucity of available high quality JAK3 crystal
structures, JAK2 cocrystal structures were used. The final
LRRK2 model (Figure 2A) was constructed from an
unpublished in-house JAK2 cocrystal structure similar to a
published structure (PDB code 3JY9).37 The JAK2 structure
was chosen based on its very high resolution (1.7 Å) and open
ATP-binding site, which would allow the docking of many
diverse compounds without introducing steric clashes. Similar
to other protein kinases, the LRRK2 KD consists of a smaller
N-terminal lobe and a larger C-terminal lobe connected by the
hinge. According to the model, the N-terminal domain consists
mostly of β sheets and an α helix, while the C-terminal domain
consists of mostly α helices. The ATP-binding site is situated at
the interface of the two lobes. The gatekeeper, Met1947, and
catalytic residues, Lys1906 and Glu1920, line the back wall of
the ATP-binding site, while residues Glu1948, Leu1949, and
Ala1950 form the side wall along the kinase hinge (Figure 2B).
The side chains of amino acids Phe1883, Leu1885, Val1893,
and Ala1904 form the ceiling of the ATP site, and those of
Ile1933, Gly1953, Ser1954, and Leu2001 form the floor. At the
opening of ATP-binding site, the side chains of Arg1957 and
His1998 partially close in on the sugar moiety of ATP. The
binding pocket is relatively closed and flat along the hinge and
becomes more open near the solvent exposed region as well as
the sugar and phosphate binding region. The ATP-binding
pocket is largely hydrophobic with the exception of the hinge
Figure 1. Chemical structures of high-throughput screening hits 1 and
2.
and approaches that would allow us to improve the selectivity
of compounds such as 1 and 2.
Homology Modeling and ATP-Binding Site Analysis.
Use of crystal structure information is critical to the success of a
structure-based drug design (SBDD) effort. However, the
structure of the kinase domain of LRRK2 has not yet been
reported.27 Therefore, construction of a LRRK2 homology
model7,10,28−30 using known kinase structures as templates was
initiated. The most crucial elements for building a homology
model are the choice of structural templates and sequence
alignment.31−33 The choice of template can be influenced by
numerous factors including sequence identity and functional
relevance to the target protein, structural quality, and purpose
of the model. Since our goal was to use the homology model
for SBDD, we had to be even more prudent with the selection
of structural templates in order to generate a model that would
allow the docking of most LRRK2 inhibitors.
Two approaches were employed to identify appropriate
structural templates. The first approach relied on using kinase
domain (KD) crystal structures of proteins that have a high
degree of sequence homology with the KD of LRRK2.34 Using
Psi-Blast,35 a panel of human protein kinases with the highest
KD sequence identity to LRRK2 was identified (Table 1),
although none of the kinases have sequence identity to LRRK2
greater than 32%. Several homology models were built with
a
Table 1. Multiple Sequence Alignment of ATP-Binding Site Residues of LRRK2 and Selected Kinases
a
b
Residue numbers are based on Swiss-Prot nomenclature.34 Cells colored in red have identical amino acids as in LRRK2. Gatekeeper residue.
B
dx.doi.org/10.1021/jm300452p | J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
Article
literature29,38 and those found in the high-throughput screen,
were initiated. Reasonable docking models could be generated
for most of the LRRK2 inhibitors either by using the automatic
docking program Glide (Schrodinger, Inc.)39 or by manually
̈
docking inhibitors in the LRRK2 homology model. The side
chain conformations of Arg1957 and Gln1961 prevented the
docking of some inhibitors, so a second LRRK2 model was
generated by moving these mobile side chains out of the
inhibitor binding site (Figure 2B). These two models were used
in lieu of a crystal structure to guide our design efforts.
Identification of molecular basis that determines kinase
selectivity40−45 is an important aspect of kinase drug discovery
research because of the highly conserved kinase domain
structures and ATP binding sites across the human kinome.46
LRRK2 KD has low sequence identity to other protein kinases
and even its closest related family member, LRRK1,28,47 has a
KD sequence identity of only 32% (Table 1). The ATP-binding
site of LRRK2, however, is more conserved and has much
higher sequence identity to other protein kinases. In an effort to
identify protein residues that could impart kinase selectivity, a
detailed binding site sequence analysis was performed (Table
1). Out of the 19 binding site residues whose side chains could
potentially interact with ATP-competitive inhibitors, six are
conserved in over 400 kinases, five are conserved in 200−300
kinases, and eight are conserved in less than 150 kinases. To
identify residues with the highest likelihood to confer selectivity
against other kinases, we analyzed the relative locations,
accessibility, and properties of the eight least conserved
residues and narrowed the list down to four residues:
Phe1883, Leu1949, Ser1954, and Arg1957.
Phe1883 is removed from the ATP-binding site and at the
interface of the solvent exposed region. This residue is
something other than phenylalanine in over 350 kinases and
could perhaps be contacted by inhibitors with larger groups.
Leu1949 presents the most attractive selectivity handle because
of its proximity to the hinge binding moiety and its shorter side
chain relative to the dominant Phe or Tyr side chains in
approximately 290 other kinases.48,49 The smaller Leu side
chain creates a small cavity in LRRK2 that is not present in
most other kinases and could be exploited to improve kinase
selectivity. Similarly, there are over 230 kinases with a side
chain larger than Ser1954, and of those, 200 have either a Glu
or Asp side chain. Arg1957 provides another residue to target
for selectivity with approximately 200 kinases containing a
negatively charged Asp or Glu side chain. We hypothesized that
Figure 2. (A) LRRK2 kinase domain homology model with the N-
terminal domain on top and C-terminal domain at bottom. The
solvent accessible surface of the ATP-binding site is colored in pink,
while the cartoon defining the protein backbone is colored in green.
(B) LRRK2 ATP-binding site in complex with ATP (colored in
yellow) looking down through the N-terminal domain. Side chains of
binding site residues are shown as sticks and colored in green. Side
chains of the four key selectivity residues (Phe1883, Leu1949,
Ser1954, and Arg1957) are colored in magenta. The alternative
conformations of Arg1957 and Gln1961 are shown in cyan. Hydrogen
bond interactions between ATP and LRRK2 are shown as yellow
dashed lines.
backbone hydrogen bond donor/acceptors, back pocket
Lys1906 and Glu1920, and the front pocket Arg1957 side
chain.
To investigate the validity of the LRRK2 model for structure-
based design purposes, extensive docking studies on a diverse
set of LRRK2 inhibitors, including those reported in the
Table 2. Matched-Pair Activity Cliffs Identified from the Triazolopyridine JAK2 Activity Cliff Analysis
C
dx.doi.org/10.1021/jm300452p | J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
Article
Figure 3. (A) Docking model of compound 1 (green) in LRRK2 binding site. Key hydrogen bonds are shown in yellow dashed lines. Solvent
accessible surface is shown in pink. (B) Docking model of compound 1 in JAK2. (C) Overlay of compound 1 (green) and 8 (magenta) docking
models in LRRK2 (colored in white with L1949 in green). The side chain of Y931 of JAK2 is also shown and colored in cyan.
targeting one or more of these four residues was a good
strategy50 to achieve general kinase selectivity and potency of
the small molecule inhibitors.
and Figure 3B. Several key interactions could account for the
excellent LRRK2 and JAK2 potency of 1. The 2-aminotriazole
core binds to the adenine site and forms a pair of strong
hydrogen bonds to the hinge backbone in addition to
numerous hydrophobic interactions with residues above and
below the core. The 3,5-dimethyl-4-phenol moiety binds in the
mostly hydrophilic back pocket, with the hydroxyl accepting
one hydrogen bond from the conserved catalytic Lys1906 while
donating one to the conserved Glu1920 in helix C of LRRK2
(residues Lys882 and Glu898 of JAK2). The phenol ring also
makes hydrophobic interactions with gatekeeper Met1947 in
LRRK2 and Met929 in JAK2. The unsubstituted aniline ring
binds to a hydrophobic channel that leads to the solvent
exposed region of the ATP site in both proteins. On the basis of
this binding mode, replacement of the 4-hydroxyl of the phenol
would be difficult, since it acts as both a donor and acceptor to
two highly conserved catalytic residues. All three matched-pair
compounds 4, 6, and 7 involved replacement of this hydroxyl
which resulted in significant loss in JAK2 potency. As this
subpocket of LRRK2 is nearly identical to that of JAK2, a
similar loss in LRRK2 potency would be expected to occur.
