Structure guided design of a series of selective pyrrolopyrimidinone MARK inhibitors

Article abstract of DOI:10.1016/j.bmcl.2016.08.068

The initial structure activity relationships around an isoindoline uHTS hit will be described. Information gleaned from ligand co-crystal structures allowed for rapid refinements in both MARK potency and kinase selectivity. These efforts allowed for the identification of a compound with properties suitable for use as an in vitro tool compound for validation studies on MARK as a viable target for Alzheimer's disease.[Figure presented]

Full text of DOI:10.1016/j.bmcl.2016.08.068

Accepted Manuscript
Structure guided design of a series of selective pyrrolopyrimidinone MARK
Inhibitors
Jason D. Katz, Andrew Haidle, Kaleen K. Childers, Anna A. Zabierek, James
P. Jewell, Yongquan Hou, Michael D. Altman, Alexander Szewczak, Dapeng
Chen, Andreas Harsch, Mansuo Hayashi, Lee Warren, Michael Hutton, Hugh
Nuthall, Hua-Poo Su, Sanjeev Munshi, Matt G. Stanton, Ian W. Davies, Ben
Munoz, Alan Northrup
PII:
S0960-894X(16)30900-3
DOI:
http://dx.doi.org/10.1016/j.bmcl.2016.08.068
BMCL 24194
Reference:
Bioorganic & Medicinal Chemistry Letters
To appear in:
Received Date:
Revised Date:
Accepted Date:
13 July 2016
19 August 2016
20 August 2016
Please cite this article as: Katz, J.D., Haidle, A., Childers, K.K., Zabierek, A.A., Jewell, J.P., Hou, Y., Altman, M.D.,
Szewczak, A., Chen, D., Harsch, A., Hayashi, M., Warren, L., Hutton, M., Nuthall, H., Su, H-P., Munshi, S., Stanton,
M.G., Davies, I.W., Munoz, B., Northrup, A., Structure guided design of a series of selective pyrrolopyrimidinone
MARK Inhibitors, Bioorganic & Medicinal Chemistry Letters (2016), doi: http://dx.doi.org/10.1016/j.bmcl.
2016.08.068
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Graphical Abstract
To create your abstract, type over the instructions in the template box below.
Fonts or abstract dimensions should not be changed or altered.
Structure Guided Design of a Series of
Selective Pyrrolopyrimidinone MARK
Inhibitors
Leave this area blank for abstract info.
Jason D. Katz, Andrew Haidle, Kaleen K. Childers, Anna A. Zabierek, James P. Jewell, Yongquan Hou,
Michael D. Altman, Alexander Szewczak, Dapeng Chen, Andreas Harsch, Mansuo Hayashi, Lee Warren,
Michael Hutton, Hugh Nuthall, Hua-Poo Su, Sanjeev Munshi, Matt G. Stanton, Ian W. Davies, Ben Munoz, and
Alan Northrup

Bioorganic & Medicinal Chemistry Letters
Structure guided design of a series of selective pyrrolopyrimidinone MARK
Inhibitors
Jason D. Katz, a Andrew Haidle,a Kaleen K. Childers,a Anna A. Zabierek,a James P. Jewell,a Yongquan
Hou,a Michael D. Altman,b Alexander Szewczak,c Dapeng Chen,d Andreas Harsch,d Mansuo Hayashi,e Lee
Warren,e Michael Hutton,e Hugh Nuthall,e Hua-Poo Su,f Sanjeev Munshi,f Matt G. Stanton,a Ian W.
Davies,a Ben Munoz,a and Alan Northrupa
a Department of Chemistry, Merck & Co., 33 Avenue Louis Pasteur, Boston, MA 02115, USA
b Department of Structural Chemistry, Merck & Co., 33 Avenue Louis Pasteur, Boston, MA 02115, USA
c Department of In Vitro Sciences, Merck & Co., 33 Avenue Louis Pasteur, Boston, MA 02115, USA
d Department of Drug Metabolism, Merck & Co., 33 Avenue Louis Pasteur, Boston, MA 02115, USA
e Department of Neuroscience, Merck & Co., 33 Avenue Louis Pasteur, Boston, MA 02115, USA
f Department of Structural Chemistry, Merck & Co., 770 Sumneytown Pike, West Point, PA 19486, USA
ARTICLE INFO
ABSTRACT
Article history:
Received
Revised
Accepted
Available online
The initial structure activity relationships around an isoindoline uHTS hit will be described.
