The Discovery of a Potent, Selective, and Peripherally Restricted Pan-Trk Inhibitor (PF-06273340) for the Treatment of Pain

Article abstract of DOI:10.1021/acs.jmedchem.6b00850

The neurotrophin family of growth factors, comprised of nerve growth factor (NGF), brain derived neurotrophic factor (BDNF), neurotrophin 3 (NT3), and neurotrophin 4 (NT4), is implicated in the physiology of chronic pain. Given the clinical efficacy of anti-NGF monoclonal antibody (mAb) therapies, there is significant interest in the development of small molecule modulators of neurotrophin activity. Neurotrophins signal through the tropomyosin related kinase (Trk) family of tyrosine kinase receptors, hence Trk kinase inhibition represents a potentially druggable point of intervention. To deliver the safety profile required for chronic, nonlife threatening pain indications, highly kinase-selective Trk inhibitors with minimal brain availability are sought. Herein we describe how the use of SBDD, 2D QSAR models, and matched molecular pair data in compound design enabled the delivery of the highly potent, kinase-selective, and peripherally restricted clinical candidate PF-06273340.

Full text of DOI:10.1021/acs.jmedchem.6b00850

Subscriber access provided by CORNELL UNIVERSITY LIBRARY
Drug Annotation
The Discovery of a Potent, Selective and Peripherally Restricted
Pan-Trk Inhibitor (PF-06273340) for the Treatment of Pain
Sarah Elizabeth Skerratt, Mark D. Andrews, Sharan K Bagal, James Bilsland, David Brown, Peter J.
Bungay, Susan Cole, Karl R Gibson, Russell Jones, Inaki Morao, Angus Nedderman, Kiyoyuki Omoto,
Colin Robinson, Thomas Ryckmans, Kimberly Skinner, Paul Anthony Stupple, and Gareth Waldron
J. Med. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.jmedchem.6b00850 • Publication Date (Web): 21 Oct 2016
Downloaded from http://pubs.acs.org on October 24, 2016
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 free 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 accessible to all
readers and 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.
Journal of Medicinal Chemistry 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 45
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
The Discovery of a Potent, Selective and Peripherally Restricted
PanꢀTrk Inhibitor (PFꢀ06273340) for the Treatment of Pain
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Sarah E. Skerratt,*a Mark Andrews,b Sharan K. Bagal,a James Bilsland,a David Brown,b Peter J.
Bungay,a Susan Cole,b Karl R Gibson,b Russell Jones,b Inaki Morao,b Angus Nedderman,b
Kiyoyuki Omoto,a Colin Robinson,b Thomas Ryckmans,b Kimberly Skinner,b Paul Stupple,b
Gareth Waldrona
aPfizer Global Research & Development, The Portway Building, Granta Park, Great Abington, Cambridge, CB21
6GS, UK, bPfizer Global Research & Development, Ramsgate Road, Sandwich CT13 9NJ, UK
ABSTRACT
The neurotrophin family of growth factors, comprised of nerve growth factor (NGF), brain
derived neurotrophic factor (BDNF), neurotrophin 3 (NT3) and neurotrophin 4 (NT4), is
implicated in the physiology of chronic pain. Given the clinical efficacy of antiꢀNGF
monoclonal antibody (mAb) therapies, there is significant interest in the development of small
molecule modulators of neurotrophin activity. Neurotrophins signal through the tropomyosin
related kinase (Trk) family of tyrosine kinase receptors, hence Trk kinase inhibition represents a
potentially “druggable” point of intervention. In order to deliver the safety profile required for
chronic, nonꢀlife threatening pain indications, highly kinaseꢀselective Trk inhibitors with
minimal brain availability are sought. Herein we describe how the use of SBDD, 2D QSAR
1
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 2 of 45
1
2
3
4
models and Matched Molecular Pair data in compound design enabled the delivery of the highly
5
6
potent, kinaseꢀselective and peripherally restricted clinical candidate PF-06273340.
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
INTRODUCTION
The neurotrophin family of growth factors, comprised of nerve growth factor (NGF), brain
derived neurotrophic factor (BDNF), neurotrophin 3 (NT3) and neurotrophin 4 (NT4), is
implicated in the physiology of chronic pain. Preclinical and clinical studies have identified a
crucial role for NGF in the pathogenesis of inflammatory pain,1 with neutralizing antibodies
demonstrating efficacy in preclinical pain models and in clinical trials for osteoarthritis, chronic
lower back pain and interstitial cystitis.2ꢀ5 BDNF has also been implicated in the pathogenesis of
chronic pain, and has been shown to be upregulated in clinical pain states.6 Given the role of
NGF and BDNF in modulating pain, and the clinical efficacy of antiꢀNGF monoclonal antibody
(mAb) therapies, there is significant interest in the development of small molecule oral
therapeutics to complement existing treatment options.
Neurotrophins signal through the tropomyosin related kinase (Trk) family of tyrosine kinase
receptors, whereby NGF signals preferentially through TrkA, BDNF and NTꢀ4 through TrkB,
and NTꢀ3 through TrkC. All neurotrophins interact with equal affinity at the p75 receptor, a
member of the TNFR superfamily. An alternative point of intervention, high in the neurotrophin
signaling cascade and amenable to small molecule therapy, could therefore be the inhibition of
intracellular Trk kinases (Figure 1).7ꢀ8
2
ACS Paragon Plus Environment