The loss in activity in the last matched pair (D), does not
involve a hydroxyl replacement but rather an addition of a
methyl to the aniline portion of the molecule. As shown in
Figure 3B, the unsubstituted aniline of compound 1 fits tightly
against the JAK2 protein surface formed by the hinge and side
chain of Tyr931. An ortho-methyl substitution on the phenyl
ring could clash with Tyr931 and cause a significant drop in
JAK2 potency. This observation is consistent with the 135-fold
drop in JAK2 activity seen in compound 8. An inspection of the
LRRK2 protein surface (Figure 3A) indicates that the methyl of
compound 8 might fit well in the pocket created by the smaller
side chain of Leu1949 in LRRK2, a key selectivity residue that
was suggested by the ATP binding site analysis. Indeed,
Identification of Selectivity Hotspot through
Matched-Pair Analysis. The triazolopyridine hit 1 was an
attractive starting point for further optimization owing to its
good LRRK2 potency (Ki = 20 nM) and high ligand efficiency
(LE = 0.42); however, it was found to inhibit JAK2 with a Ki of
3 nM. To evaluate whether JAK2 selectivity could be achieved
in this series, a substructure search of our internal compound
library using the unsubstituted version of 1 (Table 2, R1−R5 =
H) was performed. One-hundred compounds containing the
substructure were tested in a JAK2 inhibition assay, and the
data were analyzed using MMP51−56 to uncover structural
features associated with significantly reduced JAK2 potency.
For the matched-pair analysis, a Tanimoto similarity
coefficient of 0.94 was used as a cutoff to identify closely
related pairs. The cutoff JAK2 Ki for being considered inactive
was 250 nM. Out of the 100 tested compounds, 320 compound
pairs were identified that met the chemical space similarity
criteria. For a pair to be considered to have an activity cliff, the
biological activity ratio had to be at least 2 orders of magnitude.
Only four pairs met the activity ratio criteria as shown in Table
2. Three of the four pairs (A, B, and C) involved a replacement
of the hydroxyl (R4) with a hydrogen, methyl, or methyl
sulfone, all of which led to a significant loss of JAK2 activity.
The last pair (D, compounds 1 and 8) was not substituted in
the same manner as the other three pairs. An additional methyl
group on the aniline (R2) portion of compound 8 greatly
reduced the activity of the inhibitor against JAK2 in comparison
to compound 1.
Compound 1 was docked to the ATP-binding site of the
LRRK2 homology model. Models of this compound in the
ATP-binding site of LRRK2 and JAK2 are shown in Figure 3A
D
dx.doi.org/10.1021/jm300452p | J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
Article
compound 8 was 2-fold more potent (Ki ≈ 10 nM) than
compound 1 in the LRRK2 binding assay, indicating that the
pocket formed by Leu1949 can provide a selectivity handle.
In an effort to confirm that JAK2 selectivity is a good
indicator of overall kinase selectivity in this series, compound 8
was tested against an Invitrogen panel of 63 kinases
(Supporting Information Table 1). Consistent with the
hypothesis, compound 8 was very selective and only inhibited
one kinase (Abl) in addition to wild-type and mutant G2019S
LRRK2 at greater than 50% inhibition at 1 μM. These data
strongly support the proposed binding mode and the
hypothesis that general kinase selectivity can be achieved in
this series by introducing small groups in the vicinity of the
Leu1949 side chain. However, optimization of the triazolopyr-
idine scaffold was not pursued because of difficulties in
maintaining potency and good physicochemical properties
while replacing the 3,5-dimethyl-4-hydroxyphenyl moiety.
Optimization of the Diaminopyrimidines. The diami-
nopyrimidine series identified in the HTS campaign (2, Figure
1) had excellent biochemical potency and ligand efficiency. The
physicochemical properties also represented reasonable starting
points for a series designed to cross the blood−brain barrier.
However, for this scaffold to advance as a lead series, significant
improvement of the selectivity profile was needed.
1 and does not access the Leu1949 selectivity pocket; (3) the
selectivity of this class of compounds could be significantly
improved by ortho-substitution on the aniline ring. Compound
2 was tested in the JAK2 assay and found to be a potent
inhibitor of JAK2 kinase activity (Table 4). The truncated
compound 9 was also tested in both the LRRK2 and JAK2
binding assays (Table 3). Loss of the morpholino amide
Table 3. LRRK2 Activity and JAK2 Selectivity of Ortho-
Substituted Front Pocket Phenyl Diaminopyrimidines
Ki (μM)
compd
R
LRRK2
JAK2
JAK2/LRRK2 index
9
H
0.011
0.020
0.006
0.033
0.007
0.064
>3.2
6
>160
>533
>97
10
11
12
13
CH3
OCH3
Br
>3.2
>3.2
OCHF2
>3.2
>457
resulted in a 2-fold loss in potency against LRRK2 and a 9-fold
loss in potency against JAK2. These data suggested that the
molecular weight and TPSA of this series could be significantly
reduced and that this series could lead to the discovery of
compounds with properties consistent with CNS penetration.
Using knowledge of structure−selectivity relationships
gained from the triazolopyridines, a methyl group was installed
in the ortho position of the aniline moiety in the truncated
series (compound 10, Table 3). Although there was a slight loss
in potency against LRRK2, significant selectivity against JAK2
was achieved. It appeared from the homology model that the
pocket near Leu1949 was small and that large groups, such as a
phenyl, would not fit. Small substituents were tolerated to
varying degrees (Table 3). The methyl (10) and Br (12)
substitutions caused a 2- to 3-fold drop in LRRK2 potency,
while the methoxy (11) and difluoromethoxy (13) groups led
to slight improvement in LRRK2 potency. More importantly,
all of the ortho-substituted compounds, 10−13, achieved
greater than 97-fold selectivity over JAK2 in comparison to the
unsubstituted compound 9. These same trends were observed
when the same ortho-substitutions were incorporated into the
original HTS hit (14−16, Table 4). First, the morpholinoamide
analogues had the same potency rank order seen with the
truncated analogues. The slight loss in activity seen with
compound 9 was regained with the addition of the
morpholinoamide. Second, the compounds were very selective
against JAK2, with the methoxy substituent achieving over
1000-fold selectivity against JAK2. This result supports the idea
that the methoxy group fits favorably in the pocket near
Leu1949.