Information gleaned from ligand co-crystal structures allowed for rapid refinements in both
MARK potency and kinase selectivity. These efforts allowed for the identification of a
compound with properties suitable for use as an in vitro tool compound for validation studies on
MARK as a viable target for Alzheimer’s disease.
2009 Elsevier Ltd. All rights reserved.
Keywords:
MARK
kinase
Tau
Alzheimer’s Disease
Pyrrolopyrimidinone
Alzheimer’s disease (AD) represents a huge unmet medical
need that will only become more urgent with the current aging
population. In fact, the number of people suffering from this
disease has been estimated to reach 13-16 million in the United
States alone by 2050(1). However, despite significant efforts
from the scientific community, robust disease-modifying
between tau and neurological disease. As such, one therapeutic
approach that has been pursued is to decrease the
phosphorylation on tau through inhibiting tau kinases, via the
hypothesis that this would lead to less tau aggregation.
Multiple kinases are known to phosphorylate tau and many
have been examined as possible targets for AD(4). Among them
are the microtubule-affinity regulating kinase (MARK) family,
consisting of 4 isoforms (MARK1-4). MARKs phosphorylate
tau protein in its repeat domain and thereby regulate its affinity
for MTs and affect the aggregation of tau into NTFs(5). It was
found that PAR-1, the fly homolog of mammalian MARK,
therapies for AD have remained elusive(2).
A major
physiological marker of AD is the abnormal deposit of plaques in
the brain, primarily made up of either amyloid Aβ protein or
neurofibrillary tangles (NFTs). As such, many of the therapeutic
strategies approached to combat AD have focused on reducing
the severity of these lesions. Tau binds to and stabilizes
microtubules (MTs), particularly in neuronal axons, and thus
plays a crucial role in the neuronal cytoskeleton stabilization.
Tau can become abnormally hyper-phosphorylated in the AD
brain, causing it to lack affinity for MTs and self-associate into
NFTs(3). Observations of AD brains show a strong correlation
between cognitive dysfunction and cortical NFT density.
Moreover, mutations in tau have been shown to cause a form of
directly phosphorylates tau at S262 and S356.
This
phosphorylation event is a prerequisite for the action of
downstream kinases including GSK-3 and Cdk5 to phosphorylate
several other sites and generate disease-associated phospho-
epitopes(6). Active MARKs are elevated and colocalize with
NTFs in the AD brain, and MARK phosphorylation sites on tau
are elevated early in transgenic mouse models of tauopathy.
These findings suggest that inhibition of MARKs would be an
frontotemporal dementia, which provided a direct genetic link
———
 Corresponding author. Tel.: +1-617-992-2057; fax: +1-617-992-2406; e-mail: jason_katz2@merck.com

attractive therapeutic strategy for the treatment of AD by
reducing the levels of unbound tau that are available for the
formation of NTFs. In this communication, the initial structure
activity relationships around an isoindoline uHTS hit will be
described, followed by its evolution into an in vitro tool
compound. The subsequent communication details the efforts to
deliver an in vivo tool compound for validation studies on
MARK as a viable target for AD.
Conversely, an alternate sequence of synthetic steps would allow
for initial formation of the lactam 6, followed by oxidation and
SNAr to arrive at pyrrolopyrimidinones 5.
The transposed isoindolinone analog 11 was accessed using
the synthetic sequence described in Scheme 2. Commercially
available acid 8 could be transformed to the amino pyrimidine 9
using procedures described above (esterification, oxidation to the
sulfone, displacement with formanilide, and a Boc protection).
The bromide was then converted to the isoprenyl analog followed
by transformation of the ester to the pyrazolyl amide 10. This
advanced intermediate was then treated with IBX under the
conditions described by Baran and Nicolau (15, 16, 17, 18, 19)
followed by Boc deprotection to deliver the highly functionalized
isoindolinone 11. To the best of our knowledge, the IBX
methodology in reference 11 has not been previously extended to
more complex, dense hetererocyclic systems such as in 11(20).
Figure 1. Properties of uHTS hit compound 1.