Page 3 of 45
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 1. Neurotrophins NGF, BDNF, NTꢀ3, NTꢀ4/5 and their receptors TrkA, TrkB and TrkC.
A key concern in the development of Trk kinase inhibitors for the treatment of chronic pain is
the potential risk of central nervous system (CNS) adverse events. TrkB is expressed throughout
the CNS with the BDNF/TrkB axis involved in excitatory signaling, longꢀterm potentiation and
feeding behavior.9ꢀ12 TrkA is highly expressed in the cholinergic neurons of the basal forebrain,
with TrkA gene ablation leading to dysfunction in cholinergic circuitry in preclinical species.13ꢀ15
In addition, clinical CNS sideꢀeffects have been noted with 1ꢀ(5ꢀ(4ꢀaminoꢀ7ꢀisopropylꢀ7Hꢀ
pyrrolo[2,3ꢀd]pyrimidineꢀ5ꢀcarbonyl)ꢀ2ꢀmethoxyphenyl)ꢀ3ꢀ(2,4ꢀdichlorophenyl)
urea
(CEꢀ
245677), an oral panꢀTrk/Tie2 kinase inhibitor identified by Pfizer, previously in development
for the treatment of certain cancers.16 Phase I multiple dose trials were stopped due to the
development of significant CNS adverse events which included cognitive deficits, personality
changes and sleep disturbances. These effects fully resolved upon cessation of dosing.
Preclinical studies with this compound have shown it to be efficacious in multiple pain models.
However, at doses associated with significant (> 50%) CNS Trk receptor occupancy, changes in
cortical electroencephalography (EEG), cognitive function, body weight and hypothalamic
mRNA were observed.17 These findings were recapitulated with panꢀTrk inhibitors devoid of
Tieꢀ2 kinase activity (data not shown). These data confirm the potential of small molecule Trk
3
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 4 of 45
1
2
3
4
5
6
kinase inhibitors as pain therapeutics, but highlight the risk of onꢀtarget adverse events resulting
from pharmacologically relevant exposure in the CNS.
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Given the therapeutic efficacy of brain restricted antiꢀNGF mAb therapies, and the known safety
risks associated with CNS Trk receptor occupancy, a rational approach to delivering a wellꢀ
tolerated Trk kinase therapeutic is to restrict compounds to the peripheral compartment. It is
known from projects within Pfizer and the wider pharmaceutical community that it is possible to
deliver small molecules that are orally bioavailable yet peripherally restricted.18ꢀ22 The design of
compounds in physicochemical space appropriate for absorption across the gastrointestinal (GI)
epithelium (e.g. molecular weight < 500, polar surface area < 140, < 10 rotatable bonds)23 but
that are substrates for bloodꢀbrain barrier (BBB) efflux transporters such as Pꢀglycoprotein (Pꢀ
gp) and Breast Cancer Resistance Protein (BCRP) is an emerging strategy to deliver this type of
profile. According to the free drug hypothesis, unbound concentration determines a drug’s
pharmacological effect. If a drug is a substrate for BBB efflux transporters, the unbound brain
concentration (Cb,u) may be less than the unbound plasma concentration (Cp,u). An example of a
compound that exhibits significant peripheral restriction (via targeting BBB efflux transporters
Pꢀgp and BCRP) whilst retaining good oral bioavailability, cell penetration and pharmacological
activity is the tyrosine kinase inhibitor imatinib.24 Imatinib was not specifically designed to be
peripherally restricted, yet discoveries such as this engender knowledge of how to prospectively
design orally bioavailable, peripherally restricted molecules in a rational manner.
Targeting transporters in order to achieve peripheral restriction can introduce risk factors.20 For
example, the apical membrane of intestinal epithelial cells contains the same efflux transporter
4
ACS Paragon Plus Environment

Page 5 of 45
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
proteins Pꢀgp and BCRP found at the BBB. Drugs that are efflux substrates could therefore be
restricted in their passage across the intestinal epithelium resulting in poor absorption from the
gut lumen (Figure 2). Set against this risk (at commonly prescribed oral drug doses of
adequately soluble compounds (10 ꢀ 500 mg)) the intraluminal compound concentration is likely
to be in the range of hundreds of micromolar providing a high concentration gradient which, for
a compound with sufficient permeability, will mean a high driving force for flux across the
epithelium. In addition such gut luminal concentrations are likely to be high enough to saturate
efflux transporters, given that substrates of Pꢀgp typically possess Km values in the range 1 −
100 ꢁM. These considerations suggest that in most cases, the risk of poor absorption and nonꢀ
linear pharmacokinetics of orally delivered efflux substrates will be low.25
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 2. Schematic illustration of orally administered drug absorption across the intestinal epithelium and
distribution across the brain vascular endothelium. Cp,u = unbound plasma concentration, Cb,u = unbound brain
concentration and Ci,u = unbound intracellular concentration.
An additional requirement for a small molecule Trk kinase inhibitor, suitable for chronic
administration in a nonꢀlife threatening indication, is exquisite kinase selectivity. Kinases
modulate numerous physiological functions, and hence inhibiting offꢀtarget kinases can lead to
5
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 6 of 45
1
2
unwanted safety side effects.26ꢀ27 The majority of kinase inhibitors bind to the highly conserved
ATP binding site (Type I inhibitors), and hence can be nonꢀselective. A strategy that has
emerged to design kinase inhibitors with enhanced kinase selectivity is to target proteinꢀligand
interactions in less conserved areas of the ligand binding domain, outside of the ATP binding
pocket.28ꢀ31 Imatinib, a Type II kinase inhibitor, is a wellꢀestablished example of this whereby it
binds to the inactive form of the Abl protein in which the Nꢀterminal region of the activation
loop adopts a AspꢀPheꢀGlyꢀout (DFGꢀout) conformation (see supplementary information S2).
Since the deep pocket region accessible in a DFGꢀout conformation has a relatively low
sequence conservation compared to other kinases, imatinib has been shown to be a relatively
selective kinase inhibitor.30ꢀ32 Indeed, when Karaman et al. generated KINOMEscan selectivity
scores for a selection of marketed kinase inhibitors, they demonstrated the potential for Type II
inhibitors such as imatinib, sorafenib and lapatinib to display higher selectivity than Type I
inhibitors such as sunitinib and dasatinib, especially when only higher affinity offꢀtarget
interactions were considered.33ꢀ34 Whilst kinase inhibitors adopting a DFGꢀout conformation are
certainly not guaranteed to be highly kinomeꢀselective,35ꢀ36 this binding mode offers the
opportunity to access an allosteric binding pocket adjacent to the ATPꢀbinding site and target
unique inactive conformations.37
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Herein we describe the discovery of a highly potent, kinaseꢀselective, peripherally restricted,
efficacious and wellꢀtolerated series of “DFGꢀout” panꢀTrk inhibitors. The use of SBDD, 2D
QSAR models and Matched Molecular Pair data in prospective design enabled rapid
optimization of the lead chemical series and delivery of clinical candidate 14 (PF-06273340).38
6
ACS Paragon Plus Environment

