Discovery of Allosteric, Potent, Subtype Selective, and Peripherally Restricted TrkA Kinase Inhibitors

Article abstract of DOI:10.1021/acs.jmedchem.8b00280

Tropomyosin receptor kinases (TrkA, TrkB, TrkC) are activated by hormones of the neurotrophin family: nerve growth factor (NGF), brain derived neurotrophic factor (BDNF), neurotrophin 3 (NT3), and neurotrophin 4 (NT4). Moreover, the NGF antibody tanezumab has provided clinical proof of concept for inhibition of the TrkA kinase pathway in pain leading to significant interest in the development of small molecule inhibitors of TrkA. However, achieving TrkA subtype selectivity over TrkB and TrkC via a Type I and Type II inhibitor binding mode has proven challenging and Type III or Type IV allosteric inhibitors may present a more promising selectivity design approach. Furthermore, TrkA inhibitors with minimal brain availability are required to deliver an appropriate safety profile. Herein, we describe the discovery of a highly potent, subtype selective, peripherally restricted, efficacious, and well-tolerated series of allosteric TrkA inhibitors that culminated in the delivery of candidate quality compound 23.

Full text of DOI:10.1021/acs.jmedchem.8b00280

Subscriber access provided by UNIV OF NEW ENGLAND ARMIDALE
Discovery of Allosteric, Potent, Subtype Selective
and Peripherally Restricted TrkA Kinase Inhibitors
Sharan K Bagal, Kiyoyuki Omoto, David C Blakemore, Peter J. Bungay, James G Bilsland,
Philip J Clarke, Matthew S. Corbett, Ciarán N Cronin, Jingrong Jean Cui, Rebecca Dias, Neil J
Flanagan, Samantha E Greasley, Rachel Grimley, Eric Johnson, David Fengas, Linda Kitching,
Michelle L Kraus, Indrawan McAlpine, Asako Nagata, Gareth J. Waldron, and Joseph S Warmus
J. Med. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.jmedchem.8b00280 • Publication Date (Web): 19 Apr 2018
Downloaded from http://pubs.acs.org on April 19, 2018
Just Accepted
“Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted
online prior to technical editing, formatting for publication and author proofing. The American Chemical
Society provides “Just Accepted” as a service to the research community to expedite the dissemination
of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in
full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully
peer reviewed, but should not be considered the official version of record. They are citable by the
Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore,
the “Just Accepted” Web site may not include all articles that will be published in the journal. After
a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web
site and published as an ASAP article. Note that technical editing may introduce minor changes
to the manuscript text and/or graphics which could affect content, and all legal disclaimers and
ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or
consequences arising from the use of information contained in these “Just Accepted” manuscripts.
is published by the American Chemical Society. 1155 Sixteenth Street N.W.,
Washington, DC 20036
Published by American Chemical Society. Copyright © American Chemical Society.
However, no copyright claim is made to original U.S. Government works, or works
produced by employees of any Commonwealth realm Crown government in the course
of their duties.

Page 1 of 63
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Discovery of Allosteric, Potent, Subtype Selective
and Peripherally Restricted TrkA Kinase Inhibitors
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
a*
a
a
b
c
Sharan K. Bagal, Kiyoyuki Omoto, David C. Blakemore, Peter J. Bungay, James G. Bilsland,
f
d
e
e
c
Philip J. Clarke, Matthew S. Corbett, Ciaran N. Cronin, J. Jean Cui, Rebecca Dias, Neil J.
c
e
c
e
f
Flanagan, Samantha E. Greasley, Rachel Grimley, Eric Johnson, David Fengas, Linda
c
e
e
e
c
Kitching, Michelle L. Kraus, Indrawan McAlpine, Asako Nagata, Gareth J. Waldron, Joseph
S. Warmus.d
aWorldwide Medicinal Chemistry, Pfizer Global R&D UK, The Portway Building, Granta Park,
Cambridge, CB21 6GS, UK
bPharmacokinetics, Dynamics & Metabolism, Pfizer Global R&D UK, The Portway Building,
Granta Park, Cambridge CB21 6GS, UK
cPfizer Global R&D UK, The Portway Building, Granta Park, Cambridge CB21 6GS, UK
dPfizer Global R&D, Groton Laboratories, Eastern Point Road, Groton, Connecticut 06340, USA
ePfizer Global R&D, La Jolla Laboratories, 10770 Science Center Drive, San Diego, California,
92121, USA
fPeakdale Molecular, Discovery Park House, Ramsgate Road, Sandwich, Kent CT13 9ND, UK
ACS Paragon Plus Environment
1

Journal of Medicinal Chemistry
Page 2 of 63
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
KEYWORDS
Nerve growth factor receptor, NTRK1, NGF, TrkA, Tropomyosin Receptor Kinase, Pain,
Allosteric, Type III inhibitor, Kinase inhibitor
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
ABSTRACT
Tropomyosin receptor kinases (TrkA, TrkB, TrkC) are activated by hormones of the neurotrophin
family: nerve growth factor (NGF), brain derived neurotrophic factor (BDNF), neurotrophin 3
(NT3) and neurotrophin 4 (NT4). Moreover, the NGF antibody tanezumab has provided clinical
proof of concept for inhibition of the TrkA kinase pathway in pain leading to significant interest
in the development of small molecule inhibitors of TrkA. However, achieving TrkA subtype
selectivity over TrkB and TrkC via a Type I and Type II inhibitor binding mode has proven
challenging and Type III or Type IV allosteric inhibitors may present a more promising selectivity
design approach. Furthermore, TrkA inhibitors with minimal brain availability are required to
deliver an appropriate safety profile. Herein we describe the discovery of a highly potent, subtype
selective, peripherally restricted, efficacious and well-tolerated series of allosteric TrkA inhibitors
that culminated in the delivery of candidate quality compound 23.
ACS Paragon Plus Environment
2

Page 3 of 63
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
INTRODUCTION
Tropomyosin receptor kinases are a family of three tyrosine kinases (TrkA, TrkB, TrkC) activated
by peptidic hormones of the neurotrophin family: nerve growth factor (NGF), brain derived
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
neurotrophic factor (BDNF), neurotrophin 3 (NT3) and neurotrophin 4 (NT4) (Figure 1).1
Preclinical and clinical studies have identified a crucial role for NGF in the pathogenesis of
2
inflammatory pain, with neutralizing antibodies such as tanezumab demonstrating efficacy in
preclinical pain models and in clinical trials for osteoarthritis, chronic low back pain and interstitial
cystitis. 3 4-5 6 Since NGF binds to the TrkA kinase receptor this has provided clinical proof of
concept for inhibition of the TrkA pathway in pain and has also led to significant interest in the
development of small molecule modulators of TrkA.
Figure 1. Neurotrophins and their receptors. The neurotrophins exhibit specific interactions with
the three Trk receptors: NGF binds TrkA, BDNF and NT-4 binds TrkB, and NT-3 binds TrkC.
A key concern in the development of TrkA kinase inhibitors for the treatment of chronic pain is
achieving sufficient TrkA selectivity over TrkB and TrkC. TrkB is expressed throughout the body
ACS Paragon Plus Environment
3