The suggested binding mode of compound 2 in the LRRK2
homology model is shown in Figure 4A. The 2,4-diaminopyr-
Figure 4. (A) Docking model of compound 2 (green) in LRRK2
binding site. (B) Overlay of docking models of compound 11
(magenta) and 15 (green) in LRRK2 binding site. Intermolecular
hydrogen bonds are shown as dashed yellow lines, while the solvent
accessible surface is shown in pink. Leu1949 is shown in light green
sticks.
imidine core binds to the adenine site and interacts with the
kinase hinge through a pair of hydrogen bonds to the backbone
amide NH and carbonyl oxygen of Ala1950. According to our
model, the C5 chlorine forms favorable van der Waals
interactions with the Met1947 gatekeeper side chain, and the
C4 N-methyl fills a hydrophobic cavity. The carbonyl oxygen of
the amide group forms a weak hydrogen bond with the
guanidinium side chain of Arg1957, with nonideal hydrogen
bond geometry. The aniline ring binds in a flat hydrophobic
cleft along the hinge near the opening of the ATP-binding site
with the 4-morpholinoamide group pointing toward the side
chain of Phe1883.
On the basis of the docking model, we hypothesized the
following: (1) the front pocket morpholino amide could
possibly be removed without a significant impact on LRRK2
potency; (2) compound 2 might be a nonselective kinase
inhibitor, since it binds to the same hinge region as compound
To further confirm that filling the Leu1949 pocket
maintained selectivity against JAK2, several substitutions were
made at the C5 position of the pyrimidine core (Table 4). As
predicted by the binding mode of compound 2, R1 can
modulate the LRRK2 biochemical potency (17−20) through
van der Waals interactions, conformational preference, and
electronic effects, without having any obvious impact on JAK2
activity. Replacement with the smaller and more electronegative
fluorine (18) reduced LRRK2 activity by 24-fold and
substitution with a nitrile (20) by 4-fold. Interestingly,
E
dx.doi.org/10.1021/jm300452p | J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
Article
total brain to plasma AUC ratios and the unbound brain to
unbound plasma AUC ratios were 1.4 and 0.61, respectively, in
wild-type and 2.9 and 1.3, respectively, in P-gp/Bcrp knockout
mice. The half-life (T1/2) was 0.23 h. At 1 h following 30 mg/kg
intraperitoneal injection in wild-type mouse, the total and
unbound plasma concentrations of compound 15 were 8.4 and
1.1 μM, respectively, and the total and unbound brain
concentrations were 6.2 and 0.37 μM, respectively. While
these data suggest that P-gp and/or Bcrp slightly restrict the
brain penetration in mice for 15, it clearly demonstrates that
this compound achieves desirable brain penetration for use as a
tool compound in animal models of PD.
Chemistry. The target compounds were synthesized as
depicted in Scheme 1. Commercially available 2,4-dichloropyr-
imidines were reacted with 8 M methylamine in ethanol at
room temperature to produce intermediates 26−30. 2,5-
Dichloro-N-methylpyrimidin-4-amine 27 and appropriately
substituted anilines were suspended in n-butanol and subjected
to microwave irradiation to yield products 9−13.
Table 4. LRRK2 Activity and JAK2 Selectivity of Ortho-
Substituted Front Pocket Phenyl Morpholinoamide
Diaminopyrimidines
Ki (μM)
compd
R1
Cl
R2
LRRK2
JAK2
JAK2/LRRK2 index
2
H
0.006
0.007
0.003
0.013
0.001
0.071
0.001
0.013
0.007
>3.2
>3.2
>3.2
3.0
1
>457
>1067
>246
>3000
>45
14
15
16
17
18
19
20
Cl
CH3
Cl
OCH3
Br
Cl
Br
OCH3
OCH3
OCH3
OCH3
F
>3.2
>3.2
>3.2
CF3
CN
>3200
>246
The appropriately substituted 4-aminobenzoic acids were
coupled to morpholine using standard amide bond chemistry to
generate intermediates 31−34. These intermediates were then
coupled to the proper 2-chloropyrimidine, 26−30, via acid
catalyzed SNAr to yield products 2 and 14−20.
replacement with a bromine (17) or CF3 (19) increased
potency by 3-fold. Compound 15 was tested against a panel of
63 kinases at Invitrogen (Supporting Information Table 1). As
shown previously with compound 8, compound 15 inhibits
only wild-type and G2019S mutant LRRK2 at greater than 50%
inhibition at 1 μM and it does not inhibit Abl. This result
further validated the hypothesis that installation of substituents
in the small selectivity pocket near Leu1949 could impart
general kinase selectivity.
As there is a critical need for a potent, selective, and brain
penetrable LRRK2 tool compound, the ability of compound 15
to cross the blood−brain barrier in rodents was assessed based
on its favorable physicochemical properties (cLogP, TPSA, and
CNS MPO score of 1.9, 88 Å2, and 5.4, respectively). 15 was
evaluated in wild-type (FVB) and P-gp/BCRP knockout
mice57,58 following a 1 mg/kg intravenous injection. The
CONCLUSION
■
In the absence of LRRK2 KD crystal structures, we relied on
homology modeling as a viable way to provide a 3D LRRK2
model that can be used for structure-based design. The choice
of structural templates proved to be critical to yield a model
robust for structure-based design. We found that structures
with the highest sequence identity to LRRK2 (e.g., MLK1,
Tie2, Braf) were not the best choice for homology modeling
owing to factors such as paucity of high resolution cocomplex
structures, partially hindered binding sites or conformational
states not represented in target protein. What proved
particularly useful was the analysis of available selectivity data,
a
Scheme 1
a
(a) 8 M methylamine in ethanol, methanol; (b) Ar-NH2, n-butanol; (c) TFA, 2-methoxyethanol.
F
dx.doi.org/10.1021/jm300452p | J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
Article
5-Bromo-2-chloro-N-methylpyrimidin-4-amine (30). 1H
NMR (400 MHz, DMSO) δ 8.21 (s, 1H), 7.71 (s, 1H), 2.86 (d, J =
4.6 Hz, 4H); ESMS m/z 224.0 (M + 1).
which revealed an unexpected correlation between inhibition of
JAK3/2 and LRRK2. This result led to the choice of JAK2 as a
structural template for modeling despite relatively low sequence
identity (26%) between LRRK2 and JAK2. The particular
choice of JAK2 structure can also have an impact on the
robustness of the homology model. We chose a high resolution
JAK2 structure with an open ATP-binding site, which proved
very useful for docking a diverse set of LRRK2 small molecule
inhibitors. The LRRK2 homology models have been used in
conjunction with binding site sequence analysis and MMP
activity cliff analysis to identify residues that can potentially
impart general kinase selectivity. One such residue is Leu1949,
which is mostly a Phe or Tyr in other protein kinases. By
incorporation of small substituents on an otherwise non-
selective kinase inhibitor to reach the small selectivity pocket
created by Leu1949 in LRRK2, broad kinase selectivity can be
achieved. Through molecular modeling guided medicinal
chemistry efforts, the nonselective 2,4-diaminopyridine hit 2
was optimized into a promising lead series and compounds in
this series are being used to probe the link between LRRK2
kinase activity and Parkinson’s disease. Details of this work are
forthcoming.
5-Chloro-N4-methyl-N2-phenylpyrimidine-2,4-diamine (9).
2,5-Dichloro-N-methylpyrimidin-4-amine (27, 0.06 g, 0.337 mmol),
aniline (0.063 g, 0.674 mmol), and n-butanol (0.8 mL) were placed in
a 10 mL CEM microwave vial. The vial was capped and irradiated in a
CEM microwave reactor for 20 min at 150 °C followed by
concentration under reduced pressure. The residue was purified by
preparative reverse phase HPLC to give 5-chloro-N4-methyl-N2-
phenylpyrimidine-2,4-diamine (9, 0.0076 g, 9.6%) as an off white solid:
1H NMR (500 MHz, DMSO) δ 9.14 (s, 1H), 7.91 (s, 1H), 7.75 (d, J =
7.7 Hz, 2H), 7.30−7.19 (m, 2H), 7.13 (d, J = 3.9 Hz, 1H), 6.89 (t, J =
7.3 Hz, 1H), 2.92 (d, J = 4.6 Hz, 3H); ESMS m/z 235.0 (M + 1).
In a similar fashion, the following compounds were synthesized
from the appropriate starting materials using the procedure for 9.