Page 7 of 45
Journal of Medicinal Chemistry
1
2
3
4
RESULTS AND DISCUSSION
5
6
Hit identification and characterization
7
8
9
A high throughput screen of the Pfizer compound collection was conducted using TrkA and
TrkB cellꢀbased assays.38 Of the numerous chemical series of panꢀTrk inhibitors identified,
prioritization for lead optimization followꢀup was based on criteria that included LipE,23
predicted levels of kinase promiscuity,39 and predicted ATPꢀsite binding mode. Chemical series
with an anticipated DFGꢀout binding mode were prioritized as it is known that kinase inhibitors
that bind to less conserved areas of the ATP binding site can be more selective within the
kinome. Indeed, when a systematic analysis of kinase structural and kinase panel screening data
was undertaken within Pfizer, it was shown that ligands that contact ATPꢀsite residues
approaching the “back pocket” of the ATP site have a higher chance of being kinaseꢀselective.40
Assessment of compound binding mode was made via docking into a TrkA homology model
built on a set of inꢀhouse cFMS crystal structures.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Several chemical series were identified that demonstrated moderate to good LipE profiles, were
calculated to have high levels of kinase selectivity and predicted to adopt a DFGꢀout binding
mode. One of the series selected for lead optimization was the pyrrolopyrimidine urea series.41
An example from this series, 1,41 is highlighted in Figure 3 and Table 1. Compound 1 is a potent
inhibitor of TrkA, TrkB and TrkC in recombinant cellular assays.38 Trk isoform selectivity is not
anticipated within this series due to the highly homologous nature of the TrkA, B and C DFGꢀout
ligand binding sites.42 The overall kinase selectivity profile of 1 is encouraging (Figure 12), with
only 5 kinases (including TrkA) in a panel of 39 biochemical kinase assays from Invitrogen
showing > 50 percent inhibition (% I) at a compound concentration of 1 µM. The kinase
7
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 8 of 45
1
2
3
4
5
6
7
8
9
selectivity profile of 1 is represented as a Gini coefficient which describes the selectivity of a
compound by a single number between 0 and 1. Nonselective inhibitors are characterized by
Gini values close to zero (e.g. Staurosporine, Gini 0.15), whereas compounds with complete
selectivity exhibit Gini values of 1.43 Compound 1 has a Gini value of 0.59, indicative of its
moderate kinase selectivity, representing an encouraging startꢀpoint for further optimization.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Compound 1 was predicted to adopt a DFGꢀout binding pose in the TrkA homology model,
which was subsequently confirmed by Xꢀray crystallography studies. Figure 3 shows the crystal
structure of 1 bound to the TrkA kinase, and highlights the key proteinꢀligand interactions. The
4ꢀamino pyrrolopyrimidine motif of 1 makes a twoꢀpoint hydrogen bonding interaction with
backbone polar atoms of hinge residues Glu590 and Met592, whilst the central pyridyl group of
1 makes a πꢀstacking interaction with the gatekeeper (Phe589), and Phe669 of the DFG triad.
The carbonyl oxygen of the urea group of 1 makes a hydrogen bond interaction with the
backbone NꢀH of Asp 668 from the DFG triad. The NꢀH groups of the urea group of 1 make
hydrogen bond interactions with a conserved water molecule, which further interacts with
catalytic lysine (Lys544) and Glu560 from the alphaꢀC helix. The lipophilic “DFGꢀout” back
pocket accommodates the 1,3ꢀdifluorophenyl group.
8
ACS Paragon Plus Environment

Page 9 of 45
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
Figure 3: Crystal structure of 1 (green) bound to TrkA (pink). The hinge region of the TrkA ATP site is highlighted
in blue. The “DFG” triad of the activation loop of TrkA is highlighted in orange. Key proteinꢀligand binding
interactions are highlighted.
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Although potent, the lipophilic efficiency of 1 is moderate (LipE 3.4) due to the highly lipophilic
nature of the molecule (logD 4.6). The metabolic stability of 1 in human liver microsomes
(HLM) and human hepatocytes (hHep) is encouraging, and may be high enough to deliver low
metabolic clearance (CL) in humans. Also encouraging is the lack of blockade of the human
etherꢀaꢀgoꢀgo related gene (hERG) encoded potassium channel, as demonstrated by the lack of
signal in the dofetilide binding assay.44ꢀ45 Compound 1 is a weak substrate for the efflux
transporter Pꢀgp (efflux ratio (ER) in the MDCKꢀMDR1 assay = 2.1), is poorly soluble (< 0.3
µM at pH 6.5), and has low passive permeability (RRCK A to B apparent permeability (Papp) =
0.6 x 10ꢀ6 cmsꢀ1), such that limitations on oral bioavailability may be expected in the absence of
formulation enhancement.
Hit to lead: Improving solubility
After identification of the pyrrolopyrimidine series (e.g. 1), an initial round of synthesis aimed at
improving the aqueous solubility profile was undertaken. It is known that ureaꢀcontaining
compounds often have a poor solubility profile, and a commonly used strategy to try and
improve solubility is to replace the urea group with an amide unit. Indeed, a Matched Molecular
Pair (MMP) analysis of the Pfizer Pairwise Database46 showed that conversion of an urea to an
amide group (R1ꢀNHCONHꢀR2 → R1ꢀNHCOCH2ꢀR2) increased solubility (> 2ꢀfold) in 52% of
cases, maintained solubility within 2ꢀfold in 39% of cases and decreased solubility (> 2ꢀfold
decrease) in only 9% of cases, with a mean increase in solubility of 137 µM (no. of pairs = 56).47
9
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 10 of 45
1
2
3
4
5
6
7
8
9
Replacing the urea group for an amide unit in the pyrrolopyrimidine series maintained TrkA
potency and on average facilitated a > 30 fold increase in solubility. In addition, the passive
permeability of the amideꢀcontaining analogues was superior. An example, 2 (the direct
matched pair of 1) is highlighted in Table 1. Due to the superior solubility and permeability
profile, all further ligand optimization was conducted using the amide linker group.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Hit to lead: Improving kinase selectivity
Whilst the initial kinase selectivity profile of the DFGꢀout pyrrolopyrimidine series was
promising, an early design objective was to further improve the kinase selectivity to optimize the
overall safety profile. Kinase inhibitors generally form one to three hydrogen bonds with
backbone polar atoms of the kinase hinge region that connects the Nꢀ and Cꢀterminal lobes of the
catalytic domain, with an increasing number of hydrogen bonds thought to negatively affect
overall kinase selectivity.48 Indeed, when a systematic analysis of kinase scaffolds extracted
from the Pfizer crystal structure database (CSDb) was performed and combined with kinase
biological screening data, the authors demonstrated that more hydrogen bonding interactions at
the hinge region do not necessarily afford increased potency, and moreover may have deleterious
effects on kinome selectivity.49 A strategy to optimise kinase selectivity could therefore be to
minimise the number of hydrogen bonding interactions at the hinge region. Compound 3 is a
representative amino pyrrolopyrimidine from the initial round of synthesis described above in
which the urea to amide linker transformation was investigated (Table 1). Compound 3 is a
potent inhibitor of TrkA, TrkB and TrkC, with a moderate lipophilic efficiency profile (LipE
4.0). The metabolic stability of 3 in HLM and hHep is moderate and the hERG activity modest.
Compound 3 is a relatively good substrate for the efflux transporter Pꢀgp (MDCKꢀMDR1 ER =
10
ACS Paragon Plus Environment