Journal of Medicinal Chemistry
Page 4 of 63
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
with the BDNF/TrkB axis involved in excitatory signaling, long-term potentiation and feeding
behavior.7,8,9-10 Moreover, human genetic data has linked decreased function of TrkB with obesity
and developmental delay.11 BDNF has been shown to play a role in regulating the arterial
baroreceptor reflex and in modulating heart contraction force.12-14 TrkC is widely expressed in
neural and non-neural tissues and has wide physiological functions that include the development
and survival of the sympathetic nervous system.15-16
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Most small molecule kinase inhibitors are ATP competitive Type I or Type II binders, by
reference to the highly conserved aspartate-phenylalanine-glycine (DFG) motif in the beginning
of the activation loop in the C-lobe of the kinase domain. Type I inhibitors such as the JAK
inhibitor tofacitinib (recently approved for rheumatoid arthritis) binds in the ATP pocket in a DFG-
17
in conformation, whilst Type II inhibitors such as imatinib bind in the ATP pocket and extend
into a less conserved allosteric region that is formed following a conformational change to the
1
8
DFG-out structure. Whilst Type I and II inhibitors can display high levels of selectivity, TrkA,
TrkB and TrkC kinases have no residue difference in the ATP binding site, suggesting that
achieving TrkA selectivity over TrkB/C in this site may be extremely challenging. Moreover, our
previous work on the pan-Trk inhibitor program where isoform non-selective TrkA/B/C inhibitors
were progressed into clinical development, suggested that Type I and Type II binders tend to
19
exhibit pan-Trk activity rather than subtype selectivity. One known method for achieving isoform
selectivity within closely related kinases is the identification of Type III and Type IV allosteric
2
0
ligands. Type III ligands bind in an adjacent site to the ATP binding site without making any
interactions with the hinge region of the ATP binding site. Trametinib is a Type III inhibitor of
MEK1/2 kinases and is the first allosteric kinase inhibitor approved by the FDA, being accepted
as a single-agent for the treatment of patients with either B-Raf V600E or V600K mutated
ACS Paragon Plus Environment
4

Page 5 of 63
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
21
metastatic melanoma (Figure 2). Other allosteric kinase inhibitors that bind in a pocket adjacent
to the ATP binding site include allosteric Akt inhibitors such as MK-2206. MK-2206 targets a
unique allosteric pocket at the catalytic kinase domain and the regulatory pleckstrin homology
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
(
PH) domain of the inactive conformation of Akt and is currently under clinical investigation in
2
2
numerous Phase I and II oncology trials. Type IV inhibitors bind to a site that is distant from the
ATP binding site, exemplified by GNF-2 that binds the myristoyl pocket of the C-lobe of the kinase
domain of Abl (Figure 2).23-24 An advantage of Type III and Type IV inhibitors is that the
comparatively low sequence homology of allosteric sites provides unique opportunities for more
specific inhibition and minimal off-target pharmacology, for example Type I and Type II Akt
inhibitors tend to exhibit relatively poor kinome selectivity whereas allosteric Akt inhibitor MK-
206 is reported to be more than 100-fold selective over 256 kinases screened.22, 25
2
Figure 2. Allosteric inhibitors Trametinib, MK-2206 and GNF-2
Inspection of an apo TrkA protein crystal structure generated at Pfizer indicated TrkA protein
prefers to adopt a DFG-out autoinhibitory conformation. The DFG-out conformation is stabilized
ACS Paragon Plus Environment
5

Journal of Medicinal Chemistry
Page 6 of 63
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
by edge to face interactions between three phenylalanines: DFG motif Phe669, gatekeeper Phe589
and back pocket Phe575 (Figure 3). Moreover, the presence of these residues serve to illustrate the
uniqueness of TrkA since according to our understanding only five kinases exhibit the FFF motif,
three of which are the Trks. This motif combined with Leu564 from the α-C helix creates a
hydrophobic binding pocket behind the gatekeeper Phe589 that could, in theory, accommodate a
Type III allosteric ligand. Moreover this is a similar allosteric pocket to that formed by the kinases
IGFR-1R and c-Src.26-27 In these cases the kinase adopts a DFG-out conformation revealing a
hydrophobic pocket adjacent to the ATP binding site and Type III ligands have been shown to
bind to these allosteric sites by X-ray crystallography.26-27 Hence, the apo TrkA protein crystal
structure suggested that an allosteric approach utilizing a Type III binder to achieve TrkA
selectivity over TrkB/C was a reasonable proposition.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 3. Apo crystal structure of TrkA protein highlighting a hydrophobic pocket potentially
suitable for an allosteric Type III binder.
ACS Paragon Plus Environment
6

Page 7 of 63
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
The high level of expression of TrkA in cholinergic neurons of the central nervous system
CNS), and the effect of ablation of the TrkA gene that leads to dysfunction in cholinergic circuitry
(
in preclinical species raised the risk of CNS adverse events in the development of TrkA inhibitors
for the treatment of chronic pain.28-30 Furthermore, clinical studies on CE-245677, a pan-Trk/Tie2
kinase inhibitor identified by Pfizer as a potential treatment of certain cancers, identified CNS
side-effects leading to the suspension of Phase I multiple dose trials. Adverse events included
cognitive deficits, personality changes and sleep disturbances that fully resolved upon cessation of
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
31
dosing. Preclinical studies with CE-245677 demonstrated both efficacy in pain models and CNS
side-effects (changes in electroencephalography (EEG), cognitive function, body weight and
hypothalamic mRNA).32 The efficacy of TrkA kinase inhibitors is expected to be driven by target
engagement in peripheral neurons. Therefore, as safety and toleration risks are associated with
CNS Trk receptor occupancy, safety risks can be addressed by restricting TrkA inhibitors to the
peripheral compartment. Experience within Pfizer and the wider pharmaceutical community
suggests that small molecules can be developed that possess acceptable oral bioavailability but are
also peripherally restricted.33-37 This profile can be delivered by designing compounds within
physicochemical space appropriate for absorption across the gastrointestinal epithelium (e.g.
3
8
molecular weight (MW) <500, polar surface area (PSA) <140, <10 rotatable bonds), whilst
combining activity as substrates for the blood-brain barrier (BBB) efflux transporters P-
glycoprotein (P-gp) and Breast Cancer Resistance Protein (BCRP).33-35, 39
ACS Paragon Plus Environment
7

