Discovery of (10 R)-7-Amino-12-fluoro-2,10,16-trimethyl-15-oxo-10,15,16,17- tetrahydro- 2H -8,4-(metheno)pyrazolo[4,3- h ][2,5,11]- benzoxadiazacyclotetradecine-3-carbonitrile (PF-06463922), a macrocyclic inhibitor of anaplastic lymphoma kinase (ALK) and c-ros oncogene 1 (ROS1) with preclinical brain exposure and broad-spectrum potency against ALK-resistant mutations

Article abstract of DOI:10.1021/jm500261q

Although crizotinib demonstrates robust efficacy in anaplastic lymphoma kinase (ALK)-positive non-small-cell lung carcinoma patients, progression during treatment eventually develops. Resistant patient samples revealed a variety of point mutations in the kinase domain of ALK, including the L1196M gatekeeper mutation. In addition, some patients progress due to cancer metastasis in the brain. Using structure-based drug design, lipophilic efficiency, and physical-property-based optimization, highly potent macrocyclic ALK inhibitors were prepared with good absorption, distribution, metabolism, and excretion (ADME), low propensity for p-glycoprotein 1-mediated efflux, and good passive permeability. These structurally unusual macrocyclic inhibitors were potent against wild-type ALK and clinically reported ALK kinase domain mutations. Significant synthetic challenges were overcome, utilizing novel transformations to enable the use of these macrocycles in drug discovery paradigms. This work led to the discovery of 8k (PF-06463922), combining broad-spectrum potency, central nervous system ADME, and a high degree of kinase selectivity.

Full text of DOI:10.1021/jm500261q

Subscriber access provided by Maastricht University Library
Article
Discovery of (10R)-7-amino-12-fluoro-2,10,16-trimethyl-15-
oxo-10,15,16,17-tetrahydro-2H-8,4-(metheno)pyrazolo[4,3-h]
[
2,5,11]benzoxadiazacyclotetradecine-3-carbonitrile (PF-06463922), a
macrocyclic inhibitor of ALK/ROS1 with pre-clinical brain exposure
and broad spectrum potency against ALK-resistant mutations
Ted W Johnson, Paul F Richardson, Simon Bailey, Alexei Brooun, Benjamin J. Burke, Michael R. Collins,
J Jean Cui, Judith G Deal, Ya-Li Deng, Dac M. Dinh, Lars D. Engstrom, Mingying He, Jacqui E. Hoffman,
Robert L. Hoffman, Qinhua Huang, John Kath, Robert S Kania, Hieu Lam, Justine L. Lam, Phuong T. Le,
Laura Lingardo, Wei Liu, Michele A McTigue, Cynthia L. Palmer, Neal W Sach, Tod Smeal, Graham L. Smith,
Albert E. Stewart, Sergei L Timofeevski, Huichun Zhu, Jinjiang Zhu, Helen Y Zou, and Martin P. Edwards
J. Med. Chem., Just Accepted Manuscript • Publication Date (Web): 13 May 2014
Downloaded from http://pubs.acs.org on May 25, 2014
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 82
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Zou, Helen; Pfizer,
Edwards, Martin; Pfizer,
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
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 2 of 82
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
Discovery of (10R)ꢀ7ꢀaminoꢀ12ꢀfluoroꢀ2,10,16ꢀ
trimethylꢀ15ꢀoxoꢀ10,15,16,17ꢀtetrahydroꢀ2Hꢀ8,4ꢀ
(metheno)pyrazolo[4,3ꢀ
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
h][2,5,11]benzoxadiazacyclotetradecineꢀ3ꢀ
carbonitrile (PFꢀ06463922), a macrocyclic inhibitor
of ALK/ROS1 with preꢀclinical brain exposure and
broad spectrum potency against ALKꢀresistant
mutations
Ted W. Johnson,* Paul F. Richardson,* Simon Bailey, Alexei Brooun, Benjamin J. Burke,
Michael R. Collins, J. Jean Cui, Judith G. Deal, Ya-Li Deng, Dac Dinh, Lars D. Engstrom,
Mingying He, Jacqui Hoffman, Robert L. Hoffman, Qinhua Huang, Robert S. Kania, John C.
Kath, Hieu Lam, Justine L. Lam, Phuong T. Le, Laura Lingardo, Wei Liu, Michele McTigue,
Cynthia L. Palmer, Neal W. Sach, Tod Smeal, Graham L. Smith, Albert E. Stewart, Sergei
Timofeevski, Huichun Zhu, Jinjiang Zhu, Helen Y. Zou, Martin P. Edwards
1
ACS Paragon Plus Environment