Page 11 of 45
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
6.2), is a weak substrate for the efflux transporter BCRP, and has moderate passive permeability
(RRCK A to B Papp = 15.3). The aqueous solubility is 13.2 µM. The kinase selectivity profile of
3 is promising, with only 7 kinases (including TrkA) in a panel of 39 biochemical kinase assays
from Invitrogen showing > 50 % inhibition at a compound concentration of 1 µM, generating a
Gini value of 0.59 (Figure 12).
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
In order to design compounds with the minimal number of hydrogen bond contacts at the hinge
region, an analysis of the crystal structure of 3 bound to TrkA and the nonꢀliganded apo structure
was undertaken. The crystal structure of 3 bound to TrkA is shown in Figure 4. Compound 3
adopts a DFGꢀout binding mode, with the amino pyrrolopyrimidine motif making a twoꢀpoint
hydrogen bonding interaction with hinge residues Glu590 and Met592.
When 3 is
superpositioned onto the TrkA apo crystal structure, the ꢀNH2 motif of the amino
pyrrolopyrimidine group overlays directly with a conserved water molecule observed in the apo
structure (Figure 5). It was hoped that removal of the –NH2 motif as highlighted in Figure 5
would deliver a ligand that makes a single hydrogen bonding interaction with the backbone NꢀH
of hinge residue Met592, with a water molecule replacing the –NH2 group, satisfying the
backbone carbonyl oxygen at Glu590, and with the overall effect of improving kinase selectivity.
11
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 12 of 45
1
2
3
4
5
6
7
8
Figure 4: Crystal structure of 3 (green) bound to TrkA (pink). The hinge region of the TrkA ATP site is highlighted
in blue. The “DFG” triad of the activation loop of TrkA is highlighted in orange. Key binding interactions are
highlighted.
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 5: Superposition of 3 (green) onto the TrkA apo crystal structure (pink), all water molecules included. The
hinge region of the TrkA ATP apo site is highlighted in blue. The “DFG” triad of the activation loop of TrkA is
highlighted in orange. The red circle highlights the ꢀNH2 motif of 3 that overlays directly onto a conserved water
molecule observed in the apo structure
Deleting –NH2 at the 4ꢀposition affords ligands with comparable TrkA activity, and a much
improved kinase selectivity profile. An example, 4,38 is highlighted in Table 1. Compound 4 is
a potent panꢀTrk inhibitor with moderate LipE and moderate turnover in HLM and hHep.
Removal of the amino group from the pyrrolopyrimidine unit does not affect the propensity of
the chemotype to be a substrate for Pꢀgp, with 4 having an ER in the MDCKꢀMDR1 assay of 8.2.
As with previous pyrrolopyrimidine amide analogues, the passive permeability of 4 is moderate
(RRCK Papp = 9.9) and the aqueous solubility relatively low (17.5 µM). A significant difference
between 4 and its amino pyrrolopyrimidine analogue 3 is in the kinase selectivity profile.
Whereas 3 strongly inhibits (> 50% I) 7 kinases (including TrkA) in an Invitrogen panel of 39
biochemical kinase assays (Gini score 0.59), 4 is much more selective (Gini score of 0.82),
Figure 12.
12
ACS Paragon Plus Environment

Page 13 of 45
Journal of Medicinal Chemistry
1
2
3
4
The TrkA coꢀcrystal structure of desꢀamino pyrrolopyrimidine 5,38 a closeꢀin analogue of 4 in
which the pꢀCl phenyl group is replaced with a pꢀCN phenyl, is shown in Figure 6. Compound 5
adopts a DFGꢀout binding mode, and in general makes similar proteinꢀligand binding
interactions as amino pyrrolopyrimidine 3 (Figure 4). A notable difference is the interaction
network around the hinge region. The amino pyrrolopyrimidine motif of 3 makes a twoꢀpoint
hydrogen bonding interaction with backbone polar atoms of hinge residues Glu590 and Met592.
As Figure 6 highlights, the hydrogen bond interaction between the pyrimidyl nitrogen of 5 and
the backbone NꢀH of Met592 is maintained, but a conserved water molecule replaces the –NH2
group to make a hydrogen bond interaction that bridges the main chain carbonyl Glu590 and the
ketone oxygen atom of 5.
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 6: Crystal structure of 5 (green) bound to TrkA (pink). The hinge region of the TrkA ATP site is highlighted
in blue. The “DFG” triad of the activation loop of TrkA is highlighted in orange. Key proteinꢀligand binding
interactions are outlined. The red circle highlights the hydrogen bond interactions made by a conserved water
molecule bridging the main chain carbonyl Glu590 and the ketone oxygen atom of desꢀamino pyrrolopyrimidine 5.
The presence of a bridging water between the main chain carbonyl oxygen of TrkA Glu590 (or
the equivalent across other kinases) and a polar atom of a bound ligand is a relatively rare
occurrence. The available kinase crystal structures from the Protein Data Bank (PDB) and the
13
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 14 of 45
1
2
3
4
5
6
7
8
9
Pfizer crystal structure database (CSDb) were searched to identify similar waterꢀbridging
examples using Interaction Miner1; an inꢀhouse tool to search structural databases for specific
interactions and structural motifs.50 This search retrieved < 10 examples from other kinases
within the RMSD threshold of < 1.0 Angstrom to the matching atoms. Hence, as replacing the
direct Hꢀbond with a “through water” Hꢀbond maintained Trk activity and improved kinase
selectivity, followꢀup Trk analogues within pyrrolopyrimidine series were devoid of an –NH2
group at the 4ꢀposition.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Compound
1
2
3
4
Trk A/B/C IC50 2/5/4
(nM)
3/ND/ND
6/2/1
12/4/3
MW/logD
LipE
451/4.6
450/4.3
4.2
449/4.0
4.3
434/3.7
4.2
4.1
18
HLM
23
27
22
(µL/min/mg)
hHeps
12
ND
12
37
(µL/min/mill)
Dof Ki/hERG > 40/ND
IC50 (µM)
8.3/ND
10.8/7.8
5.2/ND
1
Interaction Miner is a Pfizer tool to search structural databases for specific interactions and structural motifs. Briefly, Interaction Miner
identifies matching interactions or structural features based on userꢀdefined atom and bond match criteria as well as geometric constraints. The
matching database structures are superposed to the query using RMSD overlay of the matching substructure, enabling users to visually analyze
database hits in the context of the query. The RMSD threshold is exposed as an additional filtering criterion to restrict the database hits based on
their geometric similarity to the query substructure.
14
ACS Paragon Plus Environment