Journal of Medicinal Chemistry
Page 8 of 63
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 4. Schematic illustration of absorption of an orally administered drug across the intestinal
epithelium and its distribution across the brain vascular endothelium that forms the BBB. The
presence of tight junctions between cells in both structures limits the paracellular movement of
drug molecules. C
p
= total plasma concentration, Cp,u = unbound plasma concentration,
C
b,u = unbound brain concentration and Ci,u = unbound intracellular concentration. According to
the free drug hypothesis, unbound concentrations determine the drug pharmacological effect and
at the BBB Cb,u may be less than Cp,u if the drug is a substrate for efflux transporters.
A risk inherent in this strategy derives from the presence of P-gp and BCRP in the apical
membrane of intestinal epithelial cells, the same efflux transporter proteins found at the BBB
(Figure 4). Hence, drugs acting as efflux substrates could also be restricted in transport across the
intestinal epithelium, resulting in poor absorption from the gut lumen with low and variable oral
bioavailability. However, at commonly prescribed oral drug doses of adequately soluble
compounds (10-500 mg), the gut intraluminal drug concentration is likely to be in the range of
hundreds of M. This would produce a high concentration gradient and hence a high driving force
for flux of compound across the epithelium. In addition, saturation of efflux transporters is
ACS Paragon Plus Environment
8

Page 9 of 63
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
expected to occur at such gut luminal concentrations, given that substrates of P-gp typically
possess Km values in the range 1−100 μM. These considerations suggest that in most cases, and
particularly for drugs with high intrinsic permeability, the risk of poor absorption and non-linear
pharmacokinetics of orally delivered efflux substrates may be low.40
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
RESULTS AND DISCUSSION
Hit Identification. We recently reported the discovery of a potent pan-Trk inhibitor 1 as a
development clinical candidate.19, 41 Ligand 1 is an equipotent inhibitor of all three Trk kinases
and was identified as a hit following a high throughput screen (HTS) of the full Pfizer file using
TrkA and TrkB cell based assays.19, 32, 41 It is a Type II inhibitor that binds to the inactive form of
TrkA in a DFG-out conformation (Figure 5A). In order to find TrkA subtype selective hits, the
HTS dataset was reanalyzed to focus on compounds selective for TrkA over TrkB. The selection
criteria were as follows: demonstration of some possibility for TrkA selectivity over TrkB in the
existing dataset (≥ 2-fold or not determined), chemotypes that were not known pan-Trk inhibitors
from the Pfizer pan-Trk program,32, 41 and lack of a common kinase hinge binder. Based on this
analysis a set of ~250 compounds was prioritized for test at TrkA and TrkB in cell based assays at
high concentration of 50µM.41, 43 The data generated suggested that the vast majority of these
compounds were not active at TrkA (TrkA cell IC50 >50µM) or were non-selective over TrkB (<2-
fold). However, a promising hit (2) emerged that was active at TrkA (TrkA cell IC50 3.3µM) and
42
>10-fold selective over TrkB (TrkB cell IC50 >50µM) (Figure 5). The allosteric nature of this hit
was confirmed via X-ray crystallography with a 1.93Å co-crystal structure of 2 with TrkA protein.
Compound 2 was indeed a Type III binder at TrkA with the activation loop and juxtamembrane
(JM) domain contributing to the binding site (Figure 5C). This binding site is similar to that
ACS Paragon Plus Environment
9

Journal of Medicinal Chemistry
Page 10 of 63
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
reported by Furuya and Su.44-45 In this structure TrkA adopts a DFG-out conformation similar to
the apo TrkA structure, and is reminiscent of the allosteric binders to the receptor tyrosine kinase
IGFR-1R. Figure 5 highlights the difference in binding mode between pan-Trk inhibitor 1 and
TrkA selective inhibitor 2. Whereas both compounds bind to the DFG-out conformation, the pan-
Trk ligand 1 interacts with the kinase hinge and extends into the DFG pocket whilst the TrkA
selective ligand 2 does not interact with the hinge and binds exclusively in a hydrophobic allosteric
pocket supported by the JM domain. The ligand bound overlay of 1 and 2 in Figure 5B indicates
an area of common ligand occupancy in the DFG pocket. Figure 6A highlights key protein-ligand
interactions between 2 and TrkA protein. There are hydrogen bonds between the side chain of
Asp668 on the activation loop and NH of 2, and between main chain NHs of Leu486 and Gly485
on the JM domain with the carbonyl oxygen of 2. Lys544 forms a pi-cation interaction with the
unsubstituted aryl ring. Moreover, this aryl unit participates in hydrophobic van der Waals
interactions with Phe589, Leu486, Leu564 and other nearby lipophilic regions (Figures 3 and 6).
A halogen bond may also be identified between the aromatic ortho Cl group in 2 and main chain
carbonyl of His648. The observed TrkA selectivity over TrkB/C can be explained by the allosteric
binding mode because the ligand interacts with the JM domain. The JM domain of TrkA is shorter
than TrkB/C, and TrkA has a number of residue differences in the protein-ligand interacting region
when compared with TrkB/C (Figure 7). Thus the JM conformation presented to ligand 2 by TrkA
is likely to be different to that presented by TrkB/C leading to preferential ligand binding at TrkA.
Moreover, Kania et al. have recently discussed the impact of JM domain conformation on protein-
ligand interaction with the class of VEGFR inhibitors and concluded that the greatest potency and
drug efficiency was garnered from interactions that worked with a natural autoinhibitory state of
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
ACS Paragon Plus Environment
1
0

Page 11 of 63
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
46
the enzyme. In this case the allosteric series exemplified by ligand 2 stabilizes the autoinhibitory
apo state of TrkA. Su and Furuya reported similar binding modes of TrkA inhibitors.44
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 5. A) Co-crystal structure of pan-Trk inhibitor 1 bound to TrkA protein in a DFG-out
conformation; Ligand 1 shown in CPK space filling model; some parts of the G-loop, activation
loop and JM domain are not solved. B) Overlay of the bound conformation in TrkA protein of
DFG-out binder 1 (grey) and allosteric ligand 2 (magenta) indicating a region of common TrkA
binding site occupancy. C) Co-crystal structure of 2 bound to TrkA protein in a DFG-out
conformation that is similar to the apo DFG-out conformation. Ligand 2 binds in an allosteric
pocket created by the activation loop and JM domain; Ligand 2 shown in CPK space filling model.
ACS Paragon Plus Environment
1
1