5-Chloro-N4-methyl-N2-o-tolylpyrimidine-2,4-diamine (10).
1H NMR (400 MHz, DMSO) δ 8.20 (s, 1H), 7.80 (s, 1H), 7.56 (d,
J = 7.9 Hz, 1H), 7.19−7.07 (m, 2H), 7.07−6.92 (m, 2H), 2.81 (d, J =
4.6 Hz, 3H), 2.21 (s, 3H); ESMS m/z 249.1 (M + 1).
5-Chloro-N2-(2-methoxyphenyl)-N4-methylpyrimidine-2,4-
1
diamine (11). H NMR (400 MHz, DMSO) δ 8.30 (dd, J = 7.4, 2.0
Hz, 1H), 7.92 (s, 1H), 7.56 (s, 1H), 7.22 (d, J = 4.2 Hz, 1H), 7.04−
6.97 (m, 1H), 6.93 (dd, J = 9.4, 3.8 Hz, 2H), 3.86 (s, 3H), 2.90 (d, J =
4.6 Hz, 3H); ESMS m/z 265.0 (M + 1).
N2-(2-Bromophenyl)-5-chloro-N4-methylpyrimidine-2,4-dia-
mine (12). 1H NMR (400 MHz, DMSO) δ 8.08−7.97 (m, 2H), 7.89
(s, 1H), 7.62 (d, J = 8.0 Hz, 1H), 7.35 (t, J = 7.7 Hz, 1H), 7.22 (d, J =
4.3 Hz, 1H), 7.00 (t, J = 7.6 Hz, 1H), 2.85 (d, J = 4.6 Hz, 3H); ESMS
m/z 314.9 (M + 1).
EXPERIMENTAL SECTION
■
General Methods. All solvents and reagents were used as obtained
from the commercial vendors. Intermediates were purified on a
Teledyne ISCO CombiFlash system using prepacked silica gel
columns using appropriate gradients of ethyl acetate and heptane
unless otherwise noted. All final products were purified by preparative
reverse phase HPLC using appropriate gradients of acetonitrile and
water with 0.05% formic acid as the counterion. All tested compounds
were determined to be >95% pure by high-performance liquid
chromatography under the following conditions: system, Agilent 1200
series HPLC instrument equipped with an Agilent 6140 quadrapole
mass spectrometer; column, Agilent C18 (2.1 mm × 30 mm, 1.8 μm);
flow rate 0.4 mL/min; mobile phase A, deionized water; mobile phase
B, 0.05% trifluoroacetic acid in acetonitrile; gradient, 3% B (0−0.5
min), 3−95% B (0.5−7.0 min), 95% B (7.0−7.5 min), 3% B (7.5−7.6
min). Purity was calculated as a percentage of total areas at 254 nm. 1H
NMR spectra were recorded using a Bruker Avance III 400 MHz or
Bruker Avance III 500 MHz spectrometer. Chemical shifts are in parts
per million (δ) referenced to Me4Si (0.00 ppm). Liquid chromatog-
raphy electrospray mass spectra were recorded using a Sciex 150EX
LC−MS system with a Shimadzu LC10AD liquid chromatograph.
2-Chloro-5-fluoro-N-methylpyrimidin-4-amine (26). Methyl-
amine (8 M in ethanol, 15 mL, 120 mmol) was added to 5-fluoro-2,4-
dichloropyrimidine (21, 9.0 g, 53.9 mmol) in methanol (40 mL). The
mixture was stirred at room temperature and then concentrated under
reduced pressure. The residue was purified by silica gel chromatog-
raphy using a methanol/dichloromethane gradient of 1−10% over 35
min to give pure 2-chloro-5-fluoro-N-methylpyrimidin-4-amine (26,
5-Chloro-N2-(2-(difluoromethoxy)phenyl)-N4-methylpyrimi-
1
dine-2,4-diamine (13). H NMR (400 MHz, DMSO) δ 8.17−8.12
(m, 1H), 7.95 (s, 1H), 7.90 (s, 1H), 7.27−7.17 (m, 3H), 7.14 (t, J = 74
Hz, 1H), 7.09−7.02 (m, 1H), 2.87 (d, J = 4.6 Hz, 3H); ESMS m/z
301.0 (M + 1).
(4-Amino-3-methoxyphenyl)(morpholino)methanone (33).
4-Amino-3-methoxybenzoic acid (0.50 g, 2.99 mmol), N-(3-
dimethylaminopropyl)-N′-ethylcarbodiimide hydrochloride (1.43 g,
7.48 mmol), 1-hydroxybenzotriazole (0.606 g, 4.49 mmol), N,N-
diisopropylethylamine (1.56 mL, 8.97 mmol), morpholine (2.608 mL,
29.9 mmol), and N,N-dimethylformamide (15 mL) were combined
and stirred for 18 h at room temperature. The mixture was
concentrated under reduced pressure and the residue partitioned
between ethyl acetate (50 mL) and water (50 mL). The aqueous layer
was extracted with ethyl acetate (30 mL), and the combined organic
extracts were washed with saturated sodium chloride (30 mL), dried
over MgSO4, and concentrated under reduced pressure. The residue
was purified by silica gel chromatography using an ethyl acetate/
heptane gradient of 0−100% over 15 min to give (4-amino-3-
methoxyphenyl)(morpholino)methanone (33, 0.505 g, 71.4%) as a
glassy solid. 1H NMR (400 MHz, DMSO) δ 6.85 (s, 1H), 6.80 (d, J =
8.0 Hz, 1H), 6.61 (d, J = 8.0 Hz, 1H), 5.12 (s, 2H), 3.77 (s, 3H), 3.57
(d, J = 4.2 Hz, 4H), 3.50 (d, J = 4.4 Hz, 4H); ESMS m/z 237.2 (M +
1).
In a similar fashion, the following intermediates were synthesized
from the appropriate starting materials using the procedure for 33.
(4-Aminophenyl)(morpholino)methanone (31). 1H NMR
(400 MHz, DMSO) δ 7.13 (d, J = 8.3 Hz, 2H), 6.54 (d, J = 8.3
Hz, 2H), 5.49 (s, 2H), 3.57 (d, J = 4.5 Hz, 4H), 3.48 (d, J = 4.7 Hz,
4H); ESMS m/z 207.2 (M + 1).
1
6.77 g, 77.7%) as an off white solid. H NMR (400 MHz, DMSO) δ
8.14 (s, 1H), 8.04 (d, J = 3.5 Hz, 1H), 2.85 (d, J = 4.6 Hz, 3H); ESMS
m/z 162.6 (M + 1).
In a similar fashion, the following intermediates were synthesized
from the appropriate starting materials using the procedure for 26.
1
(4-Amino-3-methylphenyl)(morpholino)methanone (32). H
1
2,5-Dichloro-N-methylpyrimidin-4-amine (27). H NMR (500
NMR (400 MHz, DMSO) δ 7.03 (s, 1H), 7.00 (d, J = 8.2 Hz, 1H),
6.58 (d, J = 8.1 Hz, 1H), 5.26 (s, 2H), 3.56 (d, J = 4.1 Hz, 4H), 3.48
(d, J = 4.2 Hz, 4H), 2.05 (s, 3H); ESMS m/z 221.3 (M + 1).
MHz, DMSO) δ 8.13 (s, 1H), 2.87 (d, J = 4.6 Hz, 4H); ESMS m/z
178.9 (M + 1).