Page 15 of 45
Journal of Medicinal Chemistry
1
2
3
4
5
6
Pꢀgp/ BCRP ER 2.1
10.9
24.1
6.2/1.2
15.3
8.2/1.5
9.9
RRCK AꢀB Papp 0.6
(x 10ꢀ6 cmsꢀ1)
7
8
9
Solubility (µM)
Gini score
< 0.3
0.59
13.6
ND
13.2
0.59
17.5
0.82
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Table 1: Structure and properties of compounds 1 to 4. ND = not determined
Peripheral restriction: Building an in vitro to in vivo correlation
The initial rounds of compound design (as discussed in the previous section) generated a number
of potent and selective panꢀTrk inhibitors that displayed a range of ER in the MDCKꢀMDR1
assay, demonstrating a spectrum of ability to act as substrates for the efflux transporter Pꢀgp. In
order to build an understanding of the in vitro Pꢀgp activity required to deliver significant
peripheral restriction in vivo, brain distribution experiments were conducted in rats. Compounds
with a Pꢀgp ER in the range of 1 ꢀ 50 were administered orally at a dose of 100 mg/kg and
terminal plasma and brain samples were taken 3 hours after dosing. Compound concentrations
in plasma and homogenized brain samples were quantified in parallel with assessment of binding
of compounds to plasma and brain tissue, such that unbound concentration ratios (Cb,u /Cp,u)
could be determined.51 The data from the brain penetration experiments are shown in Figure 7A.
Increasing Pꢀgp ER results in decreased brain penetration, such that compounds with a Pꢀgp ER
> 10 are significantly excluded from the CNS (Cb,u /Cp,u < 0.05). In addition, the occupancies of
TrkB and TrkC receptors in brain samples were determined using the chemical proteomic
technology developed by ActivX Biosciences. This approach utilizes biotinylated acyl
phosphates of ADP or ATP that transfer biotin to conserved lysine residues within the ATP
binding site of kinases. Binding of a test compound reduces binding of the probes to individual
kinases, which when combined with kinase proteomic analysis, allows receptor occupancy and
15
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 16 of 45
1
2
broad kinase selectivity in complex tissue lysates to be determined.52,53 Figure 7B shows the
kinase receptor occupancy profile of 3 and 6 in rat brain lysates following oral administration of
compound as determined using the ActivX biotinylated acyl phosphate ADP probe. The xꢀaxis
of Figure 7B represents the individual kinases detected in the rat brain lysate panel using the
ActivX ADP probe, the yꢀaxis represents the fold change in kinase receptor occupancy relative
to vehicleꢀtreated animals. Compound 6 (shown as a red circle in Figure 7B) has a high Pꢀgp
ER, is significantly restricted from the CNS (Pꢀgp ER = 27.0, Cb,u /Cp,u = 0.013) and has no
detectable TrkB or TrkC occupancy in brain lysates. Compound 3 (shown as a blue circle in
Figure 7B) has a lower Pꢀgp ER which translates into higher brain availability (Pꢀgp ER = 6.2,
Cb,u /Cp,u = 0.22). Compound 3 causes marked blockade of binding of the chemical proteomic
probe to TrkB and TrkC within the CNS (TrkA not detected due to low expression), but does not
display significant occupancy of any of the other kinases in the brain lysate panel, demonstrating
the high selectivity of this ligand for TrkB and TrkC over other kinases.
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
16
ACS Paragon Plus Environment

Page 17 of 45
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
Figure 7A: Relationship of Pꢀgp ER in the MDCKꢀMDR1 assay vs. rat Cb,u /Cp,u for a range of pan Trk
compounds from the pyrrolopyrimidine series. Figure 7B: Kinase receptor occupancy profile of 3 and 6 in rat
brain lysates determined using the ActivX biotinylated acyl phosphate ADP probe.
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Peripheral restriction: Use of 2D QSAR models to guide design
As highlighted by the rat brain availability data in Figures 7A and 7B, a compound with a
high Pꢀgp ER should be significantly restricted from the CNS. In order to guide the
prospective design of compounds with a high Pꢀgp ER on the Trk project, a 2D QSAR
regression model built on the Pfizer Pꢀgp dataset was utilised.54 The model predicts a Pꢀgp
ER value for proposed compounds, along with a confidence value that incorporates both the
chemical similarity and activity landscape of a test compound’s neighbourhood, calibrating
the value to a probability that the prediction will be within 2ꢀfold of the actual experimental
result. To assess the performance of the QSAR model and confidence metric, prospective
predictions of Pꢀgp ER over time can be evaluated by comparing the model predictions made
prior to testing of the compound vs. the actual experimental Pꢀgp data generated once tested.
The result of this analysis is shown in Figure 8. The yꢀaxis of the trellised pieꢀchart in Figure
8 shows the binned Pꢀgp ER predictions (low predicted efflux; Pꢀgp ER < 5, high predicted
efflux; Pꢀgp ER > 5) and the xꢀaxis shows the confidence values (0.0 to 1.0) of the Pꢀgp ER
model predictions. A cutꢀoff of Pꢀgp ER > 5 (cf. the stricter cutꢀoff of Pꢀgp ER > 10) was
chosen so as not to omit potentially interesting molecules from consideration given that the
model does not have perfect predictivity. The pies are coloured by real experimental Pꢀgp
ER data (cerise: Pꢀgp ER < 5, blue: Pꢀgp ER > 5). As highlighted in Figure 8, as the
confidence in the model prediction increases, the probability of achieving the predicted Pꢀgp
ER outcome increases. In the case of panꢀTrk, the design strategy was to prioritise
compounds predicted to have high efflux (with high model confidence) for synthesis,
disregard compounds predicted to have low efflux (with high model confidence), and more
17
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 18 of 45
1
2
3
4
5
6
7
8
9
sparingly synthesise compounds in the space for which the Pꢀgp ER QSAR model confidence
was not high. Utilising TrkA coꢀcrystal structure information alongside the 2D QSAR Pꢀgp
ER models to prioritise design ideas enabled an efficient triage process in which 69% of
compounds in the Hit to Lead phase met TrkA potency (< 100 nM) and Pꢀgp ER (ER > 5)
criteria.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 8: Binned 2D QSAR Pꢀgp ER prediction data vs. Binned confidence in 2D QSAR Pꢀgp ER prediction.
Pies are coloured by real experimental Pꢀgp ER data (cerise: Pꢀgp ER < 5, blue: Pꢀgp ER > 5).
Lead optimization
In addition to utilizing 2D Pꢀgp QSAR model data (vide supra) to guide design of compounds
with a high Pꢀgp ER, design efforts in the lead optimization phase focused on increasing
compound metabolic stability and lipophilic efficiency. In order to increase metabolic
stability, in vitro metabolite identification (MetID) data were used to identify metabolic “soft
spots”. Metabolite profiling over a range of pyrrolopyrimidine analogues in human liver S9
fractions (in the absence of the CYP cofactor NADP) identified Aldehyde Oxidase (AO) as a
likely source of metabolic oxidation. The metabolic liabilities of representative desꢀamino
pyrrolopyrimidines and amino pyrrolopyrimidines 5 (HLM < 9.8 µl/min/mg protein, hHep
13.0 µl/min/106 cells) and 6 (HLM 21.6 µl/min/mg protein, hHep 6.6 µl/min/106 cells) are
18
ACS Paragon Plus Environment