Journal of Medicinal Chemistry
Page 12 of 63
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
A
B
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 6. A) Co-crystal structure of TrkA selective ligand 2 bound to TrkA protein highlighting
key protein-ligand interactions with white dashed lines with distances given in Å. B) Co-crystal
structure of TrkA selective ligand 2 bound to TrkA protein highlighting locations and
thermodynamic properties of selected water molecules estimated by WaterMap. Oxygen atoms of
selected waters are highlighted as CPK spheres with thermodynamic profiles. Water coloring
scheme represents the stability of the water: unstable waters (high δG) are colored red and stable
waters (low δG) are colored blue.
Figure 7. Sequence alignment of the TrkA, TrkB and TrkC JM domains. The motif interacting
with TrkA selective allosteric ligand 2 is highlighted in red and indicates the putative origin of
TrkA selectivity over TrkB/C.
ACS Paragon Plus Environment
1
2

Page 13 of 63
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Hit to lead optimization. Allosteric hit 2 exhibited the following profile: TrkA cell IC50 3.3µM,
ligand efficiency (LE) 0.33 and lipophilic efficiency (LipE) 2.2 (LipE = -Log(TrkA cell IC50)-
LogD). Whilst this LE provided a reasonable starting point, the potency and LipE required
significant improvement. Moreover the potential for drug action at this allosteric site represented
a key question: will sufficient potency and LipE be achievable in this lipophilic binding site such
that a clinical candidate is a viable proposition? Analysis of the co-crystal structure shown in
Figure 6A determined the optimization strategy for hit 2. A number of ligands that bind to kinases
in the DFG-out conformation interact with the backbone NH of the DFG motif, for example pan-
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Trk ligand 1,19, 41 imatinib and BIRB 796. The corresponding interaction with the main chain
47
48
Asp668 NH could be targeted by building on the meta position of the dichlorophenyl unit of 2
(Figure 6A). The pyrazole methyl group has a vector towards the kinase exit which is solvent
exposed. This vector could be used to append polarity and decrease the lipophilicity of ligand 2
(LogD 3.3). Bound water molecules identified by crystallography were also analysed using
WaterMap (Figure 6B) which maps the locations and thermodynamic properties of water
molecules that solvate protein binding sites.49-50 A number of relatively unstable water molecules
calculated to have an unfavourable free energy and enthalpy relative to solvent water (coloured
red, Figure 6B) were identified in the direction of the kinase hinge. Such water molecules have
been reported to be successfully replaced by lipophilic groups.49-50 A cluster of water molecules
which appeared to be more stable (coloured blue, Figure 6B) could be identified at the pyrazole
methyl consistent with the methyl pointing towards solvent. Based upon the considerations
outlined above, ligand 2 was optimised to lead molecule 3 (Scheme 1). A series of virtual
molecules was designed based on a Suzuki coupling reaction with commercially or internally
available acids (Scheme 2). 11000 Molecules were created and docked against the X-ray structure
ACS Paragon Plus Environment
1
3

Journal of Medicinal Chemistry
Page 14 of 63
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
of 2 (see experimental section). The molecules were then prioritised based on docking score as
well as presence/absence of hydrogen bond with backbone NH of Asp668 in a docking pose. 100
molecules were selected for synthesis and tested in the TrkA and TrkB cellular assays. As a result,
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
3
showed a significant improvement in potency (TrkA cell IC50 50nM) and LipE (5), and
maintained >80-fold TrkA selectivity over TrkB/C in cell based assays (Scheme 1).41, 43 There
was no strong correlation between docking score and potency but it was clear that the hit rate of
the virtual library was enhanced. Note that ligand 3 wherein an aminopyridine group was
th
appended at the meta position of the dichlorophenyl unit of 2 was ranked 15 in the 100 molecule
set for docking score. Docking of 3 suggested that this fragment not only formed a hydrogen bond
with NH in Asp668 using the pyridyl N atom, but also formed a hydrogen bond with the main
chain C=O Asp668 via the amino hydrogen bond donor (Figure 8A-B). Furthermore, the
aminopyridine fragment displaced some of the unstable waters identified in Figure 6B (Figure 8A-
B). The methyl pyrazole was substituted with a primary carboxamide in order to decrease LogD
and interact with the side chain of Arg673 via a hydrogen bond with the carboxamide C=O (Figure
8
6
B). The amide moiety also replaced some of the moderately stable waters identified in Figure
B and interacted with the water network in this region. Lead molecule 3 exhibited a significant
improvement in potency (TrkA cell IC50 50nM) and LipE (5), and maintained >80-fold TrkA
selectivity over TrkB/C in cell based assays (Scheme 1).41, 43 The improvement in potency and
LipE achieved in 3, whilst retaining TrkA subtype selectivity suggested that design of candidate
quality molecules binding to the allosteric site of TrkA was an achievable prospect. Ligand 3 was
screened for protein kinase selectivity at ca. 400 biochemical kinase assays (including TrkA) from
Invitrogen™ at a concentration of 1µM (see supporting information). Pleasingly, 3 exhibited
excellent TrkA selectivity with >80% inhibition of TrkA and <25% inhibition of all other kinases
ACS Paragon Plus Environment
1
4

Page 15 of 63
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
observed, a selectivity profile that can be attributed to the unique nature of the allosteric pocket.51
Furthermore, wide ligand binding profiling in 83 target assays (Cerep™) at a concentration of
10µM to assess off-target liabilities indicated a relatively clean profile (4 hits with % inhibition
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
>50% at 10µM, all of which corresponded with IC50 or EC50 >6µM). Lead molecule 3 was
evaluated by surface plasmon resonance (SPR) binding methods with immobilised TrkA protein
3
8
(
residues 441-796) using Biacore instrumentation. With non-activated unphosphorylated TrkA
7
-1 -1
protein, ligand 3 demonstrated fast association and dissociation kinetics (K
a
1.2 x10 M s ; K
1s-
d
1)
and K (80nM) consistent with TrkA cell potency. When activated phosphorylated TrkA protein
D
was used in the SPR assay there was a substantial reduction in signal from ligand 3 binding,
indicating that allosteric ligands prefer to bind to the non-activated unphosphorylated state of
TrkA. This finding was further corroborated by screening in non-activated TrkA and activated
phosphorylated TrkA biochemical enzyme assays where 3 was approximately 6-fold more potent
against the non-activated TrkA construct (TrkA non-activated enzyme IC50 25nM, TrkA activated
enzyme IC50 150nM) (see supporting information).41, 43 The preference of allosteric ligands for the
inactive state of TrkA protein can be rationalised by consideration of protein conformation. In the
autoinhibitory DFG-out conformation the allosteric binding site can be formed, whereas in the
phosphorylated state the JM domain will move to allow the activation loop to access an active
DFG-in conformation. In the latter case the allosteric binding site is blocked by the activation loop
making the allosteric binding site less readily accessible.46
Further in vitro studies were carried out to understand metabolic liabilities of ligands 2 and 3.
Hit molecule 2 exhibited moderate metabolic turnover in human liver microsomes (HLM) and
human hepatocytes (hHep) (intrinsic clearance (CLint) in HLM 35L/min/mg protein, hHep
48µL/min/million cells, see experimental section). Lead molecule 3 is a unit lower in LogD when
ACS Paragon Plus Environment
1
5