Page 3 of 82
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
La Jolla Laboratories, Pfizer Worldwide Research and Development, 10770 Science Center
Drive, San Diego, California, 92121, United States.
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
ABSTRACT
Although crizotinib demonstrates robust efficacy in ALK positive NSCLC patients, progression
during treatment eventually develops. Resistant patient samples revealed a variety of point
mutations in the kinase domain of ALK, including the L1196M gatekeeper mutation. In addition,
some patients progress due to cancer metastasis in the brain. Using structure based drug design,
LipE, and physical property based optimization, highly potent macrocyclic ALK inhibitors were
prepared with good ADME, low propensity for Pgpꢀmediated efflux and good passive
permeability. These structurally unusual macrocyclic inhibitors were potent against wildꢀtype
ALK and clinically reported ALK kinase domain mutations. Significant synthetic challenges
were overcome, utilizing novel transformations to enable the use of these macrocycles in drug
discovery paradigms. This work led to the discovery of 8k (PFꢀ06463922), combining broad
spectrum potency, CNS ADME and a high degree of kinase selectivity.
INTRODUCTION
Anaplastic lymphoma kinase (ALK) is a member of the insulin receptor (IR) subꢀfamily of
receptor tyrosine kinases. It is primarily expressed in adult brain tissue and plays an important
1
role in the development and function of the nervous system. ALK knockout mice live a full lifeꢀ
span without obvious abnormalities, but are described as having an “antiꢀdepressive”
2
phenotype.
2
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 4 of 82
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
The ALK name was derived from the discovery of the driving role this kinase plays in
anaplastic large cell lymphoma (ALCL), a subtype of nonꢀHodgkin lymphoma, in which the
3
t(2;5)(p32;q35) chromosomal translocation involving the ALK gene was first described. The
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
finding that this translocation results in the fusion of nucleophosmin (NPM) to a truncated form
of anaplastic lymphoma kinase (ALK) led to the first suggestion that this fusion could result in a
4
constitutively active cytoplasmic tyrosine kinase that could be acting as an oncogenic driver.
Subsequently, other ALKꢀfusion proteins have been identified in diffuse large Bꢀcell lymphoma
5
6
(DLBCL), inflammatory myofibroblastic tumors (IMT) and nonꢀsmallꢀcell lung carcinoma
7
(
NSCLC). Over the last 20 years, more than a dozen fusion partners for ALK have been
identified in a variety of cancer types (typically rare childhood cancers). Like NPM, the fusion
partners replace the ALK protein domains responsible for ligand binding and crossing the
membrane with a domain that results in constitutive dimerization/oligomerization activation of
ALK tyrosine kinase in the cytoplasm. However, it was the discovery of a fusion of echinoderm
microtubule associated proteinꢀlike 4 with ALK (EML4ꢀALK) in 3ꢀ7% of NSCLC patients that
enabled the rapid clinical development and validation of ALK inhibitors as a viable, new cancer
7
c,7d
therapy.
In addition to chromosomal translocations leading to fusionꢀproteins, amplification
of ALK and activating point mutations in full length protein have been described in
8
9
10
neuroblastoma, inflammatory breast cancer and ovarian cancer patients. Similar to ALK, cꢀros
oncogene 1 (ROS1) chromosomal translocations define a unique, albeit smaller, subset of nonꢀ
11
smallꢀcell lung cancers.
1
2
Compound 1 (PFꢀ02341066, crizotinib), a MET, ALK and ROS1 kinase inhibitor at
clinically relevant drug exposures, was in early clinical trials when the EML4ꢀALK translocation
13
was identified. The marked efficacies from two clinical trials led to US Food and Drug
3
ACS Paragon Plus Environment

Page 5 of 82
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Administration (FDA) approval in 2011 based on an objective response rate of approximately
1
4
6
0% and progressionꢀfree survival (PFS) times of approximately 10 months. Xalkori was the
first ALK inhibitor approved by the US FDA as a firstꢀline treatment for ALKꢀpositive lung
cancer patients.
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 compound 1 demonstrates initial robust efficacy in ALKꢀpositive tumors, patients
eventually develop resistance. Emerging data from the analysis of patient tumor tissue collected
after progression suggests that resistance develops through point mutations in the ALK gene,
which may be as frequent as that seen in the EGFR setting (gatekeeper mutation frequency over
15
4
0%). Additionally, a common site of metastases in NSCLC patients is in the brain where
current standard of care (SOC) and recent experimental agents have only limited effectiveness,
thus identifying an area of unmet medical need. A sample taken at steady state from a single
patient at the recommended 250 mg BID dose of compound 1 showed a CSF to free plasma ratio
1
6
of 0.03. The high level of asymmetric distribution indicates poor bloodꢀbrain barrier
penetration of the drug.
Approximately 200,000 brain metastases (BM) are diagnosed annually in the United States,
accounting for 20% of cancer mortality. Across tumor types, lung cancers and breast cancers
1
7
constitute the majority of brain metastases, with the next most prevalent being melanoma.
Among the cohort of lung cancer patients with brain metastases, 10–25% of patients exhibit
brain metastases at diagnosis, another 40–50% during the course of disease, and more at
18
autopsy. Overall, patients presenting metastatic brain cancer have a poor prognosis with a
19
median survival of 2.5 months, which may be due partly to the poor bloodꢀbrainꢀbarrier (BBB)
permeability of the drugs developed for treating systemic disease. Treatment of peripheral
4
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 6 of 82
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
disease with CNS excluded drugs may extend progression free survival times, but could enhance
the proportion of patient mortality due to uncontrolled brain metastases.
Recently, several inhibitors of ALK have become a new and promising class of therapeutics
for treatmentꢀnaive and relapsed ALKꢀpositive NSCLC patients, highlighted by AP26113
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
0
21
(Ariad; undisclosed structure exemplified as 2), 3 (CH5424802, Roche/Chugai), 4 (ASP3026,
2
2
23
Astellas), and 5 (LDK378, Novartis) (Figure 1). The clinical impact on metastasis in the brain
and relation to preclinical CNS exposure or brain tumor models are not wellꢀestablished for these
compounds.
Figure 1. Structures of ALK inhibitors marketed or currently in the clinic.
RESULTS AND DISCUSSION
In vitro resistance. Compound 1 phosphoꢀALK IC values in NIHꢀ3T3 cell lines engineered
5
0
2
4
with mutations reported from the clinic have been previously described. The mutant cell IC
5
0
5
ACS Paragon Plus Environment