An uHTS screen of the Merck compound collection against
MARK3 identified a number of potential starting points for
further development. From this set, the team was attracted to
a series of pyrrolopyrimidinones, represented by the N-
cyclohexyl analog 1 (Figure 1). As a starting point for a CNS
program, the initial set of compounds profiled in the class
possessed a number of attractive properties, such as a
relatively low MW, as well as PSA and clogD values within
acceptable ranges(7, 8). In addition, compounds such as 1
proved not to be substrates for Pgp(9). While 1 was not
exquisitely potent, by the metrics we were tracking at the
time, it was relatively efficient (LE = 0.36) and exhibited a
moderate cell shift (28x) (10, 11, 12). Furthermore, there was
evidence of baseline selectivity against BRSK2, another
kinase capable of phosphorylating Tau(13). Throughout the
program, selectivity against a small panels of kinases,
including BRSK2 was used to monitor progress against off-
targets that could complicate interpretation of downstream
target validation experiments. At the time this work was
conducted, a key goal for the program was to identify in vitro
and in vivo tool compounds to validate the target. The main
focus of this manuscript will be to describe initial efforts in
this area.
Scheme 2. a) TMS-diazomethane, CH2Cl2/MeOH (1:1), 95%;
b) oxone, THF/MeOH/sat’d NaHCO3(aq)/water (1:1:1:1),
86%; c) formanilide, NaH, DMF, 0 °CRT, 1 h, 70%; d)
Boc2O, NEt3, THF, quant; e) isopropenyl boronic acid,
pinacol ester, Pd(OAc)2, cyHex3P, K3PO4, toluene/water (3:1),
95%; NaOH, THF/MeOH (1:1); g) 1-methyl-1H-pyrazol-4-
amine, BOP-Cl, iPrNEt2, CH3CN, 60%; h) IBX, THF/DMSO
(10:1), 90 °C, 40%; i) TFA, CH2Cl2, 86%.
The dihydro-pyrrolopyrimidine 13 was prepared via a slightly
modified route (Scheme 3). Treatment of sulfide 6a (prepared as
in Scheme 1, using cyclohexylamine) with lithium aluminum
hydride delivered the intermediate aminal, which was further
reduced to 12 in the presence of triethysilane. A further 2-step
sequence involving oxidation of the sulfide and displacement
with aniline allowed for isolation of the fully functionalized 13.
Scheme 3. a) LiAlH4, THF, 0 °CRT, 92%; b) BF3OEt2,
Et3SiH, CH2Cl2, -78 °CRT, 78%; c) oxone,
Scheme 1. a) R1NH2, DABCO, 2-ethoxyethanol, 140 °C; b)
m-CPBA, DCM, 20 °C; c) R2NH2, NaHMDS, THF, 20 °C; d)
R2NHCHO, NaH, DMF, 0 °C; e) NaHCO3, THF/MeOH/H2O,
20 °C.
The chemistry to prepare most of the analogs is described in
Scheme 1, starting from 2-thiomethylpyrimidine 2(14).
Treatment with m-CPBA provided the sulfone 3, which was then
displaced with an amine to deliver the aminopyrimidine analogs
4. Subsequent condensation with another primary amine
supplied the desired functionalized pyrrolipyrimidinones 5.
i
THF/MeOH/sat’d NaHCO3(aq) (1:1:1); d) aniline, PrNEt2,
155 °C, 6% (2 steps).
In addition to the physical properties mentioned above, 1
offered many potential regions for further elaboration. Based on
the kinase literature and preliminary modeling studies, it was
assumed that these compounds bound to the ATP binding pocket
of the kinase, where the amino pyrimidine was involved in a
bidentate hydrogen bond with the hinge directing the phenyl ring
along the hinge with vectors out toward solvent (Figure 2). In

this binding mode, it was postulated that the methyl groups could
make hydrophobic contacts with the Met129 gatekeeper, with the
carbonyl oxygen lone pair recruiting the conserved catalytic
lysine and/or aspartate. Based on this assumption, there were
four main areas for substitution that were considered. Most of
the examples presented below will focus on regions 1 and 4,
while details around regions 2 and 3 will be discussed in the
following paper.
Figure 2. Vectors for further elaboration of 1.