Journal of Medicinal Chemistry
Page 16 of 63
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
compared with 2 resulting in greater metabolic stability in human in vitro systems (HLM
<
8L/min/mg protein, hHep 18µL/min/million cells). However, the presence of an amino group
resulted in turnover in the human in vitro glucuronidation turnover assay (HLMUGT) with CLint
of 29µL/min/mg protein (see experimental section). Such metabolism via conjugation of the NH
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
2
unit was the major pathway, as determined by metabolite identification (data not shown).
Aneugeneticity can be a property presented by kinase inhibitors so ligand 3 was also assessed for
genetic toxicity via an in vitro micronucleus (IVMN) assay.52-53 Ligand 3 induced micronuclei in
Chinese hamster ovary (CHO) cultures, with the presence of kinetochore staining indicating that
the genotoxic activity was aneugenic and not clastogenic in nature (see experimental section).
TrkA inhibitors that would be restricted to the peripheral drug compartment following oral
administration were required so that safety concerns associated with TrkA receptors expressed in
the CNS could be avoided. In order to achieve the desired peripheral restriction (free brain/free
plasma ratio (Cb,u/Cp,u) ≤ 5%) TrkA inhibitors that were also substrates of P-gp and BCRP efflux
transporters expressed at the BBB were preferred.33-34, 39 In the assay format used presently to
assess P-gp and BCRP efflux, P-gp and BCRP substrates are characterized as having an efflux
ratio (ER) >2.5. Whilst hit 2 was not a P-gp substrate, ligand 3 was a substrate of both P-gp and
BCRP which is consistent with the higher MW and PSA of 3 relative to 2 (Scheme 1).33-34, 39 The
CNS penetration of 3 was assessed in vivo in rat and was confirmed to be peripherally restricted
with Cb,u/Cp,u of 0.04. Further oral and i.v. rat PK studies revealed that whilst 3 exhibited relatively
low clearance (CL) in vivo it also had low oral bioavailability (6%) (Table 1). The measured
passive permeability of 3 assessed in the RRCK transcellular flux assay was consistent with good
-6
-1 54-55
absorption (RRCK Papp 17 x10 cms ),
suggesting that low oral bioavailability could have
resulted from pre-systemic metabolism. Incubation of 3 with rat intestinal microsomes in the
ACS Paragon Plus Environment
1
6

Page 17 of 63
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
presence of UDPGA and NADPH revealed high turnover (CLint 175L/min/mg), presumably via
glucuronidation on the amino group (as identified via metabolite identification). Turnover of 3 was
also detectable in human intestinal microsomes (HIM) (CLint 4.5L/min/mg) and in HLM
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
supplemented with UDPGA (HLMUGT, CLint 29L/min/mg). The most effective method of
reducing intestinal metabolism (and hence improving oral bioavailability) whilst maintaining
TrkA potency, was the removal of the amino unit as exemplified by ligands 4-10 (Tables 1 and 2).
N-acyl derivative 4 (TrkA cell IC50 0.034µM, LogD 2.7, LipE 4.8) is likely to have a reduced
glucuronidation liability. However, the acetyl group was chemically unstable and ligand 4
underwent N-deacetylation in buffer systems at pH 7.5. N-acetyl amide stability can be improved
56
by reversing the amide, and in this case reversing the amide to yield 5 improved chemical and
metabolic stability (HLMUGT CLint <1.9µL/min/mg, hHep 3.1µL/min/million cells). However
reverse amide 5 was >1000-fold weaker at TrkA, possibly due to a repulsive interaction between
the C=O of 5 and main chain C=O of Asp668. Introduction of a polarized CH in 6 using a CHF
moiety resulted in a 20-fold decrease in potency at TrkA (TrkA cell IC50 1.1µM) when compared
with 3. Introduction of a lipophilic group on the 3-position of the aminopyridine as in 9 (R =Cl,
2
2
Table 2) improved TrkA potency by 3-fold (TrkA cell IC50 16nM) but also increased the
glucuronidation rate significantly (HLMUGT CLint 209µL/min/mg). Removal of the amino group
in 9 to give 8 reduced glucuronidation to a minimum (HLMUGT CLint <1.9µL/min/mg), and only
caused a moderate 7-fold loss in potency and 0.4 unit reduction in LipE (TrkA cell IC50 110nM,
LipE 4.2). Moreover, 8 gave no measurable turnover in HIM or RIM. Based on the promising in
vitro metabolism profile, 8 was progressed to PK studies in rat using oral and i.v. administration
which indicated low systemic CL (plasma CL 13.4mL/min/kg), and high oral bioavailability (85%)
(Table 1). The corresponding fluorinated derivative 7 had a similar excellent in vivo rat oral/i.v.
ACS Paragon Plus Environment
1
7