Page 19 of 45
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
highlighted in Figure 9 (full HPLCꢀUV chromatograms are available in supplementary
information S3 and S4). Amino pyrrolopyrimidine 6 has a small amount of oxidation on the
amino pyrrolopyrimidine hinge binding group and the central pyridine substituent.
Conversely, significant metabolism is observed at multiple sites of desꢀamino
pyrrolopyrimidine 5, the major sites being the pyrrolopyrimidine hinge binding group and the
central pyridine substituent. A very minor product of amide hydrolysis was also observed.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Incubation
of
structurally
similar
analogue
(R)ꢀ2ꢀ(4ꢀcyanophenyl)ꢀNꢀ(4ꢀ(7ꢀ(1ꢀ
hydroxypropanꢀ2ꢀyl)ꢀ7Hꢀpyrrolo[2,3ꢀd]pyrimidineꢀ5ꢀcarbonyl)pyridinꢀ2ꢀyl)acetamide (PFꢀ
05302191)55 in human liver S9 (no additional NADP) showed formation of a monoꢀoxidised
metabolite (see supplementary information S5). In the presence of 50 µM menadione (an AO
inhibitor), a dramatic reduction in the formation of this metabolite was observed, whereas
preincubation with 50 µM allopurinol (XO inhibitor) showed no change in metabolite profile.
These observations would suggest that metabolism in human liver S9 fractions in this series
is most likely due to AO.
Figure 9: Sites of AOꢀmediated oxidation of desꢀamino pyrrolopyrimidines 5 and amino pyrrolopyrimidine 6.
Over the last few years, AO has been increasingly recognized as playing an important role in
drug metabolism, with reported examples in which metabolism by this enzyme has had
significant clinical impact or led to the termination of a drug development program.56 In drug
discovery, a method of predicting CL parameters in humans with high confidence is highly
desirable to make accurate dose predictions and, whilst methods of scaling in vitro intrinsic
CL data to predict the in vivo metabolic CL of compounds are wellꢀestablished for CYP
19
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 20 of 45
1
2
3
4
5
6
7
8
9
P450ꢀdominated CL, similar methods have not yet been established for drugs metabolized by
AO. As a result, the impact of having AOꢀmediated metabolism cannot be fully established
until the compound has reached the clinic. Numerous medicinal chemistry design strategies
aimed at avoiding AO metabolism have been reported.57 As highlighted in Figure 9, the AOꢀ
mediated oxidation of the desꢀamino pyrrolopyrimidine group is more significant than for the
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
4
ꢀamino pyrolopyrimidine motif. However, as the desꢀamino pyrrolopyrimidine confers an
enhanced kinase selectivity profile, strategies to minimise AO oxidation on the more kinaseꢀ
selective desꢀamino pyrrolopyrimidine template were investigated. The first step in an AOꢀ
mediated oxidation involves a nucleophilic attack on a heterocyclic ring system by the
molybdenum pyranopterin cofactor MoCo,58 hence an initial strategy to minimize AO
liability on the desꢀamino pyrrolopyrimidine template was to increase the electron density of
that group by replacing the pyrimidine with a pyridyl motif. Gratifyingly, the pyrimidyl to
pyridyl transformation generated compounds that maintained potent Trk activity, high kinase
selectivity and were devoid of AOꢀmediated oxidation on the pyridylꢀcontaining hinge
binding moiety. An example of from the pyrrolopyridine series, 7,59 is highlighted in Table
2.
Compound 7 is a potent panꢀTrk inhibitor, with an improved lipophilic efficiency of LipE 4.9
due to a reduced logD relative to previous analogues. Compound 7 has low metabolic
turnover in HLM and hHep, is a good substrate for both the efflux transporters Pꢀgp (ER =
29) and BCRP (ER = 13), and has moderate passive permeability (RRCK Papp = 11.3 x 10ꢀ6
cmsꢀ1). The aqueous solubility of 7 is 7.5 µM. Compound 7 has an exquisitely selective
kinase selectivity profile, with a Gini score of 0.96 (Figure 12). However, the introduction of
the weakly basic pyridyl atom (pKa 6.5) as part of the pyrrolopyridyl ring engenders
increased hERG activity (functional hERG IC50 = 1.6 µM). Incubation of 7 in human S9 (no
20
ACS Paragon Plus Environment

Page 21 of 45
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
added NADP) showed no evidence of AOꢀmediated oxidation on the pyrrolopyridine hinge
binding group, and a low level of metabolism of the central pyridine substituent (HPLCꢀUV
chromatograms are available in supplementary information S6). Interestingly, AOꢀmediated
9
metabolism on the central pyridyl ring is ablated if the
Nꢀatom is located at the 2’ꢀ or 4’ꢀ
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
position (Table 2), in these cases however compounds are subject to chemical instability (if
the nitrogen atom is located at the 2’ꢀ position) or significant GSK3β kinase activity (if the
nitrogen atom is located at the 4’ꢀ position). Hence, the 3’ꢀpyridyl ring was established as the
best isomer with which to continue the lead optimization process.
Compound
5
6
7
Trk A/B/C IC50 62/42/ND
(nM)
73/8/ND
13/7/5
MW/logD
LipE
424/2.6
4.6
439/2.5
4.6
456/3.0
4.9
HLM
< 9
22
< 8
(µL/min/mg)
hHeps
13
7
12
(µL/min/mill)
Dof Ki/hERG 7.1/ND
IC50 (µM)
9.7/ND
1.4/1.6
Pꢀgp/ BCRP ER 31/ND
27/ND
12.1
29/13
11.3
RRCK AꢀB Papp 19.1
(x 10ꢀ6 cmsꢀ1)
Solubility (µM)
Gini score
5.2
4.4
7.5
ND
ND
0.96
Table 2: Structure and properties of compounds 5 to 7. ND = not determined
21
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 22 of 45
1
2
3
4
5
6
7
8
9
The pharmacokinetic and brain penetration profile of 7 was assessed in vivo. Intravenous
(i.v.) and oral (p.o.) dosing in rats was conducted at 1 mg/kg and 3 mg/kg respectively.
Compound 7 has a high systemic CL (blood CL 50 mL/min/kg, ~ 70% hepatic blood flow),
moderate volume of distribution (Vss 1.5 L/kg), short halfꢀlife (2.3 h), and a relatively low
bioavailability (F% 16). The calculated fraction absorbed (0.55) was however encouraging,
suggesting that active transport may not significantly impede oral absorption in this series.
To determine the extent of peripheral restriction, 7 was administered orally at 100 mg/kg,
with terminal plasma and brain samples taken 3 hours post dose. This study demonstrated a
very low unbound brain/plasma concentration ratio (Cb,u /Cp,u) of 0.0083. In addition, Trk
receptor occupancies in brain lysates were assessed using the ActivX biotinylated acyl
phosphate ADP probe. No measurable TrkB or TrkC occupancy was detected.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
The ability of 7 to reverse pain responses in vivo was assessed in an ultraviolet (UV)
irradiation induced hyperalgesia (UVIH) model of inflammatory pain (see supplementary
information, S7). Analysis of the associated plasma exposure of 7 indicated that statistically
significant efficacy was observed at unbound plasma concentrations of ≥ 1 x TrkA IC50 with
a maximal efficacy response at unbound plasma concentration of ~ 10 x TrkA IC50.
Compound 7 was well tolerated in a 7ꢀday rodent toxicology study up to the highest dose
tested of 400 mg/kg/day. Thus we demonstrate that 7 is a potent and selective inhibitor of
Trk tyrosine kinases, is orally absorbed and highly peripherally restricted, exhibits robust
antinociceptive effects in the UVIH model of inflammatory pain, and is well tolerated in vivo
.
Whilst the hERG liability of 7 was not optimal, the promising data generated on 7 enabled
the project team to enter candidateꢀseeking mode.
Candidate-seeking
22
ACS Paragon Plus Environment