1
1
2-Chloro-4-(methylamino)pyrimidine-5-carbonitrile (28). H
(4-Amino-3-bromophenyl)(morpholino)methanone (34). H
NMR (500 MHz, CDCl3) δ 8.31 (s, 1H), 5.74 (s, 1H), 3.34 (d, J = 5
Hz, 3H); ESMS m/z 171.7 (M + 1).
2-Chloro-N-methyl-5-(trifluoromethyl)pyrimidin-4-amine
(29). 1H NMR (400 MHz, DMSO) δ 8.37 (s, 1H), 7.91 (s, 1H), 2.90
(d, J = 4.5 Hz, 3H); ESMS m/z 212.1 (M + 1).
NMR (400 MHz, DMSO) δ 7.44 (s, 1H), 7.17 (d, J = 8.3 Hz, 1H),
6.80 (d, J = 8.3 Hz, 1H), 5.76 (s, 2H), 3.58 (d, J = 4.2 Hz, 4H), 3.50
(d, J = 4.0 Hz, 4H); ESMS m/z 285.1 (M + 1).
(4-(5-Chloro-4-(methylamino)pyrimidin-2-ylamino)phenyl)-
(morpholino)methanone (2). 2,5-Dichloro-N-methylpyrimidin-4-
G
dx.doi.org/10.1021/jm300452p | J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
Article
amine (27 0.10 g, 0.562 mmol), (4-aminophenyl)(morpholino)-
methanone (33, 0.348 g, 1.69 mmol), trifluoroacetic acid (0.1 mL, 1.35
mmol), and 2-methoxyethanol (2 mL) were combined and heated for
18 h 125 °C. The mixture was concentrated under reduced pressure
and the residue was purified by preparative reverse phase HPLC to
give (4-(5-chloro-4-(methylamino)pyrimidin-2-ylamino)phenyl)-
(morpholino)methanone (2, 0.051 g, 25.8%) as an off white solid:
1H NMR (400 MHz, DMSO) δ 9.46 (s, 1H), 7.95 (s, 1H), 7.83 (d, J =
8.6 Hz, 2H), 7.32 (d, J = 8.6 Hz, 2H), 7.25 (d, J = 4.6 Hz, 1H), 3.67−
3.40 (m, 8H), 2.94 (t, J = 7.0 Hz, 3H); ESMS m/z 348.1 (M + 1).
In a similar fashion, the following compounds were synthesized
from the appropriate starting materials using the procedure for 2.
(4-(5-Chloro-4-(methylamino)pyrimidin-2-ylamino)-3-
methylphenyl)(morpholino)methanone (14). 1H NMR (400
MHz, DMSO) δ 8.32 (s, 1H), 7.86 (s, 1H), 7.76 (d, J = 8.2 Hz,
1H), 7.23 (s, 1H), 7.19 (d, J = 8.3 Hz, 1H), 7.13 (d, J = 4.5 Hz, 1H),
3.54 (d, J = 38.0 Hz, 8H), 2.84 (d, J = 4.6 Hz, 3H), 2.26 (s, 3H);
ESMS m/z 362.1 (M + 1).
(4-(5-Chloro-4-(methylamino)pyrimidin-2-ylamino)-3-
methoxyphenyl)(morpholino)methanone (15). 1H NMR (400
MHz, DMSO) δ 8.41 (d, J = 8.2 Hz, 1H), 7.96 (s, 1H), 7.68 (s, 1H),
7.32 (d, J = 4.1 Hz, 1H), 7.05 (s, 1H), 7.01 (d, J = 8.2 Hz, 1H), 3.90
(s, 3H), 3.56 (d, J = 34.5 Hz, 8H), 2.90 (t, J = 7.3 Hz, 3H); ESMS m/z
378.1 (M + 1).
(3-Bromo-4-(5-chloro-4-(methylamino)pyrimidin-2-
ylamino)phenyl)(morpholino)methanone (16). 1H NMR (400
MHz, DMSO) δ 8.21 (d, J = 8.5 Hz, 1H), 8.06 (s, 1H), 7.94 (s, 1H),
7.67 (d, J = 1.6 Hz, 1H), 7.41 (d, J = 8.5 Hz, 1H), 7.32 (d, J = 4.4 Hz,
1H), 3.55 (d, J = 41.3 Hz, 8H), 2.88 (d, J = 4.6 Hz, 3H); ESMS m/z
428.0 (M + 1).
EDTA, and 5% DMSO. The separation conditions used a downstream
voltage of −500 V, an upstream voltage of −2350 V, and a screening
pressure of −1.4 psi. The product and substrate fluorescence was
excited at 488 nm and detected at 530 nm. Substrate conversion was
calculated from the electrophoregram using HTS Well Analyzer
software (Caliper Life Sciences, Hopkinton, MA).
Brain Penetration Protocol. For the brain penetration studies,
plasma and brain tissues were collected at 5 min and 1 h following the
intravenous injection of compound 15 at 1 mg/kg in wild-type (FVB)
and P-gp/Bcrp knockout mice (n = 3). The concentrations in the
plasma and brain samples were determined by LC−MS/MS after
sample treatment. The area under the curve (AUC) from time zero to
1 h was calculated using the trapezoidal rule. The mouse plasma and
brain tissue binding were determined using equilibrium dialysis.
Molecular Modeling. Homology models of LRRK2 were
constructed using the modeling program MOE, version 2009.10
(Chemical Computing Group, Montreal),36 and AMBER9959 force
field. The human LRRK2 sequence was retrieved from Swiss-Prot34
and aligned to template structure sequences using ClustalW60 followed
by manual fine-tuning of residues adjacent to loop regions, insertions,
and deletions. The models were further refined with bound ligand
using the Macromodel utility implemented in Maestro and OPLS2005
force field (Schrodinger, Inc., New York, NY).39 Inhibitor docking
̈
studies were carried out using docking program Glide SP with one
hydrogen bond constraint to the carbonyl oxygen of hinge residue
Ala1950. The docking poses were evaluated based on a combination of
criteria including the Glide docking score (cutoff = −6), favorable
intermolecular interactions with the hinge and other parts of the ATP-
binding pocket, low strain energy of the bound ligand (Estrain < 2 kcal/
mol), etc.
Activity Cliff Analysis. Activity cliff analysis was carried out using
a Pipeline Pilot protocol (Accelrys Inc.)61 that computes pairwise
Tanimoto similarities and activity changes between all pairs of
compounds. The Tanimoto similarities were calculated using extended
connectivity fingerprints configured with counts (ECFC_2).62 Only
pairs that met a similarity cutoff of 0.94 were analyzed for activity cliffs.
(4-(5-Bromo-4-(methylamino)pyrimidin-2-ylamino)-3-
methoxyphenyl)(morpholino)methanone (17). 1H NMR (400
MHz, DMSO) δ 8.40 (d, J = 8.3 Hz, 1H), 8.04 (s, 1H), 7.68 (s, 1H),
7.14 (d, J = 4.6 Hz, 1H), 7.05 (d, J = 1.6 Hz, 1H), 7.01 (dd, J = 8.3, 1.6
Hz, 1H), 3.90 (s, 3H), 3.56 (d, J = 34.6 Hz, 8H), 2.90 (d, J = 4.6 Hz,
3H); ESMS m/z 422.0 (M + 1).
(4-(5-Fluoro-4-(methylamino)pyrimidin-2-ylamino)-3-
methoxyphenyl)(morpholino)methanone (18). 1H NMR (400
MHz, DMSO) δ 8.46 (d, J = 8.3 Hz, 1H), 7.89 (d, J = 3.6 Hz, 1H),
7.56 (d, J = 5.5 Hz, 2H), 7.04 (s, 1H), 7.00 (d, J = 8.3 Hz, 1H), 3.90
(s, 3H), 3.58 (d, J = 19.4 Hz, 4H), 3.52 (s, 3H), 2.90 (t, J = 7.7 Hz,
3H); ESMS m/z 362.1 (M + 1).