Journal of Medicinal Chemistry
Page 18 of 63
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
PK profile (Table 1). The ability of 7 and 8 to act as substrates for P-gp and BCRP was also
assessed. Ligand 7 was not a P-gp substrate but was a moderate BCRP substrate (P-gp ER 2.1,
BCRP ER 4.8) whilst ligand 8 was a moderate P-gp substrate and not a BCRP substrate (P-gp ER
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
3
, BCRP ER 2). In order to understand the level of restriction from the CNS achievable with these
compounds, both 7 and 8 were progressed to CNS penetration studies in rat. Gratifyingly both
compounds were peripherally restricted with Cb,u/Cp,u ~0.1 and provided evidence that good oral
absorption combined with restriction from the CNS was achievable in this chemical series. By
removal of the amino group in 3 oral absorption was improved e.g. in compounds 7-8. However a
hydrogen bond between the amino group and main chain C=O of Asp668 (Figure 8) was sacrificed
resulting in a potency and LipE loss, shown, for example by comparing 8 and 9 in Table 2. In order
to improve the potency and LipE of 7 and 8 the co-crystal structure of 3 was further analyzed.
Figure 8C shows the binding site surface of ligand 3 when bound to TrkA protein, revealing a
buried lipophilic pocket which can be accessed by substitution on the 2-Cl-aryl ring and pyridyl
ring. Moreover hit molecule 2 has a p-Cl group that appears to fill this buried pocket (Figure 6).
The above analysis suggested that increasing van der Waals interactions between ligand and
protein utilizing the buried lipophilic pocket can be used to increase potency and LipE at TrkA.
To this end a Cl group was introduced on the 4-position (R
3
) of the 2-Cl-aryl unit of 3 to give 10
(Table 2). Ligand 10 had a substantial 25-fold improvement in potency and 0.8 unit increase in
LipE relative to 3 (TrkA cell IC50 2nM, LipE 5.8). The corresponding desamino derivative 11
(Table 2) was also potent with a reasonable LipE (TrkA cell IC50 14nM, LipE 4.9). In order to find
the ideal 4-aryl (R ) substituents the co-crystal structure of 3 was used to estimate the optimal
3
dihedral angle between the aryl-pyridyl rings to occupy the lipophilic pocket whilst maintaining
interactions with Asp668 and was identified as approximately 45°. A hypothetical list of 3-pyridyl
ACS Paragon Plus Environment
1
8

Page 19 of 63
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
(
R
2
), 4-aryl (R
3
) substituents was then created for triage e.g. cyclopropyl, F, methyl, phenyl,
cyclohexyl, methoxy, etc. (Table 3). For each potential synthesis target torsion scans were
performed to estimate strain energy of dihedral angle of 45° (dE). Finally the compounds were
docked into TrkA protein to assess how well the substituent occupied the pocket by visual
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
inspection. 3-pyridyl (R
2
), 4-aryl (R
3
) substituents prioritized for synthesis demonstrated the
-1
following properties: dE <1 kcalmol , and good occupancy in the lipophilic pocket without
causing a clash with TrkA protein. By application of these properties as desirable criteria the
compounds 12-25 shown in Table 3 were prepared. All compounds synthesized successfully
retained TrkA potency (TrkA cell IC50 <0.35µM) and where measured, demonstrated >30-fold
TrkA selectivity over TrkB/C in cell based assays (data not shown).43, 57
ACS Paragon Plus Environment
1
9

Journal of Medicinal Chemistry
Page 20 of 63
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
Scheme 1. Optimization of allosteric hit 2 to lead molecule 3
MW 432, PSA 129
MW 346, PSA 47
TrkA cell IC50 3.3µM
TrkA cell IC50 0.050µM
TrkB cell IC50 14µM
TrkC cell IC50 4.1µM
LE 0.34, LipE 5.0
LogD 2.3, RRCK 17
Potency x66
LipE 2.8
CNS restriction
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
TrkB/C cell IC50 >50µM
LE 0.33, LipE 2.2
LogD 3.3, RRCK 30
HLM 35, hHep 48
P-gp ER 1, BCRP ND
3
2
HLM <8, hUGT 29, hHep 18
P-gp ER 5, BCRP ER 7
Peripheral restriction: Cb,u/Cp,u~4%
Table 1. Compounds 3, 7, 8, 23 and 24: pharmacokinetic properties in rat
i.v. administration
plasma CL
p.o. administration
dose
mg/kg)
dose
T½
Vd
Tmax
(hr)
Oral F
(%)
ID
3
Cb,u/Cp,u
(
(mg/kg)
(
hr)
(mL/min/Kg)
(L/Kg)
1
1
1.6
31.5
1.3
0.041
0.10
3
3
3
2
2
0.5
4
6
7
3.2
3.2
2.7
4.1
2.8
0.46
1.1
2.8
1.6
~100
85
8
1
13.4
11.9
11.6
0.076
0.014
0.015
0.9
4
2
2
3
4
0.5
0.5
72
3
46
Compounds were administered via intravenous (i.v.) and oral (p.o.) routes to separate groups of
two male rats. Cb,u/Cp,u was determined separately following i.v. infusion in 4 male rats. There
were no adverse effects of the compounds at the doses used. The vehicle used was 0.5%
methyocellulose 0.5% HPMCAS, 20 mM Tris(hydroxymethyl)aminomethane (pH = 7.4).
Further details of PK and CNS penetration experiments are given in the Experimental Section.
ACS Paragon Plus Environment
2
0

Page 21 of 63
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
A
B
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
A
Figure 8. A) Co-crystal structure of lead
molecule 3 bound to TrkA protein highlighting
key protein-ligand interactions with white
dashed lines with distances given in Å. B) Co-
crystal structure of lead molecule 3 bound to
TrkA protein highlighting interactions with
Arg673 and Asp668 with white dashed lines.
Parts of TrkA protein omitted for clarity C)
Co-crystal structure of lead molecule 3 bound
to TrkA protein highlighting allosteric binding
site surface with brown representing
hydrophobic regions, green representing
neutral polarity regions and blue representing
polar regions.
C
ACS Paragon Plus Environment
2
1

Journal of Medicinal Chemistry
Page 22 of 63
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
Table 2. Pyridyl ring SAR and associated in vitro data
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
TrkA
HLM
UGTe
RRCK P-gp BCRP
ID
R1
R2 R3
LogDb LipEc HLMd
hHepf
g
cell IC50
P
app
ER
ER
a
(
M)
4
5
6
7
8
9
NAc
H
H
H
F
H
H
H
H
H
H
Cl
Cl
0.034
2.7
2.3
3.5
2.6
2.8
3.2
2.9
3.0
4.8
NA
2.5
4.5
4.2
4.6
5.8
4.9
ND
<8
-
-
3.1
12.0
8
10.0
6.1
8.5
19
20
20
11
CONH
2
>50
<1.9
ND
<1.9
<1.9
209
7
ND
ND
4.8
2.0
ND
ND
ND
CHF
2
1.1
11.5
<8
2.1
2.1
3.0
14
H
0.074
0.11
H
Cl
Cl
H
H
<8
13
78
23
17
14
NH
NH
H
2
0.016
0.002
0.014
10.3
19
11
1
1
0
1
2
11
11
<8
<1.9
17
4.2
a
43, 57
TrkA data were generated in PathHunter cells from DiscoveRx with at least 3 tests on 3 different assay runs.
b
c
d
LogD measured at pH 7.4. LipE= -Log(TrkA cell IC50) – LogD. Human liver microsome metabolic CLint in
e
f
mL/min/mg protein. Human liver microsome glucuronidation CLint in µL/min/mg. Human hepatocyte CLint in
µL/min/million cells. Apparent permeability (x 10 cms ) measured with low efflux MDCKII cell line (RRCK). Data
g
-6
-1
55
for LogD, HLM, HLMUGT, hHep, RRCK, P-gp and BCRP were results from at least 2 replicate determinations.
ND denotes not determined.
ACS Paragon Plus Environment
2
2