Initial structural changes to the lead were consistent with the
original docking studies. For instance, substitution within or
about the Ar ring (region 2) was rather permissive of polar
functionality, (Table 1, piperidine 14 or 3-pyridyl 15), which
would be expected with a solvent-facing region. In general,
substitution off of this aromatic ring delivered more potent
analogs, albeit with no improvements in LE. However, more
importantly this functional group tolerance allowed for fine
tuning of physical properties, which will be discussed in more
detail in the following communication. Similarly, changes to the
lactam (region 4) completely eroded MARK activity, such as
with the inverted isoindolinone 11 and the des-carbonyl
pyrrolopyrimidine 13, supporting the hypothesis that the lactam
carbonyl was involved in H-bonding interactions with the
catalytic residues.
Table 1: Initial SAR About the Pyrrolpyrimidone Core

resulted in >9x loss in potency vs. the cyclohexyl reference
compound (Table 1). As a result, much of the early SAR in this
region proceeded in a more empirical fashion. In fact, the initial
SAR indicated the core of the initial hit seemed to represent an
optimized minimal core (e.g. gem-dimethyl, carbonyl of the
lactam, N-Aryl along hinge). However, it still remained to be
determined how best to optimize the cyclohexyl fragment. In
order to understand this more completely, it was decided to
remove this piece entirely and evolve the lead back out into this
space. To this end, removing the cyclohexyl piece completely, as
with 18, had a detrimental effect on the absolute potency, but the
core still retained most of the binding efficiency (Table 2).
Subsequent step-wise growth of the carbon chain, from the N-Me
19 (compare to 14), to the ethyl and isopropyl analogs 20 and 21
respectively, hinted that it was possible to achieve a level of
binding efficiency comparable to or better than the cyclohexyl
starting point. One direction pursued was to install small,
symmetrical ring systems, such as the cyclopentyl analog 22.
While this was generally tolerated, there were no obvious SAR
breakthroughs with the initial analogs. For instance, attempts to
incorporate heteroatoms into these ring systems, exemplified by
the N-acyl piperidine 23 were accompanied with large drop-offs
in potencies.
While the initial cyclic analog approach was not fruitful,
extending out into the space with acyclic analogs proved to be
more promising. Simply going from the isopryopyl analog 21 to
the sec-butyl analog 24 resulted in a modest, but noticeable,
increase in both potency and binding efficiency. Even more
intriguing was the influence of stereochemistry at the alpha
carbon, as the sec-butyl R isomer 25 was about 6 fold less potent.
This preference for the S isomer in the acyclic analogs was
shown to be rather general (>50x for some analogs, data not
shown). As such, maintaining the alpha-methyl gearing element
and continuing to grow off of the other position, such as with the
N-3-methylbutan-2-yl 26 or the alpha-methyl-benzyl 27 resulted
in a trend toward increased MARK potency with minimal
sacrifices in binding efficiency. Attempts to explore the alpha
position a little further, such as with the gem-dimethyl analog 28
or with the difluoromethyl 29 indicated that there was little room
for further elaboration in this region. Docking studies on these
compounds suggested that the alpha methyl group should be in
close proximity to the glycine rich loop with minimal tolerance
for substitution. In general, while slow progress was being made
using this empirical SAR method, many of the more interesting
leads suffered from poor cell activity (Table 1 & 2, MARK4 cell
IC50) and insufficient off-target kinase selectivity (data not
shown), limiting their usefulness as tool compounds.
About this time, crystal structures of select compounds bound
to the MARK2 enzyme became available. One example was the
fluorobenzyl 30 (Figure 3), which confirmed many aspects of
the proposed binding mode. As suspected, the aminopyridine
was making contacts with the hinge regions and the pyridine ring
was projecting out toward solvent. The gem-dimethyl groups
were located near the Met129 gatekeeper residue and the
carbonyl of the lactam ring was within H-bonding distance of the
catalytic Lys82. While the fluorobenzyl group occupied the
ribose region as expected, there were no obvious interactions
present to explain much of the SAR in that region (vide supra).
In particular, the importance of stereochemistry of the alpha
methyl group was confounding, as the binding pocket appeared
to be capable of accommodating both enantiomers.