Page 7 of 82
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
data for compound 1 ranged from 165ꢀ3039 nM, which represented a 2 to 38ꢀfold shift relative to
wt potency (pALK cell IC 80 nM) in the same engineered line.
50
Objectives. The goal of this next generation ALK inhibitor program was to develop a small
molecule inhibitor that combined CNS exposure with broad spectrum ALK potency. A
compound with these properties has potential utility in the treatment of patients who exhibit
tumor progression, overcoming both resistance mutations and brain metastases.
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
By avoiding transporterꢀmediated efflux at the BBB and tumor cell surface, the increased CNS
availability of this next generation inhibitor could provide significant efficacy benefits. The
benefits include higher free brain and tumor intracellular concentrations for better target
modulation, critical for patients who relapse due to brain metastases or overꢀexpression of efflux
transporters in tumor cells. To mitigate cell toxicity associated with offꢀtarget activities and
improve therapeutic indices, a high degree of kinase selectivity was required to avoid potential
liabilities associated with the inhibition of closely related kinases.
Critical parameters to meet objectives: LipE and MW. Ligand interactions with most
25
proteins are substantially influenced by lipophilicity. To monitor the progress of optimization,
lipophilic efficiency (LipE = pKi (or pIC ) – log D) was used as a numerical index of binding
5
0
2
6
effectiveness per unit lipophilicity. Drug design aimed at improving LipE results in parallel
optimization for desirable ADME and greater likelihood of safe drug profiles by increasing the
ratio of potency to lipophilicity. Since log D within a series also tends to correlate with oxidative
27
clearance, increasing LipE will raise the chance for a lower dose through improvements in IC50
and clearance, assuming the introduction of structural features do not impart greater metabolic
28
liabilities.
6
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 8 of 82
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
27,29
Since molecular weight is inversely correlated with permeability within a series,
reducing
ligand size was crucial to provide the smallest, most potent ligands with the greatest chance of
possessing desired drugꢀlike properties, including CNS exposure.
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
Acyclic aminopyridine/aminopyrazine ALK inhibitors. Optimization to address emerging
2
4a,30
resistance to compound 1 was previously described.
Table 1 highlights some key
compounds from the firstꢀ and secondꢀgeneration ALK inhibitor programs. Compound 1 lacks
sufficient potency against resistant clinical mutants and shows high Pꢀglycoprotein 1 (Pgp) efflux
31
as measured by the MDR BA/AB ratio (>2.5), consistent with CNS exclusion. Structureꢀbased,
propertyꢀbased, and efficiencyꢀfocused drug design efforts culminated in broadly potent,
permeable, and metabolically stable secondꢀgeneration ALK inhibitors, highlighted by
2
4a
compound 6a (PFꢀ06439015). This compound improved wildꢀtype and gatekeeper (L1196M)
mutant potencies by 105ꢀ and 128ꢀfold relative to compound 1 and showed robust in vivo
efficacy in the same gatekeeperꢀmutant cell line. While significant improvements in potency
were achieved, the overall higher lipophilicity required introduction of polar groups to lower the
lipophilicity of the whole molecule for acceptable LipE and good metabolic stability. This added
polarity and molecular weight resulted in high efflux potential and very low rat CSF drug levels.
Other analogues were pursued that lacked excess hydrogen bond donors and polarity. Compound
6
b and 6c both had improved in vitro clearance and permeability relative to compound 1. The
MDR efflux ratios for both compounds are less than or equal to 2.5 and consistent with good
predicted brain availability in humans. Unfortunately, both wildꢀtype and L1196M ALK cellular
IC values suffered. Compound 6d also displayed a desirable, low MDR efflux ratio, but lacked
50
the required potency and in vitro clearance.
7
ACS Paragon Plus Environment

Page 9 of 82
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
The amide substituents on the phenyl ring of compounds 6e, 6f and 6h were designed to lower
lipophilicity and improve efficiency. Indeed, the amide group in 6e and 6f lowered the
lipophilicity significantly relative to the triazole 6d. Insights into the impact of the amide on
potency and efficiency were gained by highlighting matched molecular pairs. The secondary
amide 6e showed roughly a 10ꢀfold reduction in cellular potencies against both wildꢀtype and
mutant ALK and a slight improvement in LipE compared to the aminopyridine 1,2,3ꢀtriazole 6d.
Addition of a methyl group provided the tertiery amide analogue 6f, which lost more than 100ꢀ
fold potency relative to the triazole analogue 6d, presumably due to steric clash of the pyrazole
and N,Nꢀdimethylamide in the preferred bound conformation The aminopyrazine matched
molecular pair demonstrated a similar trend. Nꢀmethyl amide 6h showed a 6ꢀfold loss in mutant
cell potency but a modest gain in LipE relative to the triazole analogue 6g. Although acyclic
phenyl sulfone analogue 6i came the closest to meeting our lab objectives, it lacked the required
potency and selectivity (data not shown).
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
These acyclic compounds represented the difficulty in overlapping potency (L1196M IC <25
5
0
nM) and low MDR BA/AB ratios (<2.5) to afford the highest probability of reaching efficacious
exposures in the CNS. To overlap the required properties, new design efforts targeted neutral
inhibitors with improved lipophilic efficiency that fell within property space commensurate with
low efflux. This strategy required a novel approach, which is the focus of this manuscript.
Table 1. Potency and ADME of compound 1 and other key acyclic leads
pALK
cell IC50
ALK Ki
(nM)
MDR
BA/AB
(ratio)
(
nM)
Log
D
HLM Cl
(mL/min/kg)
b
Compound #
Structure
LipE
ALK-
a
c
d
pALK-
L1196M
cell IC50
L1196M
Ki (nM)
8
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 10 of 82
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
(
nM)
80
NH
Cl
0
.74
12.5/0.28
N
F
1
2.0 4.1
2.9 5.3
2.4 3.1
44
13
25
N
Cl
O
8.2
843
(44.5)
N
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
H2N
6
a
<0.1
0.76
6.6
30.6/2.8
(10.9)
(
PF-
6439015)
0
.2
0
4
.0
95
16.3/20.0
(0.82)
6
b
3
8
3,200
9
.0
249
27.5/11.0
(2.5)
6c
2.4 3.2
3.4 3.4
2.1 3.7
8.0
66
28
70
2,650
<0.2
14
16.5/10.9
(1.5)
6
d
e
2
.6
176
2
.8
116
12.6/0.74
(17.0)
6
3
5
1,665
22
10
18.8/2.5
(7.6)
6f
--
2.1
--
58
3
9
ACS Paragon Plus Environment