ASSOCIATED CONTENT
■
S
* Supporting Information
Table 1 listing Invitrogen kinase selectivity profiling data. This
material is available free of charge via the Internet at http://
pubs.acs.org.
(3-Methoxy-4-(4-(methylamino)-5-(trifluoromethyl)-
1
pyrimidin-2-ylamino)phenyl)(morpholino)methanone (19). H
NMR (400 MHz, DMSO) δ 8.31 (d, J = 8.2 Hz, 1H), 8.19 (s, 1H),
8.07 (s, 1H), 7.21 (d, J = 4.3 Hz, 1H), 7.07 (s, 1H), 7.02 (d, J = 8.2
Hz, 1H), 3.90 (s, 3H), 3.56 (d, J = 37.1 Hz, 9H), 2.92 (d, J = 4.3 Hz,
3H); ESMS m/z 412.1 (M + 1).
2-(2-Bromo-4-(morpholine-4-carbonyl)phenylamino)-4-
(methylamino)pyrimidine-5-carbonitrile (20). 1H NMR (400
MHz, DMSO) δ 8.33 (d, J = 5.9 Hz, 2H), 8.18 (d, J = 8.2 Hz, 1H),
7.76 (d, J = 4.3 Hz, 1H), 7.08 (d, J = 1.5 Hz, 1H), 7.04−6.97 (m, 1H),
3.88 (s, 3H), 3.56 (d, J = 37.1 Hz, 8H), 2.88 (d, J = 4.5 Hz, 3H);
ESMS m/z 369.1 (M + 1).
AUTHOR INFORMATION
■
Corresponding Author
*For H.C. and D.J.B.: phone, 650-225-1000; fax, 650-467-5155;
e-mail, chen.huifen@gene.com (H.C.) and burdick.dan@gene.
com (D.J.B).
Present Address
⊥Novartis Institutes for Biomedical Research (NIBR), 4560
Horton Street, Emeryville, CA 94608-2916.
Microfluidic Capillary Electrophoresis Assay. LRRK2 reactions
were carried out in a final volume of 10 μL per well in a 384-well
microplate. A standard enzymatic mixture, for which the reaction was
initiated by the addition of 5 μL of 2× ATP to 5 μL of 2× enzyme,
contained 10 nM G2019S, 1 μM FAM-LRRKtide (5FAM-
GAGRLGRDKYKTLRQIRQ-CONH2), 130 μM ATP, 25 mM Tris,
pH 8.0, 5 mM MgCl2, 0.01% Triton X-100, 2 mM DTT, 5 mM β-
glycerophosphate, 5 mM NaF,100 μM Na3VO4, and 1× protease
inhibitor cocktail (Calbiochem). After incubation for 120 min at
ambient temperature, the product and substrate in each reaction were
separated using a 12-sipper microfluidic chip (Caliper Life Sciences,
Hopkinton, MA) run on a Caliper LC3000 (Caliper Life Sciences,
Hopkinton, MA). The separation of product and substrate was
optimized by choosing voltages and pressure using Caliper’s optimizer
software (Hopkinton, MA). The separation buffer contained 100 mM
HEPES, pH 7.2, 0.015% Brij-35, 0.1% coating reagent 3, 10 mM
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
We thank Drs. Nicholas Skelton and Steven Magnuson for
critical reading of the manuscript and helpful discussions. We
also thank the analytical, bioanalytical, NMR, pharmacokinetic,
and biopharmacology colleagues for their assistance with
compound purification and testing.
ABBREVIATIONS USED
■
LRRK2, leucine-rich repeat kinase 2; PD, Parkinson’s disease;
JAK2, Janus kinase 2; JAK3, Janus kinase 3; MMP, matched
H
dx.doi.org/10.1021/jm300452p | J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
Article
(15) Liu, Z.; Hamamichi, S.; Lee, B. D.; Yang, D.; Ray, A.; Caldwell,
G. A.; Caldwell, K. A.; Dawson, T. M.; Smith, W. W.; Dawson, V. L.
Inhibitors of LRRK2 kinase attenuate neurodegeneration and
Parkinson-like phenotypes in Caenorhabditis elegans and Drosophila
Parkinson’s disease models. Hum. Mol. Genet. 2011, 20, 3933−3942.
(16) Lee, B. D.; Shin, J.-H.; VanKampen, J.; Petrucelli, L.; West, A.
B.; Ko, H. S.; Lee, Y.-I.; Maguire-Zeiss, K. A.; Bowers, W. J.; Federoff,
H. J.; Dawson, V. L.; Dawson, T. M. Inhibitors of leucine-rich repeat
kinase-2 protect against models of Parkinson’s disease. Nat. Med. 2010,
16, 998−1000.
molecular pair; LE, ligand efficiency; KD, kinase domain;
SBDD, structure-based drug design
REFERENCES
■
(1) Lees, A. J.; Hardy, J.; Revesz, T. Parkinson’s disease. Lancet 2009,
373, 2055−2066. Erratum in Lancet 2009, 374, 684.
(2) Zimprich, A.; Biskup, S.; Leitner, P.; Lichtner, P.; Farrer, M.;
Lincoln, S.; Kachergus, J.; Hulihan, M.; Uitti, R. J.; Calne, D. B.;
Stoessl, A. J.; Pfeiffer, R. F.; Patenge, N.; Carbajal, I. C.; Vieregge, P.;
Asmus, F.; Muller-Myhsok, B.; Dickson, D. W.; Meitinger, T.; Strom,
T. M.; Wszolek, Z. K.; Gasser, T. Mutations in LRRK2 cause
autosomal-dominant parkinsonism with pleomorphic pathology.
Neuron 2004, 44, 601−607.
(3) Paisan-Ruiz, C.; Jain, S.; Evans, E. W.; Gilks, W. P.; Simon, J.; van
der Brug, M.; Lopez de Munain, A.; Aparicio, S.; Gil, A. M.; Khan, N.;
Johnson, J.; Martinez, J. R.; Nicholl, D.; Carrera, I. M.; Pena, A. S.; de
Silva, R.; Lees, A.; Marti-Masso, J. F.; Perez-Tur, J.; Wood, N. W.;
Singleton, A. B. Cloning of the gene containing mutations that cause
PARK8-linked Parkinson’s disease. Neuron 2004, 44, 595−600.
(4) Dachsel, J. C.; Farrer, M. J. LRRK2 and Parkinson disease. Arch.
Neurol. 2010, 67, 542−547.
(17) Garcia-Serna, R.; Mestres, J. Chemical Probes for biological
systems. Drug Discovery Today 2011, 16, 99−106.
(18) Deng, X.; Dzamko, N.; Prescott, A.; Davies, P.; Liu, Q.; Yang,
Q.; Lee, J.-D.; Patricelli, M. P.; Nomanbhoy, T. K.; Alessi, D. R.; Gray,
N. S. Characterization of a selective inhibitor of the Parkinson’s disease
kinase LRRK2. Nat. Chem. Biol. 2011, 7, 203−205.
(19) Ramsden, N.; Perrin, J.; Ren., Z.; Lee, B. D.; Zinn, N.; Dawson,
V. L.; Tam, D.; Bova, M.; Lang, M.; Drewes, G.; Bantscheff, M.; Bard,
F.; Dawson, T. M.; Hopf, C. Chemoproteomics-based design of potent
LRRK2-selective lead compounds that attenuate Parkinson’s disease-
related toxicity in human neurons. ACS Chem. Biol. 2011, 6, 1021−
1028.