Page 23 of 45
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
As outlined above, the AOꢀliability of the pyrrolopyrimidine series can be mitigated by
replacing the pyrimidyl “head group” with a pyridyl motif, albeit with the introduction of
hERG activity. In order to attenuate the hERG liability associated with the pyrrolopyridine
series, the basicity of the pyridyl group was modulated via the addition of a fluorine atom at
the meta position. This pKaꢀlowering strategy delivered compounds with comparable Trk
potency (albeit with lower LipE), metabolic stability and kinase selectivity, and with an
improved hERG liability profile relative to pyrrolopyridine analogues. Table 3 highlights
example fluoroꢀpyrrolopyridine 1059 and comparator pyrimidine and pyridine analogues 838
and 9.59 MetID analysis of the fluoropyridyl ring of 2ꢀ(4ꢀchlorophenyl)ꢀNꢀ(2ꢀ(7ꢀfluoroꢀ1ꢀ
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
isopropylꢀ1Hꢀpyrrolo[3,2ꢀc]pyridineꢀ3ꢀcarbonyl)pyridinꢀ4ꢀyl)acetamide (PFꢀ06287755,59
a
structurally similar analogue of fluoropyridyl 10), in human S9 in the absence of NADP,
confirmed no AO oxidation at the fluoroꢀpyrrolopyridine motif (see supplementary
information S8).
An alternative strategy to remove AOꢀliability is to add a “blocking group” at the
metabolically labile position. In the pyrrolopyrimidine series, an ꢀNH2 group was assessed as
a blocking group at the labile 2ꢀposition. This strategy delivered compounds with
comparable or improved potency, LipE, and hERG profiles. MetID studies with human S9 in
the absence of NADP confirmed no AOꢀmediated oxidation on the 2ꢀaminopyrimidyl ring.
Table 1 highlights example 2ꢀaminopyrimidine, compound 11.38
Compound
8
9
10
11
23
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 24 of 45
1
2
3
4
AO liability on Yes
head group
No
No
No
5
6
7
Trk A/B/C IC50 18/4/ND
(nM)
17/5/ND
33/7/ND
8/5/4
8
9
MW/logD
LipE
429/2.5
5.2
428/2.7
5.0
446/3.5
4.0
444/2.5
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
5.6
8
HLM
25
35
37
(µL/min/mg)
hHeps
12
23
ND
4.8
24
(µL/min/mill)
Head
Group 4.4
6.6
4.4
pKa
Dof Ki (µM)
11.9
1.4
11.3
38.9
Pꢀgp/ BCRP ER 5.8/38.0
12.5/ND
19.8
3.8/ND
15.4
8.7/13.6
15.8
RRCK AꢀB Papp 18.4
(x 10ꢀ6 cmsꢀ1)
Solubility (µM)
Gini score
ND
ND
ND
ND
30
0.85
0.82
0.91
Table 3: Structure and property comparison of pyrimidine, pyridine, fluoroꢀpyrrolopyridine and 2ꢀ
aminopyrimidine containing Trk ligands. ND = not determined
Interestingly, the kinase selectivity selective profile of
2ꢀaminopyrrolopyrimidine 11 is
comparable to desꢀamino pyrrolopyrimidine 8 (Table 2), even though there is the potential for
an additional Hꢀbond with the conserved hinge region. However, as exemplified in Figure
10, the ꢀNH2 group of the
2ꢀaminopyrrolopyrimidine hinge binder makes an imperfect
hydrogen bonding interaction with the backbone carbonyl of hinge residue Met592. Due to
the subꢀoptimal geometry of the donor–acceptorꢀdonor arrangement (O…HꢀN bond angle is
123° cf. an ideal O…HꢀN bond angle of 180°) the additional hydrogen bond may not
significantly contribute to the free energy of binding, and hence not erode kinase selectivity.60
24
ACS Paragon Plus Environment

Page 25 of 45
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 10: Hydrogen bond interactions of 2ꢀaminopyrrolopyrimidine group with the TrkA hinge binding region
(blue). The red circle highlights the nonꢀoptimal bond angle between backbone carbonyl A592 and the ligand –
NH2.
Overall, the 2ꢀaminopyrrolopyrimidine head group conferred the best balance of properties of
the hinge binding groups assessed although the aqueous solubility was still relatively low.
Compound 12,38 an exemplar
2
ꢀaminopyrrolopyrimidine (Table 4), is a potent and selective
Trk kinase inhibitor with a high Pꢀgp ER and minimal hERG liability. As with other 2ꢀ
aminopyrrolopyrimidines, there is no AO oxidation on the aminopyrrolopyrimidine ring. The
logD is high (4.2), likely contributing to the relatively poor solubility of 12 µM. In order to
design 2ꢀaminopyrrolopyrimidine analogues with enhanced solubility and an improved LipE
profile, compounds of reduced logD were required. An analysis of available Trk coꢀcrystal
ꢀiPr motif of
2ꢀ
structures suggested polar groups would be tolerated at the
N
aminopyrrolopyrimidine compounds such as 12, as this region of the binding site is highly
ꢀiPr motif
solvent accessible. Addition of a solubilising hydroxymethylene group to
N
generated compounds such as 1338 that retained Trk potency, increased LipE due to a lower
logD, and effected an improved solubility profile. In addition, replacement of the
chlorophenyl ring of with a lower logD pyridyl motif delivered ligands such as 14 which had
further optimised LipE and solubility profiles.
25
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 26 of 45
1
2
3
4
5
Compound
12
13
14
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Trk A/B/C IC 6/2/1
(nM)
3/2/2
6/4/3
MW/logD
LipE
448/4.2
478/3.7
4.8
479/2.7
5.5
4.0
HLM
< 8.0
< 8.0
< 8.0
(µL/min/mg)
hHeps
14.0
11.2
4.0
(µL/min/mill)
Dof
Ki/hERG 16.9/ ND
> 4.3/ND
> 40.3/> 30.0
IC50 (µM)
Pꢀgp/ BCRP ER
10.9/ND
> 34.0/ND
6.4
35.7/4.0
5.4
RRCK AꢀB Papp 5.2
(x 10ꢀ6 cmsꢀ1)
Solubility (µM)
Gini score
12
29
131
0.93
0.70
0.92
Table 4: Trk pharmacology and in vitro ADME profiles of 2ꢀaminopyrimidine compounds 12, 13 and 14.
As highlighted in Table 4, 14 is a highly potent panꢀTrk inhibitor, with an excellent LipE
profile. Compound 14 has low metabolic turnover in HLM and hHep, is a good substrate for
efflux transporters Pꢀgp (ER = 35.7) and BCRP (ER = 4.0), and has moderate passive
permeability (RRCK Papp = 5.4 x 10ꢀ6 cmsꢀ1). Its oxidation by AO on the amino
pyrrolopyrimidine substituent in human S9 is minimal (see supplementary information S9),
and only a low level of metabolism by AO is detected on the central pyridyl ring. The
aqueous solubility of 14 is 131 µM, much improved over previous analogues, it is highly
kinaseꢀselective (Gini score of 0.92), and has no measurable activity at the hERG channel.
26
ACS Paragon Plus Environment