Page 11 of 82
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
<
0.2
9.7
53
5.1/3.2
(1.6)
6
g
4.0 3.3
41
0.7
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
0.85
16
33.8/7.7
(4.4)
6
h
3.0 3.5
76
10
8
.2
312
<0.2
.4
4.6
54
15.6/11.4
(1.4)
6i
3.4 3.9
1
a
b
c
Log D was measured at pH 7.4. LipE = ꢀlog(pALK L1196M cell IC ) – log D. Cl
refers to the total intrinsic
50
int,app
d
clearance obtained from scaling in vitro halfꢀlives in human liver microsomes. MDCKꢀMDR1 BA/AB efflux at 2
ꢁ
M substrate concentration and pH 7.4.
Acyclic ligands bound to the ALK kinase domain provided key structural information which
revealed protein ligand interactions and facilitated the design of ligands with improved binding
affinity and efficiency. Methyl ether 6b was one of the earliest coꢀcrystal structures obtained in
the wildꢀtype kinase domain of ALK (Figure 2, left panel). The aminopyridine core interacted
with the hinge portion of the protein via an acceptorꢀdonor motif. The benzylic fragment adopted
a conformation with the benzylic hydrogen coplanar to the methoxy substituent that minimized
Aꢀ1,3 strain. Both the benzylic methyl and fluorophenyl filled lipophilic pockets. In addition, the
aryl fluoride polarized the ortho aryl CꢀH groups, which provided close contact and electrostatic
complementarity with backbone carbonyls at Gly1269 and Arg1253. The orientation of the
triazole ring was at a 30 degree torsion with a close contact to the CH of the Gly1202 residue,
approximating a CꢀH donorꢀpi interaction. Finally, the Leu1256 residue beneath the ligand filled
the hydrophobic pocket created by the Uꢀshaped structure of the inhibitor.
1
0
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 12 of 82
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
The coꢀcrystal structure of the ALK kinase domain with amide 6h (Figure 2, right panel, cyan
structure) revealed that the amide NꢀH formed a hydrogen bond to water and the carbonyl group
of the secondary amide interacted with the carbonyl group of His1124 via a water molecule. The
amide carbonyl did not seem optimal for interaction with structural waters to Lys1150, which
would have required a more perpendicular dihedral angle. Overlay of 6h with 6i (a similar
analog having the 2ꢀtriazolꢀ2ꢀyl head group – green structure) illustrated the orientation of the
amide away from the surface of the triazole group (Figure 2, right panel). The dihedral angle of
the amide (CCCO) was 45 degrees while the angle for the triazole (CCNN) was 64 degrees,
representing an approximate 20 degree rotation difference. Consequently, the methyl group did
not point to the desired direction for a stronger hydrophobic interaction with Val1130, Gly1123
and Leu1122 in the Gꢀloop. Overall, the Gꢀloop backbone was displaced away from the Nꢀ
methyl amide 6h relative to triazole 6i.
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
Structural data in Figure 2 showcased specific interactions driving binding, but also presented
opportunities to enhance lipophilic efficiency. The bound structures of these Uꢀshaped binders
underscored the proximity of the substituted fluorophenyl head group to the heteroꢀaromatic tail,
which was directly linked to the aminopyridine or aminopyrazine hingeꢀbinding motif. Two
3
3,34
general designs were pursued to provide novel macrocyclic
templates and a means to control
conformations for further optimization. Specifically, macrocycles were formed by linking the
two groups from the triazole 4ꢀcarbon to the methoxy carbon of 6b or the pyrazole carbon and
amide nitrogen of 6e and 6f. These templates were expected to reinforce the binding
conformation and provide additional protein ligand interactions in the linker region.
Additionally, they significantly reduced the number of rotatable bonds and were more compact
with increased buried surface area than acyclic analogues, which we postulated would provide
1
1
ACS Paragon Plus Environment

Page 13 of 82
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
permeability advantages. Furthermore, we envisioned that designing macrocyclic based
inhibitors driven by LipE and MW optimization would provide the highest probability to achieve
desirable CNS ADME properties without sacrificing potency.
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. Left panel: Compound 6b coꢀcrystal structure in ALK kinase domain (purple, PDB
4
CNH) highlighting Type I Uꢀshaped binding conformation and proximity of methoxyꢀ
fluorophenyl head group and triazole tail group. Right panel: ALK coꢀxtl of 6h (cyan, PDB
CMO) overlaid with 6i (green, PDB 4CMT) showing triazole, amide and Gꢀloop
4
conformations.
Macrocyclic ALK Inhibitors. To evaluate the macrocycle design concept, a variety of 12ꢀ to
14ꢀmembered etherꢀlinked macrocycles were prepared based on the coꢀcrystal structure analysis
of acyclic methyl ether 6b and the data is summarized in Table 2. According to matched
molecular pairs, the smaller ring sizes consistently provided the most lipophilic efficient
macrocycles. For example, cyclic ethers 7a, 7c and 7e had higher LipE than their corresponding
largerꢀring matched molecular pairs 7b, 7d and 7f. The 12ꢀmembered macrocycle 7e displayed
the highest LipE (4.4) with picomolar binding affinities and good cellular potencies (ALK IC50
1
.0 nM; ALKꢀL1196M IC 20 nM). Although some of the more efficient analogues provided
50
improved potency and lipophilic efficiency relatively to the acyclic analogue 6b, the macrocyclic
1
2
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 14 of 82
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
ethers were generally too lipophilic and still lacked the required efficiency for more facile
overlap of potency, ADME and CNS availability.
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
Table 2. Etherꢀlinked macrocycles
ALK Ki (nM)
pALK cell IC (nM)
50
Compound
#
Log
D
b
Structure
LipE
4.0
ALK-L1196M Ki
pALK-L1196M cell IC50
(nM)
a
(
nM)
0
.36
22
7
a
b
3.0
3.6
3.4
3.1
3.3
3.8
1
.6
101
0
.22
0.1
15
7
3.4
2.8
2.3
4.4
3.9
<
112
3
.8
86
7
c
2
9
654
5
524
7
d
e
3
6
3,655
<
0
0.1
.57
1.0
20
7
<
0.1
0.9
21
7
f
0.62
1
3
ACS Paragon Plus Environment