(5) Satake, W.; Nakabayashi, Y.; Mizuta, I.; Hirota, Y.; Ito, C.; Kubo,
M.; Kawaguchi, T.; Tsunoda, T.; Watanabe, M.; Takeda, A.;
Tomiyama, H.; Nakashima, K.; Hasegawa, K.; Obata, F.; Yoshikawa,
T.; Kawakami, H.; Sakoda, S.; Yamamoto, M.; Hattori, N.; Murata, M.;
Nakamura, Y.; Toda, T. Genome-wide association study identifies
common variants at four loci as genetic risk factors for Parkinson’s
disease. Nat. Genet. 2009, 41, 1303−1307.
(20) Zhang, J.; Deng, X.; Choi, H. G.; Alessi, D. R.; Gray, N. S.
Characterization of TAE684 as a potent LRRK2 kinase inhibitor.
Bioorg. Med. Chem. Lett. 2012, 22, 1864−1869.
(21) Kramer, T.; Lo Monte, F.; Goering, S.; Okala Amombo, G. M.;
Schmidt, B. Small molecule kinase inhibitors for LRRK2 and their
application to parkinson’s disease models. ACS Chem. Neurosci. 2012,
3, 151−160.
(22) Lovitt, B; Vanderporten, E. C.; Sheng, Z.; Zhu, H.; Drummond,
J.; Liu, Y. Differential effects of divalent manganese and magnesium on
the kinase activity of leucine-rich repeat kinase 2 (LRRK2).
Biochemistry 2010, 49, 3092−3100.
(23) Wager, T. T.; Hou, X.; Verhoest, P. R.; Villalobos, A. Moving
beyond rules: the development of a central nervous system
multiparameter optimization (CNS MPO) approach to enable
alignment of druglike properties. ACS Chem. Neurosci. 2010, 1,
435−449.
(24) Hopkins, A. L.; Groom, C. R.; Alex, A. Ligand efficiency: a
useful metric for lead selection. Drug Discovery Today 2004, 9, 430−
431.
(25) 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.
(26) Moore, W. J.; Richard, D.; Thorarensen, A. An analysis of the
diaminopyrimidine patent estates describing spleen tyrosine kinase
inhibitors by Rigel and Portola. Expert Opin. Ther. Pat. 2010, 20,
1703−1722.
(27) Deng, J.; Lewis, P. A.; Greggio, E.; Sluch, E.; Beilina, A.;
Cookson, M. R. Structure of the ROC domain from the Parkinson’s
disease-associated leucine-rich repeat kinase 2 reveals a dimeric
GTPase. Proc. Natl. Acad. Sci. U.S.A. 2008, 105, 1499−1504.
(28) Marín, I. The Parkinson disease gene LRRK2: evolutionary and
structural insights. Mol. Biol. Evol. 2006, 23, 2423−2433.
(29) Nichols, R. J.; Dzamko, N.; Hutti, J. E.; Cantley, L. C.; Deak, M.;
Moran, J.; Bamborough, P.; Reith, A. D.; Alessi, D. R. Substrate
specificity and inhibitors of LRRK2, a protein kinase mutated in
Parkinson’s disease. Biochem. J. 2009, 424, 47−60.
(30) Liu, M.; Kang, S.; Ray, S.; Jackson, J.; Zaitsev, A. D.; Gerber, S.
A.; Cuny, G. D.; Glicksman, M. A. Kinetic, mechanistic, and structural
modeling studies of truncated wild-type leucine-rich repeat kinase 2
and the G2019S mutant. Biochemistry 2011, 50, 9399−9408.
(31) Cavasotto, C. N.; Phatak, S. S. Homology modeling in drug
discovery: current trends and applications. Drug Discovery Today 2009,
14, 676−683.
(32) Grant, M. A. Protein structure prediction in structure-based
ligand design and virtual screening. Comb. Chem. High Throughput
Screening 2009, 12, 940−960.
́ ́
(6) Simon-Sanchez, J.; Schulte, C.; Bras, J. M.; Sharma, M.; Gibbs, J.
R.; Berg, D.; Paisan-Ruiz, C.; Lichtner, P.; Scholz, S. W.; Hernandez,
D. G.; Kruger, R.; Federoff, M.; Klein, C.; Goate, A.; Perlmutter, J.;
̈
Bonin, M.; Nalls, M. A.; Illig, T.; Gieger, C.; Houlden, H.; Steffens, M.;
Okun, M. S.; Racette, B. A.; Cookson, M. R.; Foote, K. D.; Fernandez,
H. H.; Traynor, B. J.; Schreiber, S.; Arepalli, S.; Zonozi, R.; Gwinn, K.;
van der Brug, M.; Lopez, G.; Chanock, S. J.; Schatzkin, A.; Park, Y.;
Hollenbeck, A.; Gao, J.; Huang, X.; Wood, N. W.; Lorenz, D.; Deuschl,
G.; Chen, H.; Riess, O.; Hardy, J. A.; Singleton, A. B.; Gasser, T.
Genome-wide association study reveals genetic risk underlying
Parkinson’s disease. Nat. Genet. 2009, 41, 1308−1312.
(7) Mata, I. F.; Wedemeyer, W. J.; Farrer, M. J.; Taylor, J. P.; Gallo,
K. A. LRRK2 in Parkinson’s disease: protein domains and functional
insights. Trends Neurosci. 2006, 29, 286−293.
(8) Greggio, E.; Cookson, M. R. Leucine-rich repeat kinase 2
mutations and Parkinson’s disease: three questions. ASN Neuro 2009,
1 (e00002), 13−24.
(9) Gandhi, P. N.; Chen, S. G.; Wilson-Delfosse, A. L. Leucine-rich
repeat kinase 2 (LRRK2): a key player in the pathogenesis of
Parkinson’s disease. J. Neurosci. Res. 2009, 87, 1283−1295.
(10) Drolet, R. E.; Sanders, J. M.; Kern, J. T. Leucine-rich repeat
kinase 2 (LRRK2) cellular biology: a review of recent advances in
identifying physiological substrates and cellular functions. J. Neuro-
genet. 2011, 25, 140−151.
(11) West, A. B.; Moore, D. J.; Biskup, S.; Bugayenko, A.; Smith, W.
W.; Ross, C. A.; Dawson, V. L.; Dawson, T. M. Parkinson’s disease-
associated mutations in leucine-rich repeat kinase 2 augment kinase
activity. Proc. Natl. Acad. Sci. U.S.A. 2005, 102, 16842−16847.
(12) Greggio, E.; Jain, S.; Kingsbury, A.; Bandopadhyay, R.; Lewis, P.;
Kaganovich, A.; van der Brug, M. P.; Beilina, A.; Blackinton, J.;
Thomas, K. J.; Ahmad, R.; Miller, D. W.; Kesavapany, S.; Singleton, A.;
Lees, A.; Harvey, R. J.; Harvey, K.; Cookson, M. R. Kinase activity is
required for the toxic effects of mutant LRRK2/dardarin. Neurobiol.
Dis. 2006, 23, 329−341.
(13) Cookson, M. R. The role of leucine-rich repeat kinase 2
(LRRK2) in Parkinson’s disease. Nat. Rev. Neurosci. 2010, 11, 791−
797.
(14) Rudenko, I. N.; Chia, R.; Cookson, M. R. Is inhibition of kinase
activity the only therapeutic strategy for LRRK2-associated Parkinson’s
disease? BMC Med. 2012, 10, 20−27.