Page 23 of 63
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Table 3. Summary of lead optimization and associated in vitro data
TrkA cell
IC50 (M)
RRCK P-gp BCRP
ID
R1
H
H
H
H
H
H
R2
H
H
H
H
H
F
R3
LogDb LipEc HLMd hHepe
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
a
f
P
app
ER
ER
1
1
1
1
2
3
4
5
F
0.040
2.7
2.6
2.4
2.7
3.1
2.7
4.7
4.8
4.5
3.8
3.9
5.2
<8
46
11
14
33
12
7.3
36
10.0
12
2.3
ND
CH
3
0.041
0.12
4.6
5.3
5.6
7.5
3.1
ND
ND
ND
ND
ND
OCH
3
9.7
6.2
22
13
OCHF
2
0.34
13
16
0.093
0.012
ND
23
1
1
7
8
F
9.1
H
F
Cl
0.014
0.005
0.030
0.031
2.8
2.8
5.1
5.5
4.5
6.6
<8
<8
<8
9.0
9.0
10
12
13
30
23
41
12
ND
79
1
2
2
9
0
1
Cl
H
H
CF
3
3.0
12
8.5
0.72
ND
ND
Cl
0.94
3.2
2
2
2
3
H
H
Cl
Cl
0.0099
0.010
3.0
3.3
5.0
4.7
15
<8
14
11
16
8
ND
103
4.7
28
2
4
F
F
Cl
F
0.0047
0.023
3.3
3.1
5.0
4.5
<8
<8
3.7
6.1
12
17
23
14
69
25
ND
a
43, 57
TrkA data were generated in PathHunter cells from DiscoveRx with at least 3 tests on 3 different assay runs.
b
c
d
LogD measured at pH 7.4. LipE= -Log(TrkA cell IC50) – LogD. Human liver microsome metabolic CLint in
mL/min/mg protein. Human hepatocyte CLint in µL/min/million cells. Apparent permeability (x 10 cms ) measured
with low efflux MDCKII cell line (RRCK). Data for LogD, HLM, HLMUGT, hHep, RRCK, P-gp and BCRP were
e
f
-6
-1
55
results from at least 2 replicate determinations. ND denotes not determined.
ACS Paragon Plus Environment
2
3

Journal of Medicinal Chemistry
Page 24 of 63
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
When F or methyl groups were introduced at R
3
e.g. 12 and 13, TrkA potency was 2-3-fold
weaker than chloro derivative 11 with a marginally lower LipE (Table 2 and 3). Moreover, the
methyl group in 13 carried a greater metabolic liability (HLM CLint 46L/min/mg protein, hHep
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
3
6L/min/million cells). Methoxy, difluoromethoxy and cyclopropyl substituted compounds 14-
6 were less potent than 11 at TrkA (up to 24-fold weaker), with up to a unit lower LipE, perhaps
1
due to a combination of substituent size, dipole and conformational effects. The CF
was potent at TrkA (TrkA cell IC50 0.03µM) but was less LipE efficient (LipE 4.5) than the
corresponding chloro derivative 18 (LipE 5.1). Introduction of R =F in ligands 12 and 18 to give
3
analogue 20
2
fluorinated compounds 17 and 19 respectively, increased TrkA potency by ca. 3-4 fold and LipE
by ca. 0.4-0.5 units. The F group may occupy part of the binding site highlighted in Figure 8C and
also contribute to increasing the population of desired bioactive conformer in water. TrkA active
compounds can also result from R
of ligands 7 and 8 (Table 2) reveals that the smaller F derivative 7 (TrkA cell IC50 0.074µM, LipE
.5) is more potent and LipE efficient than 8 (TrkA cell IC50 0.11µM, LipE 4.2), which is
consistent with the size of the binding site pocket occupied by R in TrkA (Figure 8C). The R
2
=Cl as exemplified by 8 and 9 (Table 2). However a comparison
4
2
1
substituent had a vector towards the kinase exit and appeared to tolerate a range of substituents
such as alcohols, amines and heterocycles e.g. 18-25 (Table 3). Secondary amides were generally
more potent and LipE efficient than tertiary amides e.g. 5-[[2-chloro-5-(2-
pyridyl)benzoyl]amino]-N,N-dimethyl-1-phenyl-pyrazole-3-carboxamide, the dimethyl amide
derivative of ligand 3 (-C(O)N(CH
potent and had a lower LipE than 3 (TrkA cell IC50 0.05µM, LipE 5.0), possibly because the
additional HBD in 3 interacts with the water network (Figure 6B and 8B). Despite the R
3
2
) , TrkA cell IC50 0.095µM, LipE 4.6) was ca. 2-fold less
1
substituent SAR being relatively tolerant, the key challenge was balancing the properties required
ACS Paragon Plus Environment
2
4

Page 25 of 63
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
for a clinical candidate consistent with a once or twice daily dosing regimen. In particular, potency
-6
-1 54-55
(
TrkA cell IC50 <15nM), apparent permeability (RRCK Papp >5 x 10 cms )
P-gp and BCRP
efflux (ER >2.5) and metabolic stability (hHep CLint ≤5µL/min/million cells) were important
parameters that needed to be balanced within certain ranges to predict with reasonable probability
the desired dose and PK profile in humans, using physiologically-based pharmacokinetic
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
5
8
modelling. For example, although amines such as 21 were potent TrkA inhibitors (Table 3), they
-6
-1
often exhibited poor apparent permeability (RRCK Papp 0.72 x 10 cms ) which could limit oral
absorption.54-55 Oxetane derivative 22 was a potent TrkA inhibitor (TrkA cell IC50 9.9nM) but
appeared to be chemically unstable at low pH ≤2 which made clinical development more
complicated (Table 3). The oxetane instability was postulated to involve ring opening of the
protonated oxetane. The primary alcohols 18-19 appeared promising as they met the desired
potency, apparent permeability, and P-gp and BCRP efflux substrate profile, although the
metabolic stability in hHep was still moderate (hHep CLint~10µL/min/million cells). Several
known metabolic pathways for primary alcohols such as 18-19 involve oxidation or conjugation
of the alcohol moiety. In order to reduce metabolism involving the primary alcohol, tertiary
alcohols 23-25 were prepared. Alcohols 23-24 exhibited higher metabolic stability in hHep (hHep
CLint 3-5µL/min/million cells) than 18-19 (hHep CLint ~10µL/min/million cells) despite being
more lipophilic than 18-19 (δLogD +0.5). This suggested that a major metabolism pathway for 18-
1
9 involved the primary alcohol group and this pathway can be successfully attenuated by
5
9
introduction of the gem-dimethyl group. Ligands 23 and 24 met the desired criteria for potency
-6
-1
(
TrkA cell IC50 = 10 and 4.7nM), permeability (RRCK Papp = 16 and 12 x10 cms ), P-gp (ER =
2
8 and 23) and BCRP (ER = 103 and 69) efflux, and metabolic stability (hHep CLint = 4.7 and
6
3.7L/min/10 cells) respectively. Assuming CL is predicted accurately from metabolism in hHep,
ACS Paragon Plus Environment
2
5