Page 15 of 82
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
a
b
Log D was measured at pH 7.4. LipE = ꢀlog(pALK L1196M cell IC ) – log D.
50
The second template followed a design strategy to harness the increased LipE of the 12ꢀ
membered ether macrocycles while lowering lipophilicity with a lactam linker. Modeling (Figure
3) suggested a single carbon atom linking the amide with the proximal pyrazole moiety leading
to a proposed 12ꢀmembered lactam that would have minimal binding strain (see modeling
methods section). In addition, the Nꢀmethyl group may provide close contact with Val1130,
Leu1122, and Gly1123 of the Gꢀloop and the lactam carbonyl oxygen may act through a
structural water to stabilize Lys1150 (Figure 3). Importantly, the amide macrocycles would
significantly lower lipophilicity, placing them in optimal log D space.
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
A set of amideꢀlinked macrocycles was prepared to test the hypothesis. The coꢀcrystal structure
of the ALK kinase domain with the cyclic amide 8a supported the design hypothesis and
molecular modeling, superimposing well with the original docking result (Figure 3, right panel).
As expected, the Nꢀmethyl group in 8a was pulled tighter toward the Gꢀloop by 1.2 Å from the
open amide position, and therefore had a closer contact with the carbonyl group of Leu1122 (3.4
Å) and nearby side chains of Leu1122 (4.1 Å), Gly1123 (4.7 Å), and Val1130 (4.7 Å). The
amide carbonyl of macrocycle 8a formed a water bridge with Lys1150, and a second water
bridge to His1124. It was expected that interaction of the lactam carbonyl oxygen with waters in
the bound state may effectively lower binding desolvation penalties.
1
4
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 16 of 82
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 3. Left panel: Acyclic amide 6h coꢀcrystal structure in ALK (cyan, PDB 4CMO) overlaid
with modeled macrocycle 8a (purple) in ALK (see modeling methods section). Right panel: 8a
coꢀcrystal structure in ALKꢀ kinase domain (green, PDB 4CMU) overlaid with modeled ligand
(purple).
Improvement in key parameters was realized upon cyclization to the amide macrocyle 8a
Table 3) from acyclic analogues 6d-f (Table 1). Compound 8a improved cellular potencies
(
against ALK and L1196M ALK by 10ꢀfold relative to compound 6d and also lowered
lipophilicity by one unit, without additional molecular weight. As a result, macrocycle 8a had
more than two units higher LipE. Relative to 6e, compound 8a improved cellular potency against
the gatekeeper mutant protein by 119ꢀfold while maintaining log D, which also translated to a
two unit increase in LipE. In addition, molecular weight was low. Improvments in potency and
efficiency were even more pronounced relative to the closest matched pair 6f. A reduction of the
in vitro clearance from 56 mL/min/kg (high) to the lower limit of detection for this assay
demonstrated the inherent metabolic stability of the cyclic structure. Finally, the permeability of
8a, measured by Pgp efflux potential, was superior to the acyclic analogue 6f. Overall, there was
a dramatic and compelling improvement in overall properties for 8a relative to 6f. These results
clearly positioned the macrocyclic amides as an attractive series for the generation of ALK
1
5
ACS Paragon Plus Environment

Page 17 of 82
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
inhibitors capable of brain penetration. The higher LipE and lower MW allowed navigation to
desired CNS ADME space without sacrificing potency and safety.
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 3. Impact of macrocyclization, benzylic methyl, aminopyrazine vs. pyridine and different
tail groups on potency and efficiency
pALK
ALK Ki
(nM)
cell IC50
(nM)
MDR
BA/AB
ratio)
Compound
#
Log
D
HLM Cl
(mL/min/kg)
b
Structure
LipE
ALK-
pALK-
a
c
d
(
L1196M L1196M
Ki (nM) cell IC50
(
nM)
<
0.2
.29
0.70
14
28.3/8.1
(4.2)
8
a
b
2.2 5.7
8.6
8.0
8.4
8.0
0
1
7
--
--
34.6/5.8
(6.0)
8
2.2
--
61
<
0.2
6.4
97
25.4/4.6
(5.5)
8c
2.0 5.0
2.5 6.4
2.5 3.4
1
.9
<0.2
0.21
1.4
26.5/11.9
(2.2)
8d
0
.10
17
156
29.6/9.6
(3.1)
8
e
9.7
6
1
1,230
1
6
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 18 of 82
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
<0.2
.10
0.18
2.1
19.9/7.2
(2.8)
8
f
2.4 6.3
8.3
0
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
<
<
0.08
0.02
0.11
0.75
40.3/20.1
(2.0)
8
g
2.7 6.4
23.5
O
F
N
<
0.07
1.3
21
22.1/2.4
(9.2)
8
h
1.8 5.9
9.1
8.0
O
0.40
S
O
O
H
2
N
N
<
0.1
.12
0.34
1.9
8.0/1.2
(6.7)
8
i
2.2 6.5
0
a
b
c
Log D was measured at pH 7.4. LipE = ꢀlog(pALK L1196M cell IC ) – log D. Cl refers to the total intrinsic
int,app
50
d
clearance obtained from scaling in vitro halfꢀlives in human liver microsomes. MDCKꢀMDR1 BA/AB efflux at 2
ꢁ
M substrate concentration and pH 7.4.
Table 3 highlights important ALK protein sensitivities to subtle structural changes within the
amide macrocycle series. The stereochemical sensitivity was determined by the eutomer
distomer ratio. Both 8a and 8d improved potency relative to their enantiomeric pairs 8b and 8e
by 210 and 610ꢀfold, respectively, based on gatekeeper mutant biochemical potencies. The desꢀ
methyl analogue, 8c, lost potency and LipE relative to the more potent methyl analogue 8a.
Three matched molecular pairs highlighted the tolerance of the aminopyrazine core. Several
examples (8a/8d and 8f/8g and 8h/8i) showed an improvement in potency for the aminopyrazine
analogues, but the LipE improvement was attenuated as this change was accompanied by an
increase, rather than a decrease, in lipophilicity. The additional buried nitrogen atom was not
1
7
ACS Paragon Plus Environment