I
dx.doi.org/10.1021/jm300452p | J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
Article
(33) Cavasotto, C. N. Homology models in docking and high-
throughput docking. Curr. Top. Med. Chem. 2011, 11, 1528−1534.
(34) Bairoch, A.; Boeckmann, B.; Ferro, S.; Gasteiger, E. Swiss-Prot:
juggling between evolution and stability. Briefings Bioinf. 2004, 5, 39−
55.
(53) Wassermann, A. M.; Bajorath, J. Chemical substitutions that
introduce activity cliffs across different compound classes and
biological targets. J. Chem. Inf. Model. 2010, 50, 1248−1256.
(54) Peltason, L.; Bajorath, J. Computational analysis of activity and
selectivity cliffs. Methods Mol. Biol. 2011, 672, 119−132.
(55) Griffen, E.; Leach, A. G.; Robb, G. R.; Warner, D. J. Matched
molecular pairs as a medicinal chemistry tool. J. Med. Chem. 2011, 54,
7739−7750.
(35) Altschul, S. F.; Madden, T. L.; Schaffer, A. A.; Zhang, J.; Zhang,
̈
Z.; Miller, W.; Lipman, D. J. Gapped BLAST and PSI-BLAST: a new
generation of protein database search programs. Nucleic Acids Res.
1997, 25, 3389−3402.
(56) Seebeck, B.; Wagener, M.; Rarey, M. From activity cliffs to
target-specific scoring models and pharmacophore hypotheses.
ChemMedChem 2011, 6, 1630−1639.
(36) MOE, version 2009.10; Chemical Computing Group Inc.:
Montreal, Canada.
(57) Scherrmann, J. M. Expression and function of multidrug
resistance transporters at the blood−brain barriers. Expert Opin. Drug
Metab. Toxicol. 2005, 1, 233−246.
(37) Wang, T.; Duffy, J. P.; Wang, J.; Halas, S.; Salituro, F. G.; Pierce,
A. C.; Zuccola, H. J.; Black, J. R.; Hogan, J. K.; Jepson, S.; Shlyakter,
D.; Mahajan, S.; Gu, Y.; Hoock, T.; Wood, M.; Furey, B. F.; Frantz, J.
D.; Dauffenbach, L. M.; Germann, U. A.; Fan, B.; Namchuk, M.;
Bennani, Y. L.; Ledeboer, M. W. Janus kinase 2 inhibitors. Synthesis
and characterization of a novel polycyclic azaindole. J. Med. Chem.
2009, 52, 7938−7941.
(58) Schinkel, A. H. P-Glycoprotein, a gatekeeper in the blood−brain
barrier. Adv. Drug Delivery Rev. 1999, 36, 179−194.
(59) Wang, J.; Cieplak, P.; Kollman, P. A. How well does a restrained
electrostatic potential (RESP) model perform in calculating conforma-
tional energies of organic and biological molecules? J. Comput. Chem.
2000, 21, 1049−1074.
(60) Thompson, J. D.; Higgins, D. G.; Gibson, T. J. CLUSTAL W:
improving the sensitivity of progressive multiple sequence alignment
through sequence weighting, position specific gap penalties and weight
matrix choice. Nucleic Acids Res. 1994, 22, 4673−4680.
(61) Pipeline Pilot, version 7.5; Accelrys Inc.: San Diego, CA.
(62) Rogers, D.; Hahn, M. Extended-connectivity fingerprints. J.
Chem. Inf. Model. 2010, 50, 742−754.
(38) Covy, J. P.; Giasson, B. I. Identification of compounds that
inhibit the kinase activity of leucine-rich repeat kinase 2. Biochem.
Biophys. Res. Commun. 2009, 378, 473−477.
(39) Maestro; Schrodinger, Inc.: New York, 2009.
̈
(40) Gould, C.; Wong, C. F. Designing specific protein kinase
inhibitors: insights from computer simulations and comparative
sequence/structure analysis. Pharmacol. Ther. 2002, 93, 169−178.
(41) Caffrey, D. R.; Lunney, E. A.; Moshinsky, D. J. Prediction of
specificity-determining residues for small-molecule kinase inhibitors.
BMC Bioinf. 2008, 9, 491−505.
(42) Huang, D.; Zhou, T.; Lafleur, K.; Nevado, C.; Caflisch, A.
Kinase selectivity potential for inhibitors targeting the ATP binding
site: a network analysis. Bioinformatics 2010, 26, 198−204.
(43) Posy, S. L.; Hermsmeier, M. A.; Vaccaro, W.; Ott, K. H.;
Todderud, G.; Lippy, J. S.; Trainor, G. L.; Loughney, D. A.; Johnson,
S. R. Trends in kinase selectivity: insights for target class-focused
library screening. J. Med. Chem. 2011, 54, 54−66.
(44) Martin, E.; Mukherjee, P.; Sullivan, D.; Jansen, J. Profile-QSAR:
a novel meta-QSAR method that combines activities across the kinase
family to accurately predict affinity, selectivity, and cellular activity. J.
Chem. Inf. Model. 2011, 51, 1942−1956.
(45) Anderson, P. C; De Sapio, V.; Turner, K. B.; Elmer, S. P.; Roe,
D. C.; Schoeniger, J. S. Identification of binding specificity-determining
features in protein families. J. Med. Chem. 2012, 55, 1926−1939.
(46) Manning, G.; Whyte, D. B.; Martinez, R.; Hunter, T.;
Sudarsanam, S. The protein kinase complement of the human
genome. Science 2002, 298, 1912−1934.
(47) Korr, D.; Toschi, L.; Donner, P.; Pohlenz, H. D.; Kreft, B.;
Weiss, B. LRRK1 protein kinase activity is stimulated upon binding of
GTP to its Roc domain. Cell. Signalling 2006, 18, 910−920.
(48) Galkin, A. V.; Melnick, J. S.; Kim, S.; Hood, T. L.; Li, N.; Li, L.;
Xia, G.; Steensma, R.; Chopiuk, G.; Jiang, J.; Wan, Y.; Ding, P.; Liu, Y.;
Sun, F.; Schultz, P. G.; Gray, N. S.; Warmuth, M. Identification of
NVP-TAE684, a potent, selective, and efficacious inhibitor of NPM-
ALK. Proc. Natl. Acad. Sci. U.S.A. 2007, 104, 270−275. Erratum in
Proc. Natl. Acad. Sci. U.S.A. 2007, 104, 2025.
(49) Anderson, D. R.; Meyers, M. J.; Vernier, W. F.; Mahoney, M.
W.; Kurumbail, R. G.; Caspers, N.; Poda, G. I.; Schindler, J. F.; Reitz,
D. B.; Mourey, R. J. Pyrrolopyridine inhibitors of mitogen-activated
protein kinase activated protein kinase 2 (MK-2). J. Med. Chem. 2007,
50, 2647−2654.
(50) Huggins, D. J.; Sherman, W.; Tidor, B. Rational approaches to
improving selectivity in drug design. J. Med. Chem. 2012, 55, 1424−
1444.
(51) Leach, A. G.; Jones, H. D.; Cosgrove, D. A.; Kenny, P. W.;
Ruston, L.; MacFaul, P.; Wood, J. M.; Colclough, N.; Law, B. Matched
molecular pairs as a guide in the optimization of pharmaceutical
properties; a study of aqueous solubility, plasma protein binding and
oral exposure. J. Med. Chem. 2006, 49, 6672−6682.
(52) Peltason, L.; Bajorath, J. SAR index: quantifying the nature of
structure−activity relationships. J. Med. Chem. 2007, 50, 5571−5578.
J
dx.doi.org/10.1021/jm300452p | J. Med. Chem. XXXX, XXX, XXX−XXX