Table 2: Deconstruction of the Cyclohexyl Group
In contrast, initial attempts to implement rational design in
region 1 via modeling were less productive. For example,
transformation of the cyclohexyl to the unsaturated phenyl analog
16 or the pyrazole 17 were predicted to be well tolerated, but

Figure 3. X-Ray structure of Compound 30 bound to MARK2. (PDB
code: 5KZ7)
The structure of 30 did suggest two potential strategies to
further increase the potency of these compounds. In the first
case, the tolerance of certain cyclic analogs (e.g. 1 and 22), and
the crystal structure of 30 suggested that cyclization into a
bicyclic structure (Figure 4) might lock the ligand in the
bioactive conformation. A second approach was based on the
proximity of the 3-position of the benzyl ring to Glu136 at the lip
of the ATP binding pocket. It was possible that additional
binding affinity could be gained by the proper placement of an
acidic polar functionality at this position, as well as provide an
additional handle for further modification of physical properties.
Table 3: Optimization From the Crystal Structure of 30
Figure 4: Target Strategy Based of the Structure of Compound 30
In practice, both of these strategies proved to be effective. As
can be seen in Table 3, cyclization (Approach #1) to the 5-
membered indane 31 was well tolerated.(21) In order to explore
approach #2, the meta phenol 32 was prepared, resulting in
modest gains in MARK3 potency. However, a more substantial
gain was observed when both approaches were combined into a
single molecule, such as with the hydroxyindane 33.
A
subsequent crystal structure of 33 bound to MARK2 (Figure 5)
confirmed that the aromatic ring was placed in a similar
orientation as observed in 31 and that the phenol was making
additional interactions with Glu136. Subsequent modifications to
probe this Glu136 interaction demonstrated that in many ways,
the phenol was an optimized interaction, exemplified by a few
select analogs included in Table 3. For instance adjusting the
pKa with the aniline 34 resulted in 10x drop-off in intrinsic
potency. Similar modifications, such as with the sulfonamide 35
highlight the sensitivity at this region to combinations of pKa and
the degree of the substitution. While these gains in
potency/efficiency were exciting, it was equally satisfying to see
that they were displaying reasonable levels of MARK cell
activity.
Figure 5. X-Ray structure of Compound 33 bound to MARK2. (PDB
code: 5KZ8)
Compounds such as 33 proved to be useful tools to test the in
vitro effects of selective MARK inhibitors. As detailed in Figure
6, 33 is one of the more potent and efficient MARK inhibitors
profiled from this structural class, with excellent cell activity. In
addition, optimization of key contacts from the crystal structures
translated into selectivity across a subset of the kinome. While
not exquisitely selective, only 5 kinases out of a panel of 10
kinases had activity <100 fold.(22) However, the
pharmacokinetics of 33 are very poor, with high turnover in in
vitro systems translating into high clearances and short half lives
in vivo. In general, this was representative of this class of
compounds, which hindered their subsequent development as in

was not anticipated that meaningful differences in isoforms could
be achieved. Instead, efforts were focused on delivering a
compound with sufficient protein kinase selectivity as a tool
compound.
vivo tool compounds for the program. The efforts to overcome
this issue are described in the following communication.
13. Yoshida, H.; Goedert M. J. Neurochemistry 2012, 120, 165. DOI:
10.1111/j.1471-4159.2011.07523.x
14. Haidle, A.; Stanton, M.; Altman, M.; Konrad, K.; Zabierek, A.;
Katz, J.; Jewell, J. WO 20090152027 A1
15. Nicolaou, K.C.; Baran, P.S.; Zhong, Y.-L.; Sugita, K. J. Am.
Chem. Soc. 2002, 124, 2212. DOI: 10.1021/ja012124x
16. Nicolaou, K.C.; Sugita, K.; Baran, P.S.; Zhong, Y.-L. J. Am.
Chem. Soc. 2002, 124, 2221. DOI: 10.1021/ja012125p
17. Nicolaou, K.C.; Baran, P.S.; Zhong, Y.-L.; Barluenga, S.; Hunt,
K.W.; Kranich, R.; Vega, J.A. J. Am. Chem. Soc. 2002, 124, 2233.