Page 19 of 82
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
wellꢀsolvated and also decreased the basicity of the aminopyrazine leading to an increase in
lipophilicity. Table 3 also highlights the tolerability of both substituted 5ꢀ and 6ꢀmembered
aromatics adjacent to the aminopyridine or aminopyrazine cores. The methanesulfonylꢀbenzene
substituent on the aminopyrazine inhibitor 8i was one of the most lipophilic efficient analogues
(L1196M cell IC LipE 6.5). A comparison of sulfones 8h and 8i with the dimethylpyrazoles 8a
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
0
and 8d revealed that both inhibitors possess almost identical potencies. The increased LipE of
the sulfone is partly due to the polarity of the sulfone exposed to solvent, which maintained
potency and lowered log D without incurring additional desolvation penalties.
In addition to providing potent inhibtors of ALK within the desired range of log D (2ꢀ3), the
amide macrocycles generally diplayed low clearance and low MDR BA/AB efflux ratios. With
the initial encouraging results, lead optimization on the macrocyclic amide 2ꢀ
aminopyridine/pyrazine series was carried out to achieve the best overall balance of potency,
CNS availability, ADME and selectivity in a single compound.
Designing for selectivity. There are 27 residues in the ATP binding pocket of ALK and to
gain selectivity against other kinases, residues were targeted that differed from ALK. The ALK
Leu1198 residue is conserved in 26% of the kinome at this position and is most often Phe or Tyr
in other kinases (Figure 4). This smaller Leu residue is potentially an avenue to gain selectivity
against a majority of kinases (60%) by building into this pocket to bump into the larger Phe and
Tyr residues.
As a surrogate for kinases containing a Phe/Tyr residue at the position corresponding to ALK
3
4
Leu1198, TrkB (PDB 4AT3) and ALK proteins were aligned by superposition based on a set of
residues in the active site. Several macrocyclic amides were prepared and ALK coꢀcrystal xꢀray
structures were solved for 8a, 8j, 8k and 8m (Figure 4). The closest distance between ALK coꢀ
1
8
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 20 of 82
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
crystallized compounds and the Tyr635 atoms of TrkB was measured. Subꢀvan der Waals
distances reflect the need for protein or ligand movement to accommodate the ALK compounds
in kinases with Phe/Tyr residues at this selectivity position.
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. Left panel: 8a and ALK bound structure (green, PDB 4CMU) and TrkB (purple, PDB
4AT3) with selectivity residues highlighted. Right panel: Ligands from ALK coꢀcrystal
structures aligned with TrkB demonstrated the potential for clash between ligand and Tyr635; 8a
(
(
Me, green, PDB 4CMU), 8j (cyPr, yellow, PDB 4CTC), 8k (CN, cyan, PDB 4CLI), 8m
orange, PDB 4CTB). Distances shown are between 8a methyl carbon (shown as a ball) and
TrkB Tyr635. PDB atom names for the terminal atoms are shown on Tyr635 (OH, CZ, CE).
Substituents at the ortho position of the pyrazole in ALK structures can approach atoms in the
TrkB Tyr635 residue at distances below the sum of van der Waals radii for the pair of atoms.
The methyl on the pyrazole group of 8a in the ALK structure begins to reach the contact distance
to TrkB Tyr635 since it is 3.2ꢀ4.1 Å away from the three terminal atoms of the Tyr (terminal OH
group (OH), terminal carbon (CZ), and adjacent carbon (CE), see Figure 4, right panel) in the
aligned TrkB structure. The cyclopropyl group on 8j had similar distances ranging from 3.2 to
3.7 Å away. The cyano group on 8k pointed toward the terminal carbon (CZ) of Tyr635 and was
1
9
ACS Paragon Plus Environment

Page 21 of 82
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
2
.1ꢀ3.1 Å from the three terminal Tyr atoms. The carbon of the cyano group in 8m would
approximate the position of the terminal methyl of 8l which is estimated to be 2.0ꢀ2.3 Å from
Tyr635. The nitrile of compound 8m was measured to be 1.0ꢀ1.8 Å from the terminal two atoms
indicating a greater potential clash than with smaller substituents. Further, the macrocycle
restricts rotation and the groups are unable to rotate away from the Tyr/Phe residues as might be
hypothesized with acyclic compounds.
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
Macrocyclic amides 8a and 8j-m were tested against TrkB (Table 4). Both the pyrazoles
substituted at C3 with methyl and cyclopropyl (8a and 8j, respectively) retained picomolar
enzymatic potencies against TrkB and provided little to no selectivity based on ALKꢀL1196M Ki
(approximately 2ꢀfold). However, the cyanoꢀpyrazole 8k and the methyl substituted imidazoꢀ
pyrimidine 8l displayed significantly attenuated potency in TrkB and provided reasonable
selectivity ratios (approximately 40ꢀfold). Finally, the cyanoꢀimidazopyridine 8m afforded the
most robust selectivity ratio (93ꢀfold). This selectivity data was consistent with our previous
modeling analysis.
Table 4. Potency, ADME and selectivity of macrocyclic analogues
pALK-
MDR
ALK-
L1196M
Ki (nM)
TrkB Ki
(nM)
(selectivity)
L1196M Log HLM
Compound #
Structure
a
b
BA/AB
(ratio)
cell IC50
D
Clint,app
d
c
(
nM)
2
8.3/8.1
(4.2)
0.5
8
a
0.29
14
2.2
8.6
(1.7x)
2
0
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 22 of 82
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
1
6.3/8.0
(2.0)
0.4
8j
<0.1
0.70
2.0
5.8
21
2.9
2.3
1.7
14.6
(2.0x)
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
8
k
2
3
2
3
8.0/19.3
<8
(
PF-
6463922)
e
(
1.5)
(38x)
0
77
2.4/1.74
20.2)
8
l
365
<8
(
(
39x)
6
5
2
2.1/3.8
5.8)
8
m
0.56
45
2.4
<8
(93x)
(
a
b
Log D was measured at pH 7.4. Cl
refers to the total intrinsic clearance obtained from scaling in vitro halfꢀ
lives in human liver microsomes. MDCKꢀMDR1 BA/AB efflux at 2 ꢁM substrate concentration and pH 7.4. TrkB
Ki and ratio relative to ALKꢀL1196M Ki. RRCKꢀBCRP BA/AB is 37/28 (1.8); determined at 2 ꢁM substrate
int,app
c
d
e
concentration and pH 7.4.
The pyrazole cyano moiety in 8k was efficient for obtaining selectivity because the cyano
contains only one more heavy atom than the unselective methyl analogue (8a) and gains >35x
selectivity, similar to the bulkier methylꢀsubstituted imidazoꢀpyrimidine (8l). It is hypothesized
that the nitrile makes an unfavorable contact with the CE atom of the Tyr635 in TrkB, closer to
the midꢀpoint of the side chain rather than at the terminus (Figure 4, right panel). Unfavorable
desolvation penalties or electrostatics due to the proximity of the electronꢀrich nitrile nitrogen
atom and tyrosine may further enhance selectivity.
Designing for CNS availability. Previous publications highlighted the importance of
minimizing efflux in Pgp overꢀexpressing cell lines (MDCKꢀMDR1) to maximize both CSF to
2
1
ACS Paragon Plus Environment