Page 27 of 45
Journal of Medicinal Chemistry
1
2
3
4
Compound 14 was profiled in a series of in vitro safety assays, showing little cytotoxicity in
5
6
THLE or HepG2 cell lines (IC50 > 42 µM and > 300 µM, respectively), and was evaluated for
7
8
9
broader pharmacological activity in a panel of receptors, ion channels and enzymes. In this
broad panel, all IC50/Ki values were > 10 ꢁM except for COXꢀ1 (IC50 = 2.7 ꢁM) and
dopamine transporter assays (Ki = 5.2 ꢁM) and PDEs 4D, 5A, 7B, 8B and 11 (54 ꢀ 89%
inhibition at 10 ꢁM). Compound 14 was screened in the Invitrogen wide kinase panel of 309
kinases, all were inhibited by < 40% when tested at 1 ꢁM except the following: MUSK (IC50
53nM), FLTꢀ3 (IC50 395 nM), IRAK1 (IC50 2.5 ꢁM), MKK (90% @ 1 ꢁM) and DDR1 (60%
@ 1 ꢁM).
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 11 shows the crystal structure of 14 bound to TrkA kinase, and highlights key proteinꢀ
ligand interactions. Compound 14 adopts a DFGꢀout binding mode, with the 2ꢀamino
pyrrolopyrimidine hinge binding motif making a hydrogen bonding interaction to the
backbone NꢀH of Met592 via a pyrimidyl nitrogen and with a “through water” hydrogen
bond network between main chain carbonyl Glu590 and the ketone oxygen atom of 14. An
additional hydrogen bond exists between the ꢀNH2 group of the 2ꢀaminopyrrolopyrimidine
hinge binder and the backbone carbonyl of hinge residue A592, but with subꢀoptimal
geometry (O…HꢀN bond angle, 122°). The central pyridyl group of 14 makes a πꢀstacking
interaction with the gatekeeper (Phe589), and Phe669 of the DFG triad. The carbonyl oxygen
of the amide group makes a hydrogen bond interaction with the backbone NꢀH of Asp 668
from the DFG triad, and the amide NꢀH group makes a hydrogen bond to a conserved water
molecule which further interacts with catalytic lysine (Lys544) and Glu560 from the alphaꢀC
helix. The lipophilic back pocket accommodates the chloropyridyl group.
27
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 28 of 45
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 11: Crystal structure of 14 (green) bound to TrkA (pink). The hinge region of the TrkA ATP site is
highlighted in blue. The “DFG” triad of the activation loop of TrkA is highlighted in orange. Key proteinꢀ
ligand binding interactions are outlined.
Figure 12: Heat map representation of the Invitrogen selectivity panel data of 39 kinases for 1, 3, 4, 7 and 14, %I
values generated at a compound concentration of 1 µM. 0% I = green, 50% I = yellow, 100% I = red
Brain availability experiments with 14 in rat confirmed significant peripheral restriction, with
Cb,u /Cp,u = 0.026. In addition, no measurable TrkB or TrkC occupancy was detected from
brain lysates using the ActivX biotinylated acyl phosphate ADP probe. Compound 14 was
assessed in the UVIH model of inflammatory pain in rodents. In this study, statistically
significant efficacy was observed at unbound plasma concentrations of ≥ 1 x TrkA IC50, with
maximal efficacy response at an unbound plasma concentrations of ~ 10 x TrkA IC50 (see
28
ACS Paragon Plus Environment

Page 29 of 45
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
supplementary information, S10). The safety profile of 14 was assessed in vivo. In rats,
decreases in white blood cell count were observed from 150 mg/kg/day. At doses ≥ 250
mg/kg, increases in body weight gain and food consumption were observed, effects that could
be rationalized as being mediated by central inhibition of TrkB, agonists of which are known
to be anorexigenic in rodents. Adaptive changes in the liver were observed microscopically
and accompanied by increased liver weight at ≥ 250 mg/kg and increased cholesterol at 1000
mg/kg. Overall 14 was well tolerated up to 1000 mg/kg/day where plasma exposure
(unbound Cavg) was approximately 400 x TrkA IC50.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
In summary, 14 is a highly potent, kinaseꢀselective, peripherally restricted, efficacious and
wellꢀtolerated panꢀTrk inhibitor and was selected as a candidate for clinical development.
Compound Synthesis
The synthetic route towards 14 is described in Scheme 1.38 Iodination of 15 with NIS gave
monoꢀiodinated 16 in good yield. Compound 16 was subsequently alkylated with methyl 2ꢀ
bromoꢀ2ꢀmethylpropanoate using cesium carbonate as base to afford 17. Ester hydrolysis of
17 generated 18 in good yield, which was then reduced to give 19. The reduction step to
afford 19 proved challenging and a range of conditions were examined with DIBALꢀH and
superꢀhydride proving most effective. DIBALꢀH was utilised as the reducing agent upon
scaleꢀup. A subsequent protection of the resulting primary alcohol with tꢀbutyldimethylsilyl
chloride and imidazole in DMF (0°C to rt, 16 hr) delivered 20, the iodide coupling partner
required for the subsequent Weinreb ketone formation, in 96% yield. The additional
coupling partner for the ketone formation step, Weinreb amide 22, was synthesized via the
palladium mediated crossꢀcoupling of 21 and benzophenone imine. To elicit the Weinreb
ketone formation, isopropyl magnesium chloride was added to 20 in THF at 0°C. After
stirring at 0°C for 1 hr a solution of Weinreb amide 22 in THF was added, and the resulting
29
ACS Paragon Plus Environment