DOI: 10.1021/ja012126h
18. Nicolaou, K.C.; Montagnon, T.; Baran, P.S.; Zhong, Y.-L. J. Am.
Chem. Soc. 2002, 124, 2245. DOI: 10.1021/ja012127+
19. Nicolaou, K.C.; Baran, P.S.; Kranich, R.; Zhong, Y.-L.; Sugita,
K.; Zou, N. Angew. Chem. Int. Ed. 2001, 40, 202. DOI:
10.1002/1521-3773(20010105)40:1<202::AID-
ANIE202>3.0.CO;2-3
20. We briefly explored the generality of this transformation beyond
the scope described in ref 11 to prepare highly substituted
isoindolinones (V. Chevalier). Although the yield and scope were
proven to be somewhat limited, this protocol could be used to
access otherwise inaccessible functionality. Given the proposed
mechanism, it was not surprising that a range of aromatic groups
were tolerated about a simplified core, with yields ranging
between 25-43%. Similarly, as described by Baran, electron rich
aromatics capable of stabilizing the radical intermediate were
required, but highlights that additional groups could be considered
for this transformation.
Figure 6. Properties of Compound 33
In summary, we have described the development of an uHTS
lead into a potent and selective MARK inhibitor. This was
accomplished through a combination of targeted SAR based on
initial binding hypotheses, and further refined by the use of
insights gleaned from select X-ray structures. This ultimately
allowed for the identification of suitable in vitro tool compounds
such as 33. However, the development of in vivo tools
compounds was hindered by a complex series of PK issues that
are described in the following manuscript.
Acknowledgements
We would like to thank Vincent Chevalier for his work on the
IBX cyclization chemistry during a Co-Op Internship from
Northeastern.
References and Notes
1.
2.
3.
Querfurth, H. W.; LaFerla, F. M. N. Engl. J. Med. 2010, 362, 329.
DOI: 10.1056/NEJMra0909142
Citron, M. Nat. Rev. Drug Discov. 2010, 9, 387.
DOI:10.1038/nrd2896
Lee, V. M.-Y.; Brunden, K. R.; Hutton, M.; Trojanowski, J. Q.
Cold Spring Harb. Perspect. Med. 2011, 1. DOI:
10.1101/cshperspect.a006437
Martin, L.; Latypova, X.; Wilson, C. M.; Magnaudeix, A.; Perrin,
M.-L.; Yardin, C.; Terro, F. Ageing Res. Rev. 2013, 12, 289.
DOI:10.1016/j.arr.2012.06.003
21. The stereochemistry was also important in the cyclic analogs (e.g.
enantiomer of 32 IC50 = 1.8 M). Other ring systems were also
tolerated (e.g. racemic tetrahydronapthalene IC50 = 73 nM.
22. The full list is included as Supplemental Data. Off-target kinase
values are estimated IC50 data from percent inhibition at three
concentrations (0.1, 1, and 10 M).
4.
5.
Timm, T.; Marx, A.; Panneerselvam, S.; Mandelkow, E.;
Mandelkow, E.-M. BMC Neurosci. 2008, 9(Suppl 2), S9. DOI:
10.1186/1471-2202-9-S2-S9
6.
7.
8.
9.
Nishimura, I.; Yang, Y.; Lu, B. Cell 2004, 116, 671. DOI:
10.1016/S0092-8674(04)00170-9.
Wager, T.T.; Hou, X.; Verhoest P.R.; Villalobo, A. ACS Chem.
Neurosci. 2010;1:435. DOI: 10.1021/cn100008c
cLogD and PSA were calculated using ACD Labs. LE was
calculated as LE = -[RT*ln(pIC50)] / Nheavy
.
P-gp substrate susceptibility was assessed by measuring the
BA:AB ratio in LLC-PK1 cells expressing complementary DNA
(cDNA) encoding human MDR1 P-gp or rat Mdr1a P-gp.
10. The MARK3 activity was assessed in an HTRF-based assay
utilizing full-lenth human MARK3 and an biotin labeled peptide
substrate.
11. The MARK4 cell assay was conducted in HEK 293T cells stably
transfected with human MARK4 and Tau protein. Quantitative
change of phosphorylated tau protein at Ser 262 was measured
using Perkin Elmer's technology AlphaLisa.
12. Note that for the purposes of this publication, we will make no
distinction between any of the MARK isoforms. At the time, it