Page 23 of 82
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
35
free plasma ratios and free brain to free plasma ratios. Compounds that exhibited MDR BA/AB
efflux ratios (ER) of less than 2.5 did not demonstrate significant BBB efflux in vivo and have
higher probability for achieving pharmacologically relevant brain to plasma ratios. To elucidate
the physicochemical properties that provide low MDR efflux, we performed an analysis of the
Pfizer MDR data set that contains greater than 115,000 compounds evaluated in the same MDR
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
3
6
permeability model. A combination of experimental HPLCꢀbased and shakeꢀflask methods
were used to measure log D. Figure 5 depicts a subꢀset of this data that focuses within a range of
molecular weight and lipophilicity values to encompass the majority of the inhibitors in our ALK
program. Often authors discuss the influence of a single physicochemical descriptor in the
37
analysis of efflux and permeability. However, similar to the CNS MPO algorithm, the
dependence of a given ADME property is multidimensional and it is more appropriate to
27
consider several factors simultaneously.
Along with plotting binned values for molecular weight on the yꢀaxis and lipophilicity on the
xꢀaxis, we also examined further segmentation by hydrogen bond donor (HBD) count. The
independent influence of molecular weight, lipophilicity and hydrogen bond donor count was
clearly demonstrated in the outcome of binned MDR ratios performance, confirming these as
first principles. The chance of increased efflux was observed with elevated molecular weight or
HBD counts, and the chance of decreased efflux was seen with increased lipophilicity. This type
of data visualization also provided the opportunity to readily evaluate the influence of multiple
firstꢀprinciple operators and determine the tradeꢀoffs that can be utilized in compound design.
For example, the impact of increased HBD count on efflux within defined molecular weight and
lipophilicity bins was clearly demonstrated by the comparison of binned efflux ratios. Tradeꢀoffs
between physicochemical properties were observed as with the comparison of areas A and B
2
2
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 24 of 82
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
(
Figure 5) where for molecules with 2 HBD counts, similar distributions of binned efflux were
observed between groups of compounds with higher molecular weight and log D (A) and lower
molecular weight and log D (B). This, and other propertyꢀbased analyses, reinforced the
importance of molecular weight and lipophilicity efficiency indices as a means to drive to the
design of small molecule inhibitors within a focused log D range of 2ꢀ3 and with minimal
introduction of hydrogen bond donors.
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. Effect of log D, HBD count and MW on MDR BA/AB ratio (blue <2.5; red >2.5).
Percent of each sector shown and total count per bin highlighted in center of pie.
The selection of the aminopyridine as a kinase hinge binding motif provides a minimum
requirement of two HBDs and a constraint for inhibitor design. As outlined earlier, one of the
NH group hydrogen atoms provides a key interaction with the carbonyl of hingeꢀresidue
2
24a
Glu1197 within the proteinꢀligand complex.
However, the potential for the second NH2
2
3
ACS Paragon Plus Environment

Page 25 of 82
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
hydrogen to participate in an intramolecular hydrogen bond (IMHB) with the adjacent ether
oxygen may provide lower MDR ratios than other molecules with two HBDs and similar
3
8
physicochemical properties. In this program, it was observed that the presence of even a
moderately basic nitrogen when combined with elevated molecular weight and HBD count can
increase MDR ratios and limit CNS exposure, as observed with compound 1.
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
While the analysis of large data sets against physicochemical properties is a valuable method
to assess general probabilistic odds for defined outcomes, it can also furnish comparison sets for
new hypothesis generation. The pie charts of Figure 6 illustrate that macrocyclic inhibitors
display lower levels of binned MDR efflux when compared to acyclic molecules with similar
molecular weight, measured log D and HBD count. The macrocycles may afford the potential
benefits for permeability and efflux by reducing rotatable bonds and achieving a highly compact
shape for a given molecular weight. The box & whisker plots of Figure 6 illustrate that, on
3
9
average, solvent accessible surface areas (SASA) for macrocyclic molecules are almost 10%
smaller than nonꢀmacrocycles within defined molecular weight ranges. Replacing molecular
weight with SASA may provide a more accurate descriptor for molecular size and increase the
probabilistic assessment in the future design of macrocyclic inhibitors.
2
4
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 26 of 82
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 6. Top panel: Comparison of acyclic and macrocyclic compounds binned MDR ratios
(blue <2.5; red >2.5) for compounds within the same MW, log D and HBD count (2) bins.
Percent of each sector shown and total count per bin highlighted in center of pie. Bottom panel:
Acyclic and macrocyclic compound SASA differences within the same MW, log D, and HBD
count (2) bins.
The cohort of potent and low clearance compounds shown in Table 4 were also tested in vitro
for their Pgpꢀmediated efflux potential. Compounds 8a, 8l and 8m showed MDR BA/AB ratios
significantly above 2.5 and are predicted to be substrates of Pgp with the inability to adequately
cross the BBB. Both compounds 8j and 8k displayed MDR BA/AB ratios less than 2.5, and are
predicted to access the CNS. This data was consistent with the Pfizer comprehensive dataset
analysis describing the impact of MW (or SASA), log D, and HBD count on binned MDR ratios
(Figure 5).
To appreciate the efficiency differences between macrocyclic and acyclic analogues, two plots
were constructed. The LipE plot (Figure 7, left panel), with log D on the xꢀaxis and potency on
the yꢀaxis, showed that many of the macrocyclic compounds (colored aqua), some that display
2
5
ACS Paragon Plus Environment