Journal of Medicinal Chemistry
Page 26 of 63
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
systemic CL of 23 and 24 in human is predicted to be 0.59 and 0.74mL/min/kg, respectively
resulting in predicted human PK half-lives of approximately 30 hours in both cases. Moreover
oral/i.v. PK studies in rats indicated that both compounds possess sufficient permeability in the
intestine to achieve adequate absorption via the oral route, in spite of the presence of active efflux,
as oral bioavailability was 72 and 46% for 23 and 24, respectively (Table 1). Moreover, both
compounds displayed marked peripheral restriction that was likely to be the result of efflux via P-
gp and/or BCRP at the BBB because of the much lower unbound systemic exposure of the
compounds than prevails in the gut lumen following oral administration (Table 1). Compound 23
was progressed in preference to 24 due to its slightly higher oral bioavailability in rat and a more
favorable CYP induction profile in vitro (data not shown).60
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 9 shows a 1.94Å co-crystal structure of 23 with TrkA protein. The pyridyl N forms a
hydrogen bond with NH Asp668 as seen with lead molecule 3 (Figure 8). The 2,4-dichloro aryl
unit occupies the lipophilic pocket identified in Figure 8 using the 4-Cl group, with the dihedral
angle between the pyridyl-aryl groups being 45.8°. The terminal amide C=O interacts with the side
chain of Arg673 along with some of the structural waters (2 waters shown in Figure 9) and forms
part of the water network. The tertiary alcohol also interacts with water as it approaches regions
that are more solvent exposed.
ACS Paragon Plus Environment
2
6

Page 27 of 63
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
A
B
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 9. A) Co-crystal structure of candidate molecule 23 bound to TrkA protein highlighting
key protein-ligand interactions with white dashed lines with distances given in Å. B) Co-crystal
structure of candidate molecule 23 bound to TrkA protein highlighting interactions with Arg673
and waters with white dashed lines. Parts of TrkA protein omitted for clarity
Compound 23 was 180-fold selective over TrkB and 70-fold selective over TrkC in cell based
assays (TrkB cell IC50 1.8µM, TrkC cell IC50 0.70µM).41, 43, 57 Ligand 23 was screened for kinase
selectivity in 392 biochemical kinase assays (including TrkA) from Invitrogen™ at a concentration
of 10µM (Figure 10). Pleasingly 23 exhibited exquisite TrkA selectivity with >95% inhibition of
TrkA and <15% inhibition of all other kinases observed. 51, 61 Furthermore, wide ligand profiling
in 84 target assays at Cerep™ at 10µM concentration to assess off-target based liabilities indicated
a clean profile (no hits with inhibition >40% at 10µM and only 2 hits having 30-40% inhibition at
1
0µM). Ligand 23 was studied via SPR using non-activated unphosphorylated TrkA protein and
6
-1 -1
-1
exhibited fast association and dissociation kinetics (K
a
8.2 x10 M s ; K
d
0.03s ) with K 4.2nM
D
giving a residence time of 0.48 minutes. 23 was also screened in non-activated TrkA and activated
phosphorylated TrkA biochemical enzyme assays where it exhibited 18-fold selectivity for the
non-activated TrkA construct (TrkA non-activated enzyme IC50 4nM, TrkA activated enzyme IC50
ACS Paragon Plus Environment
2
7

Journal of Medicinal Chemistry
Page 28 of 63
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
7
4nM).41 These data were consistent with those generated for lead molecule 3. In fact,
retrospective analysis of TrkA cell IC50 data compared with non-activated and activated
phosphorylated TrkA biochemical enzyme assays indicated that for allosteric ligands the potency
in the TrkA cell aligned well with that in the non-activated TrkA enzyme assay but displayed a
rightward shift of up to 15-fold in the activated phosphorylated TrkA enzyme assay (see supporting
information). 41, 43, 57 Genetoxicity assays such as IVMN indicated 23 had a clean profile suitable
for advancement.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 10. Kinase selectivity profile of candidate molecule 23 at a concentration of 10µM when
screened at 392 kinases. Data generated at Invitrogen™, green cross indicates % inhibition <40%,
red cross indicates % inhibition >40%. Only TrkA shows inhibition >40%.
ACS Paragon Plus Environment
2
8

Page 29 of 63
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Compound 23 was shown to inhibit NGF induced phosphorylated ERK formation in native adult
rat DRGs with an IC50 ~57nM and abolished sensitization to capsaicin administration to NGF-
sensitised skin as demonstrated by laser Doppler imaging in vivo in anaesthetised rat (see
supporting information, rat non-activated enzyme IC50 6nM).
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Moreover, 23 was studied in the UVIH model of inflammatory pain in rats whereby its effect on
thermal hyperalgesia resulting from UV burn was assessed by measuring paw withdrawal latency
(Figure 11). Single oral doses of 23 (1.2, 12 and 40 mg/kg) were administered 48 hours after UV
treatment and paw withdrawal latency determined at 1, 3 and 6 hours post dose. Hyperalgesia was
found to be significantly reversed at doses of 23 of 12 and 40 mg/kg at each of the time-points
studied (p<0.01, 2-way ANOVA vs. vehicle). The positive control ibuprofen (100 mg/kg, p.o.)
also reversed thermal hyperalgesia at the 1 and 3 hour time-points. No effect was seen with any
dose on thermal withdrawal response on the contralateral hindpaw (data not shown).
Determination of unbound plasma concentrations of 23 in animals at 6 hours post dose showed
that statistically significant efficacy was observed at unbound plasma concentrations equivalent to
0
.4x (12 mg/kg) and 2.7x (40 mg/kg) TrkA cell IC50 (10nM).
ACS Paragon Plus Environment
2
9