Page 27 of 82
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
LipE values of greater than 6, outꢀperformed the acyclics in both potency and LipE. Importantly,
many macrocycles were able to access optimal log D space (2ꢀ3) coupled with improvements in
potency. Evaluation of both LipE and HACNT on the same plot (Figure 7, right panel) clearly
demonstrated that the macrocycles were able to achieve the highest lipophilic efficiency
measures in heavy atom count cohorts, indicating the economical use of each atom in exploiting
complementarity to the binding pocket of the kinase with minimal ligand strain. Consequently,
property overlap was made more facile.
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 7. Left panel: LipE plot comparing macrocyclic (aqua) and acyclic (light gray)
analogues. Right panel: LipE vs. HACNT plot with macrocyclic and acyclic analogues.
Potency, selectivity and PK of 8k (PF-06463922). Based on outstanding potency, low in
vitro clearance, good selectivity and low efflux potential, macrocycle 8k (Table 4) emerged as a
candidate for further profiling. As revealed in Table 5, 8k demonstrated significantly improved
cell activity against ALK and a large set of ALK clinical mutations compared to compound 1. In
the engineered NIHꢀ3T3 ALK wildꢀtype phosphorylation assay, the potency of 8k was improved
6
2ꢀfold (from 80 nM IC vs. wt ALK for compound 1 to 1.3 nM for 8k). In addition, it was
50
2
6
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 28 of 82
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
potent against all available clinical ALK mutants showing a 40ꢀ to 825ꢀfold potency
improvement compared to compound 1.
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
Table 5. pALK inhibition (IC ) values for compounds 1 and 8k against ALK and clinical
5
0
mutants of ALK
pALK IC (nM) in 3T3-EML4-ALK Engineered cell lines
50
Compound #
ALK F1174L C1156Y G1269A S1206Y L1196M L1152R G1202R 1151Tins
1
80
165
0.2
478
1.6
605
15
626
4.2
843
21
1026
9.0
1148
77
3039
38
8
k
1.3
To investigate kinase selectivity of 8k relative to its target kinases ALK and ROS1, it was
evaluated in biochemical kinase screening assays against a panel of 206 recombinant kinases. In
these preliminary biochemical kinase selectivity screens, only 10 offꢀtarget kinases were
identified that exhibited enzymeꢀbased activity and showed selectivity margins less than 100ꢀ
fold compared to the ALKꢀL1196M gatekeeper mutant (Table 6).
Table 6. Enzymeꢀbased selectivity of 8k
fold shift vs.
a
b
Enzyme
IC50 or Ki , nM
ALK-L1196M
<0.03
<0.10
1.0
b
ROS1
ALK
<0.02
b
<0.07
b
ALK-L1196M
LTK (TYK1)
FER
0.7
2.7
3.3
6.0
3.9
4.7
FES (FPS)
8.6
2
7
ACS Paragon Plus Environment

Page 29 of 82
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
PTK2B (FAK2)
TNK2 (ACK)
14
17
17
20
24
24
33
34
66
76
PTK2 (FAK)
b
NTRK2 (TRKB)
NTRK1 (TRKA)
NTRK3 (TRKC)
FRK (PTK5)
23
24
46
53
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
a
TM
b
IC50 values by Invitrogen Z’ꢀLYTE assays using K ꢀlevels of ATP, unless specified otherwise. Ki by offꢀchip
mobility shift assays conducted inꢀhouse.
M
Table 7 summarizes rat pharmacokinetic data. Compound 8k displayed low plasma clearance,
a moderate volume of distribution, reasonable halfꢀlife and bioavailability of 100%. This very
good preclinical PK and low protein binding provided high free plasma exposures. In addition,
CNS penetration of 8k was evaluated in rats. After a single 10 mg/kg oral dose, samples of CSF,
brain and plasma were collected at 1, 4, 7, 12 and 24 h with the PK parameters shown in Table 8.
4
0
AUC ratios of CSF/free plasma (0.31) and free brain/free plasma (0.21) suggest 8k distributes
into the CNS. The rat in vivo data was further supported by in vitro data in Table 4, which
indicated a lack of transporterꢀmediated efflux in both Pgp and breast cancer resistance protein
(
BCRP) overꢀexpressing cell lines.
In vivo efficacy data on 8k was generated in a variety of wildꢀtype and resistant ALK and
2
4bꢀd
ROS1 engineered cell lines, and was disclosed previously.
pharmacology manuscript is currently in preparation.
Additionally, a detailed
Table 7. Preclinical rat PK of 8k
c
Free
Cmax
Free
Cave
(nM)
AUC /AUC
(0-24)
CLp
Vss
a,b
Cmax
inf
Route
Tmax (hr)
T1/2 (h) F (%)
(ng/mL)
(ng*hr/mL)
(mL/min/kg) (L/kg)
(
nM)
2
8
ACS Paragon Plus Environment